05.04.2014 Views

Elektronika 2010-11.pdf - Instytut Systemów Elektronicznych ...

Elektronika 2010-11.pdf - Instytut Systemów Elektronicznych ...

Elektronika 2010-11.pdf - Instytut Systemów Elektronicznych ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

ok LI nr 11/<strong>2010</strong><br />

• MATERIAŁY • KONSTRUKCJE • UKŁADY<br />

• SYSTEMY • MIKROELEKTRONIKA<br />

• OPTOELEKTRONIKA • FOTONIKA<br />

konstrukcje technologie zastosowania<br />

MIESIECZNIK NAUKOWO-TECHNICZNY<br />

• ELEKTRONIKA MIKROFALOWA<br />

• MECHATRONIKA<br />

• ENERGOELEKTRONIKA • INFORMATYKA<br />

ZESPÓŁ REDAKCYJNY<br />

prof. dr hab. inż. Jerzy Klamka – redaktor naczelny<br />

Bożena Lachowicz – sekretarz redakcji<br />

Stali współpracownicy: mgr inż. Wiesław Jabłoński,<br />

mgr inż. Krzysztof Kowalski, mgr inż. Cezary Rudnicki<br />

Adres redakcji: ul. Chmielna 6 m.6, 00-020 Warszawa,<br />

tel./fax (022) 827 38 79; tel.: 826 65 64,<br />

e-mail: elektronika@red.pl.pl, www.elektronika.orf.pl<br />

Zamówienia na reklamę przyjmuje redakcja lub Dział Reklamy<br />

i Marketingu, ul. Mazowiecka 12, 00-950 Warszawa, skr. 1004,<br />

tel./fax (022) 827 43 66, 826 80 16, e-mail: reklama@sigma-not.pl<br />

Kolportaż: ul. Ku Wiśle 7, 00-716 Warszawa, tel. (022) 840 35 89;<br />

tel./fax: (022) 840 59 49, (022) 891 13 74<br />

RADA PROGRAMOWA<br />

prof. dr hab. inż. Władysław Torbicz (PAN) – przewodniczący<br />

prof. dr hab. inż. Leonard Bolc, dr hab. inż. Jerzy Czajkowski, prof.<br />

dr hab. Zdzisław Drozd, prof. dr hab. inż. Jerzy Frączek, dr hab inż.<br />

Krzysztof Górecki, dr inż. Józef Gromek, mgr inż. Jan Grzybowski,<br />

prof. dr hab. Ryszard Jachowicz, prof. dr hab. Włodzimierz<br />

Janke, prof. dr hab. Włodzimierz Kalita, inż. Stefan Kamiński, prof.<br />

dr hab. inż. Marian P. Kaźmierkowski, dr inż. Wojciech Kocańda,<br />

prof. dr hab. Bogdan Kosmowski, mgr inż. Zbigniew Lange,<br />

dr inż. Zygmunt Łuczyński, prof. dr hab. inż. Józef Modelski, prof. dr<br />

hab. Tadeusz Morawski, prof. dr hab. Bohdan Mroziewicz, prof. dr hab.<br />

Andrzej Napieralski, prof. dr hab. Tadeusz Pałko, prof. dr hab. inż. Marian<br />

Pasko, dr hab. inż. Ryszard Romaniuk, dr hab. inż. Grzegorz Różański,<br />

prof. dr hab. inż. Edward Sędek, prof. dr hab. Ludwik Spiralski,<br />

prof. dr hab. inż. Zdzisław Trzaska, mgr inż. Józef Wiechowski, prof.<br />

dr hab. inż. Marian Wnuk, prof. dr hab. inż. Janusz Zarębski<br />

Czasopismo dotowane przez Ministerstwo Nauki<br />

i Szkolnictwa Wyższego. Za opublikowane w nim artykuły<br />

MNiSzW przyznaje 9 punktów.<br />

SIGMA - NOT<br />

Spółka z o.o.<br />

00-950 Warszawa<br />

skrytka pocztowa 1004 ul. Ratuszowa<br />

11<br />

tel.: (0-22) 818 09 18, 818 98 32<br />

fax: (022) 619 21 87<br />

Internet<br />

http://www.sigma-not.pl<br />

Prenumerata<br />

e-mail: kolportaz@sigma-not.pl<br />

Informacje<br />

e-mail: informacja@sigma-not.pl<br />

“<strong>Elektronika</strong>” jest wydawana<br />

przy współpracy Komitetu Elektroniki<br />

i Telekomunikacji Polskiej Akademii Nauk<br />

IEEE<br />

WYDAWNICTWO<br />

CZASOPISM I KSIĄŻEK<br />

TECHNICZNYCH<br />

Redakcja współpracuje<br />

z Polską Sekcją IEEE<br />

„<strong>Elektronika</strong>” jest notowana<br />

w międzynarodowej bazie IEE<br />

Inspec<br />

Publikowane artykuły naukowe były<br />

recenzowane przez samodzielnych<br />

pracowników nauki<br />

Redakcja nie ponosi odpowiedzialności<br />

za treść ogłoszeń. Zastrzega<br />

sobie prawo do skracania i adiustacji<br />

nadesłanych materiałów.<br />

Indeks 35722<br />

Nakład do 2000 egz.<br />

Skład i druk: Drukarnia SIGMA-NOT Sp. z o.o.<br />

Spis treści ● Contents<br />

Nonlinear compact thermal model of SiC power semiconductor<br />

devices (Nieliniowy skupiony model termiczny półprzewodnikowych<br />

elementów mocy wykonanych z węglika krzemu)<br />

– K. Górecki, J. Zarębski, D. Bisewski, J. Dąbrowski . . . . . . . . . 9<br />

Krummenacher feedback analysis for high-count-rate semiconductor<br />

pixel detector readout (Analiza sprzężenia zwrotnego<br />

Krummenachera pod kątem zastosowania w szybkich<br />

układach odczytu półprzewodnikowych detektorów mozaikowych)<br />

– R. Szczygieł . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12<br />

Testing of interconnections with use of reduced-size signature-based<br />

diagnostic dictionary (Testowanie uszkodzeń<br />

w połączeniach z wykorzystaniem sygnaturowego słownika<br />

diagnostycznego o zredukowanym rozmiarze) – T. Garbolino,<br />

K. Gucwa, A. Hławiczka.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15<br />

The Time-over-Threshold based silicon strip detector readout<br />

(Odczyt do krzemowych detektorów paskowych oparty<br />

o metodę Time-over-Threshold) – K. Kasiński . . . . . . . . . . . . . 19<br />

The design of low power 11.6 mW high speed 1.8 Gb/s standalone<br />

LVDS Driver in 0.18 µm CMOS (Projekt scalonego<br />

nadajnika standardu LVDS o niskim poborze mocy 11,6 mW<br />

i 1,8 Gb/s szybkości transmisji danych) – R. Kłeczek . . . . . . . . 23<br />

CORDIC and SVD implementation in digital hardware (Realizacja<br />

algorytmów CORDIC i SVD w układzie cyfrowym) – P.M.<br />

Szecówka, P. Malinowski.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />

Lab-on-a-chip for developmental competence assessment<br />

of bovine oocytes (Lab-chip do oceny jakościowej potencjału<br />

rozwojowego oocytów bydlęcych) – R. Walczak, P. Szczepańska,<br />

J. Dziuban, B. Kempisty, M. Jackowska, P. Antosik, J. Jaśkowski,<br />

A. Chełmońska-Soyta . . . . . . . . . . . . . . . . . . . . . . . . . . 30<br />

Image sensor – based fluorescence detection for lab-on-achip<br />

(Detekcja fluorescencji w lab-chipach z wykorzystaniem<br />

czujnika obrazu) – R. Walczak . . . . . . . . . . . . . . . . . . . . . . . . . 33<br />

Scripting languages for simulations in modern SCADA systems<br />

(Wykorzystanie języków skryptowych dla symulacji<br />

w nowoczesnych systemach SCADA) – P. Marciniak, Z. Kulesza,<br />

A. Napieralski, R. Kotas . . . . . . . . . . . . . . . . . . . . . . . . . . . 36<br />

FPGA implementation of feature extraction algorithm for<br />

speaker verification (Implementacja algorytmu estymacji<br />

cech sygnału mowy systemu automatycznej weryfikacji mówcy<br />

w układzie FPGA) – M. Staworko, M. Rawski . . . . . . . . . . . . . . 41<br />

Digital implementation of a programmable type-2 fuzzy logic<br />

controller (Cyfrowa realizacja programowalnego sterownika<br />

rozmytego 2-go rzędu) – M. Bryk, A. Wielgus . . . . . . . . . . . . . . 44<br />

Universal Tool for Estimation of Programmable Logic Controllers<br />

Processing Power (Uniwersalne narzędzie do szacowania<br />

mocy obliczeniowej sterowników przemysłowych)<br />

– Z. Kulesza, S. Marasek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48<br />

XML markup language based design tool integration method<br />

in distributed design environments (Metoda integracji narzędzi<br />

w rozproszonych środowiskach projektowania wykorzystująca<br />

język znaczników XML) – M. Szlęzak, A. Pawlak,<br />

K. Wojciechowski . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51<br />

A dedicated high-level language for implementing nonrecursive<br />

filter banks and transforms (Dedykowany język wysokiego<br />

poziomu do implementowania nierekursywnych banków<br />

filtrów i transformacji) – M. Parfieniuk. . . . . . . . . . . . . . . . . . . . . 55<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Expandable GSM and GPS systems simulator (Rozbudowany<br />

symulator systemów GSM oraz GPS) – J. Pochmara, J. Pałasiewicz,<br />

P. Szablata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60<br />

Alvis approach to Hexor robot controller development (Zastosowanie<br />

języka Alvis do projektowania sterownika dla robota<br />

Hexor) – M. Szpyrka, P. Matyasik, R. Mrówka . . . . . . . . . . . . 63<br />

Model of human palm controlled by glove with micromachined<br />

accelerometers (Model ludzkiej dłoni sterowanej przez<br />

rękawicę wyposażoną w mikromaszynowe czujniki przyspieszenia)<br />

– R. Kotas, Z. Kulesza, W. Tylman, A. Napieralski,<br />

P. Marciniak . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67<br />

TECHNIKA SENSOROWA: Bezkontaktowy czujnik przemieszczenia<br />

w złożonym układzie pomiarowym charakterystyk<br />

elektro-termo-mechanicznych stopów z pamięcią kształtu<br />

(Non-contact sensor of displacement in complex measurement<br />

system for measuring the electro-thermo-mechanical characteristics<br />

of SMA) – G. Kłapyta, M. Kciuk . . . . . . . . . . . . . . . . . . . . 71<br />

TECHNIKA MIKROFALOWA I RADIOLOKACJA: Poprawa zależności<br />

poziomu listków bocznych od współczynnika kompresji<br />

dla sygnałów złożonych z małą bazą (Peak-to-sidelobe<br />

level vs. time-bandwidth product improvement for complex<br />

radar signals with small base) – M. Łuszczyk . . . . . . . . . . . . . . 75<br />

TECHNIKI INFORMATYCZNE: Między Web 2.0 i 3.0: Mobilne<br />

systemy informacyjne z rozszerzoną rzeczywistością (Between<br />

Web 2.0 and Web 3.0: Mobile information systems using<br />

Augmented Reality) – J. Brzostek-Pawłowska . . . . . . . . . . . . . 79<br />

Using multi-frequency coherent signal to measurement of frequency<br />

response of narrow-band LF and HF circuits (Zastosowanie<br />

wieloczęstotliwościowego sygnału koherentnego<br />

do pomiaru charakterystyk częstotliwościowych układów wąskopasmowych<br />

m.cz. oraz w.cz.) – J. Duchiewicz, A.S. Sowa,<br />

J.S. Witkowski, A. Francik, A.L. Dobrucki, B. Idźkowski . . . . . . . 88<br />

Radiokomunikacyjne aspekty planu implementacji strategii<br />

e-nawigacji (Radiocommunication aspects of an e-navigation<br />

strategy implementation plan) – K. Korcz. . . . . . . . . . . . . . . . . . 92<br />

Urządzenie do generacji silnych i ultrasilnych pól magnetycznych<br />

oraz sprzężony z nim układ pomiarowy (Device for the<br />

generation of strong and ultrastrong magnetic fields with the<br />

measuring system) – D. Płoch, P. Sliż, E. Szeregij . . . . . . . . . . . 97<br />

Aspekty algorytmiczne organizacji jednostki procesorowej<br />

do mnożenia liczb Cayleya ( Algorithmic aspects of Cayley<br />

numbers multiplier organization) – A. Tariov, G. Tariova . . . . . 104<br />

TECHNIKA PRÓŻNI I TECHNOLOGIE PRÓŻNIOWE: Biuletyn<br />

Polskiego Towarzystwa Próżniowego 1 (50) <strong>2010</strong>. . . . . . . . 109<br />

Spektrometr Elektronowego Rezonansu Paramagnetycznego<br />

(EPR), umożliwiający ilościowe pomiary liczby spinów<br />

w badanej próbce (Electron Paramagnetic Resonance (EPR)<br />

spectrometer for quantitative measurements of the spins number<br />

in the sample under test) – J. Duchiewicz, A. Dobrucki,<br />

A. Francik, W. Stachowicz, T. Oleś, T. Duchiewicz . . . . . . . . . . 117<br />

Moduły lokalizacji oraz testy chipów GNSS dla elementów systemu<br />

PROTEUS (The location modules and tests of GNSS receivers<br />

for the elements of PROTEUS system) – R. Darakchiev,<br />

P. Sitek, K. Brzostowski, A. Foks-Ryznar, M. Kalarus, R. Zdunek . .122<br />

Analiza możliwości zastosowania sensorów inercjalnych<br />

MEMS w projekcie PROTEUS (An analysis of the applicability<br />

of MEMS inertial sensors in the PROTEUS project) – M. Kalarus,<br />

P. Sitek, K. Brzostowski, A. Foks-Ryznar, R. Darakchiev . 127<br />

Antena mikropaskowa o poszerzonym paśmie pracy (Broadband<br />

microstrip patch antenna) – M. Pergoł, W. Zieniutycz, Ł.<br />

Sorokosz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130<br />

Properties of feature contours for audio classification tasks<br />

(Właściwości konturów cech w zadaniach klasyfikacji sygnałów<br />

akustycznych) – T. Mąka . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134<br />

Streszczenia artykułów ● Summaries of the articles<br />

GÓRECKI K., ZARĘBSKI J., BISEWSKI D., DĄBROWSKI J.: Nieliniowy<br />

skupiony model termiczny półprzewodnikowych elementów mocy<br />

wykonanych z węglika krzemu<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 9<br />

W pracy przedstawiono nieliniowy skupiony model termiczny<br />

półprzewodnikowych elementów mocy wykonanych z węglika krzemu,<br />

sformułowany na bazie sieci Cauera. Zaprezentowano opis analityczny<br />

modelu oraz metodę wyznaczania wartości parametrów tego modelu.<br />

Dokładność i przydatność zaproponowanego modelu zweryfikowano<br />

doświadczalnie dla diody Schottky’ego i tranzystora MESFET przy różnych<br />

warunkach chłodzenia tych elementów.<br />

Słowa kluczowe: model termiczny, węglik krzemu, półprzewodnikowe<br />

elementy mocy<br />

SZCZYGIEŁ R.: Analiza sprzężenia zwrotnego Krummenachera pod<br />

kątem zastosowania w szybkich układach odczytu półprzewodnikowych<br />

detektorów mozaikowych<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 12<br />

Obwód sprzężenia zwrotnego Krummenachera [1] jest często używany<br />

we wzmacniaczach ładunkowych podłączonych do detektorów półprzewodnikowych<br />

ze sprzężeniem DC. Układ ten w bardzo prostej strukturze<br />

zapewnia podwójną funkcjonalność: ciągłe kasowanie wzmacniacza ładunkowego<br />

i kompensację prądu upływu detektora. Pomimo prostej budowy<br />

analiza tego układu pod kątem szybkości działania jest dosyć złożona.<br />

Przedstawiony artykuł prezentuje analizę niniejszego obwodu w oparciu<br />

o parametry technologii 90 nm CMOS ukierunkowaną na aplikacje do<br />

odczytu półprzewodnikowych detektorów hybrydowych realizowanych<br />

w technologiach submikronowych lub 3D.<br />

Słowa kluczowe: sprzężenie Krummenachera, CSA, detektor mozaikowy<br />

GÓRECKI K., ZARĘBSKI J., BISEWSKI D., DĄBROWSKI J.: Nonlinear<br />

compact thermal model of SiC power semiconductor devices<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 9<br />

The paper deals with the nonlinear compact thermal model of SiC power<br />

semiconductor devices, based on the Cauer network. The analytical description<br />

of the model and the method of the model parameter estimation<br />

are presented. The accuracy and usefulness of the model is verified experimentally<br />

for the Schottky diode and MESFET transistor at their various<br />

cooling conditions.<br />

Keywords: Thermal model, silicon carbide, power swmiconductor devices<br />

SZCZYGIEŁ R.: Krummenacher feedback analysis for high-countrate<br />

semiconductor pixel detector readout<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 12<br />

The Krummenacher feedback circuit [1] is a well know and often used in<br />

charge sensitive amplifiers (CSA) connected to DC-coupled semiconductor<br />

detectors. The circuit provides in a very simple structure dual functionality:<br />

continuous reset and detector leakage current compensation. In spite<br />

of the simplicity the characterization of the circuit for high speed applications<br />

is not very straightforward. This article presents an analysis of the<br />

feedback circuit based on 90 nm CMOS and technology and targeted at<br />

high-count-rate semiconductor hybrid pixel detector applications realized<br />

in submicron or 3D technologies.<br />

Keywords: Krummenacher feedback, CSA, pixel detector<br />

<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Streszczenia artykułów ● Summaries of the articles<br />

GARBOLINO T., GUCWA K., HŁAWICZKA A.: Testowanie uszkodzeń<br />

w połączeniach z wykorzystaniem sygnaturowego słownika diagnostycznego<br />

o zredukowanym rozmiarze<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 15<br />

W artykule zaproponowano nową metodę redukcji rozmiaru syganturowego<br />

słownika diagnostycznego, który jest wykorzystywany do testownia uszkodzeń<br />

statycznych i opóźnieniowych w połączeniach testowanych przez<br />

rejestr pierścieniowy R-LFSR. Nowo opracowana metoda – podobnie jak<br />

w poprzednich pracach autorów – zakłada, że testowana magistrala n‐bitowa<br />

zostaje podzielona na b jednakowych fragmentów o szerokości k bitów<br />

każdy. Każdy taki fragment magistrali jest testowany przez oddzielny rejestr<br />

R-LFSR złożony z 2k przerzutników D. Procedura testowa obejmuje cztery<br />

fazy, w czasie których rejestry parzyste i nieparzyste pracują naprzemiennie.<br />

Takie podejście eliminuje zjawisko wzajemnego wpływu na siebie sąsiednich<br />

rejestrów R-LFSR, które było wadą poprzednich rozwiązań ponieważ ograniczało<br />

możliwość zmniejszenia rozmiaru słownika diagnostycznego. Nowa<br />

technika umożliwia detekcję, lokalizację oraz identyfikację wszystkich zamodelowanych<br />

uszkodzeń, mogących wystapić na n‐bitowej magistrali, oraz wymaga<br />

słownika diagnostycznego o znacznie mniejszym rozmiarze. Rozmiar<br />

tego słownika jest określony wyłącznie przez krotność uszkodzeń r w każdym<br />

k‐bitowym fragmencie magistrali, nawet gdy szerokość tej magistrali n >>k.<br />

Słowa kluczowe: testowanie połączeń; wbudowane samotestowanie; rejestr<br />

pierścieniowy; sygnatura; sygnaturowy słownik diagnostyczny<br />

KASIŃSKI K.: Odczyt do krzemowych detektorów paskowych oparty<br />

o metodę Time-over-Threshold<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 19<br />

Praca przedstawia project układu odczytowego dla krzemowych detektorów<br />

paskowych. Układ może znaleźć zastosowanie w wielokanałowych<br />

specjalizowanych układach scalonych dla potrzeb eksperymentów Fizyki<br />

Wysokich Energii. Pomiar ładunku jest oparty o metodę Time-over-Threshold<br />

pozwalającej na integrację niskomocowego przetwornika Analogowo-<br />

Cyfrowego w każdym kanale.<br />

Słowa kluczowe: wielokanałowy układ scalony, krzemowy detektor paskowy,<br />

Time-over-Threshold<br />

KŁECZEK R.: Projekt scalonego nadajnika standardu LVDS o niskim<br />

poborze mocy 11,6 mW i 1,8 Gb/s szybkości transmisji danych<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 23<br />

Niskonapięciowa transmisja różnicowa LVDS dzięki swoim licznym zaletom<br />

jest coraz częściej stosowana w układach transmisji danych. Praca<br />

przedstawia projekt modułu układu scalonego nadajnika LVDS. Układ<br />

został zaprojektowany w technologii CMOS firmy United Microelectronics<br />

Corporation, o rozmiarze charakterystycznym równym 180 nm. Opracowane<br />

rozwiązanie nadajnika jest kompatybilne ze specyfikacją IEEE. Zaprojektowany<br />

nadajnik LVDS charakteryzuje się bardzo niskim poziomem<br />

statycznego 7,5 mW, dynamicznego zużycia mocy 8,5 mW (11,6 mW),<br />

podczas transmisji danych z szybkością 400 Mb/s (1,8 Gb/s).<br />

Słowa kluczowe: niskonapięciowa transmisja różnicowa LVDS, transmisja<br />

danych, układy wejścia/wyjścia, układ scalony, technologia CMOS,<br />

system transmisyjny, przepustowość danych<br />

SZECÓWKA P.M., MALINOWSKI P.: Realizacja algorytmów CORDIC<br />

i SVD w układzie cyfrowym<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 26<br />

Algorytm SVD (Singular Value Decomposition) jest zaliczany do najbardziej<br />

efektywnych metod pozwalających odwracać macierze. Artykuł opisuje próbę<br />

sprzętowej realizacji algorytmów CORDIC i SVD. Rozważono szereg<br />

architektur – warianty bardzo oszczędne sekwencyjne, a także rozwiązania<br />

pozwalające uzyskać wysoką wydajność obliczeniową, z przetwarzaniem<br />

potokowym. Porównano także rezultaty uzyskane przy zastosowaniu arytmetyki<br />

stało- i zmiennoprzecinkowej. Koncepcje zostały zaimplementowane<br />

w języku opisu sprzętu (VHDL) poddane weryfikacji i syntezie za pomocą<br />

narzędzi XIlinx. Niektóre warianty zostały przetestowane fizycznie.<br />

Słowa kluczowe: CORDIC, SVD, układ cyfrowy, VHDL, układ FPGA<br />

WALCZAK R., SZCZEPAŃSKA P., DZIUBAN J., KEMPISTY B., JACKOW-<br />

SKA M., ANTOSIK P., JAŚKOWSKI J., CHEŁMOŃSKA-SOYTA A.: Labchip<br />

do oceny jakościowej potencjału rozwojowego oocytów bydlęcych<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 30<br />

W artykule przedstawiono nowa spektrofotometryczną metodę wykorzystującą<br />

mikrosystem typu lab-on-a-chip do jakościowej oceny oocytów<br />

bydlęcych. Przedstawiono wyniki badań z rzeczywistym materiałem biologicznym.<br />

Otrzymano zgodność pomiędzy wynikami uzyskanymi za pomocą<br />

opisanego systemu a metodą referencyjną.<br />

Słowa kluczowe: lab-on-a-chip, bydło, oocyt, zarodek, kompetencje rozwojowe<br />

GARBOLINO T., GUCWA K., HŁAWICZKA A.: Testing of interconnections<br />

with use of reduced-size signature-based diagnostic dictionary<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 15<br />

The paper presents a new method for size reduction of a signature-based<br />

diagnostic dictionary that is normally used for testing of static and delay<br />

faults in interconnections that are tested by means of an R-LFSR ring<br />

register. The newly developed method, similarly to the previous studies<br />

of the authors, assume that the n-bit bus under test is split into b fragments<br />

with their width of k bits each. Each fragment of the bus is tested with use<br />

of a separate 2k-bit R‐LFSR. The test procedure consists of four phases<br />

during which odd and even registers operate alternately. Such an approach<br />

eliminates effect of mutual impact between states of neighbouring<br />

R-LFSRs in case of shorts between feedback lines of these registers. These<br />

possible interactions were a drawback of previous solutions as they<br />

limited the possibility to reduce size of the diagnostic dictionary. Owing to<br />

application of this new technique to full detection, localization and identification<br />

of all the considered faults that may occur on an n-bit bus, the new<br />

solution needs much smaller dictionary, where its size is determined by<br />

the multiplicity r of faults within each k-bit fragment, even if the bus width<br />

n >> k.<br />

Keywords: interconnect test; BIST; interconnect BIST; IBIST; ring LFSR;<br />

signature; diagnostic dictionary<br />

KASIŃSKI K.: The Time-over-Threshold based silicon strip detector<br />

readout<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 19<br />

This work presents the project of a silicon strip detector readout circuit.<br />

The circuit is to be used in the multi-channel detector readout integrated<br />

circuit with a possible application in High Energy Physics experiments. The<br />

charge measurement is based on the Time-over-Threshold method which<br />

allows integration of the low-power ADC into each channel.<br />

Keywords: integrated cicuit, silicon strip detector, Time-over-Threshold,<br />

multichannel<br />

KŁECZEK R.: The design of low power 11.6 mW high speed 1.8 Gb/s<br />

stand-alone LVDS Driver in 0.18 µm CMOS<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 23<br />

Due to many advantages low voltage differential signaling LVDS has become<br />

a popular choice for fast data on-chip transmission, on-board/backplane or<br />

cable connections. LVDS standard offers achieving a high-speed data transmission<br />

and low power consumption at the same time. This paper presents<br />

a description of standard and design of LVDS transmitter fully compatible with<br />

IEEE specification, implemented in CMOS 180 nm UMC technology. The main<br />

driver’s functional blocks: LVDS core and common mode feedback (CMFB)<br />

are described in detail, whereas control buffer and band-gap reference source<br />

are only mentioned. Results of simulation are also presented. Designed LVDS<br />

driver characterizes a very low level of static 7.5 mW and dynamic 8.5 mW<br />

(11.6 mW) power dissipation at data rate 400 Mb/s (1.8 Gbp/s).<br />

Keywords: low-voltage differential signaling LVDS, fast data communications<br />

circuits, high speed integrated circuits (IC), input/output (I/O) drivers,<br />

low power design<br />

SZECÓWKA P.M., MALINOWSKI P.: CORDIC and SVD implementation<br />

in digital hardware<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 26<br />

Singular Value Decomposition (SVD) is classified among the most effective<br />

numeric methods of matrices inversion. The paper presents a study of<br />

hardware implementation of SVD and CORDIC algorithms. Various digital<br />

architectures were proposed and compared, including low-cost sequential<br />

and high-performance pipelined solutions. Fixed point and floating point<br />

arithmetic was considered. The concepts were implemented in VHDL, verified<br />

and synthesized with Xilinx tools. Selected approach was physically<br />

implemented and tested.<br />

Keywords: CORDIC, SVD, digital, hardware, VHDL, FPGA<br />

WALCZAK R., SZCZEPAŃSKA P., DZIUBAN J., KEMPISTY B., JACKOW-<br />

SKA M., ANTOSIK P., JAŚKOWSKI J., CHEŁMOŃSKA-SOYTA A.: Lab-ona-chip<br />

for developmental competence assessment of bovine oocytes<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 30<br />

In the paper a novel spectrophotometric methodology utilizing lab-on-achip<br />

device for assessment of quality of bovine oocytes is presented. Qualification<br />

results of living bovine oocytes are described. Good correlation of<br />

obtained results with reference methodology has been achieved.<br />

Keywords: lab-on-a-chip, bovine, oocyte, embryo, development competence<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Streszczenia artykułów ● Summaries of the articles<br />

WALCZAK R.: Detekcja fluorescencji w lab-chipach z wykorzystaniem<br />

czujnika obrazu<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 33<br />

W artykule przedstawiono krótką dyskusję na temat detekcji fluorescencji<br />

w mikrosystemach typu lab-on-a-chip. Przedstawiono nowe rozwiązanie<br />

technicznye bazujace na czujniku obrazowym współpracującym ze specjalizowanych<br />

oprogramowaniem autorskim. Przedstawiono przykład<br />

wykorzystujący opracowany układ detekcji fluorescencji w przenośnym<br />

urządzeniu do detekcji patogenów żywności wykorzystujący analizę DNA<br />

bakterii.<br />

Słowa kluczowe: detekcja fluorescencji, mikrofluidyka, lab-on-a-chip<br />

MARCINIAK P., KULESZA Z., NAPIERALSKI A., KOTAS R.: Wykorzystanie<br />

języków skryptowych dla symulacji w nowoczesnych systemach<br />

SCADA<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 36<br />

Głównym celem napisania tego artykułu jest zaprezentowanie nowoczesnych<br />

możliwości systemów SCADA (Supervisory Control And Data Acquisition).<br />

Pierwsza część skupia się na historii powstawania systemów<br />

SCADA i ukazuje, jak zmieniają się one od momentu ich pojawienia się..<br />

Zawiera także opis języków skryptowych, które są wbudowane w oprogramowanie<br />

SCADA. Druga część artykułu prezentuje zdolność wykorzystania<br />

języków skryptowych na przykładzie Visual Basic for Applications,<br />

który jest zintegrowany z oprogramowaniem iFIX Intellution HMI/SCADA<br />

Automation Software. Wszystkie przykłady zaczerpnięte są z pracy magisterskiej<br />

autora pod tytułem „Wykorzystanie pakietów SCADA w wizualizacji<br />

procesu produkcji leków”, której głównym celem było wykonanie<br />

systemu wizualizacji. System ten prezentuje realny proces produkcji, który<br />

jest wdrożony w jednej z firm farmaceutycznych w Łodzi.<br />

Słowa kluczowe: system SCADA, język skryptowy, archiwizacja danych,<br />

interfejs użytkownika<br />

STAWORKO M., RAWSKI M.: Implementacja algorytmu estymacji<br />

cech sygnału mowy systemu automatycznej weryfikacji mówcy w<br />

układzie FPGA<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 41<br />

Artykuł przedstawia układ estymacji cech sygnału mowy w systemie automatycznej<br />

weryfikacji mówcy bazujący na parametrach LFCC z wykorzystaniem<br />

nowej architektury realizującej proces uśredniania w dziedzinie<br />

częstotliwości. Proponowane rozwiązanie jest przeznaczone do implementacji<br />

w strukturach reprogramowalnych jako część systemu jednoukładowego,<br />

charakteryzuje się niskim poborem mocy oraz krótkim czasem<br />

wyznaczania parametrów LFCC z sygnału mowy.<br />

Słowa kluczowe: automatyczna weryfikacja mówcy, LFCC, FPGA, niski<br />

pobór mocy<br />

BRYK M., WIELGUS A.: Cyfrowa realizacja programowalnego sterownika<br />

rozmytego 2-go rzędu<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 44<br />

W artykule przedstawiono projekt cyfrowego układu scalonego CMOS realizującego<br />

sterownik rozmyty 2-go rzędu. Zaproponowana architektura<br />

układu umożliwia połączenie sekwencyjnego przetwarzania kolejnych reguł<br />

rozmytych z równoległym przetwarzaniem dla górnej i dolnej funkcji<br />

przynależności każdego zbioru rozmytego. Uzyskany sparametryzowany<br />

model VHDL umożliwia syntezę układu o rozmiarze wymaganym w konkretnej<br />

realizacji.<br />

Słowa kluczowe: zbiór rozmyty II rzędu; sterownik rozmyty, realizacja<br />

cyfrowa<br />

KULESZA Z., MARASEK S.: Uniwersalne narzędzie do szacowania<br />

mocy obliczeniowej sterowników przemysłowych<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 48<br />

Artykuł omawia problematykę metod szacowania mocy obliczeniowej sterowników<br />

przemysłowych. Ze względu na brak uniwersalnych metod konieczne<br />

było opracowanie dedykowanego benchmarka syntetycznego. Poprzez symulację<br />

reprezentatywnego obciążenia programowego został on wykorzystany do<br />

przetestowania szerokiego zakresu sterowników przemysłowych. Dodatkowe<br />

testy przeprowadzono w celu zweryfikowania osiągów sterowników dla różnych<br />

elementów programowych. Otrzymane wyniki zestawiono z istniejącymi<br />

wyznacznikami mocy obliczeniowej sterowników, wykazując poważne wady<br />

tych ostatnich. Zaprojektowany benchmark wykazuje cechy prawidłowego<br />

i uniwersalnego narzędzia do porównywania mocy obliczeniowej sterowników,<br />

z szansą na praktyczne wykorzystanie w przemyśle.<br />

Słowa kluczowe: PLC, sterownik, testowanie, benchmarking, moc obliczeniowa<br />

WALCZAK R.: Image sensor – based fluorescence detection for labon-a-chip<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 33<br />

In the paper a brief discussion on fluorescence detection in lab-on-a-chip<br />

is carried out. A novel, low-cost image sensor – based detection instrumentation<br />

co-working with a “clever” software is described. An example of<br />

application of the novel method in a portable device for detection of food<br />

pathogens by DNAanalyze is presented.<br />

Keywords: fluorescence detection, microfluidics, lab-on-a-chip<br />

MARCINIAK P., KULESZA Z., NAPIERALSKI A., KOTAS R.: Scripting<br />

languages for simulations in modern SCADA systems<br />

<strong>Elektronika</strong> (LI_), no 11/<strong>2010</strong>, p. 36<br />

The goal of this research is to present modern SCADA (supervisory control<br />

and data acquisition) systems abilities. The first part of this article concentrates<br />

on the SCADA history and shows how SCADA systems have changed<br />

since they appeared. This part also includes the description of scripting<br />

languages which are embedded in SCADA software. The second part<br />

of this research presents scripting languages abilities based on example<br />

of Visual Basic for Applications which are integrated with iFIX Intellution<br />

HMI/SCADA Automation Software. All examples come from the author’s<br />

master thesis which is entitled “The use of SCADA System in Visualization<br />

of Medicine Production Process”. The main point of this master thesis was<br />

to create a visualization system. This system presents a real production<br />

process which is implemented in one of the pharmaceutical companies<br />

in Lodz.<br />

Keywords: SCADA system, scripting language, data aquisition, Human<br />

Machine Interface<br />

STAWORKO M., RAWSKI M.: FPGA implementation of feature extraction<br />

algorithm for speaker verification<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 41<br />

In this paper we propose a feature extraction circuit of automatic speaker<br />

verification system based on the LFCC with novel architecture for spectral<br />

averaging. Proposed solution is optimized for implementation in programmable<br />

structures as System on Programmable Chip and significantly reduces<br />

feature extraction execution time and power consumption.<br />

Keywords: Terms-automatic speaker verification; LFCC; FPGA; low power<br />

application<br />

BRYK M., WIELGUS A.: Digital implementation of a programmable<br />

type-2 fuzzy logic controller<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 44<br />

This paper presents the design of a digital CMOS integrated circuit implementing<br />

a type-2 fuzzy logic controller. The proposed architecture is suitable<br />

for serial processing of fuzzy rules combined with parallel processing<br />

of upper and lower membership functions of type-2 fuzzy sets. The parameterized<br />

VHDL model allows to synthesize the circuit of the required size<br />

for a particular application. Moreover, on-chip programming is performed.<br />

Keywords: type-2 fuzzy set; fuzzy logic controller; digital implementation<br />

KULESZA Z., MARASEK S.: Universal Tool for Estimation of Programmable<br />

Logic Controllers Processing Power<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 48<br />

Article covers the issue of PLC processing power estimation techniques.<br />

Absence of universal methods entailed need to elaborate specialized<br />

synthetic benchmark. By simulating representative workload, it was used<br />

to test wide spectrum of controllers. Additional tests were carried out to<br />

measure controllers performance for different program items. Obtained<br />

results are confronted with existing determinants of PLC processing power.<br />

Serious drawbacks of the latter are shown. On the contrary, designed<br />

benchmark proves to be appropriate and universal tool for PLC performance<br />

comparison with promising future applications in industry.<br />

Keywords: PLC, controller, testing, benchmarking, processing power<br />

<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Streszczenia artykułów ● Summaries of the articles<br />

SZLĘZAK M., PAWLAK A., WOJCIECHOWSKI K.: Metoda integracji<br />

narzędzi w rozproszonych środowiskach projektowania wykorzystująca<br />

język znaczników XML<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 51<br />

Integracja narzędzi projektowania jest zasadniczym problemem w rozproszonym<br />

projektowaniu układów elektronicznych typu SoC. Artykuł prezentuje<br />

metody i narzędzia dla opartej o język znaczników XML integracji rozproszonych<br />

narzędzi.. Pokazano jak XML umożliwia: uniwersalny opis narzędzi,<br />

łatwe magazynowanie, wyszukiwanie i dostęp do opisów narzędzi, jak<br />

również zdalne wywoływanie narzędzi. Zaprezentowane koncepcje i metody<br />

zostały zweryfikowane w środowisku projektowania wykorzystującym narzędzie<br />

TRMS (ang. Tools Registration and Management Services). Środowisko<br />

to zawiera, m.in. narzędzie do śledzenia zmian w specyfikacjach dostępnych<br />

w sieci Internet narzędzi projektowania oraz system ontologiczny, który<br />

umożliwia wnioskowanie o relacjach pomiędzy zadaniami projektowymi<br />

a narzędziami. We wnioskach wskazano na zalety wykorzystania języka<br />

znaczników XML w procesie integracji narzędzi.<br />

Słowa kluczowe: projektowanie systemów typu SoC, projektowanie<br />

rozproszone, integracja rozproszonych narzędzi, opis narzędzi w języku<br />

XML, zarządzanie narzędziami, ontologie dla narządzi projektowania<br />

PARFIENIUK M.: Dedykowany język wysokiego poziomu do implementowania<br />

nierekursywnych banków filtrów i transformacji<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 55<br />

W artykule przedstawiono nowatorskie podejście do implementowania nierekursywnych<br />

banków filtrów i transformacji. Opracowany został dziedzinowy<br />

język, który pozwala opisywać te systemy przejrzyściej, zwięźlej i szybciej<br />

niż z użyciem MATLAB/Simulink lub SPL, istniejących narzędzi do rozwijania<br />

algorytmów cyfrowego przetwarzania sygnałów. Jego składnia jest ukierunkowana<br />

na ścisłe powiązanie kodu z grafem przepływu danych w rozpatrywanej<br />

transformacji i na umożliwienie wyspecyfikowania algorytmu w kategoriach<br />

transformacji elementarnych: obrotów planarnych, odbić, stopni „lifting”, opóźnień<br />

itp. W odróżnieniu od wymienionych platform, proponowane podejście<br />

pozwala uniknąć konstruowania skomplikowanych wyrażeń macierzowych,<br />

choć notacja macierzowa jest dostępna jako podzbiór języka MATLAB. Skojarzony<br />

kompilator przekształca opisy systemów w dosyć wydajne implementacje<br />

Java, C++ lub C, które mogą być wykorzystywane do szybkiego<br />

prototypowania aplikacji, które opierają się na podpasmowej dekompozycji<br />

sygnałów, lub do przygotowywania funkcji celu na potrzeby optymalizacji<br />

współczynników schematów obliczeniowych.<br />

Słowa kluczowe: język dziedzinowy, DSL, kompilator, generacja kodu,<br />

implementacja, bank/zespół filtrów, transformacja, cyfrowe przetwarzanie<br />

sygnałów, DSP<br />

POCHMARA J., PAŁASIEWICZ J., SZABLATA P.: Rozbudowany symulator<br />

systemów GSM oraz GPS<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 60<br />

Publikacja prezentuje propozycję implementacji prostego symulatora systemów<br />

GPS i GSM. Moduł został przystosowany do instalacji rozszerzeń zapewniających<br />

dodatkową funkcjonalność. Urządzenie pozwala na praktyczne<br />

wykorzystanie przewagi technologii bezprzewodowej w wielu złożonych<br />

zastosowaniach. Połączenie protokołu NMEA oraz języka komend AT sprawiło,<br />

że prezentowany system jest kompatybilny z najnowszymi standardami<br />

obowiązującymi w branży telekomunikacyjnej. Projekt generuje efekty<br />

spotykane w wysoko-budżetowych rozwiązaniach korzystając z podzespołów<br />

oferowanych w dostępnych cenach. Dzięki wykorzystaniu popularnego<br />

standardu krótkich wiadomości (SMS) oraz światowego systemu nawigacji<br />

(GPS), prezentowany moduł może zostać użyty w każdym projekcie, który<br />

wykorzystuje powyższe technologie. System może służyć do symulowania<br />

zachowań telefonu komórkowego, śledzenia innych urządzeń, lokalizowania<br />

nadajników i prezentowania otrzymanych danych na dowolnej mapie.<br />

Projekt sprawdzi się również w złożonych procesach obliczeniowych.<br />

Słowa kluczowe: GSM, GPS, GPRS, NMEA, standard AT, komendy AT, SMS,<br />

szerokość geograficzna, długość geograficzna, urządzenia śledzące, nadajniki<br />

SZPYRKA M., MATYASIK P., MRÓWKA R.: Zastosowanie języka Alvis<br />

do projektowania sterownika dla robota Hexor<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 63<br />

Alvis jest językiem modelowania, rozwijanym głównie z myślą o projektowaniu<br />

i weryfikacji systemów wbudowanych. Wywodzi się on z algebr procesów CCS<br />

i XCCS, ale w języku tym równania algebraiczne zostały zastąpione przez język<br />

programowania wysokiego poziomu oparty na języku Haskell. W przeciwieństwie<br />

do algebr procesów, które umożliwiają wyłącznie tekstowy opis systemów<br />

wbudowanych, w języku Alvis struktura projektowanego systemu, z punktu widzenia<br />

przepływu danych i sterowania, przedstawiana jest graficznie za pomocą<br />

diagramów komunikacji. Poniższy artykuł zawiera wprowadzenie do języka<br />

Alvis zilustrowane modelem sterownika dla robota mobilnego Hexor II.<br />

Słowa kluczowe: język modelowania Alvis, systemy wbudowane, Hexor<br />

II, modelowanie formalne<br />

SZLĘZAK M., PAWLAK A., WOJCIECHOWSKI K.: XML markup language<br />

based design tool integration method in distributed design<br />

environments<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 51<br />

Design tools integration is a key problem in distributed design of Systemson-Chip.<br />

The paper presents methods and tools for integration of distributed<br />

design tools based on the markup language XML which supports<br />

universal description of tools, straightforward storage, search and retrieval<br />

of tool descriptions, as well as remote invocation of tools. Presented concepts<br />

and methods have been verified with the design environment based<br />

on Tools Registration and Management Services (TRMS). The TRMSbased<br />

environment includes a tool for Internet-wide tracking of changes<br />

in design tool specifications and ontology for reasoning on relationships<br />

between design tasks and tools. Advantages of the approach have been<br />

summarized.<br />

Keywords: SoC design, distributed collaborative design; remote tools integration;<br />

XML-based tool wrapping; tools management; tools ontology<br />

PARFIENIUK M.: A dedicated high-level language for implementing<br />

nonrecursive filter banks and transforms<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 55<br />

This paper presents a novel approach to implementing nonrecursive filter<br />

banks and transforms. A domain-specific language has been developed<br />

that allows such systems to be described more clearly, more compactly,<br />

and faster than with either MATLAB/Simulink or SPL, the existing tools for<br />

developing DSP algorithms. Its syntax is aimed at closely linking code to<br />

the signal flow graph of a given transform and at allowing the algorithm to<br />

be specified in terms of elementary transformations: plane rotations, reflections,<br />

lifting steps, delays, etc. Unlike the mentioned platforms, our approach<br />

allows to avoid constructing complicated matrix expressions, even<br />

though matrix notation is supported via a subset of the MATLAB language.<br />

The associated compiler converts system descriptions into quite efficient<br />

Java, C++, or C implementations, which can be used to rapidly prototype<br />

applications based on subband processing of signals or to prepare objective<br />

functions for optimizing coefficients of computational schemes.<br />

Keywords: domain-specific language, DSL, compiler, code generation,<br />

implementation, filter bank, transform, digital signal processing, DSP<br />

POCHMARA J., PAŁASIEWICZ J., SZABLATA P.: Expandable GSM<br />

and GPS systems simulator<br />

Elektrponika (LI), no 11/<strong>2010</strong>, p. 60<br />

This paper proposes and implements a simple extendable GSM and GPS<br />

systems simulator. It allows user to learn advantages of modern wireless<br />

technology and use it in many complicated solutions. By combining NMEA<br />

protocol and AT commands, system is compatible with newest standards<br />

in telecommunication industry. Project uses low cost hardware to produce<br />

effects that are nowadays only seen in very expensive solutions. By utilizing<br />

popular short messages standard (SMS) with worldwide working GPS<br />

system, it is possible to make use of presented module in any project<br />

involving wireless technology. This system is useful for simulating mobile<br />

phone, tracking device, localizing beacon within map range and using<br />

module for more complex data calculating processes.<br />

Keywords: GSM, GPS, GPRS, NMEA, AT standard, AT commands, SMS,<br />

latitude, longitude, computer application, tracking device<br />

SZPYRKA M., MATYASIK P., MRÓWKA R.: Alvis approach to Hexor<br />

robot controller development<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 63<br />

Alvis is a novel modelling language defined especially for the embedded<br />

systems design and verification. The language has its origin in CCS and<br />

XCCS process algebras, but algebraic equations have been replaced with<br />

a Haskell based high level programming language. Moreover, Alvis provides<br />

communication diagrams for the visual modelling of an embedded<br />

system structure, especially from the control and data-flow point of view.<br />

This paper presents an introduction to Alvis based on a model of a controller<br />

for the Hexor II mobile robot.<br />

Keywords: Alvis modelling language, embedded systems, Hexor II Robot,<br />

formal modelling<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Streszczenia artykułów ● Summaries of the articles<br />

KOTAS R., KULESZA Z., TYLMAN W., NAPIERALSKI A., MARCI-<br />

NIAK P.: Model ludzkiej dłoni sterowanej przez rękawicę wyposażoną<br />

w mikromaszynowe czujniki przyspieszenia<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 67<br />

Celem badań zaprezentowanych w artykule było sprawdzenie możliwości<br />

konstrukcji w pełni niezależnego systemu mikroprocesorowego do wykrywania<br />

ułożenia ludzkiej dłoni przy wykorzystaniu mikromaszynowych czujników<br />

przyspieszenia. Zaimplementowany algorytm polega na pomiarze<br />

przyspieszenia grawitacyjnego. Zaprojektowany system mikroprocesorowy<br />

analizuje dane z czujników i steruje modelem ludzkiej dłoni wykonanym<br />

w naturalnej skali. Model składa się z nieruchomego nadgarstka, dwóch<br />

palców oraz przeciwstawnego kciuka. Każdy palec posiada trzy stopnie<br />

swobody. Dodatkowy stopień swobody zastosowano dla kciuka.<br />

Słowa kluczowe: model ludzkiej dłoni, mikromaszynowe czujniki przyspieszenia<br />

KOTAS R., KULESZA Z., TYLMAN W., NAPIERALSKI A., MARCI-<br />

NIAK P.: Model of human palm controlled by glove with micromachined<br />

accelerometers<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 67<br />

The aim of the research presented in this paper was to test the possibility<br />

of inventing a fully autonomous device to detect the position of the fingers<br />

with the use of micromachined accelerometers. An implemented algorithm<br />

is based on the measurement of gravity acceleration. Designed microprocessor<br />

system analyses that data and controls the model of a human palm<br />

made in a 1:1 scale. The model has a motionless wrist, two fingers and<br />

an opposing thumb. Every finger has three joints. An extra joint is made<br />

for the thumb. This paper is based on the author’s master thesis which is<br />

entitled “Data glove controlled by a microprocessor system”.<br />

Keywords: model of human palm, micromachined accelerometers<br />

KŁAPYTA G., KCIUK M.: Bezkontaktowy czujnik przemieszczenia<br />

w złożonym układzie pomiarowym charakterystyk elektro-termo-mechanicznych<br />

stopów z pamięcią kształtu<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 71<br />

Artykuł przedstawia koncepcję oraz opis realizacji praktycznej bezkontaktowego<br />

czujnika przemieszczenia, pracującego synchronicznie jako element<br />

złożonego systemu pomiarowego, służącego do wyznaczania charakterystyk<br />

elektro-termo-mechanicznych aktuatorów ze stopów z pamięcią<br />

kształtu. Odkształcenie cięgna jest mierzone za pomocą dwukierunkowego,<br />

scalonego, optycznego licznika impulsów. Pomiar odkształcenia jest<br />

realizowany synchronicznie z pomiarem innych wielkości – elektrycznych<br />

(prąd, napięcie) oraz nieelektrycznych (temperatura). Wartość odkształcenia<br />

aktuatora zamieniana jest na sygnał PWMo wypełnieniu proporcjonalnym<br />

do liczby zliczonych impulsów. Woltomierz, pracujący w synchronicznym<br />

układzie pomiarowym, zarządzanym za pomocą komputera PC<br />

poprzez sieć GPIB, służy do pomiaru napięcia, którego wartość średnia<br />

jest proporcjonalna do wypełnienia sygnału PWM. Cały układ pomiarowy<br />

jest sterowany z poziomu programu, napisanego w środowisku LabVIEW.<br />

Słowa kluczowe: czujnik przemieszczenia, stopy z pamięcią kształtu,<br />

sma, stanowisko<br />

ŁUSZCZYK M.: Poprawa zależności poziomu listków bocznych od<br />

współczynnika kompresji dla sygnałów złożonych z małą bazą<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 75<br />

Najczęściej wykorzystywanym sygnałem w radiolokacji jest sygnał złożony<br />

z wewnątrzimpulsową modulacją częstotliwości (LFM). Podstawowym<br />

parametrem tego typu sygnałów jest baza sygnału, którą wyznacza się<br />

jako iloczyn czasu trwania impulsu oraz dewiacji sygnału zmodulowanego<br />

częstotliwościowo. Oczekiwany poziom czasowych listków bocznych<br />

sygnału po kompresji dla sygnałów LFM uzyskuje się dla współczynników<br />

kompresji powyżej 100. W przypadku stosowania sygnałów LFM o mniejszej<br />

bazie (np. krótkim czasie trwania) następuje znaczna redukcja poziomu<br />

czasowych listków bocznych. Istotne obniżenia listków bocznych można<br />

osiągnąć poprzez zastosowanie nieliniowej modulacji częstotliwości<br />

(NLFM). W artykule przedstawiono algorytm syntezy sygnałów NLFM oraz<br />

zaprezentowano wyniki badań symulacyjnych. Proponowana metoda jest<br />

nowym podejściem do problemu zapewnienia wymaganej rozdzielczości<br />

odległościowej w warunkach minimalizacji strefy martwej radaru.<br />

Słowa kluczowe: sygnał złożony, nieliniowa modulacja częstotliwości,<br />

współczynnik kompresji<br />

KŁAPYTA G., KCIUK M.: Non-contact sensor of displacement in complex<br />

measurement system for measuring the electro-thermo-mechanical<br />

characteristics of SMA<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 71<br />

The paper presents idea and description of practical realisation of non-contact<br />

displacement sensor. The sensor is working synchronically in complex<br />

measurement system for determining electro-thermo-mechanical characteristics<br />

of SMA actuators. Displacement of actuator is measured using<br />

bidirectional, integrated optical impulse counter. Measurement process<br />

is realised synchronically with measurements of other values: electrical<br />

(current, voltage) and non-electrical (temperature). Displacement value of<br />

actuator is transferred to PWM signal with width proportional to number<br />

of counted pulses. Voltmeter measuring PWM voltage is controlled by PC<br />

computer using GPIB network. The whole measurement system is controlled<br />

with program in LabVIEW environment.<br />

Keywords: displacement sensor, shape memory alloys, complex measuring<br />

system, electro-thermo-mechanical characteristics of SMA<br />

ŁUSZCZYK M.: Peak-to-sidelobe level vs. time-bandwidth product<br />

improvement for complex radar signals with small base<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 75<br />

The mostly applicable radar signal is pulse signal with linear frequency<br />

modulation (LFM). The main feature such signal is time-bandwidth product<br />

which is calculated as product of pulse duration T and signal bandwidth<br />

B. Theoretical peak to sidelobe level is achievable with time-bandwidth<br />

product value grater than 100. For time-bandwidth product smaller then<br />

100 (i.e. short pulse duration or narrow deviation of FM modulation) peakto-sidelobe<br />

level (PSL) and pulse compression coefficient are significantly<br />

reduced. Non-linear frequency modulated (NLFM) radar signal for small<br />

time-bandwidth product features better PSL and pulse compression coefficient.<br />

The NLFM signal synthesis algorithm and simulation results are<br />

presented in the paper. LFM and NLFM signals with small time-bandwidth<br />

product are compared and results are discussed in aspect of radar resolution<br />

improvement.<br />

Keywords: complex radar signal, pulse compression, NLFM, time-bandwidth<br />

product<br />

BRZOSTEK-PAWŁOWSKA J.: Między Web 2.0 i 3.0: Mobilne systemy<br />

informacyjne z rozszerzoną rzeczywistością<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 79<br />

Dynamiczny od 2009 r. rozwój mobilnych systemów informacyjnych wykorzystujących<br />

technologie rozszerzonej rzeczywistości (Augmented Reality<br />

– AR) niekiedy nazywanych przeglądarkami otoczenia (reality browsers),<br />

czerpiących informacje m.in. ze źródeł społecznie tworzonego kontentu i<br />

kontekstowo, w sposób zindywidualizowany, udostępniających informacje<br />

końcowemu użytkownikowi, zachęcił Autorkę do przedstawienia zasad<br />

działania i architektury tych systemów, jak również przedstawienia raczkującego<br />

udziału polskich deweloperów w prognozowanym na najbliższe<br />

lata - świetnym rozwoju tych systemów. Systemy te są przykładem implementacji<br />

idei Web 2.0 i Web 3.0 oraz „przetwarzania w chmurze”.<br />

Słowa kluczowe: mobilne systemy informacyjne, przetwarzanie w chmurze,<br />

usługi sieciowe, portale społecznościowe, Web 2.0, Web 3.0, ciekawy<br />

punkt (Point of Interest – POI), rozszerzona rzeczywistość (Augmented<br />

Reality – AR), aplikacje mobilne, przeglądarka otoczenia, przeglądarka<br />

AR, sklepy internetowe z mobilnymi aplikacjami<br />

BRZOSTEK-PAWŁOWSKA J.: Between Web 2.0 and Web 3.0: Mobile<br />

information systems using Augmented Reality<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 79<br />

Dynamic growth can be observed from the 2009 mobile information systems<br />

with Augmented Reality technology (AR), sometimes called reality<br />

browsers. These browsers draw information including sources from social<br />

networking and contextually on an individual basis to end user. The observable<br />

technology trend encouraged the author to present the principles<br />

of operation and architecture of these systems, as well as present the<br />

beginning involvement of Polish developers with this technology, and also<br />

the forecasts for these systems which have a steep trend upwards for<br />

development. These systems are an example of implementing the idea of<br />

Web 2.0 and Web 3.0 and the “cloud computing”.<br />

Keywords: Mobile information systems, cloud computing, Web services,<br />

social networking, Web 2.0, Web 3.0, Point of Interest POI, Augmented<br />

Reality AR, reality browser, AR browser, mobile applications, app store<br />

<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Streszczenia artykułów ● Summaries of the articles<br />

DUCHIEWICZ J., SOWA A.S., WITKOWSKI J.S., FRANCIK A., DOBRU-<br />

CKI A.L., IDŹKOWSKI B.: Zastosowanie wieloczęstotliwościowego<br />

sygnału koherentnego do pomiaru charakterystyk częstotliwościowych<br />

układów wąskopasmowych m.cz. oraz w.cz.<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 88<br />

W artykule opisano metodę jednoczesnego pomiaru wielu słabych sygnałów<br />

okresowych. Częstotliwości poszczególnych sygnałów zostały tak wybrane,<br />

że ich pomiar jest dokonywany z wykorzystaniem detekcji synchronicznej<br />

(koherentnej). Opracowano specjalny odbiornik do pomiaru sygnałów w.cz.,<br />

wykorzystujący demodulator I&Q oraz dwustopniową detekcję koherentną.<br />

Dzięki wykorzystaniu wieloczęstotliwościowej detekcji koherentnej uzyskano<br />

dużą czułość odbiornika i duża dynamikę pomiaru. Podano przykład zastosowania<br />

opisanej metody do pomiaru charakterystyki częstotliwościowej<br />

filtru kwarcowego o częstotliwości środkowej 30 MHz.<br />

Słowa kluczowe: wieloczęstotliwościowy sygnał koherentny, wieloczęstotliwościowa<br />

detekcja synchroniczna<br />

KORCZ K.: Radiokomunikacyjne aspekty planu implementacji strategii<br />

e-nawigacji<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 92<br />

Przedstawiono ogólne założenia, cele i kluczowe elementy strategii e-nawigacji<br />

w żegludze morskiej. Omówiono priorytetowe potrzeby użytkowników<br />

e-nawigacji. Zaprezentowano zagadnienia radiokomunikacyjne<br />

powiązane ze wstępnym planem implementacji strategii e-nawigacji. Na<br />

koniec przedstawiono perspektywy koncepcji e-nawigacji oraz Światowego<br />

Morskiego Systemu Łączności Alarmowej i Bezpieczeństwa GMDSS.<br />

Słowa kluczowe: radiokomunikacja morska, e-nawigacja, Światowy Morski<br />

System Łączności Alarmowej i Bezpieczeństwa (GMDSS), systemy<br />

informacyjne<br />

PŁOCH D., ŚLIŻ P., SZEREGIJ E.: Urządzenie do generacji silnych i ultrasilnych<br />

pól magnetycznych oraz sprzężony z nim układ pomiarowy<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 97<br />

W artykule zaprezentowano aparaturę do wytwarzania ultra-silnych impulsowych<br />

pól magnetycznych, znajdującą się w Zakładzie Elektroniki Fizycznej<br />

Uniwersytetu Rzeszowskiego. Opisano sposób wykonania cewki<br />

roboczej oraz system akwizycji sygnałów pomiarowych. Omówiono system<br />

sterujący, wykorzystujący środowisko LabView oraz aplikacje kart analogowo-cyfrowych.<br />

Przedstawiono przykładowe wyniki eksperymentalne dla<br />

struktur półprzewodnikowych z podwójnymi studniami kwantowymi<br />

Słowa kluczowe: pole magnetyczne, LabView, rezonans magnetofononowy<br />

(MPR), źródło prądowe<br />

TARIOV A., TARIOVA G.: Aspekty algorytmiczne organizacji jednostki<br />

procesorowej do mnożenia liczb Cayleya<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 104<br />

W artykule zostały przedstawione aspekty algorytmiczne organizacji<br />

dedykowanej jednostki obliczeniowej przeznaczonej do przyspieszenia<br />

procedury wyznaczania iloczynu dwóch liczb Cayleya (oktonionów), reprezentujących<br />

obok kwaternionów rozszerzenie algebry liczb zespolonych.<br />

Atutem proponowanej struktury jest zredukowana dwukrotnie liczba<br />

bloków mnożenia względem naiwnej metody implementacji owej operacji.<br />

Przy syntezie omawianej struktury algorytmicznej została zastosowana<br />

reprezentacja macierzowa operacji mnożenia oktonionów, co pozwala<br />

przedstawić mnożenie liczb Cayleya za pomocą iloczynu wektorowomacierzowego.<br />

Uwzględnienie pewnych relacji pomiędzy elementami tej<br />

macierzy pozwala zmniejszyć liczbę operacji mnożenia niezbędnych do<br />

realizacji procedury mnożenia oktonionów.<br />

Słowa kluczowe: liczby hiperzespolone, mnożenie liczb (oktaw) Cayleya,<br />

szybki algorytm mnożenia oktonionów<br />

DUCHIEWICZ J., DOBRUCKI A., FRANCIK A., STACHOWICZ W.,<br />

OLEŚ T., DUCHIEWICZ T.: Spektrometr Elektronowego Rezonansu<br />

Paramagnetycznego (EPR), umożliwiający ilościowe pomiary liczby<br />

spinów w badanej próbce<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 117<br />

Przedstawiono budowę dwukanałowego spektrometru EPR, umożliwiającego<br />

pomiary liczby spinów względem wzorca. W skład takiego spektrometru,<br />

oprócz typowych bloków (blok mikrofalowy, odbiornik sygnału EPR oraz<br />

stabilizator pola magnetycznego) wchodzą: podwójny rezonator pomiarowy,<br />

dodatkowy układ odbiorczy sygnału EPR oraz program sterujący, zapewniający<br />

jednoczesną rejestrację sygnału pochodzącego od próbki badanej oraz<br />

od próbki odniesienia. Rozważono również możliwość przekonstruowania<br />

opracowanego spektrometru EPR na pasmo L na spektrometr nadający się<br />

szczególnie do badań dozymetrycznych napromieniowanej żywności.<br />

Słowa kluczowe: ilościowe pomiary EPR, dwukanałowy spektrometr<br />

EPR, spektrometr EPR do badań dozymetrycznych<br />

DUCHIEWICZ J., SOWA A.S., WITKOWSKI J.S., FRANCIK A., DOBRU-<br />

CKI A.L., IDŹKOWSKI B.: Using multi-frequency coherent signal to measurement<br />

of frequency response of narrow-band LF and HF circuits<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 88<br />

In this paper a method allowing simultaneous measurement of tested objects<br />

using many periodic signals and applied to a receiver of the authors’ design has<br />

been described. Frequencies of the signals were chosen in such a way that<br />

the measurement executed is the coherent measurement of all used signals.<br />

The receiver was applied effectively to measurements of HF signals utilizing<br />

an I&Q demodulator and two-stage coherent detection. Good sensitivity of the<br />

receiver and high dynamic range of the measuring set-up were obtained due<br />

to the application of coherent detection. Multiple shortening of the required time<br />

of measurement was obtained due to simultaneous multi-signal measurement.<br />

An example demonstrating the usefulness of the technology of simultaneous<br />

coherent measurement of many signals has been described.<br />

Keywords: multi-frequency coherent signal, multi-frequency synchronous<br />

demodulation<br />

KORCZ K.: Radiocommunication aspects of an e-navigation strategy<br />

implementation plan<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 92<br />

The general assumptions, goals and key elements of the marine e-navigation<br />

strategy have been presented. The priority users needs of an<br />

e-navigation was described. The radiocommunication issues concerning<br />

the preliminary plan of an e-navigation strategy implementation have been<br />

presented. At the end the future of an e-navigation concept and Global<br />

Maritime Distress and Safety System (GMDSS) have been presented.<br />

Keywords: Maritime Radiocommunication, e-navigation, Global Maritime<br />

Distress and Safety System (GMDSS), information systems<br />

PŁOCH D., ŚLIŻ P., SZEREGIJ E.: Device for the generation of strong<br />

and ultrastrong magnetic fields with the measuring system<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 97<br />

This article describes a system for generating strong pulsed magnetic<br />

fields, located in the Department of Physical Electronics, University of<br />

Rzeszow. Was presented to the charging and discharging capacitors, as<br />

well as checking the data using the LabView environment. The paper presents<br />

experimental results for semiconductor double quantum wells.<br />

Keywords: magnetic field, LabView, magnetophonon resonance (MPR),<br />

current source<br />

TARIOV A., TARIOVA G.: Algorithmic aspects of Cayley numbers multiplier<br />

organization<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 104<br />

In work the rationalized algorithmic structure of processing unit for Cayley numbers<br />

product calculating with the reduced number of multiplications is presented.<br />

Since multiplier requires much more hardware than adder, fewer multiplications<br />

imply law power. Therefore, reducing the number of multiplications in VLSI processors<br />

design is usually a desirable task. This approach allows to lower hardware<br />

expenses and creates favorable conditions for effective convolution realization<br />

in the reprogrammable platform. The computational procedure for Cayley<br />

numbers multiplication is described in matrix notation. This notation enables us<br />

to represent adequately the space-time structure of an implemented computational<br />

process and directly maps this structure into the hardware realization space.<br />

The proposed structure can be successfully applied to accelerate calculations in<br />

FPGA based platforms as well as enhance efficiency of hardware in general.<br />

Keywords: Cayley numbers, multiplications of number<br />

DUCHIEWICZ J., DOBRUCKI A., FRANCIK A., STACHOWICZ W.,<br />

OLEŚ T., DUCHIEWICZ T.: Electron Paramagnetic Resonance (EPR)<br />

spectrometer for quantitative measurements of the spins number in<br />

the sample under test<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 117<br />

The construction of the two-channel EPR spectrometer, enabling quantitative<br />

measurements of the spins number with regard to the model sample<br />

has been described. A such spectrometer consists apart of standard blocks<br />

(microwave unit, receiver of EPR signal and magnetic field stabilizer)<br />

and of the double measuring resonator, the additional EPR signal receiver<br />

and the control program enabling simultaneous recording of the EPR<br />

signal of the sample under test and of the reference sample. A rebuilding<br />

of the designed L-Band EPR spectrometer to be suitable for quantitative<br />

dosimetry of the irradiated food has been considered.<br />

Keywords: EPR quantitative measurements, two-channel EPR spectrometer,<br />

EPR spectrometer for dosimetry<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Streszczenia artykułów ● Summaries of the articles<br />

DARAKCHIEV R., SITEK P., BRZOSTOWSKI K., FOKS-RYZNAR A.,<br />

KALARUS M., ZDUNEK R.: Moduły lokalizacji oraz testy chipów<br />

GNSS dla elementów systemu PROTEUS<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 122<br />

System PROTEUS ma za zadanie stworzenie nowej jakości w zarządzaniu<br />

kryzysowym i działaniach ratowniczych. Innowacyjne projekty elementów<br />

systemu, w szczególności częściowo autonomicznych robotów mobilnych<br />

oraz modułów lokalizacji dla ludzi w budynkach wymagają nowych rozwiązań<br />

technicznych. Niniejszy artykuł prezentuje w jaki sposób najnowsze<br />

technologie mogą być ze sobą połączone w celu stworzenia jakościowo<br />

nowych modułów lokalizacji, które spełniają ścisłe wymagania środowiskowe,<br />

zasilania, wielkości i dokładności.<br />

Słowa kluczowe: PROTEUS, zarządzanie kryzysowe, robot mobilny,<br />

mobilne centrum dowodzenia, przenośny zestaw czujników, bezzałogowy<br />

statek latający, nasobny zestaw czujników, GNSS<br />

KALARUS M., SITEK P., BRZOSTOWSKI K., FOKS-RYZNAR A.,<br />

DARAKCHIEV R.: Analiza możliwości zastosowania sensorów inercjalnych<br />

MEMS w projekcie PROTEUS<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 127<br />

Artykuł prezentuje rozważania nad możliwością zastosowania sensorów inercjalnych<br />

MEMS (Micro Electro-Mechanical System) do wyznaczania zmian<br />

położenia oraz orientacji obiektów. Analiza przeprowadzona zastała w kontekście<br />

realizowanego projektu PROTEUS – zintegrowanego mobilnego systemu<br />

mającego wyznaczyć nowe standardy w podejmowaniu działań antyterrorystycznych<br />

i antykryzysowych [1]. Jako element testowy wykorzystano moduł<br />

Xsens MTiG zawierający zintegrowany system INS/GPS (Inertial Navigation<br />

System/Global Positioning System). Przedstawiono jego charakterystykę<br />

oraz możliwości wykorzystania wewnątrz budynków w warunkach braku widoczności<br />

satelitów nawigacyjnych. Otrzymane rezultaty pozwoliły stwierdzić,<br />

że wykorzystanie podsystemu INS do klasycznej nawigacji zliczeniowej (dead<br />

reckoning) nie ma praktycznie sensu z uwagi na szybko narastający błąd.<br />

Jednak INS znakomicie sprawdza się jako uzupełnienie GPS umożliwiając<br />

pomiar orientacji oraz szybkich zmian położenia obiektu. Ponadto, analizy<br />

czasowo-częstotliwościowe danych otrzymywanych z INS pozwalają w pewnym<br />

stopniu określać parametry ruchu oraz pozycję obiektu.<br />

Słowa kluczowe: nawigacja inercjalna, lokalizacja, MEMS, PROTEUS<br />

DARAKCHIEV R., SITEK P., BRZOSTOWSKI K., FOKS-RYZNAR A.,<br />

KALARUS M., ZDUNEK R.: The location modules and tests of GNSS<br />

receivers for the elements of PROTEUS system<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 122<br />

The PROTEUS system will set a new standard in crisis management and<br />

rescue operations. The innovative design of partially autonomous mobile<br />

robots and modules for people location in buildings requires new technical<br />

solutions. This article presents how the recent technologies can be<br />

combined in order to design brand-new location modules that fulfill strict<br />

environmental, power supply, size and accuracy requirements.<br />

Keywords: PROTEUS, crisis management, mobile robot, mobile command<br />

centre, mobile sensors set, unmanned aerial vehicle, personal navigation<br />

set, GNSS<br />

KALARUS M., SITEK P., BRZOSTOWSKI K., FOKS-RYZNAR A.,<br />

DARAKCHIEV R.: An analysis of the applicability of MEMS inertial<br />

sensors in the PROTEUS project<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 127<br />

This article considers the possibility of using MEMS (Micro Electro-Mechanical<br />

System) inertial sensors to determine changes in position and<br />

attitude of the objects. The analysis was carried out in the context of the<br />

ongoing project PROTEUS – the integrated mobile system which aims<br />

to set new standards for counterterrorism and rescue operations [1]. All<br />

measurements were based on the Xsens MTiG module containing an integrated<br />

INS/GPS (Inertial Navigation System/Global Positioning System).<br />

Its performance and possible use inside the buildings during GPS outages<br />

were then presented. The results revealed that the classical dead reckoning<br />

algorithm is practically meaningless due to the unlimited position error<br />

growing with time. However, the INS can supplement GPS measurements<br />

which gives the capability of tracking the rapid changes in position and<br />

attitude. In addition, the time-frequency analysis of data obtained from the<br />

INS allows to estimate some movement parameters and position of the<br />

object.<br />

Keywords: inertial navigation, localization, MEMS, PROTEUS<br />

PERGOŁ M., ZIENIUTYCZ W., SOROKOSZ Ł.: Antena mikropaskowa<br />

o poszerzonym paśmie pracy<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 130<br />

W pracy przedstawiono wyniki badań dotyczące anteny mikropaskowej<br />

o poszerzonym paśmie pracy. Poprzez zastosowanie układu dwóch<br />

szczelin sprzęgających we wspólnej warstwie masy uzyskano 29% pasmo<br />

pracy (0,96…1,25 GHz) mierzone przy poziomie WFS < 1,5. Ze względu<br />

na obserwowane w charakterystyce promieniowania minima, wykonano<br />

drugą wersję anteny ze zmodyfikowanymi wymiarami warstwy masy. Efektem<br />

tego zabiegu było uzyskanie anteny o 29% paśmie pracy charakteryzującej<br />

się symetryczną wiązką główną.<br />

Słowa kluczowe: anteny mikropaskowe, charakterystyka promieniowania<br />

MĄKA T.: Właściwości konturów cech w zadaniach klasyfikacji sygnałów<br />

akustycznych<br />

<strong>Elektronika</strong> (LI), nr 11/<strong>2010</strong>, s. 134<br />

W pracy przedstawiono technikę pozwalającą na określanie klasy sygnału<br />

dźwiękowego poprzez wykorzystanie właściwości konturów cech. W<br />

zaproponowanym podejściu zastosowano wykrywanie pików w konturach<br />

przy użyciu zmiennego progu decyzyjnego oraz fuzji atrybutów konturów.<br />

Na podstawie analizy statystycznej uzyskanego zbioru odległości między<br />

pikami dla określonych konturów cech, możliwe jest określenie klasy<br />

sygnału. W celu weryfikacji prezentowanego podejścia przedstawiono<br />

zastosowanie wyników analizy konturów cech oraz funkcji decyzyjnej pozwalające<br />

w efektywny sposób (z dokładnością 98% dla użytego zbioru<br />

testowego) dokonywać klasyfikacji segmentów dźwiękowych zawierających<br />

mowę oraz muzykę.<br />

Słowa kluczowe: kontury cech, wykrywanie pików, klasyfikacja akustyczna<br />

PERGOŁ M., ZIENIUTYCZ W., SOROKOSZ Ł.: Broadband microstrip<br />

patch antenna<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 130<br />

In the paper the broadband microstrip antenna has been presented. The<br />

antenna is supplied by two slots located in the common groundplane. Due<br />

to this special method of electromagnetic coupling the bandwidth is greater,<br />

equal to 29% (0.96…1.25 GHz) at the level of SWR < 1.5. However,<br />

the radiation pattern of designed antenna was unacceptable – some<br />

undesired minima occurred. As a consequence, the antenna has been<br />

modified, the dimension of the groundplane has got shorter, so that the<br />

radiation pattern was satisfied.<br />

Keyword: microstrip anten nas, radiation pattern<br />

MĄKA T.: Properties of feature contours for audio classification<br />

tasks<br />

<strong>Elektronika</strong> (LI), no 11/<strong>2010</strong>, p. 134<br />

A technique for classifying audio segments based on properties of feature<br />

contours is described. The proposed approach uses a simple method utilizing<br />

peaks detection procedure with adaptive thresholding and fusion of<br />

contours attributes. It is possible to determine the signal class based on<br />

statistical analysis of the distances set between peaks for selected feature<br />

contours. In order to validate presented method, results analysis of feature<br />

contours along with decision function was applied to the discrimination<br />

problem between speech and music signals. In the result, obtained classification<br />

accuracy was 98% for the considered test set.<br />

Keywords: feature contours, peaks detection, audio classification<br />

<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Nonlinear compact thermal model of SiC power<br />

semiconductor devices<br />

(Nieliniowy skupiony model termiczny półprzewodnikowych elementów<br />

mocy wykonanych z węglika krzemu)<br />

dr hab. inż. KRZYSZTOF GÓRECKI, prof. dr hab. inż. JANUSZ ZARĘBSKI,<br />

mgr inż. DAMIAN BISEWSKI, dr inż. JACEK DĄBROWSKI<br />

Akademia Morska w Gdyni, Katedra Elektroniki Morskiej<br />

Abstraction of the heat dissipated in semiconductor devices<br />

depends generally on three phenomena: conduction, convection<br />

and radiation [1, 2]. One of these mechanisms can be<br />

dominant depending on the inner temperature T j<br />

of the chip<br />

and on the difference of the temperatures between the device<br />

case and the surrounding [1].<br />

In modeling of the device cooling the microscopic or macroscopic<br />

thermal models are used [2, 3]. In the microscopic<br />

models the space distribution of the temperature in the chip is<br />

considered. Unfortunately, these models are complex, therefore<br />

they need a long time of calculations and practically they<br />

can be only used for an analysis of thermal properties of the<br />

chip of the discrete semiconductor devices [4, 5]. On the other<br />

hand, the compact thermal models are willingly used in the<br />

analysis of electronic circuits [6, 7, 8].<br />

The compact thermal models describe the difference between<br />

the device inner temperature (corresponding to one or<br />

a few selected points) and the ambient temperature T a<br />

generally<br />

expressed by the convolution of the thermal power p th<br />

dissipated in the device and the derivative of the transient<br />

thermal impedance Z(t) of the device [9]. In the static (d.c.)<br />

conditions the difference of the temperatures T j<br />

and T a<br />

is equal<br />

to the product of the power p th<br />

(the constant value) and the<br />

thermal resistance R th<br />

. The parameter R th<br />

corresponds to the<br />

value of Z(t) in the thermal steady-state.<br />

The device compact thermal model instead of the convolution<br />

integral is very often formulated as an equivalent electrical<br />

circuit of the form of the Cauer [10] or the Foster [3, 4,<br />

6, 7, 11] networks representing the device transient thermal<br />

impedance, excited by the current source of the efficiency corresponding<br />

to the power dissipated in the device [5, 7, 10].<br />

As it results from, e.g. [8, 10], both the network constructions<br />

of the thermal model (Fig. 1) are fully equivalent from<br />

the point of view of the terminal T j<br />

, but the Foster network<br />

has no direct physical interpretation, whereas the Cauer network<br />

results directly from dyscretization of the one-dimensional<br />

heat transfer equation. The values of the RC parameters<br />

existing in the considered networks can be estimated with the<br />

use of specialist methods or algorithms, e.g. the algorithm<br />

ESTYM [11].<br />

a)<br />

R 1 R 2 Foster R n<br />

T j T a<br />

p th<br />

C 1<br />

C 2 C n T a<br />

Fig. 1. Device thermal models based on the Foster (a) and the Cauer (b) network<br />

Rys. 1. Modele termiczne elementów półprzewodnikowych bazujące na sieci<br />

Fostera (a) i Cauera (b)<br />

b)<br />

The compact thermal models presented in the literature<br />

are linear models; that means the influence of the device inner<br />

temperature on the efficiency of the heat abstraction is<br />

not included in these models. On the other hand, as it results<br />

from the authors’ investigations [12–14] the thermal resistance<br />

and the transient thermal impedance of the device strongly<br />

depend on the device dissipated power – consequently on the<br />

device inner temperature.<br />

In the paper the compact nonlinear thermal model of modern<br />

SiC devices is proposed. This model was experimentally<br />

verified for SiC-MESFET and SiC-SBR (Schottky Barrier Rectifier).<br />

The form of the nonlinear thermal model<br />

As it results from the measurements of the transient thermal<br />

impedance of any device [11, 12], the course of Z(t) depends,<br />

among others on the device dissipated power, which means<br />

that the values of the parameters (RC elements) existing in<br />

the considered models have to depend on the power.<br />

This phenomenon is proved by the measurements and the<br />

use of the algorithm ESTYM [11]. So, to estimate correctly the<br />

device inner temperature the nonlinear compact thermal model<br />

with RC parameters depending on the dissipated power<br />

has to be formulated.<br />

In order to obtain it, five following stages have to be performed:<br />

a) in the first stage the device transient thermal impedance<br />

in the wide range of the dissipated power should be measured.<br />

Note, that the lowest power value should cause<br />

the device junction temperature to increase so high as to<br />

secure the correct accuracy of the measurements. On the<br />

other hand, at the highest value of the dissipated power<br />

the inner temperature cannot exceed the maximum allowable<br />

device temperature given in the catalogue;<br />

b) in the second stage of this procedure, the values of the<br />

elements R i<br />

,C i<br />

(Cauer network) are estimated with the<br />

use of the algorithm ESTYM at various values of the power.<br />

The Cauer network is chosen because of its physical<br />

origin;<br />

c) in the third stage the dependence<br />

R i<br />

(p th<br />

) and C i<br />

(p th<br />

) are drafted. Then,<br />

Cauer R' 1 R' 2 R' n<br />

T j T a<br />

p th C' 1 C' 2 C' n<br />

T a<br />

on the basis of these dependences,<br />

the proper approximation function is<br />

fitted;<br />

d) in the fourth stage the values of the<br />

parameters existing in the dependences<br />

C i<br />

(p th<br />

) and R i<br />

(p th<br />

) are estimated;<br />

e) in the last stage the proper model<br />

of the network form (see Fig. 2) is formulated<br />

and implemented to SPICE.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


e R1<br />

Tj<br />

...<br />

Fig. 2. Network representation of the model Z(t) with the parameters<br />

depending on the power<br />

Rys. 2. Reprezentacja obwodowa modelu Z(t) z parametrami uzależnionymi<br />

od mocy<br />

The efficiencies of the controlled sources existing in Fig. 2<br />

are expressed by<br />

du<br />

i<br />

i<br />

Ci<br />

=<br />

C<br />

i<br />

⋅<br />

(1)<br />

dt<br />

e =<br />

R<br />

⋅<br />

i<br />

(2)<br />

where C i<br />

and R i<br />

are described by the formulae given in Section<br />

III. The values of the derivative du/dt are calculated with<br />

the use of DDT function.<br />

Estimation of the model parameters<br />

The practical realization of the algorithm described in illustrated<br />

on the example of the MESFET transistor CRF24010F<br />

and of the Schottky diode SDP10S30. Both the considered<br />

devices have operated without any heat-sink at the ambient<br />

temperature T a<br />

= 20 o C.<br />

Ri [K/W]<br />

Ci [J/K]<br />

10<br />

p th i C1<br />

e R2<br />

100<br />

10<br />

1<br />

i C2<br />

Ri<br />

= i<br />

Ri<br />

i CN<br />

e RN<br />

0,1<br />

0 0,5 1 1,5 2 2,5 3 3,5<br />

p th [W]<br />

100<br />

10<br />

1<br />

0,1<br />

0,01<br />

0,001<br />

0,0001<br />

R 6<br />

C 6<br />

C 4<br />

C 5<br />

R 4<br />

R 5<br />

R 2<br />

R 3<br />

CRF24010F<br />

R 1<br />

CRF24010F<br />

C 1<br />

C 2<br />

C 3<br />

0,00001<br />

0 0,5 1 1,5 2 2,5 3 3,5<br />

p th [W]<br />

Fig. 3. Measured (points) and calculated (lines) dependences<br />

R i<br />

(p th<br />

) and C i<br />

(p th<br />

) for the transistor CRF24010F<br />

Rys. 3. Zmierzone (punkty) i obliczone zależności R i<br />

(p th<br />

) oraz<br />

C i<br />

(p th<br />

) dla tranzystora CRF24010F<br />

Tabl. 1. Values of the parameters existing in Eqs (3), (4) for the transistor<br />

CRF24010F<br />

Tab. 1. Wartości parametrów występujących we wzorach (3), (4) dla<br />

tranzystora CRF24010F<br />

i C i0<br />

a i1<br />

a i2<br />

b i1<br />

b i2<br />

R i0<br />

d i1<br />

d i2<br />

e i1<br />

e i2<br />

p i1<br />

p i2<br />

p i3<br />

p i4<br />

1 10 -5 10 -5 -3 10 0.2 0.3 0.7 0.5 4 0.5 2 2 1<br />

2 0.003 0.9 -0.7 1.5 3 23 -0.3 -0.7 35 78 2 3 -2 -3<br />

3 0.13 0.9 -0.7 2.5 5 1.65 0.3 0.7 1 24 2 3 2 1<br />

4 0.27 0.9 -0.7 2.5 6 2.5 3 3 0.7 3 2 3 1 1<br />

5 0.4 0.9 -0.7 2.5 6 700 -0.15 -0.85 35 90 2 3 -2 -4<br />

6 9.2 0.9 -0.7 2.5 6 163 -0.15 -0.85 35 90 2 3 -2 -4<br />

According to the point a) of the algorithm, the courses<br />

of the transient thermal impedance Z(t) at the device dissipated<br />

power in the wide range from are measured.<br />

Then, according to the point b) of the algorithm, the dependences<br />

R i<br />

(p th<br />

) and C i<br />

(p th<br />

) are obtained with the use of the<br />

algorithm ESTYM. These dependence represented by points<br />

are shown in Fig. 3 for the transistor CRF24010F.<br />

According to the point c) of the algorithm, the following description<br />

of the dependence C i<br />

(p th<br />

) and R i<br />

(p th<br />

) is proposed<br />

⎡<br />

⎛<br />

p<br />

⎞ ⎛<br />

⎞⎤<br />

th<br />

− pi<br />

1<br />

pth<br />

− p<br />

i<br />

2<br />

C =<br />

⋅ ⎢<br />

+<br />

⋅<br />

⎥ (3)<br />

⎜−<br />

+ ⋅<br />

⎣<br />

⎟<br />

⎜−<br />

⎟<br />

i<br />

Ci0 1 a exp<br />

a exp<br />

i1<br />

i 2<br />

⎝ bi<br />

1 ⎠ ⎝ b<br />

i<br />

2<br />

⎠<br />

⎦<br />

⎡<br />

⎛<br />

p<br />

⎞ ⎛<br />

⎞⎤<br />

th<br />

− pi3<br />

pth<br />

− p<br />

i<br />

4<br />

R<br />

=<br />

⋅ ⎢<br />

+<br />

⋅<br />

⎥ (4)<br />

⎜−<br />

+ ⋅<br />

⎣<br />

⎟<br />

⎜−<br />

⎟<br />

i<br />

Ri<br />

0<br />

1 d exp<br />

d exp<br />

i1<br />

i 2<br />

⎝ ei<br />

1 ⎠ ⎝ e<br />

i<br />

2<br />

⎠<br />

⎦<br />

where C i0<br />

, a i1<br />

, a i2<br />

, b i1<br />

, b i2<br />

, d i1<br />

, d i2<br />

, e i1<br />

, e i2<br />

, R i0<br />

, p i1<br />

, p i2<br />

, p i3<br />

, p i4<br />

are the<br />

model parameters. The values of these parameters, obtained<br />

by matching the measurements and calculations results (the<br />

fourth stage in the algorithm) are collected in Table 1 for the<br />

transistor CRF24010F.<br />

To estimate the correctness of the approximation procedure,<br />

the dependences R i<br />

(p th<br />

) and C i<br />

(p th<br />

) were calculated with<br />

Eqs. (3) and (4). The calculation results are shown in Fig. 3<br />

with the use of solid lines. As seen the points and the solid<br />

lines fit well.<br />

Verification of the model accuracy<br />

To verify the accuracy of the presented model, in Figs. 4 and 5<br />

the measured (solid lines) and calculated (dotted lines) courses<br />

of the transient thermal impedance of the diode SDP10S30 at<br />

five values of the device dissipated power (Fig. 4) and of the transistor<br />

CRF24010F at six values of the device dissipated power<br />

are compared. As seen, the measured and calculated results fit<br />

well, which confirms good accuracy of the proposed model.<br />

To justify the necessity of using the device nonlinear thermal<br />

model, the following experiment has been performed. The<br />

time dependences of the inside temperature T j<br />

(t) of the investigated<br />

transistor CRF24010F obtained with the use of both<br />

the nonlinear (Fig. 2) and the classical linear device thermal<br />

models of the device dissipated power of the various shapes<br />

are calculated. As it was shown in [15], the considered dependence<br />

obtained from the nonlinear model differ from the other<br />

ones even more than 30%.<br />

It is worth mentioning, that using the nonlinear thermal<br />

model in the d.c. thermal analysis is not justified, because<br />

there is only one parameter – the thermal resistance existing<br />

in the model. So, it is necessary to find the dependence R th<br />

(p th<br />

).<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Using the measurement results for the considered transistor<br />

the following empirical dependence is proposed<br />

⎡<br />

⎛ p<br />

⎞ ⎛ ⎞⎤<br />

= ⋅ ⎢<br />

⋅<br />

⎜ −<br />

th<br />

p<br />

+<br />

⋅<br />

⎥ (6)<br />

⎣<br />

⎜ −<br />

th<br />

R<br />

⎟<br />

⎟<br />

th<br />

Rth<br />

max<br />

a1<br />

exp a2<br />

exp<br />

⎝ b1<br />

⎠ ⎝ b<br />

2<br />

⎠<br />

⎦<br />

where R thmax<br />

, a 1<br />

, a 2<br />

, b 1<br />

and b 2<br />

are the model parameters.<br />

In Table 2 the values of the parameters describing Eq. (6)<br />

for various cooling conditions of the investigated devices are<br />

collected. The investigations were carried out for the diode<br />

SDP10S30 and the transistor CRF24010F operating both<br />

without any heat-sink and situated on the small heat-sink.<br />

Measurements were performed with the electrical methods<br />

described in [14].<br />

The results of the measurements – points and the calculations<br />

according to Eq. (6) – lines of the dependence of the<br />

diode and the transistor thermal resistance on the dissipated<br />

power are presented in Fig. 6.<br />

As seen, the measured and calculated results differ from<br />

each other not more than 5%, which confirms well the proposed<br />

model.<br />

In the presented figures, decreasing of the thermal resistance<br />

of both the investigated devices operating at the various<br />

cooling conditions on increasing of the device dissipated<br />

power is observed. It is also seen, that the dependence<br />

R th<br />

(p th<br />

) is stronger for devices operating without any heat-sink.<br />

Z(t) [K/W]<br />

35<br />

30<br />

25<br />

20<br />

15<br />

10<br />

5<br />

SDP10S30<br />

p th = 3.3 W<br />

p th = 4.7 W<br />

p th = 1 W<br />

p th = 2.1 W<br />

p th = 5.7 W<br />

0<br />

0,0001 0,001 0,01 0,1 1 10 100 1000 10000<br />

t [s]<br />

Fig. 4. Measured (solid lines) and calculated (dotted lines) courses<br />

of Z(t) of the diode SDP10S30<br />

Rys. 4. Zmierzone (linie ciągłe) i obliczone (linie kreskowe) przebiegi<br />

Z(t) diody SDP10S30<br />

Z(t) [K/W]<br />

100<br />

80<br />

60<br />

40<br />

20<br />

CRF24010F<br />

T a = 20 o C<br />

p th = 0.51 W<br />

p th = 1.06 W<br />

p th = 1.5 W<br />

p th = 0.21 W<br />

p th = 2.03 W<br />

p th = 2.72 W<br />

0<br />

0,0001 0,001 0,01 0,1 1 10 100 1000 10000<br />

t [s]<br />

Fig. 5. Measured (solid lines) and calculated (dotted lines) courses<br />

of Z(t) of the transistor CRF24010F<br />

Rys. 5. Zmierzone (linie ciągłe) i obliczone (linie kreskowe) przebiegi<br />

Z(t) tranzystora CRF24010F<br />

Tabl. 2. Values of the parameters in Eq. (6) for the diode SDP10S30<br />

and the transistor CRF24010F<br />

Tab. 2. Wartości parametrów w równaniu (6) dla diody SDP10S30<br />

i tranzystora CRF24010F<br />

a)<br />

Parameters<br />

Rth [K/W]<br />

b)<br />

Rth [K/W]<br />

Diode without<br />

any heat-sink<br />

Diode on<br />

the<br />

heat-sink<br />

Transistor<br />

without any<br />

heat-sink<br />

Transistor<br />

on the<br />

heat-sink<br />

R thmax<br />

78 34 93 22.2<br />

80<br />

70<br />

60<br />

50<br />

40<br />

30<br />

20<br />

10<br />

a 1<br />

0.88 0.9 0.85 0.92<br />

a 2<br />

18 25 18 45<br />

b 1<br />

0.12 0.1 0.15 0.08<br />

b 2<br />

1.5 1.5 0.7 1.5<br />

0<br />

0 1 2 3 4 5 6<br />

p th [W]<br />

Fig. 6. Calculated and measured dependences of the thermal resistance<br />

of the diode SDP10S30 (a) and the transistor CRF24010F<br />

(b) on the dissipated power<br />

Rys. 6. Obliczone i zmierzone zależności rezystancji termicznej<br />

diody SDP10S30 (a) oraz tranzystora CRF24010F (b) od mocy<br />

It probably results from the fact, that the influence of the heat<br />

conduction, radiation and convection is different in the devices<br />

situated on the heat-sink and operating without it.<br />

As it is seen, that in the considered range of device dissipated<br />

power the device thermal resistance can change even<br />

more than 20%.<br />

Conclusions<br />

SDP10S30<br />

device without any heat-sink<br />

device on the heat-sink<br />

100<br />

90<br />

CRF24010F<br />

80<br />

device without any heat-sink<br />

70<br />

60<br />

50<br />

40<br />

30<br />

20<br />

device on the heat-sink<br />

10<br />

0<br />

0 2 4 6 8 10<br />

p th [W]<br />

In the paper the compact nonlinear thermal model of two SiC<br />

semiconductor devices is proposed. The accuracy of this model<br />

is verified on the example of the power MESFET transistor<br />

CRF24010F and the Schottky diode SDP10S30. A good agreement<br />

between the measurements and the calculations with<br />

the use of the new model in the wide range of changes of the<br />

device dissipated power and for various conditions of theirs<br />

cooling is achieved.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 11


The examples included a strong influence of the power dissipated<br />

in the device on the values of its thermal parameters<br />

– the thermal resistance and the transient thermal impedance.<br />

For the device operating without a heat-sink, the changes of<br />

thermal resistance corresponding to the considered changes<br />

of the power equal to even 25% are observed. So, omitting<br />

nonlinearities in the device thermal model can be the reason<br />

of serious errors in the calculations of the device inner temperature,<br />

which is responsible among others for the device<br />

reliability.<br />

The proposed nonlinear thermal model can be used in the<br />

construction of the electrothermal models of the considered<br />

power semiconductor devices, dedicated to the analyze and<br />

design electronic circuits.<br />

References<br />

[1] Janke W.: Zjawiska termiczne w elementach i układach<br />

półprzewodnikowych. WNT, Warszawa, 1992.<br />

[2] Lisik Z.: Zjawiska w strukturach półprzewodnikowych – metody<br />

ich modelowania. Wydawnictwo Politechniki Łódzkiej, Łódź,<br />

2005.<br />

[3] Zarębski J.: Tranzystory MOS mocy. Fundacja Rozwoju Akademii<br />

Morskiej w Gdyni, Gdynia, 2007.<br />

[4] Zarębski J.: Modelowanie, symulacja i pomiary przebiegów elektrotermicznych<br />

w elementach półprzewodnikowych i układach<br />

elektronicznych. Prace Naukowe Wyższej Szkoły Morskiej<br />

w Gdyni, Gdynia, 1996.<br />

[5] Napieralski A., Grecki M.: Two-dimensional Simulation of Semiconductor<br />

Devices with Temperature Computation Inside the<br />

Silicon Structure. Electromagnetic Phenomena in Nonlinear Circuits,<br />

Poznań 1994, pp. 195–200.<br />

[6] Szekely V.: A New Evaluation Method of Thermal Transient<br />

Measurement Results. Microelectronic Journal, vol. 28, no. 3,<br />

1997, pp. 277–292.<br />

[7] Székely V.: Identification of RC Networks by Deconvolution:<br />

Chances and Limits. IEEE Transactions on Circuits and Systems-<br />

I Theory and Applications, vol. 45, no. 3, 1998, pp. 244–258.<br />

[8] Zarębski J., Górecki K.: Modelling CoolMOS Transistors in<br />

SPICE. IEE Proceedings on Cicuits, Devices and Systems, vol.<br />

153, no. 1, 2006, pp. 46–52.<br />

[9] Zarębski J., Górecki K.: Properties of Some Convolution Algorithms<br />

for the Thermal Analysis of Semiconductor Devices. Applied<br />

Mathematical Modelling, vol. 31, no. 8, 2007, pp. 1489 –1496.<br />

[10] Bagnoli P.E., Casarosa C., Ciampi M., Dallago E.: Thermal resistance<br />

analysis by induced transient (TRAIT) method for power<br />

electronic devices thermal characterization. I. Fundamentals and<br />

theory. IEEE Transactions on Power Electronics, 1998; vol. 13,<br />

no. 6, pp. 1208–19.<br />

[11] Górecki K., Zarębski J.: Parameters estimation of thermal model<br />

of semiconductor devices. Kwartalnik Elektroniki i Telekomunikacji,<br />

vol. 52, no. 3, 2006, pp. 347–360.<br />

[12] Górecki K., Zarębski J., Stepowicz W.J.: The influence of selected<br />

factors on the thermal parameters of semiconductor devices.<br />

<strong>Elektronika</strong>, no. 10, 2005, pp. 18–20.<br />

[13] Górecki K., Zarębski J.: Microcomputer system for measuring<br />

thermal parameters of a class of semiconductor devices and integrated<br />

circuits. Metrology and Measurement Systems, vol. VIII,<br />

no. 4, 2001, pp. 379–396.<br />

[14] Zarębski J., Górecki K.: A Method of Measuring the Transient<br />

Thermal Impedance of Monolithic Bipolar Switched Regulators.<br />

IEEE Transactions on Components and Packaging Technologies,<br />

Vol. 30, No. 4, 2007 pp. 627–631.<br />

[15] Górecki K., Zarębski J., Bisewski D., Dąbrowski J.: Nonlinear<br />

compact thermal model of SiC power semiconductor devices. 17 th<br />

International Conference Mixed Design of Integrated Circuits and<br />

Systems MIXDES <strong>2010</strong>, Wrocław, <strong>2010</strong>, pp. 365–370.<br />

Krummenacher feedback analysis for high-count-rate<br />

semiconductor pixel detector readout<br />

(Analiza sprzężenia zwrotnego Krummenachera pod kątem zastosowania<br />

w szybkich układach odczytu półprzewodnikowych detektorów mozaikowych)<br />

dr inż. ROBERT SZCZYGIEŁ<br />

Akademia Górniczo-Hutnicza w Krakowie, Katedra Metrologii, Kraków<br />

Semiconductor pixel detectors are widely used both in material<br />

science and medicine and in high energy physics applications.<br />

The readout electronics for such detectors usually has<br />

to match the pitch of the detector and provide fast processing<br />

of the incoming charge pulses. Applications specific integrated<br />

circuits (ASICs) are a natural solution in this field. Important<br />

constraints for these ASICs are low noise and low power consumption<br />

per unit of the area. For DC coupled pixels the frontend<br />

circuits have to provide a bias voltage and compensate<br />

detector leakage current.<br />

The feedback circuit proposed by F. Krummenacher [1] is<br />

a smart solution for charge sensitive amplifiers (CSA) providing<br />

both continuous reset and detector leakage current compensation.<br />

It is used in many applications for X-ray imaging<br />

[2], particle tracking detector systems [3–6] and other [7]. This<br />

paper presents an analysis of the feedback circuit for a high<br />

speed readout system from the designer point of view. An application<br />

in 90 nm CMOS technology dedicated for X-ray imaging<br />

is assumed.<br />

The following design targets for high speed pixel readout<br />

ASIC are of the highest importance:<br />

12<br />

– fast signal shaping – to achieve a count rate of 2 Mcps/<br />

channel the signal at the output of the readout chain should<br />

return to 1% of maximum value in less than 300 ns. Minimization<br />

of time constants is therefore one of the most important<br />

goals,<br />

– low noise – adjusted to future application. For X-ray imaging<br />

the usual requirement is clear separation of Cu alpha line from<br />

the noise. For silicon detectors of thickness in the range of 300<br />

μm this translates into the required ENC of < 200 e - rms,<br />

– low power – typical requirement is that the final system<br />

should be air-cooled. This practically translates into a power<br />

limit in the range of 200…500 mW/cm 2 and 20...50 μW<br />

for 100 × 100 μm pixel size.<br />

Another important issue which limits minimization of ASIC<br />

components is matching. The smaller process feature size the<br />

biggest is the impact of the matching on the design of the circuit.<br />

This paper presents a small signal analysis of the Krummenacher<br />

feedback assuming an implementation in 90 nm<br />

CMOS technology. Derived conclusions allow to point out the<br />

directions for optimization of the circuit for high-count-rate<br />

pixel applications.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


The circuit<br />

The scheme of the Krummenacher feedback used with a CSA<br />

is presented in Fig. 1a. A simplified analysis of the circuit was<br />

presented in [4]. It is important to point out the existence of the<br />

C d<br />

capacitance, which is essential for stable circuit operation.<br />

Fig. 2. Small signal schematic of the feedback with CSA<br />

Rys. 2. Schemat małosygnałowy układu sprzężenia zwrotnego i CSA<br />

a) b)<br />

Fig. 1. Scheme of CSA with the feedback circuit (a) and its simplified<br />

version (b)<br />

Rys. 1. Schemat CSA ze sprzężeniem zwrotnym (a) i jego uproszczona<br />

wersja (b)<br />

Two signal paths can be distinguished in the feedback:<br />

– first, high bandwidth signal path, via transistors M2 and M1,<br />

which can be modeled as an equivalence resistance of R f<br />

=<br />

2/g m1<br />

, where g m1<br />

is a transconductance of M1 and M2 (M1 and<br />

M2 have the same W/L and the same drain current I krum<br />

/2),<br />

– second via M2, which drain current is than integrated on<br />

a capacitance C d<br />

, and the voltage is than converted into<br />

a current by M3; it can be modeled with an inductance<br />

of L f<br />

= 2C d<br />

/(g m1<br />

g m3<br />

). The high value of C d<br />

significantly limits<br />

the bandwidth of this path.<br />

This simplification (see Fig. 1a) shows the main principle<br />

of the circuit operation – the leakage current of the detector<br />

connected to the input of the circuit flows into the high value<br />

inductance L f<br />

and does not gener output of the amplifier what<br />

would happen in the case of the ate the voltage offset at the<br />

current flowing through R f<br />

, which provides continuous reset<br />

for the charge amplifier.<br />

Analysis<br />

Small signal circuit behavior in low frequency range is dominated<br />

by capacitances C d<br />

and C in<br />

(see Fig. 1a), which in pixel<br />

applications are in the range of a few pF and tens to hundreds<br />

of fF respectively. The capacitances introduced by transistors<br />

M0-M4 are in the range of a few fF. Thus one can see that the<br />

majority of the capacitances in the circuit can be neglected as<br />

being parallel to C d<br />

or C in<br />

. Only the ones connected to the M1<br />

and M2 common source node remain. The total capacitance<br />

at this node can be expressed as:<br />

C 1<br />

= C jd0<br />

+ C dg0<br />

+C ds0<br />

+C gs1<br />

+C js1<br />

+C ds1<br />

+C gs2<br />

+C js2<br />

+C ds2<br />

where C jdX<br />

and C jsX<br />

are drain and source junction capacitances<br />

of transistor MX, while C dgX<br />

, C dsX<br />

and C gsX<br />

are drain-gate,<br />

drain-source and gate-source capacitances of MX<br />

respectively. Note that C gs2<br />

is the component of C 1<br />

as the amplifier<br />

has low output impedance. The scheme used for small<br />

signal calculations is presented in Fig. 2.<br />

Given the amplifier transfer function of −K v<br />

/(1 + s/ω g<br />

), solving<br />

the circuit equations leads to the fourth order transfer function<br />

R(s) with the second degree nominator.<br />

V<br />

R<br />

(<br />

s<br />

)<br />

=<br />

I<br />

out<br />

in<br />

R<br />

0<br />

(<br />

s<br />

−<br />

z<br />

1<br />

)(<br />

s<br />

−<br />

z<br />

2<br />

)<br />

=<br />

(<br />

s<br />

−<br />

p<br />

)(<br />

s<br />

−<br />

p<br />

)(<br />

s<br />

−<br />

p<br />

)(<br />

s<br />

−<br />

p<br />

)<br />

1<br />

In the discussed circuit all the transistors have similar drain<br />

current (from I krum<br />

/2 to I krum<br />

). As in deep submicron technologies<br />

transconductance of transistor depends primarily on the<br />

drain current, the assumption g mX<br />

>> g dsY<br />

is made for all further<br />

calculations (this is also true for the detector leakage current<br />

flowing into M3 if it does not exceed I krum<br />

/2).<br />

The circuit DC gain factor R DC<br />

can be than expressed as:<br />

R<br />

DC<br />

=<br />

R<br />

0<br />

z<br />

1<br />

z<br />

2<br />

p<br />

p<br />

p<br />

p<br />

1<br />

2<br />

3<br />

4<br />

≈<br />

g<br />

m<br />

3<br />

m<br />

1<br />

The R DC<br />

is related to the sensitivity of the circuit to the detector<br />

leakage current. By increasing I krum<br />

, so that g m3<br />

and g m1<br />

are increased the change of the V out<br />

due to the leakage current<br />

can be limited. However, this will increase the noise contributions<br />

from M1 and M3.<br />

A relatively high value (a few pF) capacitance of C d<br />

introduces<br />

a low frequency pole, which can be calculate using standard<br />

low pole separation technique, and expressed as:<br />

(<br />

g<br />

+<br />

g<br />

+<br />

g<br />

what is consistent with [1] and [4]. Note that the detector leakage<br />

current increases g m3<br />

and moves the pole to higher frequencies.<br />

Assuming high value of ω g<br />

(e.g. ω g<br />

> 10 7 rad/s) the high<br />

frequency pole can be separated in the similar way:<br />

⎡<br />

⎛ C<br />

⎞<br />

⎤<br />

≈ −<br />

⎢⎜<br />

⎟<br />

+<br />

f<br />

Kν<br />

2gm<br />

p4<br />

1<br />

ω<br />

g<br />

+ ⎥<br />

⎢⎣<br />

⎝ C<br />

in<br />

⎠<br />

C<br />

1<br />

⎥⎦<br />

The p 4<br />

has two components. The first is related to the amplifier<br />

pole increased by the capacitive feedback path. The second<br />

is related to the high frequency signal path in the Krummenacher<br />

feedback circuit.<br />

Before discussing the poles p 2<br />

and p 3<br />

it is worth to note the<br />

positions of the zeros in the circuit:<br />

2<br />

2<br />

g<br />

ds<br />

4<br />

+<br />

g<br />

ds<br />

2<br />

(<br />

g<br />

+<br />

g<br />

+<br />

g<br />

)<br />

g<br />

ds<br />

1<br />

+<br />

g<br />

g<br />

3<br />

ds<br />

0<br />

4<br />

2<br />

g<br />

≈<br />

)<br />

g<br />

g<br />

m<br />

1<br />

ds<br />

1<br />

ds<br />

0<br />

m<br />

3<br />

m<br />

1<br />

ds<br />

4<br />

m<br />

3<br />

p 1<br />

≈ −<br />

≈ −<br />

g<br />

m<br />

3<br />

C<br />

1<br />

+<br />

C<br />

d<br />

(<br />

g<br />

m<br />

1<br />

+<br />

g<br />

ds<br />

1<br />

)<br />

C<br />

d<br />

2<br />

g<br />

ds<br />

4<br />

+<br />

g<br />

z<br />

1<br />

≈ −<br />

2<br />

C<br />

d<br />

ds<br />

2<br />

2<br />

g<br />

z<br />

2<br />

≈<br />

−<br />

C<br />

m<br />

1<br />

1<br />

ds<br />

4<br />

+<br />

g<br />

g<br />

m<br />

3<br />

m<br />

1<br />

ds<br />

2<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 13


of the signal is related to the high frequency pole p 4<br />

, while p 2<br />

and p 3<br />

poles influence the trailing edge of the signal. In the<br />

case of C f<br />

= 7 fF a fast decaying oscillation is visible (p 2<br />

and<br />

p 3<br />

are complex conjugate poles here), when for C f<br />

= 9 fF the<br />

overshoot disappears (p 2<br />

and p 3<br />

are real numbers). The effect<br />

of the low frequency pole p 1<br />

is not visible, since it produces<br />

a very long time constant (in the range of a few microseconds)<br />

and results in a small, but very long signal overshoot.<br />

Fig. 5a presents simulated frequencies of the circuit poles<br />

as a function of the CSA feedback capacitance C f<br />

. For low values<br />

of C f<br />

the poles p 2<br />

and p 3<br />

are complex conjugate and give<br />

the same frequency, while for higher C f<br />

values they turn into<br />

real ones and the frequencies separate. Low values of C f<br />

are<br />

usually more interesting as they result in lower circuit noise.<br />

However, as it can be seen from the Fig. 4, setting C f<br />

below certain<br />

value for given feedback transistor parameters will produce<br />

oscillating timing response. As mentioned before, the second<br />

way of getting the real poles is decreasing the capacitance C 1<br />

.<br />

One of the options to achieve this goal is to limit the size of transistors<br />

M1 and M2. As it can be seen if Fig. 5b for given circuit<br />

configuration the poles stay real until certain M1 and M2 width<br />

value, above which they turn into complex conjugate.<br />

Note that in deep submicron technologies, comparing to technologies<br />

with larger feature sizes, transistors easier enter into the<br />

week inversion region where transconductance does not depend<br />

much on the transistor size [8]. This allows to change the transistor<br />

width (and so the terminal capacitances) in a wide range without<br />

significant change of its transconductance. E.g. for drain cura)<br />

Fig. 3. Simulated frequency response of the circuit. Locations<br />

of poles and zeros are indicated<br />

Rys. 3. Symulowana charakterystyka częstotliwościowa obwodu.<br />

Na wykresie wskazano lokalizacje zer i biegunów<br />

b)<br />

Fig. 4. Simulated time response of the circuit for two different<br />

values of CSA feedback capacitance C f<br />

(for I krum<br />

= 20 nA). Indicated<br />

timing constants correspond to the transfer function poles<br />

Rys. 4. Symulowany napięcia na wyjściu CSA dla dwóch różnych wartość<br />

pojemności sprzężenia zwrotnego C f<br />

(dla I krum<br />

= 20 nA). Wskazane<br />

stałe czasowe odpowiadają biegunom funkcji przejścia układu<br />

Fig. 5. Simulated pole frequencies plotted against CSA feedback<br />

capacitance C f<br />

(a) and the width of M1 and M2 (b)<br />

Rys. 5. Symulowane częstotliwości biegunów w zależności od<br />

pojemności sprzężenia zwrotnego C f<br />

(a) i szerokości tranzystorów<br />

M1 i M2 (b)<br />

By comparing p 1<br />

and z 1<br />

it can be seen that on the frequency<br />

characteristics z 1<br />

will always precede p 1<br />

by approx. two decades.<br />

Another observation refers to z 2<br />

and p 4<br />

. By decreasing C 1<br />

both are pushed in the direction of higher frequencies, and z 2<br />

cancels p 4<br />

for C 1<br />

→ 0.<br />

Depending on the transistor parameters, poles p 2<br />

and p 3<br />

can<br />

be either negative real numbers or complex conjugate numbers<br />

with negative real parts. This will lead to the timing response<br />

including either exponential or exponentially decaying sine/cosine<br />

functions. An example of frequency characteristics of the<br />

circuit with indicated all poles and zeros is presented in Fig. 3.<br />

To simplify the analysis we put ω g<br />

→ ∞, what limits the<br />

transfer function to 3 rd order. After dividing the nominator of<br />

the transfer function by s – p 1<br />

we get somewhat complex quadratic<br />

equation, which discriminant can be simplified to:<br />

This indicates two ways of achieving a non-oscillating timing<br />

response from the circuit: by increasing C f<br />

and by decreasing C 1<br />

.<br />

The Fig. 4 shows the transient simulation results of the<br />

circuit for two values of the CSA feedback capacitance C f<br />

.<br />

As it can be seen, for the C f<br />

= 7 fF the output signal has much<br />

higher overshoot than for C f<br />

= 9 fF. Increased C f<br />

efficiently removes<br />

the oscillations from the timing response, but the price<br />

to pay is the noise performance, as the output pulse amplitude<br />

is related to C f<br />

. The Fig. 4 also presents the timing constants<br />

corresponding to the transfer function poles. The leading edge<br />

14<br />

∆<br />

=<br />

C<br />

2<br />

2<br />

d g<br />

m<br />

1<br />

K<br />

ν<br />

C<br />

f<br />

−<br />

C<br />

1<br />

(C<br />

f<br />

+<br />

C<br />

in<br />

/<br />

K<br />

ν<br />

(<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Fig. 6. Simulated minimum CSA feedback capacitance C f<br />

for<br />

which the circuit has real poles only vs. the width of the transistors<br />

M1 and M2<br />

Rys. 6. Symulowana minimalna wartość pojemności sprzężenia<br />

zwrotnego CSA, dla które układ posiada wyłączenie bieguny rzeczywiste<br />

w zależności od szerokości tranzystorów M1 i M2<br />

rent of 10 nA, a transistor of W/L = 2 μm/0.5 μm has g m<br />

= 300 nS,<br />

while a transistor of W/L = 20 μm/0.5 μm has g m<br />

= 317 nS. This<br />

means the change of less than 6% in transconductance when<br />

the width of the transistor is increased ten times.<br />

Concluding previous sections the plot in Fig. 6 can be presented.<br />

It shows the minimum C f<br />

for which the circuit has real<br />

poles only against the width of M1 and M2 (which is proportional<br />

to C 1<br />

). The plot separates two areas: the top one, where the<br />

circuit parameters lead to exponential timing response, and the<br />

bottom one, where exponentially decaying sine/cosine response<br />

can be seen. Note that minimization of M1 and M2 width has<br />

its limit derived from the matching, what translates into Vout DC<br />

spread. The final notice is that usually some small overshoot<br />

can be allowed and faster signal trailing edges can be achieved<br />

by using circuit with complex conjugate poles which are “close<br />

to real” (have relatively small imaginary part).<br />

Conclusions<br />

In the article the detailed small signal analysis of a charge sensitive<br />

amplifier with well known Krummenacher feedback circuit<br />

was performed. The analysis pointed out the directions for optimization<br />

of the circuit as well as the limits for stable circuit operation.<br />

As a conclusion it was shown that a non-oscillating timing<br />

response of the circuit can be achieved by proper selection of the<br />

CSA feedback capacitance and the total capacitance of common<br />

source node of the Krummenacher feedback circuit. The tradeoffs<br />

to be kept in mind here are noise and matching performance.<br />

This work was supported by Polish Ministry of Science and<br />

Higher Education in years 2009–2011.<br />

References<br />

[1] Krummenacher F.: Pixel detectors with local intelligence: An<br />

IC designer point of view. Nucl. Instrum. Methods A vol. 305,<br />

pp. 527–532, 1991.<br />

[2] Ballabriga R., Campbell M., Heijne E. H. M., Llopart X., Tlustos L.:<br />

The Medipix3 Prototype, a Pixel Readout Chip Working in Single<br />

Photon Counting Mode With Improved Spectrometric Performance.<br />

IEEE Trans. Nucl. Sci. vol. 54, no. 5, pp. 1824–1829, Oct. 2007.<br />

[3] Hu Y.: High performance low noise charge preamplifier with DC<br />

coupling to particle silicon detectors in CMOS technology. Electronics<br />

Letters vol. 34, no. 13, pp. 1274–1275, June 1998.<br />

[4] Hu Y., Deptuch G., Turchetta R., Guo C.: A Low-Noise, Low-Power<br />

CMOS SOI Readout Front-End for Silicon Detectors Leakage<br />

Current Compensation with Capability. IEEE Trans. Nucl. Sci.<br />

vol. 48, no. 8, pp. 1022–1030, Aug. 2001.<br />

[5] Aspell P.et al.: Delta: a charge sensitive front-end amplifier with<br />

switched gain for low-noise, large dynamic range silicon detector<br />

readout. Nucl. Instrum. Methods A vol. 461, pp. 449–455, 2001.<br />

[6] Llopart X., Ballabriga R., Campbell M., Tlustos L., Wong W.:<br />

Timepix, a 65k programmable pixel readout chip for arrival time,<br />

energy and/or photon counting measurements. Nucl. Instrum.<br />

Methods A vol. 581, pp. 485–494, 2007.<br />

[7] Weng M., Mandelli E., Moses W. W., Derenzo S. E.: A High-<br />

Speed Low-Noise CMOS 16-Channel Charge-Sensitive Preamplifier<br />

ASIC for APD-Based PET Detectors. IEEE Trans. Nucl. Sci.<br />

vol. 50, no. 4, pp. 898–902, Aug. 2003.<br />

[8] Grybos P., Idzik M., Maj P.: Noise optimization of charge amplifiers<br />

with MOS input transistors operating in moderate inversion<br />

region for short peaking times. IEEE Trans. Nucl. Sci. vol. 54,<br />

no. 3, pp. 555–560, June 2007.<br />

Testing of interconnections with use of reduced-size<br />

signature-based diagnostic dictionary<br />

(Testowanie uszkodzeń w połączeniach z wykorzystaniem sygnaturowego<br />

słownika diagnostycznego o zredukowanym rozmiarze)<br />

dr inż. TOMASZ GARBOLINO, dr inż. KRZYSZTOF GUCWA,<br />

dr hab. inż. ANDRZEJ HŁAWICZKA, prof. P.Ś., Politechnika Śląska, <strong>Instytut</strong> Elektroniki<br />

There are distinguished two Interconnect Built In-Self Test<br />

(IBIST) categories: test-per-scan [1, 10] and test-per-clock<br />

[2–5, 8], which have distinctly different properties. Unfortunately<br />

test-per-scan IBIST require long testing time. This drawback<br />

does not occur in the test-per-clock IBIST technique,<br />

where the testing time is much shorter (i.e. several orders of<br />

magnitude) and where there is a potential ability to detect dynamic<br />

faults: delay faults, crosstalk and switching noise [6].<br />

A new type of the test-per-clock IBIST structure that has<br />

recently become being used for testing of interconnections represents<br />

a specific ring register R-LFSR [2, 3], where n lines of<br />

the Network Under Test (NUT) act during the test as feedback<br />

lines of the LFSR. The cells being in normal operation mode<br />

the n transmitters and n receivers of the network lines, during<br />

the test constitute a 2n-bit shift register, that together with n<br />

lines of the NUT make a R-LFSR. During the test, the R-LFSR<br />

generates consecutive 2n-bit pseudorandom test vectors. The<br />

half of each vector is a stimulus vector for the NUT.<br />

In order to detect a fault in interconnections it is mandatory<br />

to find out – for selected initial state SD – such the length m<br />

of the sequence of the register states, that the final state<br />

(signature) of such a sequence S m<br />

F 0<br />

associated with faultfree<br />

bus would differ from each state S m<br />

F j<br />

attributable to<br />

a faulty bus. In practice one has to find out the signature<br />

S m<br />

F 0<br />

for the fault-free R-LFSR unit as well as a set of signatures<br />

{S m<br />

F j<br />

} that correspond to defects that were modeled<br />

for the bus. In that way the diagnostic dictionary is obtained<br />

DD = S m<br />

F 0<br />

∪ {S m<br />

F j<br />

}.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 15


3 2 1 0 3 2 1 0 3 2 1 0<br />

I 11 I 10 I 9 I 8 I 7 I 6 I 5 I 4 I 3 I 2 I 1 I 0<br />

4 5 6 7<br />

All methods that employ the ring register and the associated<br />

diagnostic dictionary suffer from the problem that consists in<br />

fast growth of the dictionary size in pace with the increase of the<br />

number of connections under test and multiplicity of the considered<br />

faults. Diagnostic dictionary size, which is proportional to<br />

n r , for n as high as several hundreds enormously increase and<br />

it makes the method infeasible for practical application.<br />

The above-discussed problem was solved – to some extent<br />

– in [5] by splitting the n-bit bus into smaller k-bit fragments,<br />

where n >> k, with further application of many R‐LFSR units<br />

that are mutually identical but separated one from another and<br />

each unit is used to test its own fragment of the bus, where the<br />

total number of fragments is b = n/k. The example of such a solution<br />

for n = 12, k = 4 and b = 3 is shown in Fig. 1. The connections<br />

that form the bus are marked in Fig. 1 with wider lines and<br />

denoted by consecutive numbers I 0<br />

-I 11<br />

. For the test duration<br />

the bus is split into b = 3 mutually separated fragments and<br />

each fragment comprises k = 4 adjacent connections. Test procedure<br />

for each fragment employs identical R-LFSR unit that<br />

is made up of 2k = 8 D-type flip-flops. Please note that all the<br />

three R-LFSR structures in Fig. 1 are mutually disconnected.<br />

It was assumed that the 2k-bit registers R-LFSR#1,<br />

R‐LFSR#2 and R‐LFSR#3 are seeded with the same 2k‐bit initial<br />

state SD. This assumption is indispensable to enable use of<br />

the same diagnostic dictionary for all the three registers. Moreover,<br />

the same test duration of m clock cycles was assumed for<br />

all the R-LFSR units. Signatures obtained after m clock cycles<br />

in three substructures R-LFSR#1, R-LFSR#2 and R‐LFSR#3<br />

shall be denoted as S m<br />

#1, S m<br />

#2 and S m<br />

#3, respectively.<br />

Fault types that shall be considered henceforth include:<br />

stuck-at-0/1, AND/OR shorts of maximum p max<br />

= 3 adjacent lines,<br />

delay faults. Due to technological considerations the assumption<br />

was made that either AND or OR shorts may occur within the<br />

entire n-bit bus, but not the both shorts simultaneously. The maximum<br />

multiplicity of faults within the bus fragment to be tested<br />

with a single R‐LFSR should be not more than r max<br />

= 3. Individual<br />

single faults of bus lines shall be denoted with f i<br />

symbols, where<br />

i = 1, 2,...u. Consequently, F j<br />

= {fi 1<br />

, f i 2<br />

,...} represent sets of such<br />

single faults f i<br />

and bring together both single and multiple faults<br />

within every R-LFSR unit, where j = 1, 2,..., w. Cardinality of each<br />

set meets the condition 1 ≤ | F j<br />

| ≤ r max<br />

. Moreover, the set F 0<br />

= ∅<br />

corresponds to the case when the bus is free of faults. The 2k‐bit<br />

signature that corresponds to the fault represented by the set F j<br />

shall be denoted as S m<br />

F j<br />

. The signature-based dictionary that<br />

comprises 2k‐bit signatures S m<br />

F j<br />

is referred to as DD.<br />

Motivation<br />

4 5 6 7<br />

4 5 6 7<br />

R-LFSR 8 #3 R-LFSR 8 #2 R-LFSR 8 #1<br />

Fig. 1. The IBIST structure: three seprated 8-bit registers R‐LFSR<br />

Rys. 1. Struktura IBIST: trzy rozłączne 8-bitowe rejestry R-LFSR<br />

In case of any shorts between lines that belong to two neighbouring<br />

R‐LFSRs, states of these two registers mutually affect one<br />

another. Subsequently, it leads to the situation where any fault<br />

of other type that may occur in connections that are tested by<br />

one of these registers affects also values of signatures that are<br />

obtained for the both registers. Such signatures must be added<br />

to the fault dictionary, which results in its further extension. The<br />

described effect is illustrated by Example 1.<br />

Example 1: For the sake of simplicity, let us<br />

consider only two R‐LFSR registers (i th and i th + 1),<br />

each of them is 4-bit long. As it is mentioned in<br />

Section 1, the set of modeled faults covers: Sa0/<br />

Sa1, bridges between adjacent lines and delay<br />

faults. Let us assume that the circuit exhibits the<br />

fault f 1<br />

, which is an AND-type bridge and leads to<br />

a short between three lines (I 3<br />

, I 4<br />

and I 5<br />

). When<br />

the test for that fault is carried out with use of the<br />

method described in [5], the signature S m<br />

F ’ g<br />

is obtained<br />

in the register R-LFSR#(i+1) while the signature S m<br />

F ” g<br />

can be read form the register R-LFSR#i. The both signatures<br />

are assigned to the same fault.<br />

When to consider that multiplicity of fault can be as high<br />

as 3 for each register (i.e. one or two additional faults are<br />

taken into account, beside the f 1<br />

fault) the correct identification<br />

of the f 1<br />

fault shall need to store signatures for the<br />

f 1<br />

fault as well as for all possible combinations of faults on<br />

other lines of the i th R‐LFSR register (lines I 0<br />

, I 1<br />

and I 2<br />

) as<br />

well as all possible combinations of faults on other lines of<br />

the i th +1 R‐LFSR register (lines I 6<br />

and I 7<br />

). The faults that<br />

may additionally occur within the structure R‐LFSR#i include<br />

3 single faults of the Sa0 type, 3 single faults of the Sa1<br />

type as well as 12 double faults of the Sa-0/Sa-1 type. In<br />

addition, 3 single and 6 double delay faults may happen as<br />

well as 2 shorts of AND type between two lines and 1 ANDtype<br />

short between three mutually adjacent lines. Similarly,<br />

the R-LFSR#(i+1) register may experience additional faults.<br />

When to take also account for possible combinations of<br />

the mentioned faults within a single register it is necessary<br />

to consider in total z i<br />

= 45 faults for the i th R-LFSR register<br />

and z i+1<br />

=16 faults for the register i th + 1. Therefore the<br />

overall number of signatures that have to be stored for this<br />

case amounts to z i<br />

× z i+1<br />

= 720. When the width k of the<br />

bus fragment that is tested by an elementary R-LFSR register<br />

tends to increase the number of signatures that have<br />

to be stored grows at rocketing rates. The major reason for<br />

such a dramatic expansion of the diagnostic dictionary is<br />

the fact that any other fault within the register R‐LFSR#i or<br />

R‐LFSR#(i + 1) affects the sequence that is generated by<br />

the another register of the BIST structure. The above problem<br />

attracted the authors to find out another solution, allowing<br />

to reduce size of DD.<br />

The suggested solution<br />

The BIST structure shall be made up of b modules (2k-bit<br />

R-LFSRs), but both odd and even modules can work independently.<br />

It means that after seeding the even modules are used<br />

as BIST structures (Test operation) whilst the odd modules<br />

remain unaltered (Hold operation) or vice versa. The entire<br />

testing cycle comprises four phases. Each phase begins with<br />

the seeding operation and after completion of the tests the<br />

resulting signatures for specific phases can be read out. The<br />

operation modes of odd and even R-LFSRs as well as their<br />

seeds in four phases are shown in Tabl. 1.<br />

Example 2: Let us consider the same fault f 1<br />

as in Example<br />

1. Fig. 2 explains how the f 1<br />

fault is manifested during individual<br />

phases of the testing procedure due to introduction<br />

of the Hold mode: as a double fault {f 11a<br />

, f 11b<br />

} as well as single<br />

faults {f 12<br />

} and {f 13<br />

}.<br />

Tabl. 2 presents the set of signatures that are obtained<br />

for faults that are enforced by introduction of the Hold mode<br />

and occur in individual modules during subsequent phases of<br />

the testing procedure. The signature S m<br />

F 1<br />

corresponds to the<br />

double fault {f 11a<br />

, f 11b<br />

}, the signature S m<br />

F 2<br />

occurs for the single<br />

16<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Tabl. 1. Seed applied and operation mode of modules for each test<br />

phase<br />

Tab. 1. Stany początkowe I tryby pracy modułów w każdej z faz testowania<br />

Phase<br />

Odd module<br />

Even modules<br />

Seed(SD) Operation Seed(SD) Operation<br />

I 00..000 Hold SD Test<br />

II 11..111 Hold SD Test<br />

III SD Test 00..000 Hold<br />

IV SD Test 11..111 Hold<br />

phase I phase II phase III phase IV<br />

# i+1 # i<br />

f1<br />

I5 I4 0<br />

# i+1 # i<br />

I5 I4 1<br />

Fig. 2. Types of faults enforced by the Hold mode<br />

Rys. 2. Rodzaje uszkodzeń wymuszanych przez tryb Hold<br />

f1<br />

# i+1 # i<br />

f1<br />

0 0 I3<br />

fault {f 12<br />

} whereas the signature S m<br />

F 3<br />

is associated to the<br />

single fault {f 13<br />

}. Fig. 3 presents comparison between the<br />

localization process for the f 1<br />

fault when all the modules<br />

are operated at the same time [5] (Fig. 3a) and the similar<br />

process that is carried out with application of the method<br />

explained in this study (Fig. 3b). The part of the diagnostic<br />

dictionary denoted with R and marked with grey background<br />

comprises signatures for all possible combinations of<br />

faults that may occur at lines that are tested by the neighbouring<br />

R‐LFSR registers, where the set of possible<br />

faults contains at least one short between lines that are<br />

connected to different registers. For instance, these are<br />

signatures S m<br />

F ’ g<br />

and S m<br />

F ” g<br />

for Example 1.<br />

The part R presents a significant part<br />

of the diagnostic dictionary. It is the part<br />

of the diagnostic dictionary that can be<br />

eliminated if the method that is described<br />

in this study is applied (Fig. 3b).<br />

The new method needs to keep in the<br />

f1<br />

diagnostic dictionary only those signatures<br />

that correspond to not more than<br />

1 1 I3 threefold faults per each single module.<br />

This part of the diagnostic dictionary<br />

is the same (common) for the both<br />

methods and is denoted by F in Fig. 3.<br />

The bridge faults that connect the bus lines<br />

that are tested by neighbouring R-LFSR<br />

registers no longer lead to interaction between<br />

working sequences of the adjacent<br />

1<br />

0<br />

I3 registers as the neighbouring register<br />

is always in the Hold mode. It is why signatures<br />

related to faults of that type are excluded<br />

from the F part of the dictionary. This<br />

part comprises exclusively those signatures<br />

that are associated with the faults located<br />

within that fragment of the bus which<br />

is tested by the specific elementary register.<br />

# i+1 # i<br />

I5 I4 × 0 = 0 I5 I4 × 1 = I5 I4 0 × 0 ×I3 = 0 1 ×1 ×I3 = I3<br />

⇓<br />

# i+1 # i # i+1 # i<br />

⇓ ⇓ ⇓<br />

# i+1 # i<br />

# i+1 # i<br />

0<br />

1 0<br />

0<br />

I5 I4<br />

0 0<br />

Sa0 Sa0 AND Sa0 fault free<br />

f 11a f 11b f 12 f 13<br />

a) b)<br />

S F F 0<br />

= F 0 S m F 0<br />

=<br />

F j S m F j<br />

S m F j F j<br />

{f 11a , f 11b }= F 1 S m F 1<br />

m 0<br />

S m F 1 F 1 ={f 11a , f 11b }<br />

F v S m F v<br />

S m F v F v<br />

{f 12 }= F 2 S m F 2<br />

S m F 2 F 2 ={f 12 }<br />

{f 13 }= F 3 S m F 3<br />

S m F 3 F 3 ={f 13 }<br />

F v+1<br />

R {f 1 }= F g<br />

S m F v+1<br />

S m F’ g<br />

S m F” g<br />

phase<br />

phase<br />

IV<br />

III<br />

S#i,IV<br />

S#i,III<br />

F w S m F w<br />

F<br />

S#i<br />

S#(i+1)<br />

S#(i+1),I<br />

S#(i+1),II<br />

phase II<br />

phase I<br />

F<br />

Fig. 3. The dictionary‐based method for diagnosis of faults in interconnections: a) all modules work together; b) the proposed<br />

method with 4 phases of the testing procedure<br />

Rys. 3. Słownikowa metoda diagnostyki uszkodzeń w połączeniach: a) wszystkie moduły pracują równocześnie; b) proponowana<br />

metoda obejmująca 4 fazy testowania<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 17


The localization method<br />

length of 2k. In fact, it is the number of internal states that can<br />

be adopted by the register. In turn, the fourth column of Tabl. 3<br />

As it was mentioned, short-type faults are limited exclusively<br />

to AND-type shorts between adjacent lines. Each signature<br />

that can be read from a specific R-LFSR register unambiguously<br />

indicates a specific set of faults in interconnections that<br />

make up feedback lines for that register. In other words, the<br />

signature brings the information related to the I j<br />

line with specification<br />

of the fault type that could have occurred at that line<br />

or with the meaning that the line is fault free. When the signatures<br />

collected in the register R‐LFSR#(i+1) during the 1st and<br />

2nd phases (or in the R‐LFSR#i register during the 3rd and<br />

4th phases) enable identification of the same fault type at the<br />

I j<br />

line, it is the information that unambiguously determines the<br />

fault type that had happened to that line.<br />

1. If the signature<br />

lists average number α of mutually different internal states of<br />

the 2k-bit R-LFSR register that can be assigned to every fault<br />

considered in this study and that may occur at feedback lines<br />

of the register. As one can easily see, the average value of α<br />

rapidly grows in pace with increase of the k parameter. Thus,<br />

it is associated with increase of the probability that all the<br />

faults that are allowed for consideration can be detected and<br />

mutually differentiated. The columns 5 and 6 of Tabl. 3 make<br />

it possible to compare size of the diagnostic dictionary that<br />

is necessary respectively for the methods [2, 3] and [5], provided<br />

that maximum threefold faults were allowed for a k‐bit<br />

fragment of the n‐bit bus, where n ∈ {64, 128}. Moreover, in<br />

the case of solution presented in [5] it was assumed that any<br />

AND-type short between lines belonging to the neighboring<br />

obtained during modules affected only two such modules.<br />

Tabl. 2. Signatures obtained in each module<br />

for fault f 1<br />

the phase 1 (3) indicates<br />

the stuck-at Conclusion<br />

Tab. 2. Sygnatury otrzymane w każdym<br />

z modułów dla uszkodzenia f 1<br />

zero fault at the line<br />

Phase<br />

Signatures<br />

Fault f 1<br />

I j<br />

while the signature<br />

for the phase 2 (4)<br />

confirms absence of<br />

Module (i+1) Module i faults, it means that<br />

1 S the line in question is<br />

m<br />

F 1;<br />

{f 11a,<br />

f 11b<br />

} 00..000<br />

involved in the ANDtype<br />

short with a line<br />

2 S m<br />

F 2<br />

; {f 12<br />

} 11..111<br />

3 00..000 S m<br />

F 3<br />

; {f 13<br />

} or lines that belong<br />

to the neighbouring<br />

R-LFSR register.<br />

2. If the signature obtained during the phase 2 (4) indicates<br />

that a specific set L of mutually adjacent lines of the bus makes<br />

up the AND-type short while the signature for the phase 1<br />

(3) confirms that each line of that set is stuck-at zero, it means<br />

that lines of the L set are involved in the AND-type short with<br />

a line or lines that belong to the neighbouring register.<br />

The similar analysis of faults in neighbouring modules as<br />

well as familiarity with layout of mutually adjacent modules<br />

allows diagnostics of faults (determination how many lines are<br />

involved in a short between the modules under test).<br />

Size of the diagnostic dictionary<br />

Size of the dictionary for various values of the k parameter is<br />

quoted in the second column of Tabl. 3. Furthermore, the third<br />

column of that table contains number of all mutually different<br />

signatures that can be collected for an R-LFSR register with its<br />

Tabl. 3. Size of the diagnostic dictionary<br />

Tab. 3. Rozmiar sygnaturowego słownika diagnostycznego<br />

k<br />

2 3 4 5 6<br />

The proposed method enables detection, localization and identification<br />

of faults in interconnections that are organized in a bus<br />

structure. The advantage of the method consists in the fact that<br />

it guarantees substantial size reduction of the diagnostic dictionary,<br />

both in comparison with the dictionary that is used in<br />

studies [2, 3] as well as with the solution that is explained in [5].<br />

The proposed method allows reliable and dependable diagnostics<br />

of the considered faults that may be experienced for buses<br />

with their width as large as few hundreds bits when a diagnostic<br />

dictionary is used with the capacity of only about 543 thousand<br />

of 32-bit signatures. On the other hand, the drawback of the<br />

method is the test duration that is four times longer than is needed<br />

for conventional methods. The length m of a test sequence<br />

that guarantees high efficiency of the method per a single phase<br />

is only 4k > m > 2k [2, 3], so the number of test cycles only<br />

insignificantly affects the total duration of the test procedure.<br />

The most important factor that determines duration of the test is<br />

the time of seeding and time necessary to deliver the collected<br />

signatures to the ATE. However, taking account for the number<br />

of interconnections that exist in the circuits of SoC type, that<br />

time is acceptable from the point of engineering practice.<br />

References<br />

|DD| #sig α = #sig /|DD| |DD| [2, 3] |DD| [5]<br />

n = 64 n = 128<br />

4 241 256 1,06 7,04·10 37 4,96·10 75 3912<br />

8 4226 65536 15,50 6,11·10 28 3,73·10 57 387765<br />

12 19416 16777216 864,09 4,02·10 25 8,71·10 46 4813945<br />

16 53806 4294967296 79823,20 7,03·10 18 4,94·10 37 31383941<br />

20 115396 1,09951E+12 9528160,66 1,57·10 20 2,21·10 35 137706585<br />

24 212186 2,81475E+14 1326548296 8,86·10 15 7,85·10 31 461361525<br />

28 352176 7,20576E+16 2,04607E+11 4,09·10 16 4,86·10 27 1276096345<br />

32 543366 1,84467E+19 3,3949E+13 2,84·10 11 8,08·10 22 3063752325<br />

[1] Koeter J., Sparks S.: Interconnect Testing Using BIST Embedded<br />

in IEEE 1149.1 Designs. Proc. of Int. ASIC Conf., 1991, pp.<br />

P11-2.1- P11-2.4.<br />

[2] Hławiczka A. et al.: Interconnect Faults Identification and localization<br />

Using Modified Ring LFSR. Proc. of 11 th IEEE Workshop<br />

on Design and Diagnostics of Electronic Circuits and Systems<br />

– DDECS, 2008, pp. 247–250.<br />

[3] Hławiczka A. et al.: Application of Modified Ring-LFSR<br />

for Interconnect Faults Detection. Proc. of the 15th<br />

IEEE Int. Conference Mixed Design of Integrated Circuits<br />

and Systems – MIXDES, 2008, pp. 487–492.<br />

[4] Jutman A.: At-Speed On-Chip Diagnosis of Board-<br />

Level Interconnect Faults. IEEE European Test<br />

Symposium – ETS, 2004, pp. 2–9.<br />

[5] Garbolino T., Gucwa K., Hławiczka A.: How to Reduce<br />

Size of a Signature-based Diagnostic Dictionary<br />

Used for Testing of Connections. Proc. of<br />

the 13 th IEEE International Symposium on Design<br />

and Diagnostics of Electronic Circuits and Systems<br />

– DDECS, <strong>2010</strong>, pp. 201–204.<br />

[6] Attarha A., Nourani M.: Testing interconnects for<br />

noise and skew in gigahertz SoC. Proc. of Int. Test<br />

Conf., 2001, pp. 305–314.<br />

[7] Su C., Tseng W.: Configuration free SoC interconnect<br />

BIST methodology. Proc. of Int. Test Conf.,<br />

2001, pp. 1033–1038.<br />

[8] Pendurkar R., Chatterjee A., Zorian Y.: Switching activity<br />

generation with automated BIST synthesis for<br />

performance testing of interconnects. IEEE Trans.<br />

on CAD/ICS, vol.20, No 9, 2001, pp. 1143–1158.<br />

18<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


The Time-over-Threshold based silicon strip<br />

detector readout<br />

(Odczyt do krzemowych detektorów paskowych oparty<br />

o metodę Time-over-Threshold)<br />

mgr KRZYSZTOF KASIŃSKI<br />

AGH-University of Science and Technology, Department of Measurement and Instrumentation, Kraków<br />

Silicon strip or pixel detector is built of an array of reverse-biased<br />

diodes on the common silicon bulk. Doped areas create<br />

strips (in strip detectors) or pixels (in pixel detectors) which<br />

are sensing areas. The geometry of the detector determines<br />

its ability to detect the interaction location in one or two dimensions.<br />

The strip detector for example allows for acquisition of<br />

single dimension picture with the resolution defined by the<br />

strip pitch (typical 25…100 µm) [1].<br />

Certain number of electron-hole pair (related to the photon<br />

energy) is generated in the depleted region as the result of<br />

photon interaction with the detector. For example, each of the<br />

17.4 keV X-ray photon (Molybdenum) would generate 4833<br />

electron-hole pairs. In the High Energy Physics applications,<br />

the particles do not stop in the detector. As they are crossing<br />

the detector they leave some part of their energy. For example<br />

a minimal ionizing particle (MIP) generates 77 e-h pairs<br />

for each 1 µm of its path in the detector (which gives 23000<br />

e-h pairs for the 300 µm thick silicon detector). The generated<br />

charge is then acquired at the detector strips creating a nanosecond<br />

range current pulse which is then measured at the detector<br />

readout electronics [2]. The very small strip (pixel) pitch<br />

and their quantity puts significant limitations on the area and<br />

the power of each channel of the readout circuitry and forces<br />

to develop multichannel integrated circuits in deep submicron<br />

technologies.<br />

The current pulse from the detector is integrated in the<br />

charge sensitive amplifier (CSA) which is the first stage of<br />

the readout circuit. As a result, the voltage step occurs at its<br />

output with the amplitude equal to the input charge (Q in<br />

) and<br />

feedback capacitance (C fb<br />

) ratio. The polarity of this step is<br />

related to the input charge polarity.<br />

In order to prevent the pulse pile-up at the CSA output the<br />

circuits removing the charge acquired in the feedback before<br />

the next pulse arrives are necessary (reset circuits). It can<br />

be a resistor connected parallel to the feedback capacitor<br />

discharging the output voltage step in the time related to the<br />

R fb<br />

C fb<br />

– dependent time constant (Fig. 1).<br />

There are three main architectures in the field of the amplitude<br />

spectrometry circuits:<br />

– Binary readout,<br />

– Regular ADC in each or one per several channels,<br />

– Wilkinson type ADC in each channel.<br />

The binary readout allows only detecting the fact of the<br />

charge deposition over the certain threshold. The regular<br />

ADC (analog to digital converter) would be the best choice<br />

but it is always a tradeoff between the current consumption,<br />

area required, resolution achieved and the operation speed.<br />

Tight requirements often prevent from finding the good choice<br />

and sometimes require using one ADC per several channels<br />

which significantly affects the time limits of the whole<br />

circuit. The Wilkinson-type ADC (also called the Time-over-<br />

Threshold (ToT) processing) can be a possible solution. If<br />

the input charge information could be derived from the discriminator<br />

output pulse it would be possible to obtain a reasonable<br />

resolution while keeping the power consumption<br />

and the area low.<br />

Motivation<br />

The purpose of this work was to design a silicon strip detector<br />

readout chain optimized for long strips (30 pF detector<br />

capacitance). The channel functionality should cover both<br />

charge measurement (electrons and holes) and the ability<br />

to timestamp each event. The nominal input charge range is<br />

0…16 fC, recovery time for the input charge of 4 fC should<br />

be in the order of one microsecond. Each channel should be<br />

supplied by an ADC allowing high speed operation. The power<br />

budget for the whole channel is 2 mW. The channel width should<br />

not exceed 50 µm.<br />

Solutions of the ToT for the strip detectors [3–5] usually<br />

use a charge amplifier with a resistor-type reset circuit and<br />

a shaper (special filter cascade) but their characteristics (pulse<br />

width vs. input charge) is not linear. The key to the linear<br />

characteristics in the Time-Over-Threshold architecture can<br />

be a feedback circuit which discharges the feedback capacitor<br />

by a constant current. Such architectures are already used<br />

in pixel readout chips but the detector capacitances there are<br />

much smaller [6, 7].<br />

Architecture<br />

Fig. 1. Principle of the Charge Sensitive Amplifier (CSA) operation<br />

Rys. 1. Zasada działania wzmacniacza ładunkowego<br />

The channel consists of the CSA supplied with two,<br />

switchable constant-current reset circuits (Fig. 2a). The<br />

amplifier is followed by the discriminator circuit. The discriminator<br />

differential threshold is set by Th1 and Th2 lines<br />

while the actual CSA output DC level spread due to the<br />

process variation can be compensated by the 6-bit digital to<br />

analog converter (DAC) placed in each channel. The circuit<br />

provides the positive output pulse which length is related to<br />

the input charge.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 19


a) b)<br />

Fig. 2. a) Channel architecture, b) CSA core architecture<br />

Rys. 2. a) Architektura kanału, b) architektura rdzenia wzmacniacza ładunkowego<br />

Charge Sensitive Amplifier and the reset circuit<br />

The CSA module contains the amplifier itself, the feedback<br />

capacitance and two switchable feedback circuits.<br />

The PMOS-based folded cascode amplifier is a base for<br />

the CSA core (Fig. 2b). The high gain (1600 V/V) and low<br />

power (625 µW) (Tabl. 1) are the main advantages of this<br />

solution.<br />

The input transistor is a PMOS due to its better 1/f noise<br />

performance compared to the NMOS. It has been sized<br />

to achieve the best results for the 30pF detector capacitance<br />

while keeping in mind the area constraints and it operates<br />

with 500 µA current. The power supply of the M in<br />

transistor<br />

is separated and is lower than the main power supply. Since<br />

almost 90% of the total CSA current flows through the M in<br />

transistor this action significantly reduces the total CSA power<br />

consumption and additionally lowers the interference coupling<br />

from the following stages.<br />

Tabl. 1. Parameters of the csa core and summary of Channel’s power<br />

consumption<br />

Tab.1. Parametry wzmacniacza i zestawienie mocy kanału<br />

20<br />

CSA Parameter Value Circuit<br />

Voltage Gain 1600 V/V CSA + feedback<br />

circuits<br />

Power<br />

Consumption<br />

625 µW<br />

Bandwidth 760 kHz Discriminator 200 µW<br />

Power consumption 625 µW DAC 220 µW<br />

Phase Margin 80° TOTAL: 1,05 mW<br />

The voltage gain of the ideal CSA is defined by the feedback<br />

capacitance. The proposed value (20 fC) has been<br />

chosen in order to achieve the high gain while keeping the<br />

voltage step amplitude vs. detector capacitance dependence<br />

at the reasonable level.<br />

The proposed feedback circuit is simple but effective in<br />

this application. This solution has been presented by P. Fischer<br />

[8] for pixel architectures. Thanks to the constant-current<br />

operation the pulse width vs. input charge characteristic can<br />

Fig. 3. Feedback circuit architecture (for one input charge polarity)<br />

Rys. 3. Architektura układu rozładowania (dla jednej z polarności<br />

ładunku wejściowego)<br />

be linear. Due to the inability to work for both input charge<br />

polarities it was necessary to provide two switchable feedback<br />

circuits, one for each polarity. The switching is performed by<br />

four, cross-coupled CMOS transmission gates.<br />

This circuit is based on the current mirror (Fig. 3). The<br />

current from the external, off-channel reference is reflected<br />

in the mirror creating the current source I disch<br />

. This current<br />

flows all the time through the M 2<br />

transistor. Depending on<br />

the actual state of the CSA (idle or just after the charge deposition)<br />

the V ds<br />

of the M 1<br />

transistor changes polarity thus the<br />

current changes its direction. In the idle state, M 1<br />

takes the<br />

whole M 2<br />

current. When the input charge arrives, M 1<br />

starts to<br />

operate as the current mirror thus discharging the feedback<br />

capacitor.<br />

A very interesting feature of the presented feedback circuit<br />

is that the circuit keeps the linear pulse width vs. input charge<br />

characteristics even when the output of the amplifier itself is<br />

saturated which results in very high dynamic range (Fig. 4).<br />

This is thanks to the fact, that the feedback still operates correctly<br />

discharging the feedback capacitor.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Fig. 4. CSA characteristics and output pulse shapes<br />

Rys. 4. Charakterystyki i impulsy wyjściowe ze wzmacniacza ładunkowego<br />

B. Discriminator<br />

The task of the discriminator is to produce an output pulse<br />

lasting as long as the CSA output exceeds certain voltage level,<br />

called the threshold. The presented architecture consists<br />

of several blocks (Fig. 5).<br />

The first stage is the unary-gain differential amplifier<br />

(M 1<br />

–M 4<br />

). Its main task is to convert the single-ended CSA<br />

output signal into fully differential one in order to keep the<br />

operation of the next stage (actual discriminator) most effective.<br />

Additionally, this stage implements both differential<br />

threshold setting (pins Th1 and Th2) and the trimming functionality<br />

(DAC pin). The fully differential discriminator with<br />

hysteresis (M 5<br />

– M 10<br />

) has been described in details by P. Allen<br />

[9]. The transistors are sized to reach the high gain. The<br />

complex, cross-coupled load structure (M 7<br />

– M 10<br />

) introduces<br />

the hysteresis which prevents the multiple, short discriminator<br />

pulses resulting from the noise. M 11<br />

– M 14<br />

transistors provide<br />

the full swing output and increase the overall gain of the<br />

discriminator. In order to uniform the output pulse polarity no<br />

matter what is the polarization of the input charge, the pulse<br />

is inverted or buffered depending on the actual polarity line<br />

configuration.<br />

C. TrimDAC<br />

The purpose of the 6-bit, constant-current architecture Trim-<br />

DAC is to compensate for the CSA output DC level deviation<br />

(Fig. 6). Since the threshold is set globally for all the channels,<br />

the deviation changes the effective threshold setting. The reference<br />

currents for the DAC are to be set externally for both<br />

range and offset. The register supports the initialization to the<br />

default value which is equal to the half of the full range. The<br />

common-centroid layout has been used. The presented circuit<br />

keeps the linearity in the operation range and offset values<br />

exceeding the expected ones (resulting from the monte-carlo<br />

analysis).<br />

Fig. 5. Architecture of the discriminator Rys. 5. Architektura dyskryminatora<br />

Fig. 6. Simplified TrimDAC architecture (register, decoder and references are not included)<br />

Rys. 6. Uproszczony schemat układu TrimDAC (rejestr, dekoder oraz obwody referencyjne są pominięte)<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 21


a)<br />

b)<br />

c)<br />

Fig. 7. a) Discriminator pulse width vs. Input charge, b) CSA noise performance, c) DAC output voltage vs. register value<br />

Rys. 7. a) Długość impulsu dyskryminatora w funkcji ładunku wejściowego, b) wydajność szumowa wzmacniacza ładunkowego,<br />

c) napięcie wyjściowe przetwornika Cyfrowo-Analogowego w funkcji wartości w rejestrze<br />

Simulation Results<br />

The processing chain presented in this paper performs very<br />

well. According to the simulation results it provides linear response<br />

within the vast range of the input charges while keeping<br />

the noise low and power consumption very low which is<br />

very promising for the High Energy Physics application field.<br />

The channel should fit into the 50 µm x 1 mm channel slot in<br />

the integrated circuit (UMC 180 nm technology).<br />

Figure 7 contains the most important characteristics of the<br />

presented channel. Fig. 7a presents the Time-over-Threshold<br />

characteristic of the channel acquired for the certain threshold<br />

setting (approx. 2.2 fC) and discharge current configuration.<br />

The slope of this curve depends on the discharge current<br />

while its horizontal position on the threshold setting. Figure<br />

7b shows the equivalent noise charge (ENC) versus detector<br />

capacitance. The result of 710 e- for 30 pF while keeping very<br />

low power consumption and reasonable dead time (Tabl. 1) is<br />

very promising. Figure 7c depicts the DAC linearity for default<br />

external biasing resistors’ values controlling the DAC’s output<br />

offset and range.<br />

This work was supported by Ministry of Science and Higher Education,<br />

Poland.<br />

References<br />

[1] Gryboś P., Dąbrowski W.: Wykorzystanie paskowych detektorów<br />

krzemowych w dyfraktometrii promieniowania X. Kwartalnik Elektroniki<br />

i Telekomunikacji 48 (2002), z. 2, ss. 435–448.<br />

[2] Dąbrowski W.: <strong>Elektronika</strong> front-end do krzemowego detektora<br />

torów w eksperymencie ATLAS. Kwartalnik Elektroniki i Telekomunikacji<br />

48 (2002), z. 2, ss. 247–270.<br />

[3] Manfredi P. F., Leona A., Mandelli E., Perazzo A., Re V.: Noise<br />

limits in a front-end system based on time-over-threshold signal<br />

processing. Nuclear Instruments and Methods In Physics Research<br />

A, 439 (2000), 361–367.<br />

[4] Delagnes E. et al.: SFE16, a Low Noise Front-End Integrated<br />

Circuit Dedicated to the Read-Out of Large Micromegas Detectors.<br />

IEEE Transactions on Nuclear Science, 47 (2000), No. 4,<br />

1447–1453.<br />

[5] Kipnis I. et al.: A Time-over-Threshold Machine: the Readout Integrated<br />

Circuit for the BaBar Silicon Vertex Tracker. IEEE Transactions<br />

on Nuclear Science, 44 (1997), No. 34, 289–297.<br />

[6] Berg C. et al.: Bier&Pastis, a pixel readout prototype chip for<br />

LHC. Nuclear Instruments and Methods In Physics Research A,<br />

439 (2000), 80–90.<br />

[7] Huegging F. on behalf of the ATLAS Pixel Collaboration: Front-<br />

End Electronics and Integration of ATLAS Pixel Modules. Nuclear<br />

Instruments and Methods in Physics Research A, 549 (2005),<br />

157–164.<br />

[8] Rossi L., Fischer P., Wersmes N.: Pixel Detectors: From Fundamentals<br />

to Applications. Springer (2006).<br />

[9] Allen P. E., Holberg D. R.: CMOS Analog Circuit Design. Oxford<br />

University Press, 2002.<br />

Wpłata w <strong>2010</strong> roku – GwarancjĄ niŻszej ceny prenumeraty o vat!<br />

22<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


The design of low power 11.6 mW high speed 1.8 Gb/s<br />

stand-alone LVDS Driver in 0.18 µm CMOS<br />

(Projekt scalonego nadajnika standardu LVDS o niskim poborze mocy 11,6 mW<br />

i 1,8 Gb/s szybkości transmisji danych)<br />

mgr inż. RAFAŁ KŁECZEK, Akademia Górniczo-Hutnicza, Katedra Metrologii, Kraków<br />

According to Moore’s law CMOS technology has been exponentially<br />

developed over the several recent decades. This development<br />

is particularly visible in digital part where reducing<br />

the effective channel’s length is equivalent to increasing the<br />

frequency of circuits. Growth in the performance of microprocessor<br />

motherboards, optical transmission links, chip-to-chip<br />

communications has created a higher and higher speed requirements<br />

on physical layer interfaces. Due to limited performance<br />

of single link high data rates were achieved by parallelism<br />

in the past, that increased the complexity level and cost<br />

of the fabricated circuits. Nowadays data rates at the range<br />

of few hundred MHz up to or above is necessary. Reducing<br />

the power dissipation is very often one of the most critical modern<br />

circuits’ requirements.<br />

LVDS (Low Voltage Differential Signaling) standard offers<br />

high speed data transmission and low power consumption<br />

at the same time. Due to many advantages of differential<br />

operation: robustness of the link to supply lines and common<br />

mode voltage bouncing, reduction of coupling and radiated<br />

electro-magnetic interference (EMI), higher level of electromagnetic<br />

compatibility (EMC) [1–2] in conjunction with low<br />

voltage swing has become a popular choice for fast data onchip<br />

transmission, on board/backplane or cable connections.<br />

The benefits of LVDS make it valuable technology, that fits<br />

into a wide range of applications such as: PC/computing, telecommunication,<br />

automotive, etc. [3].<br />

This paper presents the design of low-power complete<br />

stand-alone LVDS driver in 180-nm CMOS technology.<br />

It offers a dense packing of transistors and implementation<br />

of mixed-mode circuits. It also occupies low area so makes<br />

the driver suitable for using in multichannel ASIC.<br />

LVDS Interface<br />

LVDS standard was specified by IEEE as a norm that based<br />

on the physical layer defined for the scalable coherent interface<br />

(SCI) [4] (separate industry standard was defined by TIA-<br />

EIA [5]). It specifies the dc and ac parameters for driver and<br />

receiver shown in Fig. 1.<br />

The most important of them are: differential voltage<br />

(V DIFF<br />

= V oa<br />

– V ob<br />

) and common mode voltage V CM<br />

= (V oa<br />

+ V ob<br />

)/2.<br />

A differential voltage is centered at a common mode voltage.<br />

Regarding a value of termination resistor R R<br />

and process, voltage,<br />

temperature (PVT) variations the ranges of this voltages<br />

are defined as:<br />

250 mV ≤ |V DIFF<br />

| ≤ 400 mV (1)<br />

1.125 V ≤ V CM<br />

≤ 1.275 V (2)<br />

For simplicity LVDS transmitter can be considered as current<br />

source with switched polarity, whereas the receiver is a comparator<br />

with hysteresis which can detect voltage signal at the<br />

level of tens of mV. The driver’s output current I D<br />

= 3.6 mA<br />

flows through differential pair lines and termination resistor<br />

R R<br />

= 100 Ω (in most cases R R<br />

= 2R L<br />

, R L<br />

is characteristic impedance<br />

of transmission line). The receiver has a high DC input<br />

impedance so the majority of driver’s output current flows<br />

across R R<br />

resistor and generates at the receiver input about<br />

360 mV, which can be detect as “one” logic state. Switching<br />

driver’s current polarity changes flow direction across R R<br />

resistor,<br />

thereby creating a valid “one” or “zero” logic state.<br />

LVDS Transmitter<br />

A stand-alone LVDS driver can be constructed of the following<br />

functional blocks (Fig. 2):<br />

• LVDS core: current source with switched polarity,<br />

• CMFB: common mode feedback circuit stabilizes the<br />

driver’s output common mode voltage V CM<br />

at the desired<br />

value, which equals V REF<br />

. Output’s voltage V CM<br />

is probed<br />

by resistive divider made of R P<br />

resistors.<br />

• Control block: block responsible for controlling the LVDS<br />

core, it buffers the input signal U DATA<br />

.<br />

• Band-gap: generates a V REF<br />

reference voltage, which should<br />

have no temperature dependence.<br />

During the simulation the receiver was modeled as a parallel<br />

combination of termination resistor R R<br />

= 100 Ω and capa-<br />

Fig. 1. Components of LVDS interface [4]<br />

Rys. 1. Elementy składowe interfejsu LVDS [4]<br />

Fig. 2. Block diagram of LVDS driver<br />

Rys. 2. Schemat blokowy nadajnika standardu<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 23


citor C R<br />

= 5 pF. In addition to complying with LVDS standard<br />

a designed LVDS driver had to be able to transmit data at speed<br />

400 Mb/s (more than 1 Gb/s for C R<br />

= 1 pF) and dissipate as<br />

low power as possible. The designed circuit was implemented<br />

in CMOS 180 nm technology with 1.8 V supply voltage.<br />

A. LVDS core<br />

To minimize dissipated power it is very important to choose<br />

appropriate core architecture. Lots of LVDS driver solutions<br />

can be found in publications [6–11].<br />

Fig. 4. Common Mode Feedback circuit [6]<br />

Rys. 4. Układ sprzężenia zwrotnego CMFB [6]<br />

Fig. 3. LVDS core: Bridged-Switched Current Sources architecture<br />

[6]<br />

Rys. 3. Rdzeń LVDS: architektura mostkowa [6]<br />

Bridged-Switched Current Sources (BSCS) architecture,<br />

shown in Fig. 3, is a typical solution [6, 8] for LVDS driver.<br />

It consist of two current sources M1, M6 and four MOS switches<br />

M2-M5 connected in bridge configuration. To simplify the<br />

core transistors’ gates M2-M4 and M3-M5 are shorted and<br />

controlled by complementary signal V IN+<br />

and V IN-<br />

. Applying logic<br />

‘1’ to V IN+<br />

and logic ‘0’ to V INswitches<br />

transistors M3-M4<br />

on and M2-M5 off. As a result a positive differential voltage<br />

V DIFF<br />

is at the input of receiver – logic ‘1’ was transmitted. In<br />

the opposite situation a logic ‘0’ is transmitted. Due to finite<br />

on-resistance of PMOS switches and level of common mode<br />

voltage V CM<br />

BSCS architecture has small headroom in the<br />

V DD<br />

supply voltage direction. That solution is more suitable<br />

for V DD<br />

voltage which is equal or above 2.5 V. For supply voltage<br />

1.8 V wider PMOS transistors are necessary. Capacitor<br />

C P<br />

= 50 pF makes the slew rate of V DIFF<br />

voltage shorter thanks<br />

to accumulated charges. The value of R P<br />

= 10 kΩ resistors used<br />

in V CM<br />

voltage probe circuit should be large compared with R R<br />

resistor and yet small enough to be able to implement in integrated<br />

circuit.<br />

B. Common mode feedback circuit<br />

Fig. 4 presents a implementation of common mode feedback<br />

circuit. The V CM<br />

common mode voltage at the driver’s output<br />

is sensed by resistive divider and compared with V REF<br />

reference<br />

voltage by the differential amplifier M1-M4. When<br />

V CM<br />

= V REF<br />

the currents flows through M1-M3 and M2-M4 branch<br />

of amplifier are equal. Current mirrors used in CMFB (V PMOS<br />

and<br />

V NMOS<br />

potentials) circuit set the output LVDS driver’s current at<br />

the desired value. In the contrary, the unequal currents flowing<br />

across M3 and M4 are mirrored to M1-M6 core’s transistors<br />

and forcing the V CM<br />

= V REF<br />

voltage. In order to minimize power<br />

consumption by CMFB a current ratio of used current mirrors<br />

(I core<br />

: I CMFB<br />

) were set as 40:1. R BIAS<br />

resistor sets the currents<br />

flowing through LVDS core and CMFB circuit so it should be<br />

implemented carefully in IC. It is also recommended to make<br />

24<br />

the connection with node of R BIAS<br />

resistor through the pad to be<br />

able to control the driver’s current by external resistor.<br />

Stability of designed negative feedback is critical. The phase<br />

margin Φ m<br />

= 95º ensures high level of negative feedback’s<br />

stability. This value is achieved by using a compensation circuit<br />

which is a serial combination of R C<br />

resistor and C C<br />

capacitor.<br />

The values: R C<br />

= 500 Ω and C C<br />

= 10 pF were obtained by<br />

AC simulation [12].<br />

C. Band-gap reference source<br />

Generation of temperature-independent V REF<br />

voltage has been<br />

achieved by using band-gap reference source [1]. Implementation<br />

of band-gap source requires an operational amplifier, with as<br />

low input offset voltage as possible. Designed op amp has a two<br />

stage architecture with Miller’s capacitance compensation. A no-<br />

Fig. 5. Corners simulation: V REF<br />

voltage dependence on temperature<br />

Rys. 5. Symulacje brzegowe: zależność napięcia V REF<br />

od temperatury<br />

Tabl. 1. Corner’s analysis settings<br />

Tab. 1. Parametry symulacji brzegowych<br />

Corner NMOS PMOS V supply<br />

[V] Temp. [ºC]<br />

TM typ typ 1.8 25<br />

WP fast fast 1.98 0<br />

WS_TMIN slow slow 1.62 0<br />

WS_TMAX slow slow 1.62 85<br />

WO_TMIN fast slow 1.98 0<br />

WO_TMAX fast slow 1.62 85<br />

WZ_TMIN slow fast 1.98 0<br />

WO_TMAX slow fast 1.62 85<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


minal V REF<br />

reference voltage of designed circuit equals 1.23 V,<br />

temperature coefficient is 0.3 ppm/ºC at T 0<br />

= 27ºC temperature,<br />

V REF<br />

voltages changes ΔV REF<br />

= 2.4 mV over the temperature<br />

range from 0ºC to 85ºC for typical conditions (TM plot in Fig. 5).<br />

V REF<br />

reference voltage sets the V CM<br />

common mode voltage at the<br />

driver’s output, so for all above cases an achieved V CM<br />

voltage is<br />

compatible with IEEE specification. Tabl. 1 presents the corners<br />

analysis settings, which were used during the simulation.<br />

D. Control block<br />

Control block circuit has been implemented as a two chains<br />

of inverters with complementary outputs, connected with the<br />

shorted gates of LVDS core’s switches. In simplify, the control<br />

blocks buffers U DATA<br />

driver’s input signal to LVDS core’s inputs.<br />

To minimize dynamic control block’s current consumption it is recommended<br />

to control the core of driver by the signal with rise/fall<br />

time around 0.5 ns. It significantly reduces the spikes of current<br />

driven from the supply lines during the switching of inverters.<br />

Simulation Results<br />

Designed LVDS driver characterizes a very low static 7.5<br />

mW and dynamic 8.5 mW power dissipation at data rate<br />

400 Mb/s (TM plot in Fig. 6). The rise/fall time of V DIFF<br />

dif-<br />

ferential voltage is 710/720 ps so the driver is capable<br />

of sending data with maximum rate about 500 Mb/s. The<br />

limits of fall/rise time are set by the output driver’s current<br />

and C R<br />

input capacitance at the receiver side (Fig. 7). According<br />

to the formula:<br />

dV<br />

DIFF<br />

I<br />

(3)<br />

D = C<br />

R<br />

dt<br />

achieving faster changes in V DIFF<br />

voltage for constant value<br />

of I D<br />

is dependent on capacitor’s value C R<br />

. Fig. 8 presents<br />

the V DIFF<br />

voltage at data rate 1.8 Gb/s and C R<br />

= 1 pF load. To<br />

increase the data rate there can be increased output driver’s<br />

current by external resistor connected with R BIAS<br />

resistor.<br />

That resistor can be also used to set value of V DIFF<br />

differential<br />

voltage when its desired level is changed by the process<br />

technology.<br />

Compatibility with IEEE specification limits the current flow<br />

across R R<br />

= 100 Ω termination resistor to maximum value<br />

4 mA. If the output driver’s current exceeds that limit it is recommended<br />

to decrease the resistance visible into receiver’s<br />

input. Achieved parameters of designed LVDS driver are<br />

shown in Tabl. 2.<br />

Fig. 6. Corners simulation: V DIFF<br />

voltage at data rate 400 Mb/s, C R<br />

= 5 pF<br />

Rys. 6. Symulacje brzegowe: napięcie V DIFF<br />

, transmisja danych<br />

400 Mb/s, C R<br />

= 5 pF<br />

Fig. 8. V DIFF<br />

voltage at data rate 1.8 Gb/s, C R<br />

= 1 pF<br />

Rys. 8. Napięcie V DIFF<br />

, transmisja danych 1,8 Gb/s, C R<br />

= 1 pF<br />

Tabl. 2. V DIFF<br />

voltage at data rate 1.8 Gb/s, C R<br />

= 1 pF<br />

Tab. 2. Napięcie V DIFF<br />

, transmisja danych 1,8 Gb/s, C R<br />

= 1 pF<br />

Fig. 7. Maximum data rate dependence on input’s receiver capacitance<br />

Rys. 7. Maksymalna szybkość transmisji danych w zależności od<br />

pojemności obciążenia<br />

Parameter<br />

V DIFF<br />

differential voltage<br />

400 Mb/s<br />

(at C R<br />

= 5pF)<br />

Data rate<br />

360 mV<br />

V CM<br />

common mode voltage 1.23V<br />

static current consumption<br />

dynamic current consumption<br />

static power dissipation<br />

4.16 mA<br />

4.16 mA/7.5 mW<br />

1.8Gb/s<br />

(at C R<br />

= 1pF)<br />

4.7 mA 6.45 mA<br />

7.5 mW<br />

dynamic power dissipation 8.5 mW 11.6 mW<br />

rise/fall time 710/720ps 180/175ps<br />

chip area 0.087 mm 2<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 25


Conclusions and Future Work<br />

This paper present s the description of standard and design<br />

of stand-alone LVDS transmitter, fully compatible with IEEE<br />

specification, implemented in CMOS 180 nm UMC technology.<br />

LVDS core’s architectures: bridged switched current sources,<br />

double current sources, open drain configuration were compared<br />

and due to project’s requirement the last one was chosen.<br />

The attention was paid about stability of common mode feedback<br />

and LVDS core. The operational amplifier used in bandgap<br />

reference source should characterize as low voltage input<br />

offset as it possible. To minimize dynamic current consumption<br />

the appropriate transistors scaling used in control blocks is recommended.<br />

The designed LVDS driver characterizes a low<br />

static (7.5 mW) and dynamic (11.6 mW) power dissipation at<br />

the data rate 1.8 Gb/s for capacitive load equal to 1pF. The<br />

transmitter will be fabricated as a module for multichannel mixed-mode<br />

ASIC.<br />

The author is very grateful to R. Szczygieł and P. Grybos for continuous<br />

technical assistance during realization of this project.<br />

References<br />

[1] Razavi B.: Design of analog integrated circuits. McGraw-Hill<br />

Companies, 2001.<br />

[2] Application Note AN6019, “Differential Signaling”, Lattice Semiconductor<br />

Corporation, 2001.<br />

[3] National Semiconductor, “LVDS owner’s manual including highspeed<br />

CML and signal coditioning”, Fourth Edition, 2008.<br />

[4] IEEE Standard for Low-Voltage Differential Signals (LVDS) for<br />

Scalable Coherent Interface (SCI), 1596.3 SCI-LVDS Standard,<br />

IEEE Std 1596.3-1996,1996.<br />

[5] Electrical characteristics of low voltage differential signaling<br />

(LVDS) interface circuits, Standard ANSI/TIA/EIA-644 1995, TIA<br />

PN-4584 Revision 1.2, 2000.<br />

[6] Boni A., Pierazzi A., Vecchi D.: LVDS I/O interface for Gb/s-perpin<br />

operation in 0.35-µm CMOS. IEEE J. Solid-State Circuits,<br />

vol. 36, no. 4, pp. 706-711, Apr. 2001.<br />

[7] Mandal G., Mandal P.: Low power LVDS transmitter with low common<br />

mode variation for 1 Gb/s-per pin operation. in. Proc. Int.<br />

Symp. on Cir. and Sys. (ISCAS), 2004, vol. 1, pp. 1120–1123.<br />

[8] Chen M., Silva-Martinez J., Nix M., Robinson M. E.: Low-voltage<br />

low-power LVDS drivers. IEEE J. Solid-State Circuits, vol. 40, no.<br />

2, pp. 472–479, Feb. 2001.<br />

[9] Jamasb S., Jelilizeinali R., Chau P. M.: A 622 MHz stand-alone<br />

LVDS driver pad in 0.18-µm CMOS. in. Proc. of IEEE Midwest<br />

Sym. on Cir. and Sys. (MWSCAS), 2001, vol. 2, pp. 610–613.<br />

[10] Tajalli A., Leblebici Y.: A power-efficient LVDS driver circuit in<br />

0.18-µm CMOS technology. Research in Microelectronic and<br />

Electronic Con. PRIME, 2007, pp. 145–148.<br />

[11] Wang Ch., Huang J., Huang J.: 1.0 Gbps LVDS transceiver design<br />

for LCD panels. in. Proc. of IEEE Asia-Pacific Con. on. Adv.<br />

Sys. Int. Cir. (AP-ASIC), 2004, pp. 236–239.<br />

[12] Hurst P. J.: Exact simulation of feedback circuit parameters. IEEE<br />

Trans. on Cir. and Sys., 1991, vol. 38, no. 11, pp. 1382–1389.<br />

CORDIC and SVD implementation in digital hardware<br />

(Realizacja algorytmów CORDIC i SVD w układzie cyfrowym)<br />

dr inż. PRZEMYSŁAW M. SZECÓWKA, Politechnika Wrocławska, Wydział Elektroniki Mikrosystemów i Fotoniki<br />

mgr inż. PIOTR MALINOWSKI, Akademia Wychowania Fizycznego we Wrocławiu<br />

Processing of matrices, especially inversion, remains a key<br />

challenge for contemporary computing machines. Very smart<br />

algorithms were proposed many years ago, by the scientists<br />

who expected rapid development of digital hardware in the<br />

future. Many of those solutions were presumed to work on<br />

futuristic parallel devices. CORDIC and Singular Value Decomposition<br />

(SVD) are good examples here [1–3]. Eventually<br />

recent years have brought the appropriate development of<br />

digital hardware and growth of programmable logic devices<br />

complexity. There is growing interest in construction of dedicated<br />

digital hardware, according to more or less classic concepts<br />

[4–7].<br />

This paper describes a study of hardware implementation<br />

of Singular Value Decomposition of matrix based on replicated<br />

CORDIC modules. The authors focus on comparison of<br />

architecture variants in the context of resource allocation, speed<br />

and accuracy. Similar works may be found in contemporary<br />

literature [8] showing growing interest in practical use of<br />

achievements of great mid XX-th century mathematicians.<br />

CORDIC and SVD overview<br />

CORDIC algorithm (Coordinate Rotation Digital Computer)<br />

was proposed by Volder in 1959 [2]. Originally it was used<br />

to transform polar to perpendicular coordinates and reverse.<br />

Nowadays it is extensively used in digital signal and data<br />

processing like DFT [7] and SVD [5]. It is quite universal tool<br />

26<br />

which may be applied in many variants and configurations. In<br />

general CORDIC consists in iterative rotations of a vector with<br />

a predefined series of constant angles. The angles decrease<br />

in a special manner forming a series: 45°, 26.7°, 14°, 7.1°,<br />

3.57° etc. Consecutive rotations are left or right depending<br />

on target and actual result. With growing number of rotations<br />

n, the increase in accuracy is obtained. This generic schematic<br />

may be applied in various modes, depending on needs.<br />

In rotation mode, a series of rotations is applied to a vector<br />

to reach defined angle. For 2-dimensional space and [x 0<br />

, y 0<br />

] T<br />

vector to be rotated by an angle of z 0<br />

, after n iterations, the<br />

new coordinates are:<br />

1<br />

x<br />

n<br />

=<br />

[ x<br />

0 cos<br />

z<br />

0<br />

−<br />

y<br />

0<br />

sin<br />

z<br />

0<br />

]<br />

(1)<br />

K<br />

n<br />

1<br />

y<br />

n =<br />

[ y<br />

0 cos<br />

z<br />

0<br />

+<br />

x<br />

0<br />

sin<br />

z<br />

0<br />

]<br />

(2)<br />

K<br />

n<br />

whilst the final rotation angle (approximately).<br />

In vector mode, CORDIC determines the angle between<br />

[x 0<br />

, y 0<br />

] T vector and X axis. After series of dummy rotations the<br />

new coordinates would be:<br />

1<br />

2<br />

2<br />

x<br />

n =<br />

x<br />

0<br />

+<br />

y<br />

0<br />

(3)<br />

K<br />

n<br />

y n<br />

= 0 (4)<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Nevertheless the product of algorithm in such case is numerical<br />

value of<br />

⎛ y<br />

z = arctg 0<br />

, estimated by cumulated sum<br />

⎠ ⎞<br />

n<br />

⎝ x 0<br />

of angles (+/- for left/right) applied for consecutive rotations.<br />

Singular Value Decomposition of a matrix consists in finding<br />

a series of singular values σ 1<br />

, σ 2<br />

..., σ l<br />

which simplify inversion<br />

of matrix. For each matrix M ∈ R m, n there exist orthogonal<br />

matrices U ∈ R m, m and V ∈ R n, n , for which<br />

T<br />

m<br />

,<br />

n<br />

U<br />

MV<br />

= Σ = diag( σ<br />

,σ<br />

K<br />

,σ<br />

)<br />

∈<br />

(5)<br />

1 2<br />

l<br />

R<br />

where l = min(m,n), and for r = rank(A) the diagonal values<br />

fulfill conditions<br />

σ<br />

≥<br />

σ<br />

≥<br />

K<br />

≥<br />

σ<br />

0<br />

(6)<br />

1 2<br />

r<br />

><br />

σ<br />

σ<br />

=<br />

K<br />

=<br />

σ<br />

0<br />

(7)<br />

r<br />

+ 1 = r<br />

+<br />

2<br />

l<br />

=<br />

CORDIC architecture<br />

Two variants of CORDIC architectures are presented in Fig. 1<br />

and 2. Both solutions are full-synchronous with single clock. In<br />

the first – sequential approach, same arithmetic modules are<br />

reused in consecutive iterations. Intermediate results are fed<br />

back via the registers and the appropriate angles are delivered<br />

to arithmetic units by the muxes. Control is provided by iteration<br />

counter. Another concept is pipelined architecture presented in<br />

Fig. 2. Schematic shows a hardware providing 3 consecutive iterations.<br />

Arithmetic blocks are replicated for each iteration, thus<br />

the data flow may form a pipeline. This solution provides much<br />

faster throughput, lower flexibility and needs more hardware resources.<br />

On the other hand the control circuitry is more simple<br />

for this solution, leading to some savings and much higher clo-<br />

nreset<br />

A pseudo-inverse matrix M + may be determined by<br />

<br />

M<br />

+<br />

+<br />

=<br />

V<br />

Σ<br />

U<br />

where Σ + is a pseudo-inverse of diagonal matrix, i.e. it is diagonal<br />

matrix formed by inverted (when non-zero) values of.<br />

SVD is currently classified among the most efficient numerical<br />

methods leading to matrices inversion σ 1<br />

, σ 2<br />

..., σ l<br />

. SVD may be<br />

performed by the appropriate rotation of a matrix. For a basic<br />

2 × 2 matrix M<br />

= ⎡<br />

a<br />

b<br />

⎤ the rotation angle is .<br />

⎣<br />

c<br />

d<br />

⎛ c + b<br />

arctg<br />

⎞<br />

⎦<br />

⎝<br />

d<br />

−<br />

a<br />

⎠<br />

This operation may be done by double use of CORDIC<br />

in two modes. First the appropriate angles are determined<br />

and then the rotations are performed. Due to the properties of<br />

CORDIC the iterations may be described by combinations of<br />

adding/subtracting and shifts of bits:<br />

x<br />

i<br />

+ 1 =<br />

x<br />

i<br />

+<br />

δ<br />

i<br />

⋅<br />

SHIFT i<br />

(<br />

y<br />

i<br />

)<br />

(8)<br />

y<br />

i+ 1 =<br />

y<br />

i<br />

− δ i<br />

⋅<br />

SHIFT i<br />

(<br />

x<br />

i<br />

)<br />

(9)<br />

+<br />

δ<br />

where δ i<br />

= +/-1 denotes left or right shift. Eventually hardware<br />

implementation of CORDIC consists of adders, subtractors<br />

and muxes<br />

T<br />

xin<br />

yin<br />

zin<br />

i<br />

i<br />

i<br />

rotation<br />

angle<br />

clk<br />

nreset<br />

clk<br />

nreset<br />

clk<br />

i<br />

shift<br />

i<br />

shift<br />

i<br />

+1<br />

"0000 "<br />

±<br />

di<br />

±<br />

di<br />

±<br />

di<br />

start, i<br />

Fig. 1. CORDIC – sequential architecture<br />

Rys. 1. Architektura CORDIC w wersji sekwencyjnej<br />

i<br />

i<br />

i<br />

nreset<br />

enable<br />

clk<br />

nreset<br />

enable<br />

clk<br />

nreset<br />

enable<br />

clk<br />

nreset<br />

clk<br />

xout<br />

yout<br />

zout<br />

i<br />

nreset<br />

nreset<br />

nreset<br />

xin<br />

±<br />

±<br />

±<br />

xout<br />

clk<br />

clk<br />

shift >> 1<br />

clk<br />

shift >> 2<br />

d<br />

d<br />

d<br />

nreset<br />

nreset<br />

shift >> 1<br />

nreset<br />

shift >> 2<br />

yin<br />

±<br />

±<br />

±<br />

yout<br />

clk<br />

nreset<br />

d<br />

clk<br />

d<br />

clk<br />

d<br />

zin<br />

nreset<br />

nreset<br />

clk<br />

rotation<br />

angle 1<br />

±<br />

clk<br />

rotation<br />

angle 2<br />

d<br />

d<br />

d<br />

Fig. 2. CORDIC – pipelined architecture Rys. 2. Architektura CORDIC w wersji potokowej<br />

±<br />

clk<br />

rotation<br />

angle 3<br />

±<br />

zout<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 27


Tabl. 1. Synthesis results for two variants of CORDIC architectures<br />

(Xilinx Virtex-5 FPGA)<br />

Tab. 1. Wyniki syntezy dwóch wariantów architektury CORDIC (układ<br />

Xilinx Virtex-5)<br />

cking speed available. The two concepts were implemented in<br />

VHDL, verified and synthesized with Xilinx ISE tools for Virtex-<br />

5 programmable device. For this comparative study fixed point<br />

arithmetic with 8-bit numbers coded in 2’complement was applied.<br />

Synthesis results summarized in Table 1. show clearly the<br />

difference between the low-cost and high-speed approach.<br />

SVD architecture<br />

Sequential<br />

Number of Slice Registers 56 208<br />

Number of Slice LUTs 151 243<br />

Pipelined<br />

Clock frequency 257 MHz 428 MHz<br />

Levels of Logic 10 2<br />

Delay 3.891 ns 2.336 ns<br />

Delay on Logic 1.612 ns (41.4%) 0.659 ns (28.2%)<br />

Delay on Route 2.279 ns (58.6%) 1.677 ns (71.8%)<br />

General concept of SVD architecture based on CORDIC modules<br />

is presented in Fig. 3. The input is a basic 2x2 matrix.<br />

The primary output are two singular values, secondary output<br />

a<br />

b<br />

c<br />

d<br />

SVD 2×2<br />

CORDIC<br />

CORDIC<br />

SHIFT-SUM<br />

SHIFT-SUM<br />

SHIFT-SUM<br />

SHIFT-SUM<br />

Fig. 3. Basic SVD architecture composed of CORDIC blocks<br />

Rys. 3. Podstawowa architektura SVD wykorzystująca moduły<br />

CORDIC<br />

σ 1<br />

σ 2<br />

θ p<br />

θ l<br />

are rotation angles. This module, either replicated or reused<br />

may be applied for construction of digital hardware working<br />

with bigger matrices. Detailed schematic of vector rotation<br />

block is presented in Fig. 4. It is a synchronous machine based<br />

on a single CORDIC element reused in consecutive iterations.<br />

The CORDIC output is fed back to the input via the register<br />

until the final value is obtained and latched. Rotation angle<br />

is delivered by the module shown in Fig. 5. Arithmetic block<br />

is reused again for consecutive iterations, thus the output is<br />

fed back. The appropriate angles for elementary rotations are<br />

stored in a memory. Control of data flow in these two modules<br />

is provided by the Finite State Machine working together with<br />

iteration counter. Activation of the strobe signal forces calculation<br />

of the angle and then the following steps of processing<br />

– left or right rotations and correction the output values scale,<br />

disturbed during iterative approximations.<br />

For this part of study, two kinds of number formats and<br />

arithmetic were applied. In the first approach the floating point<br />

numbers compatible with IEEE 754 standard [9] were used. In<br />

this format the bit vector consists of a sign bit, 8-bit, 2-complement<br />

coded exponent and 23-bit significand (non-negative).<br />

Another approach was fixed point arithmetic with 25-bit, 2-<br />

complement coded vectors. For constant angles specific format<br />

was chosen – fixed point with 2 bits reserved for integral<br />

part and the rest left for fractions (the possible angle values<br />

when scaled in radians do not exceed 2). CORDIC module<br />

described in previous section was redesigned twice for these<br />

two formats<br />

Tabl. 2. Synthesis results for 2 variants of SVD architecture (Xilinx<br />

Virtex-5 FPGA)<br />

Tab. 2. Wyniki syntezy dla dwóch wariantów architektury SVD (układ<br />

Xilinx Virtex-5)<br />

32-bit IEEE<br />

floating point<br />

25-bit<br />

fixed point<br />

Clock frequency 35 MHz 148 MHz<br />

Levels of Logic 74 35<br />

Delay 28,602 ns 6.738 ns<br />

Number of Slice 337 (1%) 314 (1%)<br />

Registers<br />

Number of Slice LUTs 4648 (14%) 2609 (7%)<br />

nreset<br />

nreset<br />

x 1<br />

d<br />

clk<br />

iteration, FSM state<br />

nreset<br />

y 2<br />

shift–sum<br />

CORDIC<br />

shift–sum<br />

iteration,<br />

FSM state<br />

enable<br />

clk<br />

nreset<br />

Out 1<br />

c<br />

y 1<br />

Out 2<br />

iteration, FSM state<br />

clk<br />

iteration,<br />

FSM state<br />

enable<br />

clk<br />

Fig. 4. SVD architecture – vector rotation block Rys. 4. Architektura SVD – blok obracania wektora<br />

28<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


nreset<br />

angle R<br />

z 1<br />

angle L<br />

nreset<br />

zero<br />

clk<br />

iteration, FSM state<br />

ROM 24×29<br />

rotation_ angle 1<br />

±<br />

di<br />

Z 2<br />

–<br />

iteration,<br />

FSM state<br />

enable<br />

clk<br />

angle<br />

rotation_ angle 2<br />

…<br />

rotation_ angle 23<br />

rotation_ angle 24<br />

Fig. 5. SVD architecture – calculation of rotation angle Rys. 5. Architektura SVD – blok wyliczania kąta obrotu<br />

|∆σ1/σ1|<br />

1,4x10 -6<br />

1,2x10 -6<br />

1,0x10 -6<br />

8,0x10 -7<br />

6,0x10 -7<br />

4,0x10 -7<br />

2,0x10 -7<br />

0,0<br />

10 -23 10 -13 10 -3 10 7 10 17 10 27 10 37<br />

σ1<br />

Fig. 6. Relative error of singular value determination for two<br />

kinds of arithmetic approach – 25-bit fixed point (lower) and<br />

-32-bit floating point floating point (upper plot)<br />

Rys. 6. Względne błędy wyliczania wartości osobliwych dla<br />

dwóch wariantów arytmetyki – 25-bitowej ze stałym przecinkiem<br />

(u dołu) i 32-bitowego ze zmiennym przecinkiem (u góry)<br />

SVD architecture with two variants of arithmetic was implemented<br />

in VHDL and synthesized for Xilinx Virtex-5 device.<br />

Synthesis results are summarized in Table 2. If to compare<br />

allocation of resources there is no huge difference in number<br />

of registers used. On the other hand the floating point variant<br />

consumes much more combinatorial logic. There is huge difference<br />

in maximum clock speed – 148 MHz for fixed point<br />

version point and only 35 MHz for floating point approach.<br />

Arithmetic operations on floating point numbers require long<br />

chains of combinatorial logic which require more time to transfer<br />

signal from one register to another.<br />

The two variants were simulated in Xilinx ISE environment<br />

for several sample matrices. The results were sent to a file,<br />

converted and compared with the ones given by SVD algorithm<br />

run in a computer (Octave tools). Fig. 6 shows two plots of relative<br />

errors obtained for two architectures. It is a bit surprising<br />

that fixed point arithmetic delivers substantially better results.<br />

Conclusions<br />

A study of digital hardware dedicated to Singular Value Decomposition<br />

was performed. The motivation was authors interest in<br />

construction of specialized computing machines performing ope-<br />

rations on matrices in highly parallel way. Significant effort was<br />

devoted to CORDIC algorithm which was used for SVD but may<br />

be treated as separate issue as well. The results lead to conclusion<br />

that contemporary FPGAs are very close to enable construction<br />

of machines dealing with huge computational complexity.<br />

Presented results, limited to small matrices are a good basis<br />

for further work, but at this stage deliver quite reasonable comparative<br />

material about architecture and arithmetic variants. In<br />

this context the results obtained for fixed and floating point are<br />

very interesting. As it was expected, fixed point approach provides<br />

higher processing speed and lower logic resources allocation.<br />

Surprising result was higher precision obtained with fixed<br />

point. Shall be noted however that 25-bit representation was<br />

selected after very careful considerations and estimations.<br />

Further research will focus on construction of devices dealing<br />

with matrices of higher dimension, perhaps with processing<br />

decomposed to basic 2x2 elements, so the described<br />

modules may be used without any redesign. An advantage of<br />

this approach is a chance to develop a methodology of processing<br />

matrices of unlimited dimension with limited number<br />

of basic SVD/CORDIC units. That would enable optimal utilization<br />

of currently available resources with at least partial<br />

independence on input complexity.<br />

References<br />

[1] Eckart C., Young G.: The approximation of one matrix by another<br />

of lower rank. Psychometrika, vol. 1, no. 3, 1936.<br />

[2] Volder J.E.: The CORDIC Trigonometric Computing Technique.<br />

IRE Transactions on Electronic Computers, 1959.<br />

[3] Golub G., Kahan W.: Calculating the singular values and pseudo-inverse<br />

of a matrix. J. SIAM Numerical Analysis, Ser. B, vol.<br />

2, no. 2, 1965, pp. 205–224.<br />

[4] Brent R.P., Luk F.T., Van Loan C.F.: Computation of the singular<br />

value decomposition using mesh-connected processors. Journal<br />

for VLSI Computer Systems, vol. 1, no. 3, 1985, pp. 243–270.<br />

[5] Cavallaro J.R., Luk F.T.: CORDIC Arithmetic for a SVD Processor.<br />

Journal for Parallel and Distributed Computing, vol. 5, 1988,<br />

pp. 271–290.<br />

[6] Andraka R.: A Survey of CORDIC Algorithms for FPGA based<br />

computers. In FPGA ‘98: Proc. of sixth international symposium<br />

on Field programmable gate arrays ACM/SIGDA, 1998, pp. 191–<br />

200.<br />

[7] Deprettere F. (ed.): SVD and signal processing. Algorithms, applications<br />

and architectures. Department of Electrical Engineering,<br />

Delft University of Technology, Elsevier Science Publishers<br />

B.V., Amsterdam, 1988.<br />

[8] Wang H., Leray P., Palicot J.: A CORDIC-based dynamically<br />

reconfigurable FPGA architecture for signal processing algorithms.<br />

URSI 08, The XXIX General Assembly of the International<br />

Union of Radio Science, Chicago IL, 2008.<br />

[9] Floating-point arithmetic, IEEE Std No. 754, 2008.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 29


Lab-on-a-chip for developmental competence<br />

assessment of bovine oocytes<br />

(Lab-chip do oceny jakościowej potencjału rozwojowego oocytów bydlęcych)<br />

RAFAŁ WALCZAK, PhD; PATRYCJA SZCZEPAŃSKA, MSc; JAN DZIUBAN, Prof.<br />

Wrocław University of Technology, Faculty of Microsystem Electronics and Photonics<br />

BARTŁOMIEJ KEMPISTY, MD, University of Medical Sciences, Department of Histology and Embryology, Poznań<br />

MARTA JACKOWSKA, MSc; PAWEŁ ANTOSIK, MD; JĘDRZEJ JAŚKOWSKI, Prof.<br />

Poznań University of Life Sciences, Faculty of Animal Breeding and Biology<br />

ANNA CHEŁMOŃSKA-SOYTA Prof,<br />

Institute of Immunology and Experimental Therapy of Polish Academy of Science<br />

30<br />

The recent development of assisted reproductive technologies<br />

is based on apply of advanced molecular biology and<br />

cell biology techniques, including analysis of genes expression,<br />

transcriptional profile and proteomics research of oocytes<br />

and embryos. Currently by using of new molecular biology<br />

techniques it is possible to define the quality of oocytes<br />

in relation to its ability to maturation, fertilization and proper<br />

embryo development in preimplantation stages. However, applying<br />

of molecular genetics methods in gametes and embryos<br />

quality assessment are invasive and lead to partially or<br />

completely loss of cell viability. On the other hand, the popular<br />

veterinarian method of qualitative selection of animals oocytes<br />

for artificial fertilization is based on morphological aspects<br />

of reproductive cells. In case of oocytes quality and maturity<br />

of ovarian follicles (more mature gives better oocytes) and<br />

complex of the cumulus cells surrounding particular oocyte<br />

(greater number of cells is a positive factor) are two most important<br />

morphological factors to be taken into account prior<br />

in vitro fertilization. Qualification and selection is done by<br />

a trained veterinarian specialist observing under a microscope<br />

oocytes “flowing” in a buffer. As long as the cell has the<br />

surrounding cumulus, it is relatively easy to divide oocyte into<br />

boundary classes. However, there is no sharp criteria dividing<br />

the oocytes into intermediate classes. What more, before in<br />

vitro fertilization, cumulus of the oocyte is removed. It causes,<br />

that on the pre-fertilization step it is very hard to classified<br />

the “naked” oocytes by microscope observation only. This is<br />

unsatisfactory selection method and one of the weakest points<br />

of agriculture breading industry because introduces large<br />

personal factor and can not be automated. For example, only<br />

in Poland at least ten thousand artificial fertilizations of bovines<br />

oocytes is done yearly, what introduces proportional number<br />

of qualification tests – often with non satisfactory results. The<br />

classification problem concerns significantly more embryos,<br />

where classification criteria are more subjective and depends<br />

on the observer experience. Therefore it is very important to<br />

develop new noninvasive classification methods and criteria,<br />

by using which it will be possible to analyze several parameters<br />

of cells without its destabilization and loss of viability.<br />

Lab-on-a-chip systems, based on microfluidic technology<br />

and recent developments in microengineering, compose<br />

chemical or biochemical laboratory in micro-scale. Applying<br />

of lab-on-a-chip system in analysis and quality assessment<br />

of farm animals (especially bovine and porcine) oocytes and<br />

embryos is a fusion of reproductive biology and lab-on-a-chip<br />

techniques. Reported in literature lab-on-a-chips for examinations<br />

of oocytes/embryos are based on indirect methodologies<br />

or may cause destruction of the cell during measurement<br />

[1–3]. Assessment of development competence of oocytes<br />

and embryos based on lab-on-a-chip system, which analyzed<br />

the spectral characteristic of the cells, is an important element<br />

in research on assisted reproductive technologies. Therefore,<br />

veterinarian service are still looking for noninvasive methodology<br />

and miniature instrument for fast, cheep and quantitative<br />

assessment of oocytes and embryos of pigs, cows and other<br />

breeding animals.<br />

Microspectrophotometry of a single living<br />

cell<br />

A novel methodology of reproductive cell characterization is<br />

based on on-chip spectrophotometric measurements of a single<br />

cell (Fig. 1). In this method, an investigated biological material<br />

is introduced between two optical fibers with core diameter<br />

tailored to the size of the material. The oocyte or embryo completely<br />

fulfils space between fibers. It ensures measurement<br />

of optical transmission/absorbance spectra of the single cell.<br />

It is possible only when lab-on-a-chip techniques are applied<br />

due to dimensional compatibility of the biological material with<br />

microchannels of the lab-on-a-chip.<br />

It has been assumed that VIS/NIR spectra of the cell will<br />

be measured by the application of wide-spectrum light source<br />

and miniaturized spectrometer co-working with the lab-on-achip<br />

The spectra will be analyzed by specialized software to<br />

extract changes of absorbance/transmittance and/or wavelength<br />

shift of the local minima/maxima of the spectra. This<br />

Fig. 1. Scheme of an idea of spectrophotometric characterization<br />

of single oocyte or embryo in lab-on-a-chip<br />

Rys. 1. Schematyczne przedstawienie idei spektrofotometrycznej<br />

charakteryzacji pojedynczego oocytu lub embriony w lab-onchipie<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Fig. 2. Block scheme of the device for quality assessment<br />

of oocyte by spectrophotometric characterization<br />

Rys. 2. Schemat blokowy urządzenia do jakościowej oceny oocytów<br />

metodą spektrofotometryczną<br />

methodology will give an information about optical properties<br />

of the characterized material. Obtained data will be in form of<br />

numbers giving to an operator objective assessment of the<br />

development competence of the reproductive cell. On the stage<br />

of investigations, results of the analyze will be compared<br />

to a reference methodology of classification of the biological<br />

material based on the morphological criteria determined during<br />

observation under optical microscope.<br />

To realize presented above scenario a novel system utilizing<br />

spectrophotometric methodology and lab-on-a-chip must<br />

be built. Block scheme of the system is shown on Fig. 2.<br />

Heart of this system is the lab-on-a-chip enabling optical<br />

characterization of the single cell.<br />

Lab-on-a-chip<br />

Lab-on-a-chip contains measurement cell, net of microchannels<br />

and passive valves for steering of fluid and oocyte/embryo<br />

flow, and inlet/outlet holes for biological material introduction/extraction<br />

and steering port (Fig. 3). The biological material is introduced<br />

into the measurement cell by inlet 1, than passes through<br />

set of passive Tesla valves (1 st check valves). Next, characterized<br />

oocyte flows into the measurement cell by sucking of the fluid<br />

by a pipette connected to the steering port of the chip. Topology<br />

of the measurement cell ensures mechanical immobilization<br />

of the cell between two optical fibers. After characterization the<br />

material is flashed back to the outlet 1 by passing through the<br />

second set of Tesla valves (2 nd check valve). Proposed configuration<br />

enables steering of the fluid flow and examined biological<br />

material transport with separation of the inlet and outlet.<br />

The microfluidic channels and channels for optical fibers (all<br />

140 mm depth) are etched simultaneously in DRIE (Deep Reaction<br />

Ion Etching) process in the 380 mm – thick monocrystalline<br />

silicon wafer (Fig. 4a). After etching, 0,3 mm – thick thermal<br />

silicon oxide is formed to passivate chemically surface of the<br />

chip. Next, the wafer is anodically bonded (450 0 C, 1,5 kV) to<br />

a borosilicate glass (Borofloat Schott, Germany) with previously<br />

mechanically drilled inlet and outlet via – holes. Following, optical<br />

fibers with outer diameter of 125 mm and 100 mm core<br />

(Ocean Optics, USA) are mounted. Fronts of the fibers are perfectly<br />

aligned each to other thanks to high precision of DRIE<br />

etching. Fibers are aligned to the edge of microfluidical channel,<br />

ensuring immobilization of the oocyte without its mechanical destruction<br />

(Fig. 4b). Fibers are fixed by use of UV epoxy hard glue<br />

(NOA 61, THORLABS, Sweden). Off-chip ends of both fibers<br />

are finished with standard SMA 905 connectors compatible with<br />

optical connections of the lamp and the spectrometer.<br />

The steering port inlet is finished with glued precise plastic<br />

connector (UpChurch, USA) enabling tight positioning of<br />

an end of the pipette. Assembled silicon-glass chip is placed<br />

on a PCB board and than in a metal package (Fig. 5). The<br />

metal package is small and shock resistant. It ensures stable<br />

positioning under microscope, safe transport and performing<br />

measurements outside laboratory, for example in a farm.<br />

Fig. 3. Layout of the lab-on-a-chip with integrated passive valves<br />

Rys. 3. Topologia lab-on-chipa ze zintegrowanymi zaworami pasywnymi<br />

Fig. 5. Assembled in the metal package lab-on-a-chip for optical<br />

characterization of the oocyte<br />

Rys. 5. Lab-on-chip po montażu w metalowej obudowie<br />

a) b)<br />

Fig. 4. Lab-on-a-chip: a) view of the silicon chip after DRIE etching in comparison to a mosquito (left picture) and enlarged view of<br />

the central part of the chip with etched measurement cell and net of Tesla valves (right picture), b) view of the measurement cell with<br />

mounted optical fibers<br />

Rys. 4. Lab-on-chip: a) widok chipa krzemowego po trawieniu DRIE w porównaniu do komara (lewe zdjęcie) oraz powiększony obszar<br />

centralnej części chipa z wytrawioną komorą pomiarową oraz siecią zaworów typu Tesla (prawe zdjęcie), b) widok komory pomiarowej<br />

z zamontowanymi światłowodami<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 31


Experiments<br />

The measurement setup consists of VIS/NIR light source (a halogen<br />

lamp by OceanOptics, USA), developed by us lab-on-achip,<br />

miniature spectrometer (Ocean Optics, USA) and a personal<br />

computer with specialized software (Fig. 6). Light transmitted<br />

from the halogen lamp by fiber 1 passes through the measured<br />

material and is collected by fiber 2 which is connected to the<br />

spectrometer. The spectral characteristics are recorded, normalized<br />

and processed under Origin 6,0 (USA) software.<br />

In the experiments total number of 60 bovine oocytes classified<br />

into four classes according to the size and quantity of<br />

the cumulus complex cells surrounding oocytes are characterized.<br />

Class 1 is the best. It is assumed that this class of<br />

oocytes gives the highest probability of successful artificial<br />

fertilization.<br />

The single oocyte is introduced into lab-on-a-chip by manual<br />

pipeting followed by sucking by capillary forces. View of the<br />

a)<br />

measurement cell with immobilized bovine oocyte is shown<br />

on Figure 7. After short measurement (circa 5 seconds) of the<br />

optical spectra, the biological material is flushed-back to the<br />

second outlet and then to a sterile transporting container for<br />

further treatment.<br />

Special attention mast be paid to conditioning of the obtained<br />

spectra data. Due to very short transmission optical<br />

path in the measurement cell (~150 mm) and according to<br />

Bouger-Lambert-Beer law, absorbance of the light will be very<br />

low. Therefore, changes in transmittance will be also very low.<br />

To emphasize this small changes, the raw data obtained from<br />

spectrometer software are normalized and subtracted from<br />

background (halogen lamp) spectral characteristic. Thus differential<br />

normalized intensity (DNI) spectral characteristic is<br />

obtained. It is assumed that to find a correlation between the<br />

reference method and proposed here, a shift of a local minima<br />

or maxima position will be investigated as main factor.<br />

Results<br />

Normalized and conditioned transmission spectra of 15 oocytes<br />

of each class have been measured, average characteristics<br />

for each class are shown on Figure 8.<br />

b)<br />

Fig. 8. Normalized transmission (DNI) spectral characteristics<br />

of bovine oocytes classified into four quality classes<br />

Rys. 8. Znormalizowane charakterystyki spektralne transmisji światła<br />

dla oocytów bydlęcych kwalifikowanych metodą referencyjną<br />

Fig. 6. Scheme (a) and view (b) of the measurement setup<br />

Rys. 6. Schemat (a) i wygląd (b) układu pomiarowego<br />

Fig. 7. View of the bovine oocyte immobilized in the measurement<br />

cell of the lab-on-a-chip<br />

Rys. 7. Widok lab-on-chipa z unieruchomionym w komorze pomiarowej<br />

oocytem bydlęcym<br />

Fig. 9. Shift of the local minima around 580 nm towards longer<br />

wavelengths correlated with reference method<br />

Rys. 9. Przesuniecie lokalnego minimum charakterystyki DNI położonego<br />

około 580 nm w kierunku dłuższych długości fali – zależność<br />

od klasy oocytu<br />

32<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Good correlation between position of the local minimum<br />

near 580 nm and class of the oocyte has been found (Fig. 9)<br />

Further experiments confirmed non-destructive nature of<br />

microspectrophotometric oocyte characterization. Successful<br />

artificial fertilization after lab-on-a-chip examination has been<br />

carried out. This result opens a way toward oocytes selection<br />

for artificial fertilization of ovine oocytes.<br />

Summary<br />

Developed microspectrophotometric methodology and labon-a-chip<br />

– based instrumentation for quality assessment of<br />

bovine oocytes have been successfully tested. Obtained optical<br />

characterization results in most cases are in good correlation<br />

to the standard reference classification method. However,<br />

some discrepancies between results of optical characterization<br />

and predicted class of the oocyte by reference method<br />

were also observed. It confirmed subjective classification of<br />

the oocyte by veterinary service. In the future investigations<br />

on applicability of the developed methodology and lab-on-achip<br />

for development competence assessment of bovine and<br />

porcine embryos will be performed. The real test in a breeding<br />

farm will be also carried out to confirm new methodology “in<br />

the filed tests”.<br />

The work are financed by POIG 01.03.01-00-014/08-02 subproject<br />

2B APOZAR of MNS-DIAG project. Authors would like to thank<br />

to colleagues from Institute of Electron Technology in Warsaw<br />

for fabrication of the silicon part of lab-on-a-chip under frame of<br />

APOZAR project.<br />

References<br />

[1] Murayama Y., Constantinou C.E., Omata S.: Micro-mechanical<br />

sensing for the characterization of the elastic properties of the<br />

ovum via uniaxial measurement. Journal of Biomechanics 37,<br />

pp. 67–72, 2004.<br />

[2] Urbanski J.P., Johnson M.T., Craig D.D., Potter D.L., Gardner<br />

D.K., Thorsen T.: Noninvasive metabolic profiling using microfluidics<br />

for analysis of single preimplantation embryos. Analitical<br />

Chemistry 80, pp. 6500–6507, 2008.<br />

[3] Zeggari R., Wacogne B., Pieralli C., Roux C., Gharbi T.: A full<br />

micro-fluidic system for single oocyte manipulation including an<br />

optical sensor for cell maturity estimation and fertilization indication.<br />

Sensors and Actuators B 125, 2007, 664–671.<br />

Image sensor – based fluorescence detection<br />

for lab-on-a-chip<br />

(Detekcja fluorescencji w lab-chipach z wykorzystaniem czujnika obrazu)<br />

RAFAŁ WALCZAK, PhD<br />

1<br />

Wrocław University of Technology, Faculty of Microsystem Electronics and Photonics<br />

2<br />

Institute of Electron Technology, Warszawa<br />

Fluorescence detection is commonly used in analytical chemistry<br />

as an optical sensing method. Well known fluorescencebased<br />

methodologies developed for “large-scale” analytical<br />

devices are transferred to micro-world of microfluidical chips<br />

dedicated for chemical and biochemical analyzes. These chips<br />

are commonly called lab-on-a-chips (LOC). LOC utilizing fluorescence<br />

detection are successfully applied at laboratory level<br />

in many fields of life-sciences – for example in separation<br />

of biological material by electrophoresis [1] or amplification of<br />

genetic material by polymerase chain reaction (PCR) with realtime<br />

analyze of amplification kinetics (real-time PCR) [2]. However,<br />

the detection issues arise while analytical systems are<br />

miniaturized. Reduction of analyzed volume means reduction<br />

in detection volume, decrease of analyte available for detection<br />

and making it more difficult to detect [3]. Therefore, only high<br />

sensitivity and expensive photodetectors co-working with sophisticated<br />

electronics are applied to ensure proper optical signal<br />

detection from nano- and picoliter volumes of analyte being<br />

detected in LOC. On the other hand, it is obvious that “conventional”<br />

laboratory fluorescence detection systems are too bulky<br />

and expensive to be applied in portable devices utilizing LOC.<br />

Therefore, new methodologies and technical solutions must be<br />

used to fulfill requirements of LOC – based portable devices.<br />

In this paper a brief review of the state-of-the-art of systems<br />

for fluorescence detection in LOC is presented. Following, description<br />

of a novel image sensor-based fluorescence detection<br />

unit co-working with LOC is presented. Finally, example<br />

of miniaturized device utilizing described here sensing system<br />

co-working with LOC is shown.<br />

Fluorescence detection in LOC<br />

Scheme of the typical optical path for fluorescence induction<br />

and detection from LOC is shown on Figure 1. Commonly used<br />

integrated apparatus for fluorescence readout is an epifluorescence<br />

microscope or devices with configuration of the optical<br />

paths similar to applied in this microscope (Fig. 1). In this<br />

device, VIS light comes usually from arc lamp, light emitting<br />

diode (LED) or laser. The light is restricted to a narrow range<br />

of wavelengths that can effectively excitate a fluorochrome<br />

and be strongly excluded by the detection channel. The nar-<br />

Fig. 1. Scheme of the typical optical path for fluorescence excitation<br />

and detection co-working with lab-on-a-chip<br />

Rys. 1. Schemat typowego układu optycznego do wzbudzania<br />

i detekcji fluorescencji współpracujący z mikrosystemem typu<br />

lab-on-chip<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 33


Fig. 2. Scheme of the typical path<br />

of analog fluorescence signal conditioning<br />

Rys. 2. Schemat typowej ścieżki obróbki<br />

analogowego sygnału fluorescencji<br />

row wavelength range<br />

is ensured by one<br />

or more interference<br />

filter(s) and dichroic<br />

mirror(s). Fluorescence<br />

light emitted by the<br />

fluorochrome is collected<br />

by the microscope<br />

objective and guided<br />

to a photodetector<br />

by passing through<br />

filter(s) and dichroic<br />

mirror(s) to exclude<br />

the excitation light.<br />

Common detectors<br />

are photomultiplier tubes<br />

(PMT) [4], semiconductor<br />

photodiodes<br />

[5] and cooled charge coupled devices (CCD) as sensing matrix<br />

in video cameras [6] or lines in spectrophotometers [7].<br />

In spite of sensitivity of the photodetector itself, important<br />

issue in highly sensitive detection of fluorescence in LOC<br />

is conditioning of an electrical signal generated by the photodetector.<br />

The role of the conditioning electronics is to amplify<br />

electrical signal with simultaneous reduction of background<br />

noises to ensure high signal-to-noise ratio (SNR). Most of<br />

the conditioning electronics is realized by the use of analog<br />

circuits (Fig. 2). Configuration of these circuits is sophisticated<br />

and only the highest quality (low noise) elements can be<br />

used.<br />

Although, fluorescence detection is widely used in LOC for<br />

many years, configuration of the detection apparatus is based<br />

on solutions developed over 30 years ago. In most cases, these<br />

solutions are technically and dimensionally incompatible<br />

with microfluidical chips. Therefore, rapid development of LOC<br />

must be followed by development of novel methodologies and<br />

technical solutions surrounding the chips and leading towards<br />

successful application of LOC in the point-of-care devices.<br />

Novel image sensor – based fluorescence<br />

detection<br />

In the novel concept of optical instrumentation for fluorescence<br />

excitation and detection, application of recent developments<br />

in optoelectronics and computer sciences is proposed.<br />

Development of low-cost optoelectronic components and<br />

devices observed in latest two decades enables fabrication<br />

a) b)<br />

of cheap and miniaturized semiconductor lasers as well as<br />

miniature CCD image sensors. The laser can be used as fluorescence<br />

excitation light source, whereas the image sensor<br />

can be a part of image – based fluorescence detector.<br />

Currently there is a lot of miniature semiconductor lasers<br />

working at visible light spectrum (for example around 408,<br />

532 and 635 nm) with optical power varying from 1 mW to<br />

hundreds of mW. These wavelength regions and powers are<br />

enough to effectively excitate fluorescence of many fluorochromes<br />

applied in life-sciences. Narrow-spectrum of laser<br />

light eliminates application of emission filter. What more low<br />

power consumption enables long-term battery operation<br />

or power supplying by computer USB port. In the novel concept,<br />

a collimated laser light is introduced directly into detection<br />

area or microchamber of LOC by edge coupling to a light<br />

guiding side wall of the chip (Fig. 3a). The chip is “observed”<br />

by an analog minicamera equipped with low-cost non-cooled<br />

CCD image sensor and miniature objective with integrated<br />

long-pass interference filter. The detection unit is positioned<br />

perpendicularly in relation to the surface of chip and laser<br />

beam. This configuration enables geometrical separation of<br />

the laser excitation light and fluorescence signal without application<br />

of dichroic mirror. It significantly simplifies configuration<br />

of the optical instrumentation co-working with LOC. View<br />

of the detection area of LOC containing fluorescence signal<br />

is collected by CCD–based minicamera detection unit. The<br />

non-conditioned “raw” analog video output signal from the<br />

minicamera is digitalized by one-channel low-cost video frame<br />

grabber connected to a personal computer. The computer<br />

stores images in any type of memory with simultaneous analyze<br />

of the data. Specialized software carries out analyze of<br />

the captured video images to give numerical values on fluorescence<br />

intensity. The crucial feature of the software – based<br />

analyze is that the images are analyzed only in selected areas<br />

where fluorescence image is present. Rest part of the image<br />

which may contain artificial optical signals are not taken into<br />

account. Thus, digital conditioning of the fluorescence signal<br />

by the software-base image analyze in spite of analog conditioning<br />

is applied in the novel solution (Fig. 3b). What more,<br />

storing of the images in computer memory enables re-analyze<br />

of the fluorescence images when it is necessary. This<br />

is unique feature of the novel method which is non-available<br />

in typical instrumentation with “non-imagining” photodetectors<br />

(PMT or photodiode) when an operator has only “one<br />

shoot” during analyze.<br />

Example of applications of novel<br />

fluorescence readout instrumentation<br />

– detection of food pathogens<br />

by real-time PCR DNA analyze<br />

Fig. 3. Scheme of the novel fluorescence excitation and detection method<br />

utilizing miniature optoelectronical components co-working with LOC (a)<br />

and schematic presentation of novel, digital path of fluorescence signal<br />

conditioning (b)<br />

Rys. 3. Schemat nowej metody wzbudzania i detekcji fluorescencji wykorzystującej<br />

miniaturowe komponenty optoelektroniczne i współpracujący<br />

z mikrosystemem typu lab-on-chip (a) oraz schemat nowej ścieżki<br />

cyfrowej obróbki sygnału fluorescencji (b)<br />

34<br />

Described here optical instrumentation became<br />

a part of device for food pathogens detection developed<br />

under European 6. Framework Programme<br />

OPTOLABCARD [15]. The goal of the project was<br />

to develop LOC-based compact instrumentation enabling<br />

detection of Salmonella spp. in human samples<br />

and Campylobacter j. in broiler chicken farms, by utilizing<br />

real-time PCR technique [8, 9].<br />

In the device, the disposable LOC (1 × 1 cm 2 ) with<br />

integrated heater and temperature sensor [2] is placed<br />

in a plastic chip holder (2.8 × 2.8 × 0.5 cm 3 ) with<br />

integrated electrical contacts to the chip and some<br />

electronics for temperature management. The chip<br />

holder has miniature electrical connection to a spe-<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Fig. 4. Portable real-time PCR DNA analyzer: a) LOC on author’s<br />

finger, b) view of the docking station ready to work<br />

Rys. 4. Przenośny analizator DNA PCR czasu rzeczywistego:<br />

a) mikrosystem typu lab-on-chip w porównaniu do monety<br />

10-centowej, b) widok stacji dokującej gotowej do pracy<br />

a)<br />

Fluorescence intensity [au]<br />

a)<br />

b)<br />

b)<br />

100<br />

Normalized fluorescence intensity<br />

[au]<br />

80<br />

60<br />

40<br />

20<br />

0<br />

100 140 180 220 260 300 340 380 420<br />

Time [s]<br />

0,6<br />

0,5<br />

0,4<br />

0,3<br />

0,2<br />

0,1<br />

0<br />

Fluorescence signal<br />

Temperature profile<br />

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34<br />

PCR cycle<br />

Fig. 5. Real-time PCR of Campylobacter j. DNA: a) fluorescence<br />

intensity change following PCR temperature profiling, b) S-curve<br />

– like bar graph describing kinetics of DNA amplification<br />

Rys. 5. PCR czasu rzeczywistego DNA bakterii Campylobacter j.:<br />

a) zmiany intensywności fluorescencji w czasie cykli temperaturowych<br />

reakcji PCR, b) krzywa typu S opisująca kinetykę namnażania<br />

materiału genetycznego<br />

cialized PCR temperature controller connected to the computer.<br />

The holder with ready to use LOC is positioned in the<br />

docking station (15 × 5 × 7 cm 3 ) in the way ensuring laser light<br />

introduction into PCR microchamber and fluorescence light<br />

collection [10]. Than, PCR temperature profiling and fluorescence<br />

signal acquisition are started. The laser light does not<br />

illuminate whole PCR microchamber but it does only in part<br />

of the chamber which corresponds to a laser light distribution<br />

cone. Therefore, the fluorescence is induced and emitted from<br />

a volume in picoliter range.<br />

View of the chip and hand-held docking station with positioned<br />

chip holder just before start of the PCR process<br />

is shown on Fig. 4.<br />

The pre-validation tests of LOC – based system for detection<br />

of Campylobacter j. were carried out with 48 chicken<br />

fecal samples [9]. All the steps – from sample preparation to<br />

final result – were performed in the single chip with 2,5 μl volume<br />

of reagents. The ratio of PCR efficiencies between onchip<br />

and on-tube (reference) was up to 300%. The sensitivity<br />

of on-chip PCR was determined as 0,7–7 ng/ml of template<br />

DNA – similar result were obtained for on-tube amplification.<br />

The LOC real-time PCR process took 30 min – at least 4 times<br />

shorter than PCR on-tube.<br />

An example of the fluorescence intensity change following<br />

temperature profiling during PCR process is shown on Fig. 5a.<br />

S-curve graph of the real-time PCR process has been compiled<br />

on the base of average fluorescence intensity during<br />

extension step of each PCR cycle (Fig. 5b). Obtained S-curve<br />

is comparable to the characteristic achieved by the use of the<br />

chip observed under epifluorescence microscope equipped<br />

with PMT.<br />

Carried out pre-validation tests confirmed usefulness<br />

of the developed optical instrumentation, as well as the whole<br />

LOC-based system for real-time PCR detection of Campylobacter<br />

j.<br />

Summary and conclusions<br />

Presented here idea and technical realization of novel fluorescence<br />

readout has been successfully tested and applied<br />

in various LOC-based instrumentation. The main advantage<br />

of presented here solution is a ratio of the price of optical readout<br />

components to the sensitivity of the whole unit. It has been<br />

obtained by application of low-cost components and “intelligent”<br />

software for conditioning of collected data. As the result, sensitivity<br />

of the novel system is comparable to sophisticated “large<br />

scale” solutions with 100-times lower consumables costs. The<br />

novel image sensor – based fluorescence detection enables development<br />

of point-of-care devices with advanced fluorescence<br />

detection with sensitivity level comparable to standard laboratory<br />

equipment. Currently, there are works on further development of<br />

the CCD image sensor – based fluorescence detection method<br />

and instrumentation toward multiwavelength detection. Preliminary<br />

works indicated also that developed optical instrumentation<br />

can be successfully used also in portable cocaine detector<br />

for monitoring of professional driver’s cocaine abuse [11],<br />

in LOC-based gel electrophoresis DNA analyzer or microcytometer<br />

for optical characterization of bio- samples [12].<br />

The works were/are financed by 6. FP OPTOLABCARD, 7 FP.<br />

LABONFOIL and Statutory Grants of PWr. Author would like<br />

to thank to J. Dziuban, A. Górecka-Drzazga, P. Knapkiewicz,<br />

P. Szczepańska and W. Kubicki from Wrocław University of Technology,<br />

J. Koszur and B. Latecki from Institute of Electron Technology<br />

in Warsaw, D. D. Bang from Danish Technical University and J.<br />

M. Ruano-Lopez from Ikerlan in Spain for fruitful co-operation.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 35


References<br />

[1] Griebel A. et al.: Integrated polymer chip for two-dimensional<br />

capillary gel electrophoresis. Lab on a Chip, 2004, vol. 4,<br />

18–23.<br />

[2] Agirregabiria M. et al.: Concentration, lysis and real-time PCR<br />

on a SU-8 lab on a chip for rapid detection of Salmonella spp. in<br />

faeces. Proc. of μTAS 2007, Paris, France, 584–586.<br />

[3] Manz A. et al.: Miniaturized total chemical analysis systems:<br />

a novel cocnept for chmical sensing. Sensors and Actuators, B1<br />

(1990), 244–248.<br />

[4] Cady N. C. et al.: Real-time PCR detection of Listeria monocytogens<br />

using integrated microfluidic platforms. Sensors and Actuators<br />

B 107, 2005, 332–341.<br />

[5] Cho Y.-K. et al.: Clinical evaluation of micro-scale chip-based<br />

PCR system for rapid detection of hepatitis B virus. Biosensors<br />

and Bioelectronics 21 (2006), 2161–2169.<br />

[6] Xiang Q. et al.: Miniature real time PCR on chip with multi-channel<br />

fiber optical fluorescence detection module. Biomedical<br />

Microdevices 9, 2007, 443–449.<br />

[7] Lee D.-S. et al.: A novel real-time PCR machine with a miniature<br />

spectrometer for fluorescence sensing in a micro liter volume<br />

glass capillary. Sensors and Actuators B 100 (2004), 401–410.<br />

[8] Ruano-Lopez Jesus M. et al.: The SmartBioPhone, a point of<br />

care vision under development trough two European projects:<br />

OPTOLABCARD and LABONFOIL. Lab on a Chip. 2009, vol. 9,<br />

iss. 11, 1495–1499.<br />

[9] Walczak R. et al.: Miniaturowy system do prowadzenia reakcji<br />

PCR czasu rzeczywistego do taniego i masowego wykrywania<br />

patogenów żywności. <strong>Elektronika</strong>, 49, 6, 2008, 242–244.<br />

[10] Bang D. D. et al.: A total integrated lab-on-a-chip systems for<br />

rapid detection of Campylobacter spp. in broiler chicken Validation<br />

trial. Proc. of mTAS2009, 345–348.<br />

[11] Walczak R. et al.: Toward portable instrumentation for quantitative<br />

cocaine detection with lab-on-a-paper and hybrid optical readout.<br />

Procedia Chemistry 2009, vol. 1, iss. 1, 999–1002.<br />

[12] Bargiel S.et al.: Microcytometer with simultaneous photometric<br />

and fluorometric characterization of biosamples. Proc. of Eurosensors<br />

2008 Conference, Dresden, Germany, 7–10 September<br />

2008, 517–519.<br />

Scripting languages for simulations in modern<br />

SCADA systems<br />

(Wykorzystanie języków skryptowych dla symulacji w nowoczesnych<br />

systemach SCADA)<br />

mgr PAWEŁ MARCINIAK, mgr ZBIGNIEW KULESZA, prof. dr. hab. ANDRZEJ NAPIERALSKI,<br />

mgr RAFAŁ KOTAS<br />

Politechnika Łódzka, Katedra Mikroelektroniki i Technik Informatycznych<br />

Introduction and brief history of scada<br />

systems<br />

Fast development in the field of technology and science (computer<br />

science, automation, electronics, robotics) made a significant<br />

impact on the appearance of contemporary factories<br />

and industrial plants. In modern manufacturing and industrial<br />

processes, mining and metallurgy industries, private and public<br />

factors leisure and security industries telemetry is often<br />

needed to connect equipment and systems separated by large<br />

distances (thousands of kilometers). New technology is being<br />

applied to almost every industrial process which demands<br />

an advanced control. SCADA systems provide real-time monitoring<br />

and control of your equipment.[1]<br />

SCADA systems have been around as long as there have<br />

been control systems. The first SCADA systems utilized data<br />

acquisition by means of panels of meters, lights and strip chart<br />

recorders. The operator had to operate manually various control<br />

knobs to handle supervisory control. These devices are<br />

still used to do supervisory control and data acquisition in<br />

small factories, plants and power generating facilities.[1]<br />

Modern SCADA systems refer to the combination of telemetry<br />

and data acquisition. SCADA involves the collecting<br />

of the information from PLCs (Programmable Logic Controllers,<br />

transferring it back to the central site, carrying out any<br />

necessary analysis and then displaying that information on<br />

operator screens, displays and personal computers. The required<br />

control actions are then sent back to the process. In<br />

practice SCADA is an industrial control system which encompasses:<br />

HMI (Human Machine Interface), computer system<br />

monitoring, data acquisition and processing, advanced visualization.[2]<br />

36<br />

Scripting languages<br />

Scripting languages allow to carry out several functions which<br />

are not embedded in SCADA software. The incorporation<br />

of scripting languages into SCADA systems increased considerably<br />

their possibilities and enabled to perform functions<br />

which were unavailable before.[1, 3]<br />

The most important virtue of using scripting languages in<br />

SCADA systems is their possibility to simulate a wide range<br />

of industrial processes. Scripting languages help to simulate<br />

processes through ability to generate signals exactly the same<br />

as signals in the real industrial processes. Furthermore, such<br />

systems create other opportunities: animating objects in the<br />

pictures, creating pictures and registers; data processing, giving<br />

access to relational database by ODBC (Open Database<br />

Connectivity), starting other applications, making their own security<br />

system, creating dialog windows and dialog boxes.[3]<br />

The majority of the SCADA vendors equips their software<br />

with following scripting languages: Visual Basic for Application<br />

(iFIX Intellution HMI/SCADA Automation software); Quick-<br />

Script (Wonderware InTouch software); CitectVBA and Cicode<br />

(CitectHMI/SCADA software).<br />

Applying Visual Basic for Application<br />

in scada sytems<br />

iFIX Intellution HMI/SCADA Automation Software uses Visual<br />

Basic for Applications as a scripting language. This event-based<br />

programming language is embedded in most Microsoft<br />

Office applications. VBA is closely related to Visual Basic, but<br />

can only run a code within a host application rather than as<br />

a standalone application. VBA can be applied to control one<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


application from another one by using OLE Automation. VBA<br />

is based on Object Oriented Programming (OOP), a programming<br />

paradigm which uses objects to design applications, animations<br />

and computer programs.<br />

All examples in the following chapter come from the<br />

author’s master thesis which is entitled “The use of SCADA<br />

System in Visualization of Medicine Production Process”.<br />

This thesis concerns making visualization of two existing production<br />

processes (drying in fluid bed dryer and granulating<br />

in vertical granulator). The practical part of the Master’s dissertation<br />

was done in iFIX software [3].<br />

A. Signal generators<br />

iFIX software has a built-in simulation driver which gives an opportunity<br />

to check the work of the system without connecting to<br />

the existing production process. This simulation driver allows to<br />

generate six functions (registers: RA, RD, RE, RF, FG, RH):<br />

• RA – in this register sawtooth wave is generated;<br />

• RD,RE,RF – in this registers three sinusoid are generated<br />

(the first one is shifted 90 degree from the second one and<br />

180 degree from the third one);<br />

• RG – in this register random value is generated;<br />

• RH – in this register triangle wave is generated (frequency<br />

is defined by the register RJ).<br />

Registers RA, RD, RE, RF have the same frequency (defined<br />

by register RY). It means that all signals which use the<br />

simulation driver have always the same frequency (it is a considerable<br />

restriction).<br />

Visual Basic for Applications allows to generate signals defined<br />

by the user. In the practical part of author’s master thesis<br />

the VBA script was written and it generates:<br />

Fig. 1. Square waveform Rys. 1. Przebieg prostokątny<br />

Fig. 2. Signal which follows the square waveform<br />

Rys. 2. Sygnał podążający za zdefiniowaną funkcją prostokątną<br />

1) Square wave – this signal is responsible for shaking filters<br />

in the production process (this script is performed every<br />

50 ms):<br />

2) A signal which follows the defined inlet exhaust temperature<br />

and determines the actual temperature value in the<br />

production process. If the defined inlet exhaust temperature<br />

is established for a long time, the actual temperature<br />

fluctuates around the defined one (this script is performed<br />

every 500 ms).<br />

3) Finite state machine (FSM) – a model of behavior composed<br />

of a finite number of states, transitions (between<br />

those states), and actions. This machine defines which<br />

events are expected to happen, what actions the process<br />

is allowed to take and how it will answer to those events.<br />

FSMs represent sequential logic circuit (it is a equivalent<br />

of a Boolean function in combinational logic circuit). It may<br />

happen that sequential logic circuit contains combinational<br />

logic circuits. The goal of FSMs is to describe a circuit with<br />

inputs and outputs.<br />

Visual Basic for Applications also could be used to implement<br />

a finite state machine. There are two types of finite<br />

state machine (Moore machine and Mealy machine). Choice<br />

of a model depends on the application, execution means and<br />

personal preferences of a designer or programmer. Both Moore<br />

machine and Mealy machine can be designed using VBA).<br />

Moore machine – the output depends only on the internal<br />

state – since the internal state only changes on a clock<br />

Fig. 3. The state diagram of Moore machine Rys. 3. Schemat blokowy automatu Moore’a<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 37


Tabl. 1. Definitions of signals and states (Moore machine)<br />

Tab. 1. Definicje sygnałów i stanów automatu Moore’a<br />

Input: States Outputs<br />

X: {X 1<br />

,X 2<br />

,X 3<br />

,X 4<br />

,X 5<br />

,X 6<br />

,X 7<br />

,X 8<br />

,X 9<br />

,X 10<br />

,X 11<br />

,X 12<br />

,X 13<br />

,X 14<br />

,<br />

X 15<br />

,X 16<br />

,X 17<br />

,X 18<br />

,X 19<br />

}<br />

{START,A,B,C,D,E,F,G,H,I,J,K} Y:{Y 1<br />

,Y 2<br />

,Y 3<br />

,Y 4<br />

,Y 5<br />

,Y 6<br />

,Y 7<br />

,Y 8<br />

,Y 9<br />

,Y 10<br />

}<br />

X 1<br />

– request to empty the product container gasket<br />

X 2<br />

– request to seal the product container gasket<br />

X 3<br />

– request to empty the extension chamber 1 gasket<br />

X 4<br />

– request to seal the extension chamber 1 gasket<br />

X 5<br />

– request to pull out the product container<br />

X 6<br />

– request to seal the product container<br />

X 7<br />

– request to empty the extension chamber 2 gasket<br />

X 8<br />

– request to seal the extension chamber 2 gasket<br />

X 9<br />

– request to pull out the extension chamber<br />

X 10<br />

– request to seal the extension chamber<br />

X 11<br />

– request to empty the product filter 1 gasket<br />

X 12<br />

– request to seal the product filter 1 gasket<br />

X 13<br />

– request to empty the product filter 2 gasket<br />

X 14<br />

– request to seal the product filter 2 gasket<br />

X 15<br />

– allow to change filters<br />

X 16<br />

– request to lower the product filter 1<br />

X 17<br />

– request to raise the product filter 1<br />

X 18<br />

– request to lower the product filter 2<br />

X 19<br />

– request to raise the product filter 2<br />

START – permission to change the actual<br />

state of product container gasket<br />

A – permission to change the actual<br />

state of extension chamber 1 gasket<br />

B – permission to change the position<br />

of product container<br />

C – permission to change the actual<br />

state of extension chamber 2 gasket<br />

D – permission to change the position<br />

of extension chamber<br />

E – permission to change the actual<br />

state of product filter 1<br />

F – permission to change the actual<br />

state of product filter ;<br />

G – permission to change filters<br />

H – permission to lower filters<br />

I – permission to loner the product filter ;<br />

I – permission to lower the product filter 1<br />

K – filters are lowered<br />

Y 1<br />

– The product container gasket has<br />

been empted<br />

Y 1<br />

– The product container gasket has<br />

been empted<br />

Y 2<br />

– The extension chamber gasket has<br />

been empted<br />

Y 3<br />

– The product container has been<br />

pulled out<br />

Y 4<br />

– The extension chamber gasket 2 has<br />

been empted<br />

Y 5<br />

– The extension chamber has been<br />

pulled out<br />

Y 6<br />

– The product filter gasket 1 has been<br />

empted<br />

Y 7<br />

– The product filter gasket 2 has been<br />

empted<br />

Y 8<br />

– Allow to change filters<br />

Y 9<br />

– The product filter 1 has been lowered<br />

Y 10<br />

– The product filter 2 has been lowered<br />

edge, the output only changes on a clock edge too. In Moore<br />

machine outputs are assigned to states or places. Following<br />

diagram illustrates Moore machine which is used to change<br />

product filters in the fluid bed dryer. This FSM has 11 states,<br />

19 inputs and 10 outputs. Transition between two sates is<br />

possible in two ways (two inputs are always responsible for<br />

changing sates). State K is the last one and in this state operator<br />

is able to change product filters.<br />

A special userform (Changing filters) was created (by using<br />

Visual Basic for Application) to operate this state machine. On<br />

this userform there are 18 command buttons and one check<br />

box – they are inputs of the state machine.<br />

An operator can control the process of changing filters<br />

through the userform. The person who controls the system<br />

is able to choose only two from all buttons (the rest of the<br />

buttons are unavailable) in each state (except A and H). In the<br />

first state there is no possibility to select buttons (only one button<br />

is available – Product container sealing OFF). The state H<br />

has three options – the checkbox and two buttons (lower the<br />

filter 1 and lower the filter 2) are available.<br />

The following figure show the fluid bed dryer. They present<br />

visualization of filters changing. Green and blue circles symbolize<br />

gaskets (the green color means that the gasket is full,<br />

the blue color determines an empty gasket). This figure has<br />

four parts (each part describes one states) – part 1 – state A,<br />

part 2 – state C, part 3 – state E, part 4 – state K.<br />

Mealy machine – this machine generates an output based<br />

on its current state and input. In Mealy machine outputs are<br />

associated with transitions. The use of a Mealy FSM leads<br />

often to a reduction of the number of states (according to Moore<br />

machine). Following diagram illustrates. Mealy machine<br />

which is used to administer pharmaceutical processes in the<br />

fluid bed dryer and vertical granulator. This FSM has 2 inputs,<br />

3 outputs and only 4 states,. Transition between two sates is<br />

possible in two ways (by switching of the process or by changing<br />

active machine).<br />

Fig. 4. Userform to change states of the machine and visualization of filters changing in fluid bed dryer<br />

Rys. 4. Panel użytkownika i wizualizacja zmiany filtrów w suszarni fluidalnej<br />

38<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Fig. 5. The state diagram of Moore machine Rys. 5. Schemat blokowy automatu Moore’a<br />

Tabl. 2. Definitions of signals and states (Mealy machine)<br />

Tab. 2. Definicje sygnałów i stanów automatu Mealy’ego<br />

Input: States Outputs<br />

X:{X 1<br />

,X 2<br />

} {START,A,B,C } Y:{Y 1<br />

,Y 2<br />

,Y 3<br />

}<br />

X 1<br />

– request to start fluid bed dryer (it is equivalent to start one<br />

of following functions: unloading from container,- unloading<br />

from a vertical granulator, heating, cooling, spraying (FBD)<br />

X 2<br />

– request to start vertical granulator (it is equivalent to start<br />

one of following functions: filling with air, filling with nitrogen,<br />

warming up, loading, mixing, spraying, granulating, loading<br />

fluid bed dryer, loading intermediate bulk container<br />

START– both machines don’t<br />

work<br />

A – only fluid bed dryer works<br />

B – only vertical granulator<br />

works<br />

C – both machines work<br />

Y 1<br />

– one of fluid bed dryer processes is on<br />

Y 2<br />

– one of vertical granulator processes<br />

is on<br />

Y 3<br />

– two processes are on (one VG process<br />

and one FBD process<br />

B. Writing data to Excel worksheet<br />

Visual Basic for Applications supports OLE (object linking and<br />

embedding) and allows to paste objects from one application<br />

to another one. This very important feature let the author of<br />

the master thesis create a database file (as Excel sheet) in<br />

which significant events are collected. Every turning on and<br />

turning off of the process functions are written to Excel worksheet.<br />

For each entry several additional information are collected<br />

- function parameters, starting time or ending time, kind<br />

of ending function (auto or manual);<br />

With the help of Visual Basic for Applications most Microsoft<br />

Excel worksheet functions can be used. VBA can modify<br />

cells, columns, fonts etc. which enables to create exactly the<br />

same worksheet as by using Microsoft Excel software.<br />

8<br />

Czas<br />

zdarzenia<br />

Nazwa<br />

zdarzenia<br />

Rodzaj<br />

zdarzenia<br />

9 13:47:45 Grzanie OFF<br />

Stopień otwarcia<br />

klapy<br />

załadunku<br />

K203010<br />

Temperatura<br />

pow.<br />

wlotowego<br />

Przepływ<br />

powietrza<br />

woltowego<br />

Ciśnienie<br />

powietrza<br />

atomizującego<br />

Czas<br />

wytrząsania<br />

filtrów<br />

Czas przerwy<br />

w wytrząsaniu<br />

filtrów<br />

Rodzaj<br />

wytrząsania<br />

10 13:47:59 Chłodzenie ON 25 1677 2 2 asynchroniczne<br />

11 13:48:20 Chłodzenie OFF<br />

12 13:48:35 Grzanie ON 55 150 5 5 synchroniczne<br />

13 13:49:03 Grzanie OFF<br />

14 13:49:06 Chłodzenie ON 25 1677 2 2 asynchroniczne<br />

15 13:49:37 Chłodzenie OFF<br />

16 13:49:46 Grzanie ON 45 150 5 5 synchroniczne<br />

17 13:50:24 Grzanie OFF<br />

18 13:50:26 Chłodzenie ON 25 1677 2 2 asynchroniczne<br />

19 13:50:49 Chłodzenie OFF<br />

Użytkownik<br />

Jan<br />

Kowalski<br />

Jan<br />

Kowalski<br />

Jan<br />

Kowalski<br />

Jan<br />

Kowalski<br />

Jan<br />

Kowalski<br />

Jan<br />

Kowalski<br />

Jan<br />

Kowalski<br />

Jan<br />

Kowalski<br />

Jan<br />

Kowalski<br />

Jan<br />

Kowalski<br />

Jan<br />

Kowalski<br />

Fig. 6. The database in Excel worksheet Rys. 6. Archiwizacja danych do arkusza kalkulacyjnego<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 39


C. Data processing and report making<br />

VBA allows to write functions which are very useful for making<br />

reports. Data can be taken from database file, processed and<br />

written to the report file. This report includes following information:<br />

• Number of starts of each function;<br />

• Time in which the function was used;<br />

• Summary.<br />

Tabl. 3. The daily report table<br />

Tab. 3. Tabela z dziennym raportem<br />

Nazwa funkcji<br />

Liczba<br />

uruchomień<br />

Całkowity<br />

czas<br />

Ilość zużytego<br />

środka<br />

natryskującego<br />

Załadunek z VG 3 00:01:45 0<br />

Załadunek z IBC 4 00:03:12 0<br />

Grzanie 5 00:08:33 0<br />

Chłodzenie 4 00:09:48 0<br />

Natryskiwanie 4 00:05:25 2,863<br />

Suma 20 00:28:43 2,863<br />

D. Advanced system protection<br />

Visual Basic for Applications enables developing embedded<br />

system protection into advanced and well adapted (for user)<br />

system protection. Those system can have hierarchical structure.<br />

It means that, programmers divides protection mechanism<br />

into some access levels. System could have for example<br />

three levels:<br />

• Administrator level access – users which are assigned to<br />

this level can do everything without limitations.<br />

• Operators level access – users on this level can only control<br />

process and he cannot modify settings and source<br />

code. They could use process functions to control the production<br />

process.<br />

• Quests level access – this level only allows to watch.<br />

Conclusion<br />

By using SCADA systems the operational costs can be minimized<br />

by means of providing direct information of system performance,<br />

thus improving system efficiency. The use of SCA-<br />

DA systems increases equipment life and reduces labor costs<br />

required for troubleshooting or servicing the equipment.<br />

The performed application (practical part of the author’s<br />

master thesis) presents the practical usage of the scripting<br />

languages in SCADA systems. The whole system was designed<br />

by using scripting languages instead of iFix Software<br />

embedded functions. This system carries out the monitoring<br />

process function, data acquisition, files presentation and protection<br />

against people who are not entitled. It shows that scripting<br />

languages have huge capabilities and allow to execute<br />

the highly specialized functions.<br />

Paweł Marciniak and Rafał Kotas are a scholarship holders<br />

of project entitled “Innovative education ...” supported by European<br />

Social Fund.<br />

References<br />

[1] Bailey D., Wright E.: Practical SCADA for Industry. Elsevier and<br />

Newnes, Burlington 2003.<br />

[2] Supervisory Control and Data Acquisition (SCADA) Systems.<br />

Technical Information Bulletin 04-1, National Communication<br />

System, October 2004,<br />

[3] Marciniak P.: Wykorzystanie pakietów SCADA w wizualizacji<br />

procesu produkcji leków.<br />

[4] AutomatykaB2B.pl – Portal branżowy dla Automatyków, marzec<br />

<strong>2010</strong>, http://automatykab2b.pl,<br />

Polskie serwery w organizacji europejskiej<br />

ACTION SA jest jednym z największych dystrybutorów oraz<br />

producentów sprzętu komputerowego w Polsce. Spółka jest<br />

notowana na Giełdzie Papierów Wartościowych w Warszawie<br />

od 2006 r.; jest członkiem Stowarzyszenia Emitentów Giełdowych.<br />

Ostatnio firma wygrała prestiżowy przetarg o wartości<br />

13,8 mln zł na dostawę serwerów własnej marki (Actina Solar)<br />

do Europejskiej Organizacji Badań Jądrowych w Genewie<br />

(CERN). W ciągu ośmiu miesięcy do Genewy w sześciu transportach<br />

trafiło 1132 serwery (952 serwery Actina Solar820 S4<br />

oraz 180 serwerów Actina Solar240 S4). Urządzenia były<br />

poddawane bardzo rygorystycznym testom i sprostały wymaganiom<br />

stawianym przez specjalistów. Przyczynią się więc do<br />

pomnażania dorobku naukowego Europy. W CERN pracuje<br />

obecnie około 200 osób z Polski – wybitnych naukowców<br />

i specjalistów.<br />

Akronim CERN pochodzi od pierwotnej, francuskojęzycznej<br />

nazwy Europejska Rada Badań Jądrowych (fr. Conseil<br />

Européen pour la Recherche Nucléaire) i mimo jej zmiany<br />

po konwencji w 1953 r. został zachowany.<br />

Serwery Actina Solar funkcjonują w ramach potężnej infrastruktury<br />

informatycznej CERN, wśród niemal 7 tys. urządzeń<br />

40<br />

tego typu. Klaster obliczeniowy złożony z dostarczonych serwerów<br />

jest w stanie wykonywać 75 bilionów działań matematycznych<br />

na sekundę.<br />

Jednym z najbardziej znanych z licznych ultranowoczesnych<br />

urządzeń wykorzystywanych przez naukowców w CERN<br />

jest Wielki Zderzacz Hadronów (LHC – Large Hadron Collider).<br />

Zderzenia hadronów są śledzone i rejestrowane za pomocą<br />

sześciu zespołów detektorów cząstek, a serwery obliczeniowe<br />

analizują, przetwarzają i zapisują wszystkie informacje<br />

powstające w wyniku każdego z eksperymentów. Szacuje się,<br />

że w momencie największego natężenia zderzeń, w ciągu sekundy<br />

dochodzi do 600 milionów kolizji protonów, co powoduje<br />

ogromną ilość danych liczoną w pB/s (petabajtach (biliardach<br />

bajtów) na sekundę)<br />

Za stosowanie innowacyjnych rozwiązań w serwerach<br />

Actina Solar oraz za dynamikę ich rozwoju, firma ACTION<br />

została dwukrotnie nagrodzona przez firmę Intel podczas<br />

dorocznych konferencji branżowych „Intel Solutions Summit”<br />

(kwiecień 2007 r. w Paryżu oraz kwiecień 2008 r. w Rzymie).<br />

(cr)<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


FPGA implementation of feature extraction<br />

algorithm for speaker verification<br />

(Implementacja algorytmu estymacji cech sygnału mowy<br />

systemu automatycznej weryfikacji mówcy w układzie FPGA)<br />

mgr inż. MICHAŁ STAWORKO, dr inż. MARIUSZ RAWSKI<br />

Politechnika Warszawska, <strong>Instytut</strong> Telekomunikacji, Wydział Elektroniki i Technik Informacyjnych<br />

Biometric systems are usually implemented on personal computers<br />

equipped with high-performance microprocessors.<br />

This is because of the computation complexity of applied algorithms,<br />

as well as their high confidential levels of security.<br />

General purpose processors contain floating-point units able<br />

to carry out millions of operations per second at frequencies<br />

in the GHz range, what allows to resolve the complex algorithms<br />

in just a few hundred of milliseconds. However, in the<br />

low-cost consumer market, such factors as price, power consumption<br />

and size determine the viability of a product. Since<br />

the main drawback of microprocessors based systems are the<br />

cost, and the necessary space required to incorporate their<br />

external associated peripherals, the use of an FPGA (Field<br />

Programmable Gate Array) becomes a suited way to implement<br />

systems that require a high computational capability at<br />

affordable prices. Additionally the FPGA allows to divide and<br />

implement algorithm as parallel parts, what allows to perform<br />

computations at lower operational circuit frequency which requires<br />

less power consumption.<br />

FPGA circuits can be programmed by the user and adapted<br />

to perform the particular task. Term “programming” in case<br />

of a FPGA architecture means changing interconnections in<br />

its internal structure and can be repeated many times. This<br />

mechanism that allows for FPGA programming on the one<br />

hand decreases operating frequency of FPGA chip in comparison<br />

to ASIC, on the other hand provides the possibility to<br />

tune-up the system to the specific parameters of implemented<br />

algorithm.<br />

The speaker identification problem can be roughly divided<br />

into two issues: speech analysis (feature extraction) and<br />

classification. Feature extraction methods are responsible for<br />

reducing the resources required to describe speech samples<br />

accurately. In case of speech analysis various digital signal<br />

processing (DSP) algorithms are used to detect desired features<br />

of speech signal. Most common are Linear Prediction<br />

Coding (LPC), Mel-scale Filterbank Cepstrum Coefficients<br />

(MFCC) and Linear-scale Filterbank Cepstral Coefficiens<br />

(LFCC). MFCC is recognized as the best known and most<br />

popular. However LFCC algorithm is often used in speaker<br />

identification application, since it produces results of comparable<br />

quality [1–3].<br />

There are some publications dealing with FPGA implementations<br />

of speaker verification systems. In [4] authors show the<br />

main features of an implementation of a SVM (Support Vector<br />

Machines) speaker verification system for Match-on-Card.<br />

The paper presents a FPGA implementation of the matching<br />

stage using a kernel based on an exponential function. The<br />

system is able to carry out the matching between the model<br />

and the feature vector 50 times faster than a software-based<br />

solution running on a Pentium IV at 1.3 GHz. In [5] the implementation<br />

of a whole SVM speaker verification system based<br />

on dedicated hardware is presented. The system consists<br />

of several stages dedicated to calculate the feature vectors,<br />

based on mel-frequency cepstral coefficients. In [6] and [7]<br />

there are presented only hardware implementations of some<br />

specific part of algorithms for speech recognition or speaker<br />

identification, that allow a significant acceleration of the processing<br />

time.<br />

This paper presents the feature extraction solution based<br />

on LFCC with the architecture specially optimized for implementation<br />

in FPGA structures. Discussed system is designed<br />

to be implemented in FPGA device as System-on-Programmable-Chip<br />

(SOPC).<br />

Feature Extraction Circuit<br />

State of the art speaker verification systems are based on<br />

the estimation of the spectral envelope of the short term signal,<br />

e.g., Mel-scale Filterbank Cepstrum Coefficients, Linearscale<br />

Filterbank Cepstrum Coefficients, or Linear Predictive<br />

Cepstrum Coefficients (LPCCs). All the features examined are<br />

based on spectral information derived from a short time-windowed<br />

segment of speech. They differ mainly in the detail of<br />

the power spectrum representation. The mel-scale filter bank<br />

centers and bandwidths are fixed to follow the mel-frequency<br />

scale, giving more detail to the low frequencies, whereas<br />

the linear filter bank provides equal details to all frequencies<br />

[1, 2, 3].<br />

Linear Frequency Cepstral Coefficients (LFCC) are used in<br />

speech and speaker recognition applications and give similar<br />

results as MFCC, which is recognized as the state of art. [8].<br />

Methods used to compute the MFCC and LFCC parameters<br />

differ in construction of filter bank that is applied for spectral<br />

averaging. To compute the MFCC the mel-scale wrapped filter<br />

bank is used, in which windows located in lower frequency<br />

are far more slimmer than located in high freq. That gives precise<br />

description of low frequencies and rough image of high<br />

frequencies. This non linear mapping corresponds to human<br />

frequency perception.<br />

To compute the LFCC parameters evenly distributed in<br />

frequency domain windows are used. That describes both,<br />

high and low frequency with the same precision. If comparing<br />

LFCC and MFCC filter banks the former gives better image<br />

of high frequency. In [9] it is concluded that a rich amount of<br />

speaker individual information is contained in the higher frequency<br />

band, and it is useful for speaker recognition. Thus<br />

LFCC may be consider as better speech signal estimators,<br />

since it provides better high-frequency resolution.<br />

Spectral Averaging Filter Bank<br />

The spectral averaging returns the acoustic energy distribution<br />

in bands limited by the averaging windows and significantly<br />

reduces the speech signal window size for further processing.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 41


Fig.1. First 9 windows of the bank of 31 cepstral filters<br />

Rys. 1. Pierwsze 9 okien z banku 31 filtrów cepstralnych<br />

With the reduction of the number of averaging windows the<br />

amount of input data for computationally expensive logarithm<br />

is reduced, as well as the number of multiplication operations<br />

during DCT computation. The main computational problem in<br />

implementation of spectral averaging filter bank is overlapping<br />

of filter windows what does not allow to perform straight<br />

multiply accumulate operation. The regularity of LFCC allows<br />

to significantly simplify computation process. The linear filter<br />

bank, in comparison to MFCC, requires also less memory for<br />

storing window coefficients, since it has linear distribution in<br />

frequency domain and every filter has the same coefficient.<br />

It is possible to find such set of averaging windows linearly<br />

distributed in frequency domain, that the coefficients of filters<br />

are linear combination of powers of 2. That allows to eliminate<br />

computational expensive multiplication operation and replace<br />

it by bit shifts and additions only. The example of such filter<br />

bank is presented on the Fig. 1.<br />

In [2] it has been shown that error rate of automatic speaker<br />

verification algorithm based on LFCC and DTW decreases<br />

when the number of filters grows. Thus such filter bank should<br />

be chosen that has coefficients being linear combination of powers<br />

of 2 and has required trade-off between the performance<br />

(computation costs) and quality (the algorithm error rate).<br />

Hardware Implementation<br />

Fig. 2. presents the block diagram of LFCC processor where<br />

function implemented in hardware components are shown.<br />

The processor was designed to be implemented in FPGA<br />

structure and to work at 50 MHz frequency.<br />

In the pre-processing module speech signal is filtered with<br />

pre-emphasis filter and buffered in dual-port RAM memory.<br />

It is then split into 256 samples frames with 160 samples<br />

overlapping. Next Hamming window function is applied to the<br />

signal. Fourier transform is implemented as FFT library core.<br />

Fig. 3. The architecture of spectral averaging circuit with 31 triangular<br />

windows, equally distributed in frequency domain<br />

Rys. 3. Architektura układu uśredniającego widmo, 31 trójkątnymi<br />

oknami równomiernie rozłożonymi w dziedzinie częstotliwości<br />

The spectral averaging (filter channels) may be the most computationally<br />

extensive part of the feature extraction unit [5].<br />

Using the linear filter bank the computation cost may be significantly<br />

reduced up to 45 times. The Fig. 3 presents the concept<br />

of hardware implementation of spectral averaging with use of<br />

31 equally distributed windows.<br />

Such cepstral averaging component does not introduce<br />

any delay into the circuit. Thanks to this it may operate at low<br />

frequency equal to the input signal sample rate, what significantly<br />

reduces the power consumption, that is proportional to<br />

the operational frequency.<br />

The complexity of given application depends on the number<br />

of spectral averaging filters used. If less filters are used<br />

the number of coefficients of each filter increases. Thus the<br />

size of multiplexers and the number of combinational circuit<br />

for bit shifts and additions increases proportionally to the number<br />

of filter coefficients.<br />

The logarithm module is implemented as FSM according<br />

to the method described in [10]. Discrete Cosine Transform<br />

is implemented in MAC fashion. The values of twiddle factors<br />

are stored in ROM memory. The size of coefficients is selected<br />

in a way that ensures most efficient utilization of embedded<br />

memory blocks of FPGA structure and can be changed using<br />

software application created by the authors.<br />

The implementation of the LFCC processor was verified<br />

with the bit accurate fixed point model presented in [11]. With<br />

the dynamic time warping pattern matching procedure and<br />

presented LFCC features processor the automatic speaker<br />

verification machine has error rate of 1,5%.<br />

Results<br />

Fig. 2. LFCC processor hardware setup scheme<br />

Rys. 2. Schemat sprzętowej implementacji procesora do wyznaczania<br />

parametrów LFCC<br />

42<br />

The LFCC processor described in this paper has been modeled<br />

in VHDL and implemented in low cost Cyclone II<br />

EP2C35F484C8 FPGA [12]. Quartus II 7.2 CAD tool was used<br />

with optimization set to “balanced ”.<br />

The 12 cepstral coefficients are estimated from one 256<br />

sample speech signal frame just in 1,211 clock cycles. The<br />

next speech frame is processed when new 180 samples are<br />

collected in buffer. The presented solution allows to real time<br />

LFCC features extraction at the frequency only 10 times greater<br />

than speech signal sampling frequency.<br />

Tabl. 1 presents the performance comparison of implementation<br />

presented in this paper and solution described<br />

in [5]. Solution described by Ramos-Lara et al. was implemen-<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Tabl. 1. Frame execution time comparison<br />

Tab. 1. Zestawienie czasu obliczeń okna sygnału mowy<br />

Method presented in [5]<br />

Pentium IV at<br />

1.5 GHz<br />

FPGA<br />

at 50 MHz<br />

Our method<br />

FPGA<br />

at 50 MHz<br />

Pre-processing 17.25 μs 55.96 μs 5.12 μs<br />

FFT 63.36 μs 30.22 μs 10.54 μs<br />

Filter Channels 45.45 μs 116.48 μs 2.56 μs a<br />

Logarithm 8.41 μs 53.78 μs 12.40 μs<br />

DCT 102.57 μs 26.46 μs 7.44 μs<br />

Delta coefficients 1.73 μs 2.54 μs –<br />

Total 238.77 μs 285.44 μs 24.22 μs<br />

a) including squared module computation, filter bank processing and normalization<br />

ted on Pentium IV general purpose processor and on a Xilinx<br />

Spartan 3XCS2000 FPGA. It can be noticed that presented<br />

solution performs very well. It is 10 times faster than both solution<br />

designed by Ramos-Lara et al.: FPGA based, as well as<br />

Pentium based. The presented solution does not estimate delta<br />

coefficients, because they may be estimated offline during<br />

the pattern matching. Additionally it was shown in [2] that the<br />

dynamic parameters do not introduce significant improvement<br />

in quality of speaker recognition (for LFCC speech features<br />

and DTW pattern matching) and introduce big computational<br />

effort in pattern matching procedure, since with dynamic<br />

parameters the feature vector grows two (when using velocity<br />

parameters, the first derivative of estimated parameters)<br />

or three times (when using velocity and acceleration parameters,<br />

1 st and 2 nd derivative of parameters)<br />

The efficiency of described solution allows to estimate speech<br />

signal features in real time even if the presented LFCC<br />

processor works at frequency just 10 times faster than the<br />

speech signal sampling frequency. This is very power efficient<br />

solution and it is suitable for battery powered devices like mobile<br />

phones, netbooks and other personal PDAs.<br />

Tabl. 2. presents the resources necessary to implement<br />

LFCC processor in Cyclone II EP2C35F484C8. The columns<br />

falling under LE, FF, RAM, MAC labels describe number<br />

of logic elements, flip flops and bits of on-chip RAM memory<br />

respectively. Column falling under f max<br />

presents the<br />

maximal frequency that given component can work with.<br />

For comparison solution described in [5] implemented in Spartan<br />

3XCS2000 requires 3817 slices (6138 4 input LUTs) 4659<br />

flip-flops and 13 multipliers.<br />

Tabl. 2. Implementation results of LFCC processor in FPGA<br />

Tab. 2. Wyniki implementacji procesora LFCC w układzie FPGA<br />

LFCC component<br />

FPGA resources<br />

LE FF RAM MAC<br />

Conclusions<br />

The presented hardware implementation of LFCC processor<br />

is very efficient. It allows to estimate speech features more<br />

than 10 times faster than similar solution presented in [5]. The<br />

described module for spectral averaging optimized for implementation<br />

in FPGA architecture significantly improves the<br />

overall performance. It solves the most computational extensive<br />

part of cepstral feature extraction. The presented solution<br />

allows to estimate speech signal features in real time even<br />

if the presented LFCC processor works at frequency only 10<br />

times greater than the speech signal sampling frequency. This<br />

makes it very power efficient solution suitable for battery powered<br />

devices like mobile phones, netbooks and other personal<br />

PDAs that require voice security facilities. The presented<br />

feature extraction module occupies less than 20% resources<br />

of the low cost Cyclone II EP2C35 FPGA. This allows to integrate<br />

it with other components of speaker verification tool<br />

in form of the System on Programmable Chip and easily fit<br />

it into one FPGA.<br />

This work was partly supported by the Ministry of Science and<br />

Higher Education of Poland – research grant no. N N516 418538<br />

for <strong>2010</strong>–2012.<br />

References<br />

Pre-processing 85 52 12 288 2 133<br />

FFT 3 428 3 468 14 870 24 105<br />

[1] Reynolds D. A.: Experimental Evaluation of Features for Robust<br />

Speaker Identification. IEEE Trans. Speech and Audio Processing,<br />

2(4), 1994, 639–643.<br />

[2] Kaczmarek A., Staworko M.: Application of Dynamic Timer Warping<br />

and Cepstrograms to Text-Dependent Speaker Verification.<br />

SPA 2009, Poznań, 24–26 September: conference proceedings,<br />

2009, 169–174.<br />

[3] Charbuillet C. i in.: Optimizing feature complementarity by<br />

evolution strategy: Application to automatic speaker verification.<br />

Speech Communication, vol. 51, no. 9, September, 2009,<br />

724–731.<br />

[4] Choi, W-Y. i in.: SVM-Based Speaker Verification System for<br />

Match-on-Card and its Hardware Implementation. ETRI Journal,<br />

vol. 28, no. 3, 2006, 320–328.<br />

[5] Ramos-Lara R., i in.: SVM speaker verification system based on<br />

a low-cost FPGA. Field Programmable Logic and Applications,<br />

2009, 582–586.<br />

[6] Nedevschi S., Patra R., Brewer E.: Hardware Speech Recognition<br />

for User Interfaces in Low cost, Low Power Devices. 43nd<br />

Design Automation Conference, 2005, 684–689.<br />

[7] Melnikoff S., Quigley S.F., Rusell M. J.: Implementing a Simple<br />

Continuous Speech Recogniton System on an FPGA. Proceedings<br />

of the 10th Annual IEEE Symposium on Field-Programmable<br />

Custom Computing Machines, Napa, California, USA,2002,<br />

275–276.<br />

[8] Campbell J. P.: Speaker Recognition: A Tutorial, Proc. IEEE, vol.<br />

85, no. 9, 1997, 1437–1462.<br />

[9] Hayakawa S., Itakura F.: Text-dependent<br />

speaker recognition using the information<br />

in the higher frequency band, ICASSP, vol.<br />

1, Acoustics, Speech, and Signal Processing,<br />

1994, 137–140.<br />

[10] Majithia J. C, Levan D.: A note on base-2<br />

logarithm computations. Proc. IEEE, 1973,<br />

1519–1520.<br />

[11] Staworko M., Rawski M.: A Data-accurate<br />

Model of Automatic Speaker Verification<br />

Algorithm for Implementation as<br />

System-on-Programmable-Chip in FPGA<br />

Structure. II International Interdisciplinary<br />

Technical Conference of Young Scientists,<br />

20-22 May, Poznan Poland, 2009,<br />

207–211.<br />

[12] Altera Corporation, Cyclone II Device Handbook,<br />

Volume 1. 2009, http://www.altera.<br />

com/literature/hb/cyc2/cyc2_cii5v1.pdf<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 43<br />

f max<br />

[MHz]<br />

Spectral averaging a 511 151 0 4 86.39<br />

Filter bank 332 107 0 0 111<br />

Logarithm 119 68 0 2 85.30<br />

DCT 127 90 5 120 2 85.07<br />

Total 4 774 (14%) 3 773 (11%) 32 278 (7%) 34 (49%) 81.25<br />

a) including squared module computation, filter bank processing and normalization


Digital implementation of a programmable<br />

type-2 fuzzy logic controller<br />

(Cyfrowa realizacja programowalnego sterownika rozmytego 2-go rzędu)<br />

mgr inż. MICHAŁ BRYK, dr inż. ANDRZEJ WIELGUS<br />

Politechnika Warszawska, <strong>Instytut</strong> Mikroelektroniki i Optoelektroniki<br />

Fuzzy systems are widely used in many applications concerning<br />

control. Most of them utilize classical models of fuzzy<br />

systems. Every fuzzy system is associated with some kinds<br />

of uncertainties which affect model’s accuracy. Since all fuzzy<br />

systems are based on some expert knowledge, and it is common<br />

truth that words can mean different things to different<br />

people, it seems clear that uncertainties are an integral part of<br />

such systems. Type-2 fuzzy logic systems are capable of handling<br />

and model such uncertainties, thus improving model’s<br />

accuracy. Practical applications of type-2 fuzzy logic controllers<br />

(FLCs) are mostly based on software implementations for<br />

general purpose processors. While the classical fuzzy controllers<br />

are often implemented in hardware as digital or analogue<br />

electronic circuits, only few hardware implementations of<br />

type-2 FLCs have been proposed [4, 11, 12]. All of them use<br />

the simplified type of type-2 fuzzy sets, called interval type-2<br />

fuzzy sets, and approximation method of type reduction. Some<br />

papers [11, 12] present FPGA based implementations. In [4]<br />

highly parallel architecture of a pipelined fuzzy logic processor<br />

is presented. It reaches high inference speed over 3 MFLIPS,<br />

but the area cost and power consumption are also very high.<br />

In this paper the authors propose a different architecture<br />

of a type-2 FLC, which was designed for serial processing of<br />

fuzzy rules combined with parallel processing of upper and<br />

lower membership functions of type_2 fuzzy sets. Hardware<br />

implementation was based on CMOS 90 nm standard cell library.<br />

Such a solution allowed us to lower significantly the area<br />

and power consumption of the circuit while keeping reasonable<br />

inference speed of about 280 KFLIPS.<br />

Type-2 Fuzzy Sets<br />

The following definition describes a classical fuzzy set.<br />

Definition 1.<br />

Fuzzy set A (type-1 fuzzy set) is a set of pairs defined on a<br />

universe of discourse X,<br />

Fig. 1. Membership function of a type-2 fuzzy set<br />

Rys. 1. Funkcja przynależności zbioru rozmytego 2-go rzędu<br />

All the fuzzy operations on type-2 sets must be redefined<br />

and become much more complicated because membership<br />

functions are now 3-dimentional, as shown in Fig. 1. Thus,<br />

hardware implementation of fuzzy processing based on the<br />

general definition of the type-2 fuzzy sets seems to be really<br />

difficult, resource and time consuming. To make the task easier,<br />

a simplified version of a type-2 fuzzy set was proposed in<br />

[9]. The definition is given below.<br />

Definition 3.<br />

A type-2 fuzzy set in which:<br />

µ<br />

~ (<br />

x<br />

,<br />

u<br />

)<br />

1,<br />

∀<br />

x<br />

∈<br />

X<br />

,<br />

∀<br />

u<br />

∈<br />

J<br />

⊆ [0,1]<br />

(3)<br />

A<br />

= x<br />

is called an interval type-2 fuzzy set.<br />

In such a case, the third dimension of a type-2 membership<br />

function is no longer needed, since it carries no new informa-<br />

A = {(µ (1)<br />

A ( x<br />

), x<br />

)},<br />

∀ x<br />

∈ X<br />

,<br />

where µ A<br />

is a membership function that assigns any element<br />

x ∈ X membership grade µ A<br />

(x) in A in which µ A<br />

(x) ∈ [0,1].<br />

In the case of a type-2 fuzzy set, all the membership grades<br />

are themselves classical fuzzy sets instead of the crisp<br />

values.<br />

Definition 2.<br />

A type-2 fuzzy set, denoted Ã, is characterized by a type-2<br />

membership function µ Ã<br />

(x, u), where x ∈ X and u ∈ J x<br />

⊆ [0, 1],<br />

i.e.,<br />

~<br />

A {((<br />

x<br />

,<br />

u<br />

),<br />

µ<br />

~ (<br />

x<br />

,<br />

u<br />

)),<br />

∀<br />

x<br />

∈<br />

X<br />

,<br />

∀<br />

u<br />

∈<br />

J<br />

⊆ [0,1]},<br />

(2)<br />

in which µ Ã<br />

(x,u) ∈ [0,1].<br />

44<br />

= A<br />

x<br />

Fig. 2. Membership function of an interval type-2 fuzzy set<br />

Rys. 2. Funkcja przynależności interwałowego zbioru rozmytego<br />

2-go rzędu<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


tion about a set. That is why we can define an interval type-2<br />

fuzzy set with upper and lower bounds of function defined on<br />

(x,u) plane (a shaded area in Fig. 1). They are called respectively:<br />

upper and lower membership functions and both are<br />

type-1 membership functions. An example is shown in Fig. 2.<br />

All the fuzzy operations on interval fuzzy sets are defined<br />

by using well known classical fuzzy logic [9,13]. Type-2 fuzzy<br />

sets bring an additional degree of freedom comparing to type-<br />

1 fuzzy sets. In the case of interval fuzzy sets, designers don’t<br />

loose that degree of freedom. Because of that, and the fact<br />

that general type-2 fuzzy sets require complex computations,<br />

interval type-2 fuzzy sets are widely used in practical applications.<br />

Type-2 fuzzy logic controller<br />

The general structure of a type-2 fuzzy logic controller, depicted<br />

in Fig. 3, consists of four basic modules: fuzzification, fuzzy inference<br />

with the rule base, type reduction and defuzzification.<br />

In the fuzzifier block crisp values from inputs are mapped<br />

into type-2 fuzzy sets. As a result, for each input signal two<br />

membership levels are obtained, one called lower membership<br />

level and the other called upper membership level, which<br />

are computed from the lower and upper membership functions<br />

respectively.<br />

In the inference block type-2 fuzzy antecedents sets are<br />

combined and computed to the type-2 fuzzy consequent sets.<br />

These operations are based on a set of fuzzy rules which are<br />

stored in the rule base. In this paper the following form of the<br />

rules was used:<br />

R if<br />

x<br />

is<br />

X<br />

and<br />

x<br />

is<br />

X<br />

...or<br />

x<br />

is<br />

X<br />

...<br />

then<br />

y<br />

is<br />

Y<br />

,<br />

y<br />

is<br />

Y<br />

... (4)<br />

k<br />

: 1 1<br />

j<br />

2<br />

2<br />

l<br />

i<br />

ij<br />

1<br />

1<br />

j<br />

2<br />

2<br />

l<br />

where x i<br />

and y i<br />

are input and output linguistic variables (eg.<br />

speed, height, temperature), X k<br />

and Y l<br />

represents labels of<br />

interval type-2 fuzzy sets (eg. slow, fast, low, medium, high).<br />

The connectives and, or represent operations on fuzzy sets:<br />

intersection and union, respectively, defined for type-1 fuzzy<br />

sets as stated in the previous section. Several assumptions<br />

have been made on fuzzy reasoning process:<br />

• Mamdani method of fuzzy reasoning is employed,<br />

• “and” connective is implemented as min,<br />

Fig. 3. General block diagram of a type-2 fuzzy logic controller<br />

Rys. 3. Schemat blokowy sterownika rozmytego 2-go rzędu<br />

• “or” connective is implemented as max,<br />

• rules connective is implemented as max.<br />

The type reduction module represents mapping from interval<br />

type-2 fuzzy sets into classical interval fuzzy sets. Since<br />

we operate on interval sets, it computes an interval [y L<br />

, y R<br />

].<br />

To date two algorithms have been proposed for interval type-<br />

2 fuzzy sets: the Karnick-Mendel iterative procedure [5] and<br />

the Wu-Mendel closed forms [7]. The former method provides<br />

an exact computation, the latter provides an approximation.<br />

Although the KM algorithm is computationally more costly, we<br />

decided to implement this solution. As mentioned earlier, the<br />

aim of this algorithm is to find an interval. In order to do so,<br />

centroids of certain type-1 membership functions are computed.<br />

The algorithm looks as follows.<br />

Algorithm 1. The Karnick-Mendel procedure of type reduction.<br />

Step 1. Initialization: set µ i<br />

and calculate y’ for each sample of<br />

the output type-2 fuzzy set (i indicates consecutive samples):<br />

s<br />

−<br />

1<br />

∑<br />

x<br />

i *<br />

µ<br />

i<br />

µ<br />

+<br />

= 1<br />

i<br />

µ<br />

i<br />

i<br />

y<br />

' =<br />

,<br />

µ<br />

= , i = 0,1,..., s −1 s−1<br />

i<br />

, (5)<br />

2<br />

µ<br />

∑<br />

i<br />

=<br />

1<br />

i<br />

Step 2. Find index k (1 ≤ k ≤ s-1) such that x k-1<br />

≤ y’ ≤ x k+1<br />

.<br />

Fig. 4. An example of type-2 fuzzy reasoning<br />

Rys. 4. Przykład wnioskowania rozmytego opartego na zbiorach rozmytych 2-go rzędu<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 45


Step 3. Set new values of µ i<br />

and calculate y ” :<br />

s−1<br />

⎪⎧<br />

µ<br />

i<br />

, i ≤ k⎪⎫<br />

∑ x *<br />

for calculation<br />

i<br />

µ µ<br />

i<br />

= ⎨ ⎬ y<br />

i<br />

L<br />

,<br />

i 1<br />

i<br />

i k<br />

y<br />

,<br />

⎪⎩<br />

µ ><br />

=<br />

′′ =<br />

⎪⎭<br />

(6)<br />

s−1<br />

⎪⎧<br />

i<br />

, i > k⎪⎫<br />

∑ µ<br />

µ<br />

i µ<br />

i<br />

= ⎨ ⎬ for yR<br />

calculation<br />

i=<br />

1<br />

⎪⎩<br />

µ<br />

i<br />

, i ≤ k⎪⎭<br />

Step 4. Check if y’ = y’’. If yes: stop y L<br />

= y” (or y R<br />

= y” ). If no:<br />

set y’ = y” and go to step 2.<br />

The defuzzification module computes crisp output value<br />

from the output interval fuzzy set. In case of an interval set, it<br />

is computed as an arithmetic mean of the interval’s boundaries<br />

y L<br />

and y R<br />

.<br />

Fig. 4 presents an example of the whole process of type-2<br />

fuzzy reasoning for the following set of two fuzzy rules:<br />

R : if<br />

1<br />

R : if<br />

2<br />

x is X<br />

1<br />

1<br />

11<br />

x is X<br />

12<br />

and<br />

and<br />

x is X<br />

Fuzzification of two input variables x 1<br />

and x 2<br />

is made for<br />

corresponding input fuzzy set to obtain two membership grades<br />

for each of them. Antecedents sets (stated in one rule)<br />

are combined together with MIN operators and mapped into<br />

consequent sets Y 1<br />

and Y 2<br />

also with MIN operators .<br />

The resulting<br />

fuzzy sets from both rules are then aggregated with MAX<br />

operator to obtain the output type-2 fuzzy set. Next, type reduction<br />

with KM algorithm is performed on this set to produce<br />

the interval [y L<br />

, y R<br />

] and finally the crisp output value.<br />

Proposed architecture of digital type-2<br />

FLC<br />

Parallel processing of fuzzy rules, as in [4], is suitable mainly<br />

for timing critical applications. However, high speed is usually<br />

accomplished with the high costs concerning area and power,<br />

which might not be acceptable in many specific applications.<br />

In order to lower the area and power consumption while keeping<br />

the speed on a reasonably high level, it was decided to<br />

2<br />

2<br />

21<br />

x is X<br />

22<br />

then<br />

then<br />

y is Y<br />

1<br />

y is Y<br />

2<br />

employ serial processing of fuzzy rules and try to minimize the<br />

number of switches. A new architecture was design that allows<br />

to combine parallel processing of both bounds of interval<br />

fuzzy sets with serial processing of rules. The general block<br />

diagram, shown in Fig. 5, presents the main idea of proposed<br />

combined processing. As each interval type-2 fuzzy set is described<br />

by the upper and lower membership functions, there<br />

are two parallel processing paths, including fuzzification and<br />

fuzzy inference blocks, that perform calculations regarding the<br />

upper and lower bounds, respectively.<br />

The proposed architecture of our digital implementation<br />

of a type-2 fuzzy logic controller is presented in Fig. 6. The<br />

circuit can operate in two modes. In the programming mode<br />

the following fuzzy data is stored in the memory unit:<br />

• input and output interval type-2 fuzzy sets,<br />

• rule base.<br />

Each membership function of input or output interval sets<br />

is stored in memory as a set of characteristic points that represents<br />

singleton, triangle or trapezoid.<br />

In the normal operation mode the FLC circuit starts to<br />

process fuzzy rules. The fuzzification module computes the<br />

grades of membership of all input signals in those input fuzzy<br />

sets, which are involved in the currently processed rule. It consists<br />

of a number of F blocks. Such a block calculates a grade<br />

of membership of an input signal in a classical fuzzy set. So,<br />

two such blocks are required for each input signal to consider<br />

the upper and lower membership functions. The inference<br />

module contains two MIN blocks that aggregate the grades<br />

of rule premises coming from the fuzzification module. The other<br />

two MIN blocks calculate the grades of the rule consequence.<br />

There are also two F blocks to give the grades of membership<br />

in output fuzzy sets. The consequences of rules are aggregated<br />

by two MAX blocks for the upper and lower membership functions,<br />

respectively. The resulting type-2 fuzzy set is stored in<br />

the cache memory in the discrete form of lookup table.<br />

Basically, there are two possible ways to process the information<br />

in the inference block. In the first one all the rules are<br />

sequentially taken into account in order to compute a grade of<br />

membership for a single element from the universe of discourse<br />

of the resulting output fuzzy set. The procedure is repeated<br />

for the rest of elements, so each rule need to be processed<br />

several times. The other way is to process each rule only once<br />

Fig. 5. Block diagram of a fuzzy logic controller with two parallel processing paths for the upper and lower bounds of interval type-2<br />

fuzzy sets<br />

Rys. 5. Schemat blokowy sterownika rozmytego z dwoma równoległymi ścieżkami przetwarzania dla dolnej i górnej funkcji przynależności<br />

interwałowego zbioru rozmytego 2-go rzędu<br />

46<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Fig. 6. Architecture of a VLSI circuit implementing a fuzzy logic controller based on interval typ-2 fuzzy sets<br />

Rys. 6. Architektura układu VLSI realizującego sterownik rozmyty oparty na interwałowych zbiorach rozmytych 2-go rzędu<br />

and compute the component grades for all the elements<br />

in the universe of discourse. Taking the next<br />

rule causes the resulting membership function of<br />

the output fuzzy set to be updated. This kind of<br />

processing allows to save up to 2*i*n i *2 r switches<br />

in the circuit, where i, n and r are the numbers<br />

of inputs, input fuzzy sets and accuracy bits, respectively.<br />

The algorithms of type reduction and defuzzification,<br />

mentioned in the previous section, were<br />

implemented as a one module since they are essentially<br />

based on the same set of operations and<br />

therefore share the same building blocks: adders,<br />

multiplication and division circuits. The aim of type<br />

reduction is to map the output interval type-2 fuzzy<br />

set, obtained in the inference module, into a classical<br />

interval fuzzy set. The iterative KM algorithm has to be<br />

executed twice per each output to find the left y L<br />

and the right<br />

y R<br />

bounds of the output interval type-1 set. The ΣΠ and Σ blocks<br />

compute iteratively the numerator and denominator of (6),<br />

respectively. After that, the result of division is stored in the<br />

register, selected by the FSM depending on the stage of algorithm.<br />

At the end of each pass of algorithm, one of the output<br />

registers R L<br />

or R R<br />

is loaded with the appropriate value of y L<br />

or<br />

y R<br />

bound, respectively. The crisp output value is finally computed<br />

in the Σ/2 block as their arithmetic mean.<br />

Experimental results<br />

The proposed type-2 FLC circuit was implemented in the<br />

standard cells approach using CMOS 90 nm technology. The<br />

VHDL model is highly parameterized which means, that depending<br />

on purpose, one can specify: the number of input and<br />

output signals, the number of input and output fuzzy sets, the<br />

type of fuzzy sets (singleton, triangle, trapezoid) and finally the<br />

number of accuracy bits.<br />

Tabl. 1. Design parameters of the FLC circuit in CMOS 90 nm technology<br />

Tab. 1. Parametry układu FLC w technologii CMOS 90 nm<br />

No of<br />

inputs<br />

No of<br />

outputs<br />

No of input<br />

fuzzy<br />

sets<br />

Tabl. 2. Comparison of the parameters of the FLC circuits<br />

Tab. 2. Porównanie parametrów układów FLC<br />

Circuit<br />

No of output<br />

fuzzy<br />

sets<br />

No of cells<br />

Accuracy<br />

Core area<br />

[μm 2 ]<br />

Processing speed<br />

[MFLIPS]<br />

pipelined FLC [4] 470 000 3.125<br />

our FLC 20 000 0.280<br />

Power<br />

[mW]<br />

4 2 4 4 4 38586 0.98<br />

2 4 4 4 4 37230 1.04<br />

4 4 4 4 4 47211 1.19<br />

2 2 4 4 4 28631 0.82<br />

2 2 4 8 4 28631 0.83<br />

2 2 4 4 8 319715 10.16<br />

2 2 8 4 8 319888 10.13<br />

Several experiments with various sets of parameters were<br />

performed in order to check their influence on the area and<br />

power consumption of the FLC circuit. The examples were<br />

chosen in part from practical control applications and reflect<br />

their typical complexity. The circuits were synthesized and<br />

then simulated with the timing parameters of standard cells<br />

from the library.<br />

The results of synthesis and simulation let us estimate the<br />

major parameters of the circuits, which are summarized in<br />

Table 1. The biggest influence on circuit size and power has<br />

the number of accuracy bits, as the sizes of all building blocks<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 47


and the power of switching elements depend directly on that<br />

parameter. It is also worth of notice that increasing number<br />

of input or output fuzzy sets does not influence the core area,<br />

nor the power consumption. We also compared our design<br />

with the inference processor of pipelined architecture [4] for<br />

the same design parameters. Since direct comparison of chip<br />

area is pointless due to the fact they were implemented in different<br />

technologies, in Table 2 the numbers of logic cells are<br />

compared instead.<br />

Conclusions<br />

In this paper the design and digital CMOS implementation<br />

of a type-2 fuzzy logic controller was presented. The parameterized<br />

VHDL model allows us to synthesize the circuit of the<br />

required size for a particular application. The proposed architecture<br />

is suitable for serial processing of fuzzy rules combined<br />

with parallel processing of upper and lower membership<br />

functions of type-2 fuzzy sets. It also implements the Karnik-<br />

Mendel iterative type reduction algorithm. Hardware implementation<br />

was based on CMOS 90nm standard cells library.<br />

The inference speed of the circuit with 2 inputs and 2 outputs<br />

can reach up to 280 KFLIPS.<br />

References<br />

[1] Mendel J.: Type 2 fuzzy sets and systems. An overwiew. IEEE<br />

Computational Intelligence Magazine, February 2007, pp. 20–29.<br />

[2] Zadeh L.: The concept of a linguistic variable and its application to<br />

approximate reasoning. Inform. Sci., vol. 8, 1975, pp. 199–249.<br />

[3] Mendel J. and John R.: Type-2 fuzzy sets made simple. IEEE<br />

Trans. Fuzzy Syst., vol. 10, no. 2, 2002, pp. 117–127.<br />

[4] Huang S. and Chen Y.: VLSI Implementation of Type-2 Fuzzy<br />

Logic Inference Processor. IEEE International Symposium on<br />

Circuits and Systems, vol. 4, 2005.<br />

[5] Wu D. and Mendel J.: Enhanced Karnik-Mendel Algorithms for<br />

Interval Type-2 Fuzzy Sets and Systems. North American Fuzzy<br />

Information Processing Society, pp. 184–189, NAFIPS ‘07 Annual<br />

Meeting of the 24-27 June 2007.<br />

[6] Karnik N. and Mendel J.: Introduction to Type-2 Fuzzy Logic Systems.<br />

IEEE FUZZ Conf., Anchorage, AK, May 1998.<br />

[7] Wu D. and Mendel J.: Uncertainty bounds and their use in the design<br />

of interval type-2 fuzzy logic systems. IEEE Trans. on Fuzzy<br />

Systems, vol. 10, 2002, pp. 622–639.<br />

[8] Hani H.: Type-2 FLCs: A New Generation of Fuzzy Logic Controllers.<br />

IEEE Computational Intelligence Magazine, February<br />

2007.<br />

[9] Liang Q. and Mendel J.: Interval Type-2 Fuzzy Logic Systems:<br />

Theory and Design. IEEE Transactions on Fuzzy Systems, vol.<br />

8, no. 5, 2000, pp. 535–550.<br />

[10] Hani H.: A Hierarchical Type-2 Fuzzy Logic Control Architecture<br />

for Autonomous Mobile Robots. IEEE Transactions on Fuzzy<br />

Systems, Vol. 12, no. 4, august 2004.<br />

[11] Melgarejo M., Garcia R. and Pefia-Reyes C.: Pro-Two: A Hardware<br />

Based Platform Fuzzy Inference For Real Time Type-2.<br />

Proceedings 2004 IEEE International Conference on Fuzzy Systems,<br />

Part 2, 2004, pp. 977–982.<br />

[12] Lynch C. and Hagras H.: Parallel Type-2 Fuzzy Logic Co-Processors<br />

for Engine Management. IEEE Fuzzy Systems Conference,<br />

July 2007, pp. 1–6.<br />

[13] Mendel J., John R., Liu F.: Interval Type-2 Fuzzy Logic Systems<br />

Made Simple. IEEE Transactions on Fuzzy Systems, vol.14,<br />

no. 6, pp. 808–821, Dec. 2006<br />

Universal tool for estimation of programmable<br />

logic controllers processing power<br />

(Uniwersalne narzędzie do szacowania mocy obliczeniowej sterowników<br />

przemysłowych)<br />

mgr inż. ZBIGNIEW KULESZA, mgr inż. STEFAN MARASEK<br />

Politechnika Łódzka, Katedra Mikroelektroniki i Technik Informatycznych<br />

Processing power is a key feature of the contemporary computer<br />

systems of all kinds. Quality of computer exploitation is<br />

directly conditioned by the ability to perform operations in a rapid<br />

way. Processing power is therefore the principal criterion<br />

of computer system design and choice. The aim of technology<br />

progress is to achieve the highest performance for the lowest<br />

possible cost [2].<br />

In recent years, specialized industrial computers known as<br />

programmable logic controllers, literally revolutionized control<br />

engineering by providing a flexibility of PC, simultaneously<br />

conforming industrial standards and expectations. PLCs, being<br />

the key element of contemporary automation systems, are<br />

subject to rapid development [3].<br />

Over the last two decades a vast body of knowledge, dealing<br />

with the performance evaluation of computer systems,<br />

has accumulated. Sophisticated measurement tools, called<br />

benchmarks, were elaborated to estimate computer systems<br />

processing power [4]. Simultaneously, no progress in this<br />

area is observed for industrial controllers. Producers as well<br />

as clients usually focus their attention on price, expansion modules<br />

available, additional function blocks, memory size and<br />

48<br />

extra features. What is more, the choice of industrial controller<br />

is often dictated by contractor or control engineers, who tend<br />

to keep to one family of PLC, since other choice require additional<br />

training [3].<br />

Since there is no representative and explicit technique<br />

of processing power estimation for programmable logic controllers<br />

available and no considerations relating benchmarks<br />

for PLC are present in scientific literature, there is a need to<br />

elaborate benchmark to enable comparison of industrial controllers.<br />

General design approach<br />

Design process was substantially influenced by benchmarks<br />

available for personal computers. Benchmark for PLC, which<br />

was created, is conceptually similar to Whetstone test [5].<br />

Certainly, the philosophy of programming PC and PLC differs,<br />

that is why elaborated algorithm renders the specific features<br />

of PLC hardware and software.<br />

Yet, as in case of Whetstone, six modules are iterated in loops,<br />

each representing different group of program items. The<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


final test result is related directly to the number of program<br />

scans that are performed in 100 second period and no float<br />

point variables are used, which is Dhrystone-like behavior [6].<br />

The weight of each module, i.e. the number of its iterations<br />

in one program scan is determined by the statistical analysis<br />

of different PLC applications, like in Dhrystone. Remark also<br />

that Whetstone modules are task-oriented, whereas PLC test<br />

modules are item-oriented.<br />

It is impossible to predict in which language the algorithm<br />

is to be implemented on a specific platform. Certain<br />

controllers do not allow some languages, most of the elder<br />

ones may be programmed only in ladder logic. Although LD<br />

popularity ceases with controllers technology development,<br />

still it is the only language present on the vast majority of<br />

architectures. That is why LD was chosen as default benchmark<br />

source code language, although created algorithm is<br />

easily portable to other languages defined by IEC 61131-3<br />

standard.<br />

The procedure of creating benchmark for industrial controller<br />

may be briefly summarized in the following steps:<br />

1) Creation of six item groups, to which the majority of program<br />

items may be assigned. Created groups are coherent<br />

with item groups present in the most relevant PLC programming<br />

tools available on the market (Step7, Syswin,<br />

Concept, RSLogix). Item groups are as follows: contacts<br />

and coils, edge detectors and latches, counters, timers,<br />

mathematical computations, variable comparisons.<br />

2) Design of six modules rendering the behavior of item<br />

groups. Each module is representative for all elements of<br />

the group, e.g. timer is represented not only by single timer<br />

block, yet also by typical wiring and additional control<br />

items.<br />

3) Analysis of PLC applications from different domains. Four<br />

sophisticated applications developed for industrial systems<br />

were analyzed.<br />

4) Computation of statistical item distribution. In such a manner<br />

percentage of each module was calculated in the representative<br />

way. Then the weights of modules were determined<br />

to render the statistical program behavior.<br />

5) Elaboration of application interface and result counter.<br />

User interacts with a program via simplified interface enabling<br />

test actuation, reset and indicating test completion.<br />

Test results have to be computed basing on the number of<br />

elapsed scans, read directly from PLC registers.<br />

6) Implementation of algorithm on different platforms. There<br />

is no universal language to program PLC. Each producer<br />

offers his own tools, even controllers of one producer<br />

differ in instruction set. That is why benchmark algorithm<br />

needs to posses cross-platform properties to implement it<br />

easily on arbitrary architecture. Seven PLC devices were<br />

tested, which comprised five different algorithm implementations.<br />

7) Tests and comparisons of PLCs. Seven PLC controllers<br />

were tested and compared by means of designed benchmark.<br />

Not only overall performance was compared, but<br />

also behavior on each of test block.<br />

8) Verification of test importance. Obtained results were confronted<br />

with bit instruction processing time. Relation between<br />

both measures was discussed. In such a way the<br />

sense of testing PLCs with benchmark was assessed.<br />

Algorithm description<br />

The general algorithm of synthetic PLC benchmark is presented<br />

on Fig. 1. General algorithm for elaborated synthetic PLC<br />

benchmark.<br />

Fig. 1. General algorithm for elaborated synthetic PLC benchmark.<br />

LOOP1 – contacts and coils, LOOP2 – edge detectors and<br />

latches, LOOP3 – counters, LOOP4 – timers, LOOP5 – mathematical<br />

computations, LOOP6 – variable comparisons<br />

Rys. 1. Ogólny algorytm benchmarka syntetycznego dla sterowników<br />

przemysłowych. LOOP1 – styki i cewki, LOOP2 –detektory<br />

zbocz i przerzutniki, LOOP3 – liczniki, LOOP4 – timery, LOOP5<br />

– obliczenia matematyczne, LOOP6 – porównania zmiennych<br />

User controls the application by START input. For majority<br />

of controllers it is the first available digital input. When the<br />

START button is pressed, clock measuring test duration time<br />

is initiated. As long as timeout is not reached, variable result1,<br />

storing test result, is incremented each program scan. To<br />

avoid variable overflow, when result1 reaches value of 1000,<br />

it is reset and variable result2, storing thousands of the result,<br />

is incremented.<br />

Each scan, program goes through 1000 iterations of test<br />

blocks, which are called loops. According to its weight, each<br />

test loop may be iterated arbitrary times, yet total number of<br />

iterations must equal 1000. E.g. if only performance of pure<br />

LOOP2 is to be tested, it has a weight of 1000, whereas other<br />

loops are not iterated at all. If a general performance of<br />

a given PLC is to be tested, we set loop weights according to<br />

representative statistical distribution.<br />

When test duration time elapsed, the number of elapsed<br />

program scans is stored in result1 and result2 variables. Consequently,<br />

STOP digital output is energized to indicate test<br />

completion. To read the benchmark result, it is essential to<br />

read the variables values from the memory via animation software.<br />

START button may be switched off either during test or<br />

after test completion. In both cases timer and result variables<br />

will be reset. The final test result is equal to number of LD<br />

items processed per one second – LIPS.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 49


PLC benchmark program scan includes not only test loops,<br />

whose processing power requirements are directly responsible<br />

for final result, but also code responsible for user interface,<br />

time measurement and program scan counting. Source code<br />

was designed in a way to isolate test loops from other ladder<br />

rungs during a test. To achieve this, conditional jumps were<br />

used. As a consequence, the negative influence of auxiliary<br />

code is minimized. What is more, regarding the fact that 1000<br />

of test iterations are performed each scan, additional ladder<br />

rungs contribution to the final result is negligible.<br />

Test results<br />

Elaborated PLC synthetic benchmark was implemented on<br />

several architectures to perform a set of testes. To increase<br />

the measurement accuracy, each test was performed five times<br />

and then averaged. For each controller, KLIPS benchmark<br />

test was performed with seven different configurations,<br />

namely statistical loop weights and separately for each test<br />

loop. Statistical density of each test lo op was estimated by<br />

analysis of four professional PLC applications source codes<br />

developed for Polish industry. Fig. 2 visualizes representative<br />

benchmark outcomes in KLIPS units for all tested controllers.<br />

Assuming test correctness, the result in KLIPS units is directly<br />

proportional to processing power. Created benchmark<br />

enabled objective PLC comparison with a single parameter.<br />

Obtained results are coherent with empirical expectations. Apparently,<br />

processing power of Siemens controllers increase<br />

for higher order families, i.e. S7-400 is approximately 10 times<br />

powerful than S7-300 and the latter 10 times powerful than<br />

S7-200. Accordingly, Quantum 534, fitted ted with better processor<br />

than Quantum 434 exhibits better performance.<br />

Fig. 2. KLIPS PLC benchmark final results<br />

Rys. 2. Wyniki benchmarku dla PLC wyrażone w KLIPS<br />

Fig. 4. Comparison of processing power in KLIPS and MIPS<br />

Rys. 4. Porównanie wartości mocy obliczeniowej w KLIPS<br />

i MIPS<br />

Although explicit research of performance to price ratio<br />

was not performed, it is obvious that 534 unit is more expensive<br />

than 434 one, as well as S7-400 is priced higher than S7-<br />

300 and, respectively, S7-200. Paying more for a controller,<br />

better performance is expected, and PLC benchmark validates<br />

these expectations.<br />

Nevertheless, word” better” ought to be proved by certain<br />

measurable feature. The only processing power determinant<br />

quoted officially by producers is bit operations processing<br />

time. The name of this unit varies among producers, yet technically<br />

it is an equivalent of MIPS unit. Such a representation<br />

is conceptually coherent both with KLIPS units and simplifies<br />

comparison. Fig. 3 presents processing power of PLCs in<br />

MIPS, according to producers’ information [7–12].<br />

A brief analysis of presented official results proves that<br />

they do not render real performance. E.g., benchmark test<br />

shown a substantial difference of processing power between<br />

two Quantum controllers, whereas they are said to have the<br />

same MIPS value. Moreover, Siemens S7-300, which performance<br />

is an order of magnitude lower, also has the same official<br />

processing power indicator.<br />

Apparently, in the market competition reality, it is convenient<br />

to legitimate PLC with high MIPS value, particularly that<br />

this feature is almost not verifiable. Next investigation is aimed<br />

at finding correlation between designed benchmark and<br />

producer-claimed performance. Fig. 4 shows the quotient of<br />

measured and published processing power, as KLIPS over<br />

MIPS ratio, for all tested controllers.<br />

There is no doubt that there is no correlation between both<br />

performance indexes. Ratio calculated for AVR-based controller<br />

may be treated as reference value. In this case the processor<br />

frequency was known (4 MHz quartz oscillator was used<br />

as external system clock) and true MIPS value was calculated.<br />

Therefore controllers that have KLIPS/MIPS ratio higher or<br />

approximately equal to 100, have reliable official performance<br />

indexes. On the other hand, performance index MIPS is<br />

overestimated for weaker PLCs, namely CQM1, S7-300 and<br />

S7-200. This highlights need and importance of elaborated<br />

PLC benchmark.<br />

Algorithm drawbacks<br />

Fig. 3. PLC processing power in MIPS according to producers<br />

Rys. 3. Moc obliczeniowa sterowników wyrażona w MIPS wg producentów<br />

50<br />

The principal goal of PLC benchmark design was to achieve<br />

the desired benchmark features [1]. Yet, certain simplifications<br />

and disadvantages should be taken into consideration. First<br />

of all, interruptions are ignored. Read/write of I/O is treated<br />

as read/write of I/O registers, updated each program cycle.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


I/O other than digital, i.e. analog, fast counters, etc., are not<br />

tested. Benchmark is designed for integer numbers. No other<br />

data types are present. Issues concerning communications<br />

are ignored. No sub-programs or sub-routines are implemented.<br />

Auxiliary commands, used to iterate test blocks, measure<br />

time and results as well as interface with the user, even if minimized,<br />

put undesired burden on CPU. Representative loop<br />

weights have inaccuracy caused by insufficient size of analyzed<br />

data set. Fraction of non-assigned elements is relevant.<br />

Finally, created test blocks do not cover the whole spectrum<br />

of LD items.<br />

Conclusion<br />

It is possible to create universal tool for industrial controllers<br />

processing power estimation. PLC benchmark was successfully<br />

created and utilized for single-unit comparison of controllers<br />

delivered by different producers. For this purpose a brand<br />

new unit was proposed – kilo ladder logic instructions per second,<br />

abbreviated KLIPS. Created benchmark conforms to all<br />

desired benchmark features [1]. Benchmark is explicit, since<br />

scope of tests is precisely defined as measuring processing<br />

time of six instruction groups: contacts and coils, triggers and<br />

edge detectors, counters, timers, mathematical operations and<br />

variable comparisons. Benchmark result is unambiguous, clearly<br />

represented by KLIPS unit, which is a direct determinant<br />

of processing power. Scalability and portability of benchmark<br />

is apparent, it was implemented on wide range of industrial<br />

controllers, from Siemens S7-400 to microcontroller-based<br />

one, in different programming languages. KLIPS benchmark<br />

results are representative, since simulated workload renders<br />

accurately real-life performance due to statistical investigations<br />

made. Testing algorithm itself is published in universal<br />

ladder logic representation with full description, anyone can<br />

judge its principles and create his own implementation. This<br />

proves benchmark high accessibility. Reproducibility is ensured<br />

by precisely defined conditions of testing, i.e. input configuration<br />

and no interfering communications.<br />

References<br />

[1] Fernandez J. and Garcia J.M.: Representative benchmarks for<br />

commercial workloads. X Jornadas de Paralelismo, pp. 308–318,<br />

September 1999.<br />

[2] Jain R.: The Art of Computer Systems Performance Analysis, 1991.<br />

[3] Parr E. A.: Programmable Controllers. An Engineers Guide, 2003.<br />

[4] Fortier P. J. and Michel H. E.: Computer Systems Performance<br />

Evaluation and Prediction, 2003.<br />

[5] Tiberghien J.: Evaluation of Computer Performance, 2005.<br />

[6] Weicker R. P.: Dhrystone: A synthetic programming benchmark.<br />

Communications of the ACM, 27(10):1013-1031, October 1984.<br />

[7] Omron Electronics LCC, Programmable Controller CQM1H, Cat.<br />

No. R301-E3-01, 2004.<br />

[8] Schneider Electric, Modicon Quantum Automation Series 140<br />

CPU 434 12A Instruction Sheet, Version 3.0, 2002.<br />

[9] Schneider Electric, Modicon Quantum Automation Series 140<br />

CPU 534 14A Instruction Sheet, Version 3.0, 2002.<br />

[10] Siemens, The SIMATIC S7-200 Micro System, Rev. 3/99, 1999.<br />

[11] Siemens, S7-300 CPU 31xC and CPU 31x, Technical Specifications,<br />

Rev. A5E00105475-07, 2006.<br />

[12] Siemens, S7-400 Automation System, CPU Specifications, Rev.<br />

A5E00850746-02, 2006.<br />

XML markup language based design tool integration<br />

method in distributed design environments<br />

(Metoda integracji narzędzi w rozproszonych środowiskach projektowania<br />

wykorzystująca język znaczników XML)<br />

dr inż. MAREK SZLĘZAK, MELEX A&D Tyszkiewicz Sp.J., Mielec<br />

dr inż. ADAM PAWLAK, Silesian Univ. of Technology, Institute of Electronics, Gliwice<br />

prof. dr hab. inż. KONRAD WOJCIECHOWSKI<br />

Silesian Univ. of Technology, Institute of Computer Science, Gliwice<br />

Distributed collaborative engineering [1] is emerging as a new<br />

engineering paradigm that integrates dispersed engineers<br />

and their tools into a virtual team that develops a new complex<br />

heterogeneous product. Distributed design processes that are<br />

realized in a collaborative network [2] have numerous advantages<br />

and possibilities in situations where co-located collaboration<br />

isn’t feasible, like:<br />

– Grouping of diverse and dispersed experts into design<br />

teams. Experts need however to be supported in their collaboration<br />

by an integrated network of required tools,<br />

– Searching for required tools in the network and automatic<br />

tracking of changes in their specifications independently<br />

from their network location,<br />

– More effective administration of specialized tools,<br />

– Relative easiness of deploying new design technologies<br />

enabled by new tools due to proliferation of the Internet.<br />

Many R&D efforts have been undertaken in recent years<br />

that target various aspects of collaborative engineering<br />

[1, 4, 12]. The addressed aspects of this new paradigm of engineering<br />

work were among others:<br />

– Collaborative working environments with both synchronous<br />

and asynchronous communication among dispersed<br />

team members,<br />

– Interoperability of remote tools that often belong to different<br />

organizations [14],<br />

– Assurance of security in distributed collaboration being<br />

a critical requirement in industrial applications.<br />

The paper is organized as follows, after a short introduction<br />

to the problem of remote design tools integration; the main<br />

requirements on an integration environment are formulated.<br />

Provide explanations on XML-based description of design<br />

tools and their use in the Internet context. In the following,<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 51


an ontology-based tool integration method and Ontotool are<br />

presented. Finally, experiments used to validate the introduced<br />

methods are pointed to, as well as the advantages of the<br />

approach are summarized.<br />

Distributed design processes based on<br />

remote tools<br />

Required design competencies, tools, and various additional<br />

resources, like libraries of virtual electronic components, may<br />

not be available locally with rising complexity and heterogeneity<br />

of new designed systems. Additionally, the time-to-market<br />

pressure calls for more concurrent engineering approaches,<br />

where more distributed engineers are involved in the design<br />

process. Tool integration and management, as well as secure<br />

transport of design data are key challenging issues for distributed<br />

collaborative engineering environments that are used<br />

by dispersed design teams. Tool integration [14] was addressed<br />

in several approaches to engineering environments for<br />

Electronic Design Automation (EDA), like: TES (Tool Encapsulation<br />

Specification) standard language [19] and the CAD<br />

framework encapsulations [19], and in Web-based Electronic<br />

Design – WELD [4]. Various workflow management tools [18]<br />

were proposed for design flow management, like dedicated<br />

ones: ASTAIR (C-Lab, Univ. Paderborn), and Reuben (OmniFlow)<br />

[3].<br />

Tools that are dispersed in the Internet space need to be<br />

discovered. Fig. 1 presents a dynamic management of such<br />

Internet-wide dispersed tools that is based on a lookup service.<br />

This tool searching service is somehow central to the<br />

presented approaches. Only tools that are registered in the<br />

lookup service may be discovered and later invoked with appropriate<br />

design data by a design workflow executed on a client<br />

– designer’s workstation.<br />

Application of the tool look-up services and workflow management<br />

in engineering design processes needs appropriate<br />

enabling technologies. Below, a short list of major requirements<br />

concerning realization of tools integration and workflow<br />

management in distributed design environments is given.<br />

Distributed tools management which includes: remote configuration<br />

and invocation of tools, control of access rights, tool<br />

registration and search in the network.<br />

User (designer) management: access rights, user group<br />

privileges, user authentication and authorization.<br />

Workflow management and execution: integration of tools<br />

into workflows, tools enactment, cross-organization and<br />

cross-border workflows.<br />

Design data management. Design data should be moved<br />

securely among tools with version control. Secure transport of<br />

Fig. 1. Dynamic Tool Management in a Distributed Environment<br />

[17]<br />

Rys. 1. Dynamiczne zarządzanie narzędziami w rozproszonym<br />

środowisku<br />

52<br />

design data won’t be addressed in this paper. Our approach<br />

has been presented in [10].<br />

Communication over the Internet that enables: firewalls<br />

crossing, enabling proxies, assurance of a robust Internet<br />

connection.<br />

From the above list of issues related to integration of distributed<br />

tools it should be visible that tools ought to be precisely<br />

described. Portable XML-based tool descriptions proved during<br />

last years to be a right format for formalization of tool’s<br />

characteristics.<br />

Tool descriptions in XML<br />

As a result of research and experiments in remote tools invocation<br />

performed within the EU project E-COLLEG (www.ecolleg.org)<br />

an XML schema (ecollegML) enabling general tool<br />

and design task descriptions has been defined. Remote tool<br />

invocation methods were further investigated and developed<br />

in the MAPPER project (mapper.eu.org).<br />

XML schema for tool and design task description<br />

Tool or design task descriptions based on the ecollegML<br />

XML schema (www.ecolleg.org/trms) and general XML rules<br />

have a modular structure of tool and task descriptions expressed<br />

in XML [11]. This modular structure simplifies processing<br />

of selected parts of the whole XML document. An<br />

example of such a module is a description of tool properties<br />

which is always processed during a search for a particular<br />

tool. Processing of a selected module reduces searching<br />

time and data fulfilling the formulated search criteria are accessed<br />

with a singly query only. A tool description comprises<br />

all allowed parameter definitions and default locations<br />

of source data. A task description constitutes a refinement of<br />

a tool description, i.e. a specification of required parameters<br />

and source data that are necessary for correct execution of<br />

the task.<br />

Internet-based searching for tools<br />

In a distributed design context, Internet-based searching for<br />

tools that are appropriate for a given design task is an essential<br />

element of a design process that influences quality of the<br />

resulting design. In design environments, like WELD [4], TurboTester<br />

[9], or JavaCAD [6] tools are selected directly from<br />

a list of available tools. Such a solution requires a priori knowledge<br />

on the required tool.<br />

The presented method offers a possibility for tool searching<br />

in the tool descriptions data base according to the<br />

user defined criteria, what predominantly improves matching<br />

of a tool to a given design task. The method has two steps.<br />

Firstly, data from the node “description” are searched through<br />

only, which reduces the overall searching time. Secondly,<br />

in case a user has already decided which description fulfils<br />

imposed criteria, a retrieval of the complete document takes<br />

place based on the description identifier contained in the “description”<br />

node. This identifier plays at the same time a role of<br />

an identifier of the document that is stored in the data base.<br />

This solution enables a more efficient search process, and<br />

minimizes a volume of transmitted data, what distinguishes<br />

this solution.<br />

A prototype tool using the XPath query language for searching<br />

data on tools and tasks has been developed. The<br />

use of the Xpath query language despite the speed in receiving<br />

results requires an additional processing of the search<br />

results, e.g. for their visualization. The improvement in results<br />

formatting of the Xpath query for tools has been achie-<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


ved by the use of the Xquery (XML Query Language), and<br />

more precisely its FLOWR expression (for-let-order-wherereturn)<br />

that allows direct transformation of results into the<br />

required format. Below, a method using Xquery queries for<br />

improvements in accessing the tool description based on<br />

the criteria, like: name, keywords, info, as well as a method<br />

for tracking changes in a tool’s description have been shortly<br />

introduced.<br />

Internet-based tool search<br />

In order to facilitate access to stored tool descriptions, a WWW<br />

page has been established that contains a search engine<br />

(Fig. 2). This enables a designer to access and learn on the<br />

characteristics of the available tools. Queries in the XQuery<br />

language in combination with the XSL sheet (Extensible Style<br />

sheet Language) have been used to realize the search and<br />

visualization process. The solutions known from the literature<br />

Fig. 2. Tool search and access through a Web browser<br />

Rys. 2. Wyszukiwanie i dostęp do narzędzi poprzez przeglądarkę WWW<br />

Fig. 3. Tool registration, search and invocation environment (TRMS)<br />

Rys. 3. Środowisko do rejestracji, wyszukiwania oraz zdalnego uruchamiania narzędzi<br />

(TRMS)<br />

haven’t the tool searching function, but a choice from a predefined<br />

list only. In the CHINOOK service [5] information on new<br />

tools is disseminated through a peer-to-peer network, and the<br />

network applet enables a selection of a tool from a predefined<br />

list only.<br />

Verification of the approach in the TRMS<br />

environment<br />

The tool description and search method has been verified in<br />

the TRMS environment (www.ecolleg.org/trms) [8, 10], as well<br />

as with the use of the www server. Fig. 3 below presents the<br />

architecture of the environment that includes three main components:<br />

Global Tool Lookup Server (GTLS), Tool Server (TS)<br />

and the Client.<br />

GTLS manages and provides surveillance over tools and<br />

users. GTLS main tasks are administration of users, as well as<br />

registration and monitoring of the tools’ use in the TRMS environment.<br />

The WWW server through a browser<br />

visualizes results of queries in the tools<br />

data base in case of a direct tool search. Also<br />

Tool Tracking brings results through a browser.<br />

TRMS applicability to integration of distributed<br />

tools in various design processes has<br />

been addressed in [13].<br />

Tool search in resources description<br />

documents (RDF)<br />

Queries on the syntactic level depend on the<br />

document structure what in case of documents<br />

based on graphs, like RDF (Resource<br />

Description Framework) documents limits<br />

their functionality. This considers primarily<br />

a situation when one tries to acquire knowledge<br />

on relations between described resources<br />

without a complete knowledge on the<br />

used syntax.<br />

The model of relations between objects<br />

and resources provides a simple semantics<br />

which may be noted with the use of the XML<br />

syntax. It should be noted that the RDF document<br />

syntax is based on a graph concept<br />

what significantly simplifies presentation<br />

of data and relationships in a graphical form.<br />

This is in contrary to XML documents where<br />

data are represented as a hierarchical structure.<br />

A part of the tool RDF description has<br />

been presented in Fig. 4 above. In the presented<br />

tool description each triple represents<br />

a statement, e.g. type(tool, software) what<br />

equals a sentence: Tool type software.<br />

A tool description has been created in order<br />

to verify data search in the RDF model<br />

instances. The used RDQL (Resource Description<br />

Query Language) language enabled<br />

search through resources in the developed<br />

model of notions and relations. The defined<br />

RDQL queries enabled to determine a template<br />

that needs to match an existing graph.<br />

A result of this search constitutes a set of pairs:<br />

name-value. Queries have been realized<br />

in the dedicated JENA programming environment<br />

(jena.sourceforge.net). A deficiency of<br />

the structural level query language is a lack<br />

of a possibility for receiving answers on queries<br />

that refer to notions that are not represented<br />

in the document.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 53


Fig. 4. Tool description diagram in RDF<br />

Rys. 4. Diagram opisu narzędzia w formacie RDF<br />

Such an approach enables a correct interpretation<br />

of a meaning of relations that are between<br />

tools and data. Ontology instances represent<br />

a knowledge base of notions related to tools,<br />

data they process and functionality supporting<br />

integration of tools. The OntoTool ontology is<br />

accessible at: www.ecolleg.org/trms/ontology.<br />

html. The tool integration environments that<br />

have been investigated in the literature didn’t<br />

develop a similar relationship between design<br />

tools and tasks.<br />

OntoTool ontology<br />

The experiments conducted have revealed limitations<br />

in reasoning on the tools use in respect to design tasks. The<br />

process of matching between design tasks and tools clearly<br />

needed to be improved, and that was brought by the use<br />

of OWL (Web Ontology Language). OWL has been deployed<br />

in order to define a relationship between a design task realization<br />

and a particular distributed tool (OntoTool). For the<br />

developed OntoTool ontology in a first step classes and relations<br />

between them have been defined, as well as their types<br />

and constraints. In the following step a class hierarchy has<br />

been defined. Finally, a knowledge base has been created<br />

that comprises OntoTool instances. In the established ontology,<br />

notions and relations like: tool, parameter, task, project,<br />

manage have been defined. Fig. 5 illustrates the OntoTool<br />

ontology.<br />

The ontology can reason on tools as the ontology taxonomy<br />

has been enriched with reasoning rules and a possibility to<br />

execute simple logical operations on statements that may be<br />

assigned “true” or “false” attributes.<br />

Experiments have been conducted with the use of dedicated<br />

services for knowledge base search in order to verify the<br />

ontology. The query OWQL (Ontology Web Query Language)<br />

language for semantic documents has been applied. The<br />

short example below presents a result of a query to OntoTool.<br />

For the query „what is a compiler” we receive the answer that<br />

a resource compiler is a “tool”.<br />

In this manner, data which have a well defined meaning<br />

may constitute a reference point for data that are required<br />

for realization of the task with the use of distributed tools.<br />

Conclusions and further research<br />

The presented method for description, storage, and search of<br />

tools has been deployed and practically checked in the TRMS<br />

environment (www.ecolleg.org/trms/). Design experiments<br />

that have been conducted so far [13, 16] confirm the approach<br />

taken for transformations of tools descriptions and the query<br />

language. Correctness of the tool tracking method has been<br />

verified by experiments taken with the Tool Tracking tool [15]<br />

and additionally by the portal dobreprogramy.pl.<br />

The XML-based tool integration method reduces involvement<br />

of a designer in tedious tasks of tool integration (writing scripts,<br />

wrapping of tools) while increasing robustness of the integration<br />

process. Furthermore, a workflow-based design process appears<br />

as a designer-friendly one; as it enables an engineer to concentrate<br />

on a design process itself leaving tool integration issues to<br />

the services of TRMS [8]. In more general terms, the use of the<br />

XML markup language enables reuse of engineering knowledge<br />

related to tool integration (tool tracking, design tracking).<br />

Fig. 5. Graphical representation of the OntoTool ontology part and the tool description class<br />

Rys. 5. Graficzna reprezentacja części ontologii OntoTool oraz klasa opisu narzędzia<br />

54<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


The innovative use of ontology in description of relations<br />

between tools and design tasks brings more semantics in<br />

a format which can be processed. Achieved results have confirmed<br />

the utility of the developed ontology and the OWQL<br />

query language. The presented tool integration method constitutes<br />

a contribution towards automation of the distributed<br />

tools integration process. Further R&D efforts are however<br />

required in order to enhance elements of the collaborative engineering<br />

technology that applies this tool integration process.<br />

The following research directions are particularly justified:<br />

– Enhancement of the workflow editor and engine in order to<br />

simplify management of distributed design processes;<br />

– Improvement of support for both asynchronous and synchronous<br />

communication among distributed engineers; and<br />

– Elaboration and standardization of network-based business<br />

transactions on design services that allow remote engineers<br />

to take full advantage of tremendous engineering resources<br />

that are available through the network.<br />

References<br />

[1] Bergman R., Baker, J.D.: Enabling collaborative engineering and<br />

science at JPL. Advances in Engineering Software 31 (2000), pp.<br />

661–668.<br />

[2] Camarinha-Matos, L., Afsarmanesh H.: The emerging discipline<br />

of collaborative networks. Luis M. Camarinha-Matos (Ed.) “Virtual<br />

Enterprises and Collaborative Networks”, Kluwer Academic<br />

Publishers, Boston, 2004, pp. 3–16.<br />

[3] Brglez F., Lavana H.: A Universal Client for Distributed Networked<br />

Design and Computing. Proc. 38th DAC, Las Vegas, 2001.<br />

[4] Chan F., Spiller M., Newton R.: WELD – An Environment for Webbased<br />

Electronic Design. Proc. DAC, San Francisco, 1998.<br />

[5] Chinook User Documentation, Montgomery, S. 2003. http://www.<br />

bcgsc.ca/project/bomge/chinook<br />

[6] Dalpasso M., Bogliolo A., Benini L.: Specification and validation<br />

of distributed IP-based designs with JavaCAD. Proc. DATE<br />

Conf., pp. 684-688, Munich, Germany, 1999.<br />

[7] Director S., Sutton, P.: Framework Encapsulations: A New Approach<br />

to CAD Tool Interoperability. Proc. of 35th Design and<br />

Automation Conference, San Francisco, 1998.<br />

[8] Pawlak A., Fraś P., Penkala P.: Web services-based collaborative<br />

system for distributed engineering. PRO-VE’08 9th IFIP<br />

Working Conference on Virtual Enterprises, Poznan, Poland,<br />

8–10 Sept. 2008, in Pervasive Collaborative Networks, Edited by<br />

Luis M. Camarinha-Matos and Willy Picard, Springer.<br />

[9] Paomets, P.: An Open and Dynamic User Interface to the CADsystem<br />

Turbo Tester. MS Thesis, Tallinn, 1998.<br />

[10] Szlęzak M., i in.: P. Security Issues in Tool Registration and Management<br />

System (TRMS). Proc E-Colleg Workshop on Challenges<br />

in Collaborative Engineering (CCE03), 15–16.04.2003,<br />

Poznań, Poland.<br />

[11] Szlęzak, M.: Markup Language in Realization of Design Tasks.<br />

Proc. 3rd Workshop on Challenges in Collaborative Engineering,<br />

Sopron, Hungary, 2005.<br />

[12] Schattkowsky T., Müller W.: Distributed Engineering Environment<br />

for the Design of Electronic Systems. Proc. E-Colleg Workshop<br />

on Challenges in Collaborative Engineering, 15–16.04.2003,<br />

Poznań.<br />

[13] Siekierska K. i in: Distributed collaborative design of IP components<br />

in the TRMS environment. Microelectronics Reliability, Elsevier<br />

Journal, vol. 46 (2006), 5–6, pp. 1019–1024.<br />

[14] TIS 2003, Workshop on Tool Integration in System Development,<br />

ESEC/FSE 2003, in conjunction with 9th European Software Engineering<br />

Conf. Helsinki, Finland, 1–5.09.2003.<br />

[15] Tool Tracking home page: www.ecolleg.org/trms/tooltracking.<br />

html.<br />

[16] Sakowski W. i in.: Mixed-signal USB IP Core Design Using Distributed<br />

Collaborative Approach. MIXDES, 2009.<br />

[17] Schattkowsky T., Mueller W., Pawlak A.: Workflow Management<br />

Middleware for Secure Distance-Spanning Collaborative<br />

Engineering. In L. Fischer (ed.) The Workflow Handbook 2004,<br />

WfMC, Lighthouse Point, FL, USA, 2004.<br />

[18] Kokoszka A. i in.: Are Workflow Management Systems useful for<br />

Collaborative Engineering?, 2002 ECPPM, eWork and Business<br />

in AEC, Portorož, Slovenia,Sept. 9–11, A.Balkema Publishers.<br />

[19] Tool Encapsulation Specification of the CAD Framework Initiative,<br />

http://www.si2.org/<br />

A dedicated high-level language for implementing<br />

nonrecursive filter banks and transforms<br />

(Dedykowany język wysokiego poziomu do implementowania<br />

nierekursywnych banków filtrów i transformacji)<br />

dr inż. MAREK PARFIENIUK, Politechnika Białostocka, Katedra Systemów Czasu Rzeczywistego<br />

Even though nonrecursive filter banks and transforms are<br />

widely used in digital signal processing (DSP) [1–3], not<br />

many tools exist that are aimed at implementing such systems.<br />

Obviously, general-purpose programming languages,<br />

code generators, development environments can be<br />

employed, but they do not take into account that these algorithms<br />

have peculiar properties, and thus facilitations<br />

like convenient syntax or predefined code library could be<br />

provided.<br />

The only two exceptions known to the author are Matlab/<br />

Simulink and SPL/SPIRAL [4, 5], which however have some<br />

shortcomings. The main problems are that they do not allow<br />

for describing transforms efficiently and compactly, a user can<br />

feel overwhelmed by numerous options, and Java code generation<br />

is neglected. Moreover, being mature, general, and<br />

proprietary, they cannot be significantly adapted to meet the<br />

needs of filter bank developers.<br />

In this paper, the tools are presented that we have developed<br />

in order to facilitate implementing transforms and filter<br />

banks by making it possible:<br />

• to describe a system clearly, compactly, and quickly in<br />

a way independent of implementation,<br />

• to convert such descriptions into quite efficient implementation<br />

code, i.e. code without excessive both data movements<br />

and subroutine calls, and with loops unrolled and<br />

partial results computed in advance,<br />

• to have generated code accessible, customizable and<br />

easy to incorporate into different applications: it should not<br />

depend on nonstandard libraries and should use descriptive<br />

identifiers,<br />

• to easily change data types (fixed- vs. floating point numbers)<br />

and access methods to data (pointers vs. one- and twodimensional<br />

arrays) of input, output, and coefficient vectors,<br />

• to generate Java code, not only C/C++ code.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 55


These goals have been achieved by introducing an innovative,<br />

domain-specific Transform Description Language (TDL),<br />

whose syntax tightly links code to the signal flow graph of<br />

a system. In particular, it requires user to specify both an operation<br />

and the data paths it acts on, which allows a transform<br />

or filter bank to be compactly and clearly described in terms of<br />

the elementary transformations. Moreover, the language has<br />

syntax similar to Matlab, so that persons familiar with the latter<br />

can use it without much effort.<br />

The TDL language is accompanied by a compiler written<br />

in Java. It converts descriptions into optimized Java or C/C++<br />

implementation code in accordance with specified requirements.<br />

A developer is only responsible for describing a system<br />

and requirements correctly.<br />

Even though our tools cannot compete with the aforementioned<br />

platforms, in some situations they are very useful.<br />

Being simple, and thus easy to use, they allow for quickly<br />

obtaining quite good implementations of a transform in<br />

popular programming languages. Thus, applications that use<br />

filter banks can be rapidly prototyped without compromising<br />

efficiency much. Especially, we are now able to easily prepare<br />

objective functions for optimizing transforms (their computational<br />

schemes and coefficients), which is the main area of<br />

our research [6, 7].<br />

Computational properties of nonrecursive<br />

filter banks<br />

Both filter banks and its special case, i.e. transforms, decompose<br />

signals into subbands [1–3]. From this point of view, it is<br />

convenient to identify such a system with a set of independent<br />

filters, as illustrated in Fig. 1a. For design and implementation<br />

purposes, however, the equivalent scheme of Fig. 1b is much<br />

more useful, in which a vector of input samples is multiplied<br />

by a matrix of scalars, for block transforms, or polynomials,<br />

for filter banks (also called lapped transforms), respectively. In<br />

the latter case, it is said that a system is represented using its<br />

polyphase matrix [1].<br />

For design purposes and for adaptive processing, a system<br />

matrix can be parameterized. Usually, its elements are<br />

indirectly related to a smaller number of system coefficients,<br />

which are controlled by an external algorithm. This is done<br />

by factorizing the matrix into elementary transformations like<br />

rotations, reflections [8], lifting steps [9], quaternion multiplications<br />

[7], gains, and delays. As computations can be saved<br />

in this way, matrix factorizations also form the basis for most<br />

efficient implementations.<br />

From our point of view, it is important that factorizations<br />

define signal flow graphs in which data go from input to output<br />

56<br />

a) b)<br />

Fig. 1. Representations of a filter bank: (a) direct and (b) polyphase<br />

Rys. 1. Reprezentacje banku filtrów: (a) bezpośrednia i (b) polifazowa<br />

Fig. 2. Polyphase matrix as a data flow<br />

Rys. 2. Macierz polifazowa jako przepływ danych<br />

through parallel paths on which elementary transformations<br />

act, as illustrated in Fig. 2. The range of the transformations<br />

is quite narrow. Most of them can be described by simple matrices<br />

and affect only 1–2 data paths. There are neither conditional<br />

nor iterative actions, and access to the internals of<br />

such a scheme is possible only via external coefficients, on<br />

which transformations may depend. Because of these facts,<br />

it is appealing to develop a declarative language that allows<br />

developers to explicitly assign operations to data paths, to use<br />

built-in standard elementary transformations, and to define<br />

their own extensions.<br />

Existing tools for developing transforms<br />

Having syntax aimed at matrix computations, MATLAB seems<br />

naturally suited to studying the systems under consideration.<br />

Indeed, offering rich programming language, advanced visualization<br />

methods, and a lot of specialized toolboxes, it is widely<br />

used to develop DSP algorithms. However, in some situations<br />

it is not as useful as it could be. Especially, representing a filter<br />

bank as a composition of elementary transformations requires<br />

both unclear and inefficient Matlab code, behind which are<br />

excessive data movements and unnecessary computations,<br />

which are related to multiplying sparse matrices and can be<br />

only partially avoided using involved matrix indexing. The<br />

related overhead is negligible in interactive experiments but<br />

becomes troublesome if a transform routine is called frequently<br />

by a coding algorithm or by a program for numerical optimization<br />

of system coefficients. Moreover, it seems generally<br />

impossible to easily convert a Matlab script into efficient code<br />

in another programming language.<br />

Simulink allows systems to be modeled as schemes, which<br />

are composed of simpler building blocks via drag-and-drop.<br />

After successful simulation, a model can be converted into an<br />

implementation using one of numerous code generators that<br />

handle wide range of real-time platforms. However, graphical<br />

development is not so convenient if there are many blocks<br />

and connections, plenty of options make controlling code generation<br />

difficult, and resulting implementations are not intended<br />

for further customization. To the best of the author’s<br />

knowledge, there is no support for Java. It is also annoying<br />

that there is no possibility to convert a Simulink scheme into<br />

an ordinary Matlab routine. Finally, both tools are expensive<br />

commercial software.<br />

The Signal Processing Language (SPL) [4, 5] offers<br />

a specialized syntax for describing matrix factorizations and<br />

algorithms that are based on computing matrix-by-vector products.<br />

It is very flexible and functional, and a nice compiler is<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


publicly available. However, a closer look reveals two issues<br />

that are annoying in some cases. Firstly, the SPL syntax is<br />

different from the MATLAB language and seems generally<br />

lead to quite long and unclear code. Secondly, the compiler<br />

generates only C and FORTRAN code. Java and C++ are not<br />

supported.<br />

It should be noted that the SPL has became a part of SPI-<br />

RAL, a huge project aimed at automatization of optimally implementing<br />

transforms and linear DSP algorithms on modern<br />

microprocessors [5]. Although the contributors claim that the<br />

developed design environment supports different transforms,<br />

only the tools and results related to the Discrete Fourier Transform<br />

(DFT) and simple filters are publicly available. Thus, for<br />

a person from outside the project it seems impossible to evaluate<br />

how the SPIRAL program generation system helps with<br />

developing filter banks.<br />

A conclusion is that there are clear reasons for thinking<br />

about new tools which supplement the existing ones in particular<br />

situations. On the other hand, developing a language<br />

and code generator is a challenging task, interesting from<br />

both engineering and scientific points of view.<br />

TDL: Transform Description Language<br />

Developed to solve practical issues, the TDL has been specifically<br />

tailored to the task. Its syntax is aimed at compactly, clearly,<br />

and quickly describing computational schemes with the<br />

properties described. In particular, being dedicated to such<br />

a specific purpose, the language lacks procedural-programming<br />

features: conditional statements, comparison operators,<br />

loops, types, and even variables. There are also no functions<br />

in the usual sense, and some constraints are put on the code<br />

structure. Especially, a description file comprises three parts,<br />

which are delimited by keywords, as shown in Fig. 3.<br />

The first part is the place for specifying transform parameters<br />

like the number of inputs, number of outputs, and number<br />

of coefficients. They have no connection with implementation,<br />

so should be tied to descriptions. In the second section, identifiers<br />

can be assigned to (possibly parameterized) matrix expressions<br />

in order to make it possible to reuse code without<br />

explicitly copying it in several places in a file. In addition to saving<br />

work, this makes descriptions clearer and prevents bugs.<br />

The last and main part is a description of the computational<br />

scheme. It consists of statements, each of which begins with<br />

specifying the indexes of scheme paths, and then, separated<br />

by the “>” symbol, the operation follows that has to act<br />

on these paths. The statement ordering determines the order<br />

in which computational stages follow in the system scheme,<br />

starting from its input. Skipping a path specification is equivalent<br />

to using that from the previous statement. Another coding<br />

simplification is that all paths can be referenced via “:”. Of course<br />

definitions of the second section can be used to simplify<br />

path specifications.<br />

In the current version of TDL, three operations on samples<br />

are available that are listed in Table 1. The first of them<br />

is multiplication of a vector of patch samples by a matrix of<br />

scalars. Such matrices are sufficient to describe most of the<br />

elementary transformations. Only delaying samples needs to<br />

be denoted using a dedicated identifier and to be differently<br />

handled by the code generator. Permuting samples among<br />

paths can essentially be represented as multiplication by the<br />

appropriate permutation matrix, but it is tedious to specify<br />

such matrices. A separate notation saves a lot of work and<br />

makes code clearer.<br />

Matrices can be specified using the means of Table 2,<br />

numeric literals, and previously defined identifiers. The most<br />

practical elements of the MATLAB language have been adopted<br />

into the TDL, e.g. colon-based generation of number<br />

sequences, bracket-based matrix concatenation, and starting<br />

comment lines with “%”. The MATLAB-like syntax is accompanied<br />

by transform-related extensions. Namely, we have<br />

predefined constant and parameterized matrices that describe<br />

the elementary transforms that are most often used in<br />

the computational schemes under consideration. Especially,<br />

there is no need for declaring by oneself both rotations and<br />

reflections [8], lifting steps [9], and quaternion multiplications<br />

Fig. 3. The general structure of TDL descriptions<br />

Rys. 3. Ogólna struktura opisów TDL<br />

Tabl. 1. Sample modifications in TDL<br />

Tab. 1. Modyfikacje próbek w TDL<br />

Notation<br />

matrix expression<br />

permute(path index<br />

list)<br />

delay(number of<br />

samples)<br />

Description<br />

multiplying the vector of path<br />

samples by a given matrix<br />

(result is assigned to the same<br />

paths)<br />

(rearranging paths)<br />

delaying samples in paths<br />

Tabl. 2. Matrix notation in TDL<br />

Tab. 2. Notacja macierzowa w TDL<br />

Category<br />

arithmetic operators<br />

(element-wise)<br />

Examples<br />

+, -, .* (multiplication), ./ (division), .ˆ (power)<br />

constant matrices ones(size), zeros(row count, column count)<br />

concatenation operators<br />

sequence vector<br />

creation<br />

element rearranging<br />

functions (elementwise)<br />

predefined<br />

constant matrices<br />

(range of sizes)<br />

predefined parameterized<br />

transformation matrices<br />

[ matrix , . . . ] (horizontal),<br />

[ matrix ; . . . ] (vertical),<br />

diag(vector) (diagonal)<br />

start : step : stop<br />

start : stop (step = 1)<br />

matrix’<br />

flipud(matrix)<br />

fliplr(matrix)<br />

(transposition)<br />

(up-down flip)<br />

(left-right flip)<br />

sin(matrix) , cos(matrix) , sqrt(matrix)<br />

I4<br />

J3<br />

PI<br />

rotation(α)<br />

reflection(c, s)<br />

liftu(α)<br />

liftd(α)<br />

quatl(a, b, c, d)<br />

(4×4 identity matrix)<br />

(3×3 counter-identity or reversal matrix)<br />

(the π number)<br />

(2-dimensional)<br />

(2-dimensional)<br />

(lifting step)<br />

(inverse lifting step)<br />

(quaternion multiplication matrix)<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 57


a)<br />

b)<br />

Fig. 5. TDL-based development framework of applications<br />

Rys. 5. Rozwijanie aplikacji z wykorzystaniem TDL<br />

Fig. 4. Loeffler’s scheme for computing the DCT: (a) signal flow<br />

graph and (b) its TDL description<br />

Rys. 4. Schemat Loefflera do obliczania DCT: (a) graf przepływu<br />

sygnału i (b) jego opis TDL<br />

[7]. It should be also noted that our approach prevents excessive<br />

unnecessary use of matrix notation, unlike both Matlab<br />

and SPL, which require either constructing sparse matrices or<br />

involved indexing to combine an elementary transform with a<br />

polyphase matrix of greater size.<br />

Fig. 4 shows both the Loeffler’s scheme [10] for computing<br />

the DCT and the corresponding TDL description, which demonstrates<br />

that the language works well in practice.<br />

TDL-based development framework<br />

The new language allows for implementing transforms using<br />

the approach illustrated in Fig. 5. Using the TDL, a system of<br />

interest is described in one file, in a way independent of implementation.<br />

Desirable properties of implementation code are<br />

specified in a separate file. Both files are input to a TDL compiler,<br />

which generates source code that implements the transform.<br />

In order to assembly an application, the resulting files are<br />

attached to the rest of its source code, and a platform-specific<br />

compiler is used to produce executables. At the time of writing,<br />

we are working on Java and C/C++ code generation, but adding<br />

support for Matlab and VHDL is planned for the future.<br />

The complier exempts developers from tedious programming<br />

tasks and protects them from making bugs. They only<br />

have responsibility for describing systems correctly and lucidly,<br />

i.e. using optimal TDL constructions. It is assumed that machine-level<br />

optimization is done by the Java/C/C++ compiler,<br />

and thus only most essential code optimization is automatically<br />

performed. Loops are unrolled, i.e. arrays are explicitly<br />

58<br />

processed element-by-element, which is recommended by<br />

processor vendors but very wearisome to write for a human.<br />

Values of multiplication/addition operands are analyzed so as<br />

to omit unnecessary operations or replace them with simpler<br />

ones if possible. Finally, partial results which can be determined<br />

in advance and are referenced to from several places<br />

of a computational scheme are only once computed as early<br />

as possible and stored in auxiliary variables.<br />

It is noteworthy that design environments exist that allow<br />

for converting a floating-point implementation of a DSP algorithm<br />

into a refined fixed-point code [11–13]. Using them in<br />

conjunction with our tools seems very advantageous.<br />

One can claim that good Java or C/C++ libraries could<br />

simplify transform implementation as much as our language.<br />

We estimate that both solutions require similar programming<br />

effort, but the former solution has evident shortcomings. Firstly,<br />

the most effective code optimization strategies cannot be<br />

employed: computing partial results in advance, avoiding procedure<br />

calls and memory allocation, and loop unrolling. Secondly,<br />

source code that calls library functions cannot be as<br />

simple and clear as a description in a dedicated syntax.<br />

Code generation details<br />

It seems best to demonstrate features of the TDL compiler<br />

via an example. For brevity reason, let us consider the very<br />

simple filter bank whose scheme is shown in Fig. 6a. Its core<br />

is composed of one lifting step, one delay, and one rotation,<br />

whose angle is an external coefficient. The corresponding TDL<br />

description is shown in Fig. 6b. For the option file in Fig. 6c,<br />

the compiler produces the Java code in Fig. 6d. The transform<br />

is implemented as a class with three methods: “resetDelays”<br />

initializes variables used to implement delays, “init” initializes<br />

variables related to the coefficient, and “process” is responsible<br />

for processing of a current input vector.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


a)<br />

b)<br />

c)<br />

d )<br />

Fig. 6. Code generation example: (a) system scheme, (b) TDL description of the polyphase matrix, (c) option file, and (d) resulting<br />

Java implementation<br />

Rys. 6. Przykład generacji kodu: (a) schemat systemu, (b) opis macierzy polifazowej w TDL, (c) opcje generacji i (d) wynikowa implementacja<br />

w Javie<br />

It is assumed that “process” is repeatedly called from<br />

higher-level code, each time with different input data. In the<br />

scheme, subsequent input vectors for the core are produced<br />

by the stage composed of the delay and decimators. In a program,<br />

moving over the sequence of input samples can easily<br />

be realized by modifying pointers or array indexes that are<br />

passed to “process”.<br />

It could be noted that the TDL compiler generates accessible<br />

identifiers, so that the resulting code can easily be refined<br />

by hand, if somewhat outside the proposed functionality<br />

is required.<br />

Compiler internals<br />

The compiler source consists of about 80 Java classes, which<br />

belong mainly to four packages: parser, matrix, buffer, and<br />

coefficient. The top-down recursive-descent parser analyzes<br />

TDL descriptions without explicitly creating a parse tree, which<br />

is modeled as a sequence of function calls. Only its subtrees<br />

related to matrix expressions are constructed in memory, as<br />

they are necessary to resolve symbols and to determine gains<br />

of data flows among buffers. Analyzing line-by-line the main<br />

part of a description, the parser builds subsequent layers of<br />

signal flow graphs. Simultaneously, flow gains are associated<br />

with nodes of the expression tree, which represent system<br />

coefficients, numeric constants, and their functions. All gains<br />

with the same value are tied to a single node of the tree, which<br />

forms the foundation for sharing partial results. On the other<br />

hand, in the signal flow graph, edges are detected that can be<br />

combined so as to avoid superfluous computations and data<br />

movements.<br />

There are several phases of code generation. After determining<br />

an appropriate structure of implementation code and<br />

preparing function headers, “init” is produced as a result of traversing<br />

the expression tree. Auxiliary results are assigned to<br />

variables, whose identifiers are then referenced by the generator<br />

module that converts the signal flow graph into sequence<br />

of constructs in the target programming language.<br />

Conclusion<br />

Obviously, the language and compiler we have developed are<br />

neither as universal nor as functional as the existing mature<br />

tools. Nevertheless, in some practical situations they make<br />

implementing transforms much easier and quicker than when<br />

using the latter. From the point of view of non-mathematicians,<br />

an additional advantage is that systems can be described based<br />

on its signal flow graph, with limited use of matrix notation.<br />

The TDL is expected to develop further in several directions.<br />

It is planned to extend its syntax so that more system<br />

classes and more elementary operations could be described.<br />

On the other hand, code generation could support other target<br />

languages, especially MATLAB/Simulink and VHDL. Finally,<br />

two additional tools could be developed. The first is a GUI-based<br />

editor for interactively creating descriptions, and the second<br />

is a code generator that produces a TDL description of<br />

the inverse transform based on that of a forward transform.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 59


This work was supported by Bialystok University of Technology<br />

under the grant W/WI/6/<strong>2010</strong>. Currently, an improved version<br />

of the language and compiler is developed as a part of the project<br />

“Platforma Informatyczna TEWI”<br />

References<br />

[1] Vaidyanathan P. P.: Multirate Systems and Filter Banks. Englewood<br />

Cliffs. NJ: Prentice-Hall, 1993.<br />

[2] Strang G., Nguyen T. Q.: Wavelets and Filter Banks. Wellesley.<br />

MA: Wellesley-Cambridge Press, 1996.<br />

[3] Piotrowski A., Parfieniuk M.: Cyfrowe banki filtrow: analiza, synteza<br />

i implementacja dla systemow multimedialnych. Białystok:<br />

Wydawnictwo Politechniki Białostockiej, 2006, p. 389.<br />

[4] Xiong J. i in.: SPL: a language and compiler for DSP algorithms.<br />

ACM SIGPLAN Notices, vol. 36, no. 5, pp. 298–308, May 2001.<br />

[5] Püschel M. i in.: SPIRAL: Code generation for DSP transforms.<br />

Proc. IEEE, vol. 93, no. 2, pp. 232–275, Feb. 2005.<br />

[6] Parfieniuk M.: Shortening the critical path in CORDIC-based approximations<br />

of the eight-point DCT. In Proc. Int. Conf. Signals Electronic<br />

Systems (ICSES), Cracow, Poland, 14–17 Sep. 2008, pp. 405–408.<br />

[7] Parfieniuk M., Petrovsky A.: Inherently lossless structures for<br />

eight- and six-channel linear-phase paraunitary filter banks<br />

based on quaternion multipliers. Signal Process., vol. 90, no. 6,<br />

pp. 1755–1767, June <strong>2010</strong>.<br />

[8] Vaidyanathan P. P.: On coefficient-quantization and computational<br />

roundoff effects in lossless multirate filter banks. IEEE Trans.<br />

Signal Process., vol. 39, no. 4, pp. 1006–1008, Apr. 1991.<br />

[9] Chen Y. J., Amaratunga K. S.: M-channel lifting factorization of<br />

perfect reconstruction filter banks and reversible M-band wavelet<br />

transforms. IEEE Trans. Circuits Syst. II, vol. 50, no. 12, pp.<br />

963–976, Dec. 2003.<br />

[10] Loeffler C., Lightenberg A., Moschytz G.: Practical fast 1-D<br />

DCT algorithms with 11 multiplications. in Proc. IEEE Int. Conf.<br />

Acoust., Speech, Signal Process. (ICASSP), vol. 2, Glasgow,<br />

Scotland, 23–26 May 1989, pp. 988–991.<br />

[11] Kim S., Kum K., Sung W.: Fixed-point optimization utility for<br />

C and C++ based digital signal processing programs. IEEE<br />

Trans. Circuits Syst. II, vol. 45, no. 11, pp. 1455–1464, Nov.<br />

1998.<br />

[12] Coors M., Keding H., Lüthje O., Meyr H.: Design and DSP implementation<br />

of fixed-point systems. EURASIP J. Appl. Signal<br />

Process., vol. 2002, no. 9, pp. 908–925, 2002.<br />

[13] Bhattacharyya S., Leupers R., Marwedel P.: Software synthesis<br />

and code generation for signal processing systems. IEEE Trans.<br />

Circuits Syst. II, vol. 47, no. 9, pp. 849–875, Sep. 2000.<br />

Expandable GSM and GPS systems simulator<br />

(Rozbudowany symulator systemów GSM oraz GPS)<br />

dr inż. JANUSZ POCHMARA, JAKUB PAŁASIEWICZ, PIOTR SZABLATA<br />

Poznan University of Technology<br />

Wireless communications systems of the next generation are<br />

expected to provide its users with new services. Supporting<br />

this problem requires new engineers which will project new<br />

technological solutions for mobile communications. We observe,<br />

this is in the fact, that many people in the not so distant<br />

future will start to see the mobile phone as an alternative to<br />

a PC. The mobile phone will become their digital life.<br />

The main problem is education of the future engineers and<br />

depends on the good education. This task is difficult in practical<br />

realization. Roughly half of the world’s population already<br />

has some type of mobile phone, making it the most wide<br />

spread technology and most common electronic device in the<br />

world. Un-tapping this ubiquitous technology creates a wide<br />

array of educational possibilities.<br />

In our opinion one of the best ways to solve this problem<br />

is learning future engineers with new technologies with more<br />

practices and exercises. Therefore we construct a simulator of<br />

the GSM and GPS system based on integrated sophisticated<br />

module. In our solution with simulator we will learn: (1) about<br />

the different mobile phone technologies; (2) how to design learning<br />

activities for mobile phones; (3) pedagogical and theoretical<br />

frameworks for mobile learning.<br />

This paper is organized as follows. In Sections “System<br />

Model” and “Software Specification” we describe the system<br />

model. Section “Computer Application” is devoted to the<br />

detailed description of the investigated. Finally, in Section<br />

“Conclusion” we present results, which have been obtained<br />

for the GSM and GPS systems. We conclude our paper with<br />

a discussion on results.<br />

Hardware specification<br />

Device uses integrated GSM/GPRS+GPS module. Small and<br />

functional device is compatible with industry standards. This<br />

makes it suitable for using it with other hardware parts (for<br />

example: microcontroller, PC). It was certified by MET Laboratories,<br />

Global Certification Forum, CNAS and CTIA Authorized<br />

Test Lab.<br />

GSM/GPRS technology makes it easy to communicate<br />

with storehouse from everywhere and anywhere [1]. Satellite<br />

navigation combined with GSM standard enables to track<br />

every object. Small dimensions (Fig.1) of this all-in-one module<br />

makes it handle and stand-alone device, which would<br />

surely find many possible solutions on the market. Low power<br />

consumption (about 70 mA while in sleep mode) ensures long<br />

working time.<br />

Fig. 1. GSM\GPRS+GPS module<br />

Rys. 1. Moduł GSM/GPRS+GPS<br />

60<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


GSM\GPRS+GPS module is using hardware interface<br />

compatible with RS-232 standard to communicate with PC via<br />

serial port, which also handles connections with other devices,<br />

modules or even external battery pack. It is possible to<br />

expand it with microcontroller, which will make module work<br />

as an independent device. It can also be controlled by a computer<br />

or (with the right software) via GSM network.<br />

System model<br />

Project uses up to date hardware to present combination of<br />

GSM and GPS technologies. With usage of standardized coding<br />

languages and current computer software it shows vast<br />

possibilities to utilize the idea of the project. Integration of both<br />

modules allows user to track anything from anywhere in anytime<br />

with easy to integrate all-in-one solution.<br />

Acquiring actual location of the bacon is realized by interpreting<br />

NMEA protocol messages presented by GPS module.<br />

By using Global Positioning System Fixed Data [2] our<br />

interpreter is able to create a group of variables and present<br />

them in one long string type variable, which is ready for use in<br />

a message compatible with GSM format.<br />

We can use up to 12 separate connections with satellites to<br />

estimate the most precise location of the simulator. Additional<br />

parameters are used to calculate possible errors and apply them<br />

in process of updating coordinates to show proper values.<br />

Fast transfer of data is realized by using Short Message<br />

Standard, best known from mobile phones. By using this technique<br />

it is possible to send up to 160 characters of data in<br />

a few seconds. It’s more than it is needed to transfer coordinates,<br />

precision parameters and additional variables to the<br />

receiver.<br />

Establishing connection to the cell phone carrier is necessary<br />

for GSM module to work properly. Realization of authentication<br />

and logging to GSM network process are implemented<br />

in the project.<br />

Fig. 2. Project schematic<br />

Rys. 2. Schemat działania<br />

urządzenia<br />

Solution schematic (Fig. 2):<br />

• GPS module downloads coordinates and additional information.<br />

• Parser seeks for specific variables presented in NMEA<br />

messages and converts them into one long string type variable.<br />

• Variable is transferred to GSM module.<br />

• Data is included into SMS type message and sent by using<br />

AT commands (GSM 07.07 compatible).<br />

• Software interpreter reads data and creates group of variables<br />

used further in software application.<br />

• Coordinates are calculated by application.<br />

• Position and additional data are shown on the screen [3,<br />

4]. Location is presented on the map [5, 6].<br />

Software Specification<br />

AT Standard<br />

Originally AT commands were used to control the modem.<br />

They were introduced to unify equipment connected to a computer.<br />

When GSM technology appeared, AT was adapted to<br />

handle modems built into mobile phones. Today each device<br />

(terminal) GSM has a built-in AT command interpreter.<br />

AT commands allow you to perform all tasks such as making<br />

calls, sending SMS, handling the SIM card, etc.<br />

Both the format and the returned type of the commands<br />

are standardized [7]. AT command starts with “AT” (except “A/”<br />

and “+++”) and ends with Carriage Return char. One line of AT<br />

command can contain up to 128 characters (during command<br />

interpretation white spaces are ignored). When command reaches<br />

terminal adapter, it returns “OK” or “ERROR”, sometimes<br />

also some additional information.<br />

All AT commands can be split into three categories syntactically:<br />

• Basic – writes command with arguments (arguments are<br />

optional and a default will be used if missing).<br />

• S parameter – sets value of the S register.<br />

• Extended:<br />

o Test command – returns list of parameters and value<br />

ranges related to the possible settings of specific command.<br />

o Read command – returns the currently set value of the<br />

parameter or parameters.<br />

o Write command – sets the user-definable parameter<br />

value(s).<br />

o Execution command – result is depending of default<br />

setting at current phone. In module it reads non-variable<br />

parameters [8].<br />

Module is compatible with GSM 07.05 and 07.07 standards.<br />

After starting the module, serial port default setting should<br />

look like this: 8-digit data bit, 1-digit stop bit, no parity check,<br />

no CTS/RTS, data rate 115200 bps.<br />

NMEA Protocol<br />

GPS receiver uses SiRFstar III chipset with ARM7 processor<br />

and is compatible with NMEA-0183 version 3.0 standard [9].<br />

Project takes advantage of the newest protocol and its data<br />

checksum feature used in debugging process to prevent errors<br />

at early stage of analysis.<br />

Project uses NMEA output messages and SiRF manufacturer<br />

sentences to prepare one long string type variable for<br />

further calculations in software application. Specially created<br />

parser seeks for coordinates and additional data included<br />

in GGA, GLL and GSA messages of NMEA standard [10]. Values<br />

are constantly changing and usage of searching engine<br />

is necessary.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 61


Manufacturer sentences are mainly used to customize<br />

the way the chip behaves. There are five input sentences in<br />

SiRF messages group. For example the PSRF103 line is used<br />

to control which NMEA sentences are sent and how often.<br />

Simulator is able to make warm start with usage of initialization<br />

data loaded from software.<br />

Example line of NMEA standard code, consisting of parameters<br />

regarding coordinates, units, used satellites and additional<br />

indicators:<br />

$GPGGA,161229.487,3723.2475,N,12158.3416,W,1,07,1.0,9.0,<br />

M,,,,0000*18<br />

Example line of SiRF message standard code, sentence<br />

104 is used to state initial values for GPS module:<br />

$PSRF104,37.3875111,-121.97232,0,95000,237759,922,12,3*3A<br />

Computer application<br />

Fig. 3. Application window<br />

– work in progress<br />

Rys. 3. Okno aplikacji<br />

w trakcie pracy<br />

62<br />

Application is still under development.<br />

Presented screenshots<br />

were taken from early<br />

alpha version with limited functionality<br />

[11].<br />

Software is needed to control<br />

the entire hardware layer. It<br />

provides support for the GSM<br />

module, GPS navigation and<br />

simulates the stand-alone device<br />

on PC. Application is userfriendly<br />

and gives full control of<br />

the device.<br />

GSM Support<br />

AT instructions allow user to operate<br />

the module in such a way<br />

as any mobile phone. Only difference<br />

is that keyboard and<br />

display aren’t included with module,<br />

to maintain the functionality, from the user perspective, an<br />

additional interface is needed. In short, user writes commands<br />

by clicking buttons or (in advance mode) directly as a text. Application<br />

opens the serial port and writes each line of command<br />

to terminal adapter. Terminal returns an alphanumeric string to<br />

inform about success or failures. Parser interprets this information<br />

and writes it in understandable to user text form.<br />

GPS Support<br />

The program reads NMEA data exported by GPS and uses<br />

a parser to extract the individual variables, then it displays<br />

them in a manner understandable to user. Extracted geographical<br />

coordinates allow us to find tracked object on the<br />

map. In addition, information is saved in history, so it’s possible<br />

to observe path of object movement.<br />

Automated work of application<br />

Program reads data exported by GPS and parser extract geographical<br />

coordinates. In the next step they are sent to the<br />

storehouse via GSM protocol. The program repeats this operation<br />

in a loop. To save energy and space required to store<br />

data there is a period of time between the loops. In addition, to<br />

reduce the energy consumption and cost of data, program can<br />

save a few measurements and send them only when length<br />

of the information is optimal in terms of the maximum length<br />

of SMS message.<br />

Fig. 4. Application schematic<br />

Rys. 4. Schemat blokowy działania aplikacji<br />

The whole program is automated and does not require<br />

usage of any interface (with the exception of the on/off functions).<br />

This part of application can be used to program the<br />

microcontroller connected with a module, equipped with antenna<br />

and battery. This idea allows engineers to create standalone<br />

device used for tracking purposes.<br />

In the storehouse, data are collected and archived. This<br />

allows not only to find the current location of the observed<br />

object on the map, but also to display the history of his movement.<br />

During the further development of the project, program<br />

features will be extended.<br />

Conclusion<br />

Current solution presented in this project can serve as a base<br />

for more complicated systems with GSM and GPS technologies<br />

implementations. Modules are prepared for linking with<br />

additional hardware parts. Possibilities of further develop<br />

of this simulator are almost endless.<br />

Simple microcontroller (ATmega8 standard) or ARM9/<br />

ARM11 processor could raise practical value of presented<br />

solution. Programming self-efficient hardware for more complicated<br />

problems is one of possible ways of expanding this<br />

project.<br />

Fig. 5. Project available expansions<br />

Rys. 5. Możliwości poszerzenia projektu<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Set of LCD screen, keyboard and speaker would allow<br />

changing simulator into proper stand-alone mobile phone solution<br />

with user friendly interface. Adding battery and using<br />

GSM module ability of charging it, could lead to creation of<br />

long time working unit used for tracking solutions.<br />

This project makes as a great base for simulating most<br />

popular wireless technologies. Nowadays investing resources<br />

into solutions taking advantages from both GPS and GSM<br />

standards is profitable and project combining both of them in<br />

one module is just the hardware that telecommunication industry<br />

needs. Low production cost and user friendly application<br />

are making this project even more valuable.<br />

The authors would like to acknowledge the contribution<br />

of S.K.I.M. department at Poznan University of Technology in<br />

making valuable suggestions in improving the content of this<br />

paper.<br />

References<br />

[1] Ashby N.: Relativity and the Global Positioning System. Physics<br />

Today, May 2002.<br />

[2] Kingsley-Hughes K.: Hacking GPS. Chapter 8, 2005.<br />

[3] Davis S.: Google Maps API Second Edition, 2006.<br />

[4] Brown M.: Hacking Google Maps and Google Earth. Part I, 2006.<br />

[5] Purvis M., Sambells J. and Turner C.: Beginning Google Maps<br />

Applications with PHP and Ajax. Part I and II, 2006.<br />

[6] Pimpler E.: Mashup Mania with Google Maps. January 2009.<br />

[7] Guthery S. and Cronin M.: Mobile Application Development.<br />

Chapter 2, 2002.<br />

[8] Waldemar N.: Komputerowe Systemy Pomiarowe. Chapter 4,<br />

2002.<br />

[9] United States of America Department of Defense, GPS SPS<br />

Performance Standard, 4th Edition, September 2008, http://pnt.<br />

gov/public/docs/2008/spsps2008.pdf<br />

[10] Ardalan A. and Awange J.: Compatibility of NMEA GGA with GPS<br />

Receivers Implementation. Volume 3, Number 3, January 2000,<br />

pp. 1–3.<br />

[11] Gross C.: Beginning C# 2008 From Novice to Professional, 2007.<br />

Alvis approach to Hexor robot controller development<br />

(Zastosowanie języka Alvis do projektowania sterownika dla robota Hexor)<br />

dr hab. MARCIN SZPYRKA, dr inż. PIOTR MATYASIK, dr inż. RAFAŁ MRÓWKA<br />

AGH Kraków, Katedra Automatyki<br />

The aim of the paper is to present an introduction to Alvis modelling<br />

language based on a real example. For this purpose,<br />

we have chosen a model of a controller for the Hexor II mobile<br />

robot.<br />

The direct ancestors of Alvis are the XCCS [1-3] and CCS<br />

process algebras [4-6]. CCS (Calculus of Communicating Systems)<br />

is one of the most famous process calculi. It provides<br />

a tool for the high-level description of interactions, communications,<br />

and synchronizations among agents and also algebraic<br />

laws to analyse agents properties. One of the main<br />

disadvantages of CCS is the lack of a suitable graphical language<br />

for modelling concurrent systems. A designer is forced<br />

to use the textual (algebraic equations) form of a system description.<br />

XCCS (eXtended CCS) is a graphical extension of the<br />

CCS process algebra. XCCS provides a graphical modelling<br />

language for the description of interactions among agents, but<br />

still preserves algebraic equations to describe the behaviour<br />

of individual agents.<br />

Alvis, as well as XCCS, uses two layers, a graphical and<br />

a textual one. The Alvis graphical layer, called communication<br />

diagrams, is a significant enhancement of XCCS diagrams.<br />

Communication diagrams are hierarchical constructions and<br />

reflect the structure of the system under consideration more<br />

precisely (active and passive agents, two-way ports, etc.)<br />

than XCCS diagrams. The XCCS algebraic layer has been<br />

replaced with a completely new Alvis code layer. Instead of<br />

algebraic equations, Alvis uses a high level programming language<br />

based on the Haskell syntax. Moreover, Haskell is used<br />

to define data types for parameters and to define functions for<br />

data manipulation.<br />

Alvis provides a possibility of the formal verification of models.<br />

An Alvis model is transformed into a labelled transition<br />

system (LTS) encoded using the Binary Coded Graphs (BCG)<br />

format. Then, the CADP toolbox [7] is used to verify its properties.<br />

Hexor microcontroller model<br />

Hexor II (Fig. 1) is an autonomous 6-legged mobile robot developed<br />

by the Stenzel company for educational purposes<br />

[10]. An Alvis model of the ATmega128 microcontroller [11] is<br />

considered in the paper. The tasks of that chip are as follows:<br />

scanning sensors (sonar, tentacles, infrared), generating signals<br />

for servos (PWM), executing the movement algorithm,<br />

communicating with the host computer, and executing higher<br />

level algorithm (obstacle avoidance, obstacle search, etc.).<br />

Fig. 1. Hexor mobile robot<br />

Rys. 1. Robot mobilny Hexor<br />

Because of some software platform limitations [10, 11]<br />

of the original design, a new Hexor’s internal controlling software<br />

architecture has been developed. The basic program<br />

with one control loop and interrupt routines has been replaced<br />

by a real-time embedded operating system. Each subsystem<br />

of Hexor robot is managed by its own task (see Fig. 2). One<br />

of the main advantages of these modifications is a more clear<br />

and easy to understand source code. This simplifies the parent<br />

control layer and enhances response time of a robot.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 63


Fig. 2. Hexor task model<br />

Rys. 2. Model zadań robota Hexor<br />

Fig. 3. Communication page<br />

Rys. 3. Strona Communication<br />

The complete Hexor system architecture consists of three<br />

elements: AvrX micro-kernel, HexorNG software, and High level<br />

intelligence. Figure 2 presents a model of tasks and communication<br />

in the proposed new system. Ovals represent tasks,<br />

arrows represent FIFO’s and data flow direction, and the<br />

square represents ISR (Interrupt Service Routine). The tasks<br />

are used for the following purposes:<br />

• Control & Communication task is the main system task<br />

responsible for setting up hardware, interchanging data<br />

between other tasks, communication services and basic<br />

intelligence. The last feature means simply: do not run into<br />

obstacles.<br />

• Sonar task provides the most recent sonar sensor readings.<br />

• IR and Tentacles tasks report any changes in sensors state.<br />

• Movement task controls servo positions and executes the<br />

movement algorithm on demand.<br />

• Watchdog task simply resets the hardware watchdog chip<br />

every second.<br />

• ISR is responsible for generating proper PWM (Pulse Width<br />

Modulation) signals for servos. It is also used for context<br />

switching and timer execution for AvrX micro kernel.<br />

Communication diagrams<br />

The Alvis graphical layer is used to define interconnections<br />

among agents and takes the form of a hierarchical communication<br />

diagram. The term agent stand for any distinguished<br />

part of the model under consideration with its own identity.<br />

The graphical layer shows all communication channels among<br />

agents. It is composed of a set of pages, i.e. non-hierarchical<br />

parts of the diagram. A page may contain active agents, passive<br />

agents and connections among them.<br />

There are two kinds of agents in Alvis. Active agents<br />

(drawn as ovals) perform some activities and are similar to<br />

tasks in Ada programming language [12–14]. Each of them<br />

can be treated as a thread of control in a concurrent system.<br />

On the other hand, passive agents (drawn as rectangles) do<br />

not perform any individual activity, and are similar to protected<br />

objects (shared variables). Passive agents provide mechanism<br />

for the mutual exclusion and data synchronisation.<br />

Moreover, Alvis provides mechanisms for the description of<br />

interrupt handling routines in the form of agents. An interrupt<br />

service is represented by a single active agent called interrupt<br />

agent. Such an agent is drawn using dashed lines.<br />

The Hexor controller communication diagram contains<br />

four pages. Communication page is shown in Fig. 3. The page<br />

contains four agents. Two of them are simple FIFO queues<br />

for storing frames to be send or to be received. Another two<br />

64<br />

Fig. 4. ServoController page<br />

Rys. 4. Strona ServoController<br />

agents are interrupt service routines. They provide interrupt<br />

driven UART communication. The Receiver task is executed<br />

every time a byte is ready to be taken from shift register. When<br />

it receives a complete frame it is stored in the queue. The<br />

Transmitter task is executed every time a byte was shifted<br />

from UART register. It takes frames from the output queue<br />

and sends them byte by byte.<br />

An agent can communicate with other agents through ports.<br />

Ports are drawn as circles placed at the edges of the corresponding<br />

oval or rectangle. Two connected ports form a communication<br />

channel. Communication diagrams provide two<br />

kinds of connections: one-way (see Fig. 4) and two-way (see<br />

Fig. 3) ones. A one-way connection contains an arrowhead that<br />

points out the input port, but such a port is treated as an input<br />

port only for this particular connection. It can play another role<br />

for its other connections.<br />

Figure 4 presents the ServoController page. It consists of two<br />

major elements. The first is a five channel software PWM implementation<br />

that generates signals for servos. The second element<br />

is the MoveController agent that executes the movement algorithm.<br />

It sets desired servo positions according to the selected movement<br />

mode (stop, forward, backward, left or right). Both agents<br />

are exchanging information via the ServoSig passive agent, which<br />

has also an input port for setting the camera position.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


This section presents a few pieces of Alvis code that demonstrate<br />

the most important parts of the language. The code<br />

layer of an Alvis model contains two parts: the preamble and<br />

implementation. Encoded in pure Haskell, the preamble contains<br />

definitions of types, constants and functions used to manipulate<br />

data in a model. The implementation contains definitions<br />

of the agents’ behaviour and is encoded using native<br />

ABD language statements.<br />

Listing 1. Part of the model preamble<br />

Listing 1. Część preambuły<br />

Fig. 6. Hexor page<br />

Rys. 6. Strona Hexor<br />

A part of a preamble is shown in Listing 1. It contains the<br />

enumerated data type FrameType and the composite data<br />

type Frame (pairs: a frame type and an integer).<br />

Listing 2. Agent Sonar implementation<br />

Listing 2. Implementacja agenta Sonar<br />

Fig. 7. Sensors page<br />

Rys. 7. Strona Sensors<br />

Alvis communication diagrams are hierarchical graphs.<br />

Pages are combined using the so-called substitution mechanism.<br />

An active agent at one level can be replaced by a page<br />

on the lower level, which usually gives a more precise and<br />

detailed description of the activity represented by the agent.<br />

Such an substituted agent is called hierarchical one. Hierarchical<br />

agents are indicated by black triangles. It should be underlined<br />

that all ports of an hierarchical agent must appear on<br />

the corresponding subpage as external (unconnected) ones.<br />

The communication diagram for the controller contains two<br />

levels. The first (higher) level is shown in Fig. 5. Three agents<br />

with black triangles icons stand for three different modules of<br />

the system under consideration. Agent Controller represents<br />

the main control process of the system. The last page (Fig. 7)<br />

contains agents that represent the robot sensors.<br />

Code layer<br />

To describe the behaviour of individual agents, Alvis uses a high<br />

level programing language based on the Haskell [15] syntax called<br />

Alvis Behaviour Description Language [16]. ABD language<br />

is used to define data types used in the model under consideration,<br />

functions for data manipulation, and behaviour of<br />

individual agents.<br />

Let us consider the definition of the agent Sonar presented<br />

in Listing 2. The agent Sonar uses two integer parameters<br />

value and oldVal (with the initial value equal to 0). Two statements<br />

are used for the communication. The in instruction for<br />

collecting data and out for sending. Each of them takes a port<br />

name as its first argument and optionally a parameter name<br />

as the second. Parameters are not used for pure synchronisations.<br />

The in statement assigns the collected value to its<br />

parameter, while the out statement sends the value of its parameter<br />

(or constant). The Sonar agent collects data through<br />

the port sonar and sends data through the port getSonar.<br />

Most statements can be followed by a guard. A guard is an<br />

additional constraint, which must be fulfilled before the corresponding<br />

statement is executed. Guards are logical expressions,<br />

written in Haskell, placed inside round brackets after<br />

the statement name.<br />

In the implementation part, the = symbol stands for the<br />

assignment operator and is a part of the exec statement. The<br />

statement is the default one in the ABD language and therefore,<br />

it can be omitted if no guard is used. For example, the<br />

Sonar agent uses the exec statement to assign a new value<br />

to the oldVal parameter.<br />

A communication through a port can be a pure synchronisation<br />

i.e. a communication without sending values of parameters.<br />

Such a communication only synchronises two agents.<br />

For example, the PWM agent (see Listing 3) synchronises<br />

with the clock.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 65


Summary<br />

Defined for the embedded systems design, Alvis seems to be<br />

more accessible for engineers than classical formal methods,<br />

but still preserves a possibility of formal verification with CADP<br />

toolbox [9]. To verify or simulate an Alvis model a third layer<br />

is introduced. It gathers information about all agents in the<br />

model and their states. The meta-data layer is generated automatically.<br />

Each agent is described using the so-called agent<br />

description block (ADB) and a list of its parameters values.<br />

The meta-data layer keeps the current state of a model and is<br />

used for the LTS generation. To verify a model, the generated<br />

LTS is encoded in the BCG (Binary Coded Graphs}) format.<br />

Then, the CADP Evaluator is used to check whether the generated<br />

LTS satisfies its requirements defined using µ-calculus<br />

[17] formulas. The CADP toolbox and BCG format allow us to<br />

analyse models up to 30 milion states.<br />

The paper is supported by the Alvis Project funded from<br />

2009–<strong>2010</strong> resources for science as a research project.<br />

Listing 3. Agent PWM implementation<br />

Listing 3. Implementacja agenta PWM<br />

In order to allow for the description of agents whose behaviour<br />

may follow different alternative paths, the ABD language<br />

offers the alt statement. The statement is similar to the basic<br />

select statement from Ada programming language. Each occurrence<br />

of the alt keyword starts a new branch. Each branch<br />

may have attached a guard. A branch is called open, if it does<br />

not have a guard attached or its guard evaluates to true.<br />

Otherwise, a branch is called closed. When an alt statement is<br />

to be executed, all guards are evaluated to determine, which<br />

branches are open. If more than one branch is open, the choice<br />

between them is indeterministic.<br />

The alt statement is also used for description services provided<br />

by passive agents. Let us consider the implementation<br />

of the FQ agent presented in Listing 4. The agent represents<br />

a FIFO queue. The two alt branches provide the queue interface.<br />

The first one describes collecting the queue head, while<br />

the second describes including a new element to the queue.<br />

Listing 4. Agent FQ implementation<br />

Listing 4. Implementacja agenta FQ<br />

66<br />

References<br />

[1] Szpyrka M., Matyasik P.: Formal modelling and verification<br />

of concurrent systems with XCCS. in Proceedings of the 7th International<br />

Symposium on Parallel and Distributed Computing<br />

(ISPDC 2008), Krakow, Poland, July 1–5 2008, pp. 454–458.<br />

[2] Balicki K., Szpyrka M.: Formal definition of XCCS modelling language,<br />

Fundamenta Informaticae, vol. 93, no. 1–3, pp. 1–15,<br />

2009.<br />

[3] Matyasik P.: Design and analysis of embedded systems with<br />

XCCS process algebra. Ph.D. dissertation, AGH University<br />

of Science and Technology, Faculty of Electrical Engineering, Automatics,<br />

Computer Science and Electronics, Kraków, Poland,<br />

2009.<br />

[4] Milner R.: Communication and Concurrency. Prentice-Hall,<br />

1989.<br />

[5] Fencott C.: Formal Methods for Concurrency. Boston, MA, USA:<br />

International Thomson Computer Press, 1995.<br />

[6] Aceto L., Ingófsdóttir A., Larsen K., and Srba J.: Reactive Systems:<br />

Modelling, Specification and Verification. Cambridge, UK:<br />

Cambridge University Press, 2007.<br />

[7] Garavel H., Lang F., Mateescu R., and Serwe W.: CADP 2006:<br />

A toolbox for the construction and analysis of distributed processes.<br />

in Computer Aided Verification (CAV’2007), ser. LNCS,<br />

vol. 4590. Berlin, Germany: Springer, 2007, pp. 158–163.<br />

[8] HexorII Robot Manual, Stenzel, 2006.<br />

[9] 8-bit AVR Microcontroller with 128K Bytes In-System Programmable<br />

Flash ATmega128 ATmega128L, Rev. 2467g-avr-09/02<br />

ed., Atmel, 2002.<br />

[10] Matyasik P. and Nalepa G. J.: Knowledge-based control of reactive<br />

systems with multi-layer architecture. in Proc. of Mixdes<br />

2007, the 14th International Conference Mixed Design of Integrated<br />

Circuits and Systems, Ciechocinek, Poland, June 21–23<br />

2007, pp. 667–672.<br />

[11] Matyasik P., Nalepa G. J., and Zięcik P.: Prolog-based real-time<br />

intelligent control of the hexor mobile robot. in Advances in Artificial<br />

Intelligence: Proceedings of the 30th Annual German Conference<br />

on AI, KI 2007, ser. LNAI, vol. 4667, Osnabruck, Germany,<br />

10-13 September 2007, pp. 485–488.<br />

[12] Ada Europe, Ada Reference Manual ISO/IEC 8652:2007(E) Ed.<br />

3, 2007.<br />

[13] Barnes J.: Programming in Ada 2005. Addison Wesley, 2006.<br />

[14] Burns A. and Wellings A.: Concurrent and real-time programming<br />

in Ada 2005. Cambridge University Press, 2007.<br />

[15] O’Sullivan B., Goerzen J., and Stewart D.: Real World Haskell.<br />

Sebastopol, CA, USA: O’Reilly Media, 2008.<br />

[16] Szpyrka M., Matyasik P., and Mrówka R.: Introduction to Alvis<br />

internal language syntax. AGH-UST, Kraków, Poland, CSL Technical<br />

Report 1, <strong>2010</strong>, http://cslab.ia.agh.edu.pl/en:csltr.<br />

[17] Emerson E. A.: Model checking and the Mu-calculus, in Descriptive<br />

Complexity and Finite Models. ser. DIMACS Series in Discrete<br />

Mathematics and Theoretical Computer Science, American<br />

Mathematical Society, 1997, vol. 31, pp. 185–214.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Model of human palm controlled by glove<br />

with micromachined accelerometers<br />

(Model ludzkiej dłoni sterowanej przez rękawicę wyposażoną<br />

w mikromaszynowe czujniki przyspieszenia)<br />

mgr RAFAŁ KOTAS, mgr ZBIGNIEW KULESZA, dr WOJCIECH TYLMAN,<br />

prof. dr hab. ANDRZEJ NAPIERALSKI, mgr PAWEŁ MARCINIAK<br />

Politechnika Łódzka, Katedra Mikroelektroniki i Technik Informatycznych<br />

Human body is a model for imitation for scientists for many<br />

ages. It is so complex and complicated that only in a few<br />

areas machines have a similar functions and construction<br />

to human body. First mention about using prosthesis (artificial<br />

limb) appears to be in antiquity. They were invented<br />

for a special purpose which is replacing human arm or leg.<br />

Unfortunately many people have lost their arm or leg as<br />

a result of accidence or disease. Engineers decided to help<br />

these people. They created many inventions. In the beginning<br />

these prosthesis had only similar shape as the human<br />

hand, arm or palm. Other constructions were too heavy for<br />

real applications. And other were so complex and complicated<br />

that they were too expensive and out of ordinary people’s<br />

reach.<br />

The second field of technology connected with this research<br />

is the field of manipulators. In every century people were<br />

improving tools to make their work and life easier. Because<br />

of these new inventions people could work easier. They were<br />

replaced in many factories by manipulators. These machines<br />

are build as a model of human arm. Their task is to move heavy<br />

and dangerous things with high precision.<br />

The principle of operating<br />

The operation of this device could be divided into four main<br />

stages (Table 1). At first virtual glove measures the gravity acceleration.<br />

This measurement is implemented with the use of<br />

six micromachined accelerometers which are appropriately<br />

spaced on three fingers (thumb, forefinger and middle finger)<br />

and on the top of a palm. Virtual glove is shown in Fig. 2.<br />

Tabl. 1. Block diagram of the operating principle<br />

Tab. 1. Schemat blokowy zasady działania urządzenia<br />

1. 2.<br />

3. 4.<br />

Fig. 1. Demonstrating model of a human palm<br />

Rys. 1. Model demonstracyjny ludzkiej dłoni<br />

The object of this project was to build a model of human<br />

palm consisted of three fingers that could be used in<br />

many applications. It is shown in Fig. 1. The main goal was<br />

to test the possibility of using micromachined accelerometers<br />

to detect the position of human fingers and using „virtual<br />

glove” to control the model. The second goal was to<br />

achieve a precise control system for a specified mechanical<br />

construction [1].<br />

Fig. 2. Virtual glove<br />

Rys. 2. Wirtualna rękawica<br />

Secondly microprocessor system based on ATmega32L<br />

analyses the acceleration signal with the use of novel algorithm.<br />

Next stage of this process is parallel 7 PWM signals<br />

generation. A complex programmable logic device (CPLD)<br />

turned out to be the most appropriate device for this task.<br />

The last stage of operation is a movement of the human palm<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 67


model made in a 1:1 scale – the joints driven by seven servomechanisms.<br />

The model has a motionless wrist, two fingers<br />

and an opposing thumb. Every finger has three joints. An extra<br />

joint is made for the thumb.<br />

The use of micromachined accelerometers<br />

One of the most important problems was to detect positions<br />

of human fingers with high accuracy. The author proposed<br />

to measure static acceleration with the use of three axis accelerometers.<br />

After the literature studies the author choose<br />

the Freescale accelerometer MMA7360LT. It is a surfacemicromachined<br />

integrated-circuit accelerometer. The device<br />

consists of a surface micromachined capacitive sensing cell<br />

(g-cell) and a signal conditioning ASIC contained in a single<br />

package [4]. The most important question was how many<br />

sensors have to be used to unambiguously determine the position<br />

of human fingers. Conducted research shows that only<br />

six accelerometers are essential to achieve this goal. They<br />

are appropriately spaced on the glove. It is shown in Fig. 3.<br />

According to the manufacturer specification of these accelerometers<br />

the author noticed that there is no need to measure<br />

static acceleration in all three axis, but there is one condition<br />

– virtual glove has to be used in vertical position. In this<br />

case author decided to use measurement in one axis for four<br />

accelerometers (on the back of the hand, on the thumb, in<br />

the middle of fore – and middle finger) and in two axis for two<br />

remaining accelerometers.<br />

The used accelerometer is very precise and sensitive but<br />

its principle of operation is very specific. Within the supply range<br />

of 2.2 and 3.6 V, the device operates as a fully calibrated<br />

linear accelerometer. The output voltage of each axis varies<br />

between 0.85 and 2.45 V. The principle of operation is shown<br />

below in Fig. 4.<br />

The output voltage depends on a current position of the<br />

accelerometer. As it is shown in the above diagram the use<br />

of one axis of the accelerometer do not give a strict position.<br />

On account of this the author proposed to measure static acceleration<br />

in two axis at the fingertips. It means that the exact<br />

position of forefinger and middle finger is recognized by a pair<br />

of accelerometers. In the next stage of the signal processing<br />

it is converted into digital value by an internal A/D converter<br />

(ADC) of ATmega32L. On account of the principle of operation<br />

of used accelerometers the novel algorithm was developed.<br />

The particular combination of digital values representing the<br />

positions of two accelerometers (e.g. a pair spaced on the<br />

forefinger) are recognized by the algorithm. And as the result<br />

ATmega32L generates pairs of specific values on Data_Out<br />

bus and Address bus. These pairs are recognized by CPLD<br />

which generates PWM signals and sets the model in appropriate<br />

motion. This solution leads to unstable operation of the<br />

invented device. It will be developed in the future.<br />

Microprocessor system based<br />

on ATmega32L<br />

The electronic system must be able to monitor the changes<br />

in the positions of human fingers in real time. Block diagram<br />

of the system is shown in Fig. 5. The first module is Power supply.<br />

The idea of this project was to invent a fully autonomous<br />

device. Four rechargeable batteries connected in series set<br />

the output voltage at 4.8 V. The author decided to use low-dropout<br />

linear regulator LM1117-3.3V which sets the final output<br />

voltage at 3.3 V. It has current limiting and thermal shutdown<br />

[6]. According to this all of the electronic elements were chosen<br />

to work correctly at 3.3 V.<br />

The second module is the User interface. It consists<br />

of Virtual glove (described in previous paragraph) and Joystick.<br />

A joystick is an input device consisting of a stick that pivots on<br />

a base and reports its direction to the device it is controlling.<br />

Joysticks are often used to control video games, and usually<br />

have one or more push-buttons whose state can also be read<br />

by the system. It allows to control the model of human palm as<br />

a simple manipulator. UP and DOWN directions of joystick are<br />

used to control the forefinger. LEFT and RIGHT directions are<br />

used to control the middle finger. The thumb is controlled by two<br />

push-buttons. This technique is very effective and precise.<br />

Fig. 3. Schematic diagram of virtual glove<br />

Rys. 3. Schemat wirtualnej rękawicy<br />

Fig. 4. Diagram of the accelerometers principle of operation<br />

Rys. 4. Schemat zasady działania czujników przyspieszenia<br />

Fig. 5. Block diagram of an electronic system<br />

Rys. 5. Schemat blokowy układu elektronicznego<br />

68<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


The main module of the electronic system is the microprocessor<br />

system. It is based on ATmega32L. The ATmega32 is<br />

a low-power CMOS 8-bit microcontroller based on the AVR<br />

enhanced RISC architecture.<br />

The ATmega32 features a 10-bit successive approximation<br />

ADC. The ADC converts an analog input voltage to<br />

a 10-bit digital value through successive approximation. The<br />

minimum value represents GND and the maximum value represents<br />

the voltage on the AREF (connected to VCC) pin<br />

minus 1 LSB [3]. In the beginning of the process data from<br />

all eight channels are collected. Only when it is finished the<br />

process continues. The novel algorithm analyses these 10-bit<br />

digital values and sets appropriate values on 5-bit data bus<br />

and 3-bit address bus.<br />

It is read by CPLD – XCR3128XL. The CoolRunner XPLA3<br />

XCR3128XL device is a 3.3 V 128 macrocell CPLD targeted<br />

at power sensitive designs that require leading edge programmable<br />

logic solutions. Pin-to-pin propagation delays are as<br />

fast as 5.5 ns with a maximum system frequency of 175 MHz<br />

[5]. The CPLD was programmed with the use of VHDL (VH-<br />

SIC Hardware Description Language, where VHSIC stands<br />

for Very High Speed Integrated Circuits). It generates 7 parallel<br />

PWM signals.<br />

The last module of the electronic system is a system of 7<br />

servomechanisms. A servomechanism, or servo is an automatic<br />

device that uses error-sensing feedback to correct the<br />

performance of a mechanism. The term correctly applies only<br />

to systems where the feedback or error-correction signals<br />

help control mechanical position or other parameters. An appropriate<br />

PWM signal is used to control the servomechanism.<br />

Examples of PWM signals controlling servomechanisms are<br />

shown in Table 2.<br />

Tabl. 2. Examples of PWM signals controlling servomechanisms<br />

Tab. 2. Przykładowe sygnały PWM sterujące serwomechanizmami<br />

Fig. 6. Block diagram of the programmed internal structure of<br />

XCR3128XL<br />

Rys. 6. Schemat blokowy struktury wewnętrznej XCR3128XL<br />

Conclusions and plans of future research<br />

The virtual glove and the model have been developed to demonstrate<br />

the possibility of imitating human fingers movements<br />

by the model of the palm (artificial limb) with the operating<br />

principle based on accelerometers. This project could be developed<br />

in many areas. The author plans equipping model<br />

with strain gauge at the end of every finger to avoid destroying<br />

grasped objects. The algorithm also needs to be improved.<br />

The goal to achieve is to use the virtual glove not only in vertical<br />

position. It would lead to larger number of acceleration<br />

signals and as a result to more complex microprocessor system.<br />

At the end of the research designed device was tested.<br />

The results of these tests are shown on the example picture<br />

(Fig. 7).<br />

Every model finger is moved by two servomechanisms.<br />

The seventh servomechanism is used to change the position<br />

of the thumb. This option allows the user to set the model of<br />

the human palm in the flat position and also in the position<br />

with an opposing thumb.<br />

PWM signal generation on XCR3128XL<br />

Fig. 7. Example of use of the virtual glove and the model<br />

Rys. 7. Przykład działania wirtualnej rękawicy oraz modelu ludzkiej<br />

dłoni<br />

Generation of 7 PWM signals was implemented on CPLD<br />

XCR3128XL. An internal structure of this device was designed<br />

with the use of VHDL. It is a hardware description language<br />

used in electronic design automation to describe digital and<br />

mixed-signal systems such as field-programmable gate arrays<br />

and integrated circuits [2]. The block diagram of the prgrammed<br />

internal structure of XCR3128XL is shown in Fig. 6.<br />

The structure could be divided into three modules: frequency<br />

divider, address decoder, PWM signal generator.<br />

This invention could find practical application in two fast developing<br />

branches: prosthetics and medical robotics. It could<br />

also be implemented as a manipulator used in dangerous<br />

environment (high temperature, toxic chemicals, explosives,<br />

radiation), sign language translator, computer interface.<br />

Rafał Kotas and Paweł Marciniak are a scholarship holders<br />

of project entitled „Innovative education...” supported by European<br />

Social Fund.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 69


References<br />

[1] Kotas R.: Data glove controlled by a microprocessor system.<br />

MSc thesis, Technical University of Lodz 2009.<br />

[2] Zwoliński M.: Projektowanie układów cyfrowych z wykorzystaniem<br />

języka VHDL. Wydawnictwa Komunikacji i Łączności,<br />

Warszawa 2002.<br />

[3] Manufacturer datasheet: ATMEL, ATmega32L 8-bit Microcontroller<br />

with 32K Bytes In-System Programmable Flash. Strona<br />

internetowa producenta ATMEL: http://www.atmel.com/avr, USA<br />

2006.<br />

[4] Manufacturer datasheet: MMA7360LT: XYZ Axis Accelerometer<br />

±1,5g, ±6g. Strona internetowa producenta Freescale Semiconductor:<br />

http://www.freescale.com, USA 2007.<br />

[5] Manufacturer datasheet: XCR3128XL 128 Macrocell CPLD. Strona<br />

internetowa producenta Xilinx: http://www.xilinx.com/, USA<br />

2006.<br />

[6] Manufacturer datasheet: LM1117 800 mA Low-Dropout Linear<br />

Regulator. Strona internetowa producenta National Semiconductor:<br />

http://www.national.com, USA 2004.<br />

Ogólnopolska konferencja „Bezpieczna firma”<br />

Śląskie Centrum Telemarketingu i Reklamy Marketing Maestry<br />

(znane na rynku jako Marketing Maestry) powstało<br />

w 2001 roku. Od 2006 r. organizuje konferencje poświecone<br />

tematyce informatycznej i marketingowej. Są one sygnowane<br />

nazwami takimi jak ISO-Silesia, ISO-Poland, IT-Medica Silesia<br />

IT-Medica Poland, ERP Silesia, Telekom Silesia, Sec-<br />

Info Silesia, EOD Silesia, Zasoby Ludzkie Silesia, Projektory<br />

Silesia i in. Spółka Marketing Maestry zorganizowała 50 imprez<br />

konferencyjnych, które zgromadziły łącznie ok. 10 tys.<br />

uczestników.<br />

Zapewnienie bezpieczeństwa informacji w firmie to zagadnienie<br />

nie tylko dla informatyków. Musi ono budzić konieczność<br />

ochrony informacji w codziennej pracy. Za bezpieczeństwo<br />

informacji odpowiadają wszyscy pracownicy. Systemy<br />

informatyczne wspomagają realizację większości procesów.<br />

Wiele działań marketingowych korzysta się ze stron WWW,<br />

stosuje mailingi, standardem stała się także internetowa<br />

sprzedaż. Takie platformy komunikacyjne gwarantują łatwy<br />

dostęp do informacji i kontakt z partnerami, ale niosą za sobą<br />

wiele różnych zagrożeń, takich jak nieuzgodniona modyfikacja<br />

i zniszczenie lub kradzież danych.<br />

Z pewnością najpowszechniejszym zagrożeniem są dzisiaj<br />

masowe infekcje komputerów przez instalowanie na komputerach<br />

użytkowników, często nawet niewielkiego oprogramowania,<br />

umożliwiającego przejmowanie kontroli przez osoby<br />

niepowołane. Wg danych amerykańskich służb specjalnych,<br />

w roku 2009, oszacowano, że światowe dochody z działalności<br />

cyberprzestępczej przekroczyły dochody z handlu narkotykami.<br />

Daje to wyobrażenie o dochodach jakimi dysponują<br />

przestępcy. Prowadzi to do możliwości przejęcia kontroli nad<br />

komputerem, co jest najpoważniejszym zagrożeniem. Niesie<br />

ze sobą ryzyko zarówno utraty danych jak i ryzyko utraty reputacji<br />

przedsiębiorstwa.<br />

Realne rozwiązania są dwa. Jedno to zbudowanie systemu<br />

bezpieczeństwa, w którym są wykorzystywane mechanizmy<br />

ochrony zasobów przedsiębiorstwa (Firewall, IPS, Antywirus),<br />

uzupełnione o analizę treści ochraniającą użytkowników<br />

(Antyspam, Filtracja treści). Drugie to cykliczne szkolenia<br />

pracowników w zakresie podstaw bezpiecznego poruszania<br />

się w sieci, jak i procedur związanych z szeroko rozumianą<br />

poufnością.<br />

Istotne jest również bezpieczeństwo przesyłanych informacji.<br />

Jeżeli firma nie dysponuje łączami prywatnymi, w których<br />

można stosować skuteczną kompresję i szyfrowanie<br />

danych, to ważne jest, aby konta pocztowe na serwerze firmowym<br />

miały certyfikaty urzędu certyfikacji. Dodatkowo, skutecznym<br />

sposobem zabezpieczenia przesyłanych informacji<br />

jest wykorzystanie podpisu elektronicznego obsługiwanego<br />

przez popularne programy pocztowe (m.in. Outlook Express,<br />

Thunderbird).<br />

Zdaniem Jacka Leguta z firmy Veracomp SA tworzenie<br />

kopii bezpieczeństwa posiadanych danych jest bardzo ważne,<br />

ale nie jest archiwizacją. Kopie bezpieczeństwa danych<br />

wykonuje się na wypadek konieczności odtworzenia w razie<br />

ich utraty (awaria systemu, umyślne bądź nieświadome wykasowanie,<br />

uszkodzenie fizyczne dysków/serwerów, katastrofy).<br />

Natomiast archiwizacja danych to ich zapisanie w innym miejscu<br />

– np. na tańszych dyskach/macierzach dyskowych – w taki<br />

sposób aby użytkownik, który będzie chciał z nich skorzystać,<br />

nie miał problemów z ich odnalezieniem i dostępem.<br />

Komercyjnych rozwiązań do archiwizacji danych, a w tym<br />

plików, baz danych, poczty elektronicznej jest na rynku wiele.<br />

W przypadku archiwizacji poczty elektronicznej i systemów<br />

plików godnym polecenia jest Enterprise Vault firmy Symantec.<br />

Do archiwizacji samej poczty elektronicznej zdecydowanie<br />

ciekawym rozwiązaniem jest sprzętowe podejście firmy<br />

Artec IT czyli Email Archiving Appliance, a w przypadku archiwizacji<br />

baz danych na uwagę zasługują rozwiązania firmy HP<br />

takie jak Database Archiving Software i HP Medical Archiving<br />

Solution.<br />

Do innych przyczyn utraty danych należy zaliczyć klęski<br />

żywiołowe. Zdecydowana większość przypadków to efekty<br />

klęski powodzi. Komputer powinien być on przesłany jak<br />

najszybciej do specjalistycznego laboratorium. Przyczyn<br />

związanych z utratą danych może być znacznie więcej, a katastrofy<br />

naturalne stanowią wśród nich jedynie 2%. Aż 26%<br />

jest efektem błędów człowieka – nieumyślnego skasowania<br />

danych lub nieostrożnego obchodzenia się z komputerami.<br />

W przypadku awarii sprzętu (56% przyczyn), głównym winowajcą<br />

w miesiącach letnich jest przegrzanie komputerów na<br />

skutek bardzo wysokich temperatur lub ich spalenie podczas<br />

powodujących silne wyładowania elektryczne burz atmosferycznych<br />

i przepięć w gniazdkach. Pozostałe okoliczności to<br />

błędy oprogramowania i wirusy.<br />

Niestety obecnie wielu użytkowników nie stosuje nawet<br />

najprostszego sposobu ochrony swoich danych, jakim jest<br />

tworzenie kopii bezpieczeństwa. Bardzo często okazuje się<br />

także, że kopie bezpieczeństwa nie są aktualizowane. W przypadku<br />

firm kluczową kwestią jest wprowadzenie odpowiedniej<br />

polityki bezpieczeństwa danych, w ramach której powinno<br />

znaleźć się wiele działań uwzględniających cała strukturę organizacji<br />

i przesyłania dokumentów. (cr)<br />

70<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Bezkontaktowy czujnik przemieszczenia w złożonym<br />

układzie pomiarowym charakterystyk elektro-<br />

-termo-mechanicznych stopów z pamięcią kształtu<br />

dr inż. GRZEGORZ KŁAPYTA, mgr inż. MAREK KCIUK<br />

Politechnika Śląska, Wydział Elektryczny, Katedra Mechatroniki, Gliwice<br />

Materiały z pamięcią kształtu SMA (Shape Memory Alloy),<br />

należące do grupy materiałów inteligentnych (Smart), cieszą<br />

się coraz większą popularnością i rosnącą liczbą zastosowań<br />

[1]. Cechą charakterystyczną stopów z pamięcią<br />

kształtu jest szczególna ich wrażliwość na zmiany temperatury<br />

i zachodząca w nich przemiana martenzytyczna [2].,<br />

zmieniająca strukturę krystalograficzną materiału, a wraz<br />

z nią właściwości makroskopowe. Materiały te pod wpływem<br />

wzrostu temperatury i/lub siły mechanicznej potrafią zmieniać<br />

swój kształt. Dzięki temu można je stosować zarówno<br />

do konstrukcji czujników temperatury, jak i do termicznie<br />

sterowanych elementów wykonawczych (mogą także pełnić<br />

obydwie te funkcje równocześnie). Na uwagę zasługuje niezwykle<br />

duży współczynnik uzyskiwanej siły elementu napędowego<br />

na jednostkę masy, co ma szczególne znaczenie<br />

w konstrukcji układów napędowych, np. w robotyce mobilnej.<br />

Materiały SMA są już dobrze poznane i opisane, jednak<br />

z uwagi na duże nieliniowości oraz występowanie pętli histerezy<br />

w charakterystyce przemiany martenzytycznej, dobór<br />

właściwości materiałów SMA jest problemem trudnym [3].<br />

Przemiana martenzytyczna zaczyna zachodzić po przekroczeniu<br />

pewnej temperatury progowej (A s<br />

), zależnej od naprężenia<br />

mechanicznego. W trakcie przemiany stosunek ilości<br />

cząsteczek w fazie martenzytu do ilości cząsteczek w fazie<br />

austenitu maleje wraz ze wzrostem temperatury i po przekroczeniu<br />

kolejnej granicznej temperatury (A f<br />

), praktycznie cały<br />

materiał przechodzi w fazę austenitu. Przemiana odwrotna<br />

zachodzi przy innych progowych wartościach temperatur (M s<br />

i M f<br />

), tworząc na charakterystyce termomechanicznej (rys. 1)<br />

charakterystyczną pętlę histerezy. Należy zwrócić uwagę, że<br />

podczas przemiany równocześnie ulegają zmianie parametry<br />

mechaniczne (długość, średnica, twardość), jak i elektryczne<br />

(rezystywność) materiału. Jest to jedna z podstawowych<br />

trudności ograniczających zakres zastosowań SMA −<br />

ze względu na olbrzymie trudności, związane z precyzyjnym<br />

sterowaniem przebiegiem przemiany. W literaturze światowej<br />

przedstawiono wiele modeli matematycznych, opisujących<br />

przemiany zachodzące w stopach SMA. Modele te skupiają<br />

się na jak najdokładniejszym odzwierciedleniu zjawisk<br />

fizycznych, a w celu ich weryfikacji, wymagane jest posiadanie<br />

wielu danych materiałowych.<br />

W jednej z grup układów wykonawczych, sterowanych za<br />

pomocą elementów SMA, są stosowane wspomniane modele<br />

matematyczne. Układy te, ze względu na stopień skomplikowania<br />

modeli, wymagają dużych mocy obliczeniowych.<br />

Istnieją także prostsze metody sterowania – np. sterowanie<br />

dwupozycyjne. Nie wymaga ono skomplikowanych obliczeń,<br />

a jedynie oszacowania maksymalnego odkształcenia przy<br />

określonej sile obciążenia mechanicznego. Sterowanie w tej<br />

metodzie odbywa się na zasadzie załącz-wyłącz, ale zapewnia<br />

ono tylko dwustanową kontrolę nad aktuatorem. Rozwinięciem<br />

tej metody jest sterowanie segmentowe (Segmented<br />

Binary Control – SBC) opisane w [4, 5]. Osobną grupę układów<br />

sterujących stanowią układy wykorzystujące logikę rozmytą<br />

oraz sieci neuronowe [6]. Nie istnieje, jak dotąd, prosta<br />

i uniwersalna metoda dokładnego pozycjonowania elementu<br />

SMA w całym zakresie ich ruchu.<br />

Najważniejszą charakterystyką, opisującą zachowanie<br />

aktuatorów SMA, jest charakterystyka termomechaniczna,<br />

przedstawiająca stopień przejścia z fazy austenitu do fazy martenzytu<br />

(ξ ) w funkcji temperatury (rys. 1). Na podstawie tej charakterystyki<br />

wyznacza się także charakterystyki odkształcenia<br />

w funkcji temperatury przy znanym obciążeniu mechanicznym.<br />

Rys. 1. Charakterystyka termomechaniczna SMA ξ = f (T) [7]<br />

Fig. 1. Thermomechanical characteristic of SMA ξ = f(T) [7]<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 71


Rys. 2. Wpływ naprężenia wewnętrznego na parametry przemiany<br />

martenzytycznej<br />

Fig. 2. Influence of internal stress on parameters of martensitic<br />

phase trabsformation<br />

W przypadku, gdy do nagrzewania stosuje się prąd elektryczny<br />

przepływający przez aktuator, można również wyznaczyć<br />

charakterystykę elektromechaniczną – odkształcenie<br />

w funkcji prądu. Wszystkie te charakterystyki są nieliniowe<br />

i zależne od dodatkowego parametru – naprężenia wewnętrznego<br />

w materiale. Zwiększanie naprężenia wewnętrznego,<br />

wynikające np. z mechanicznego obciążenia, powoduje „przesunięcie”<br />

się charakterystyki w stronę temperatur wyższych.<br />

Zjawisko to zostało przedstawione na rys. 2. W pracy [7] zauważono<br />

także, że wraz ze wzrostem obciążenia może zmieniać<br />

się szerokość pętli histerezy na charakterystyce elektromechanicznej.<br />

Stanowisko pomiarowe<br />

Podstawowym założeniem stanowiska pomiarowego realizowanego<br />

przez autorów było umożliwienie dokonania serii<br />

zautomatyzowanych pomiarów cięgien SMA o różnej długości<br />

i grubości, w celu poznania ich właściwości<br />

i parametrów. Przyjęto, że pomiary wszystkich<br />

wielkości powinny odbywać się w tej<br />

samej chwili czasowej, zarówno w trakcie<br />

pomiaru statycznego, jak i dynamicznego.<br />

Pomiar statyczny jest rozumiany jako pomiar<br />

parametrów próbki po osiągnięciu ustalonego<br />

stanu termicznego. Pomiary dynamiczne<br />

miały na celu wyznaczenie charakteru<br />

zmian mierzonych wielkości po skokowych<br />

zmianach wymuszenia prądowego.<br />

Podstawowe założenia projektowe do<br />

konstrukcji stanowiska:<br />

– grzanie oporowe cięgna z wymuszeniem<br />

prądowym,<br />

– nastawialne obciążenie mechaniczne<br />

o stałej sile (grawitacyjne),<br />

– pomiar temperatury przy użyciu kamery<br />

termowizyjnej,<br />

– optyczny pomiar odkształcenia,<br />

– jednoczesny pomiar prądu, napięcia, odkształcenia i temperatury,<br />

– maksymalna długość cięgna: 900 mm,<br />

– minimalne obciążenie mechaniczne: 75 g,<br />

– maksymalne odkształcenie: 80 mm,<br />

– automatyczny pomiar i akwizycja wyników.<br />

Konstrukcja stanowiska (rys. 3) umożliwia zamocowanie<br />

cięgien SMA (1) pomiędzy sztywną poprzeczką (2) a ruchomym<br />

tłokiem (3). Grawitacyjne obciążenie mechaniczne może<br />

być zwiększane za pomocą dodatkowych obciążników i zapewnia<br />

stałą siłę obciążenia w trakcie sesji pomiarowej. Podstawa<br />

(4), rama (5) oraz poprzeczka (2) wykonane są z profili<br />

aluminiowych. Kamera termowizyjna (6) jest zamocowana do<br />

podstawy na stałej wysokości. Mocowanie kamery umożliwia<br />

jednak przesuwanie kamery w poziomie w celu regulacji ostrości<br />

obrazu. Układ elektryczny jest odseparowany galwanicznie<br />

od aluminiowej konstrukcji stanowiska za pomocą dystansów<br />

wykonanych z przezroczystej płyty PCV. W celu poprawienia<br />

wyników pomiaru termograficznego, część stanowiska została<br />

osłonięta czarnym rękawem wykonanym z tektury. W oknie na<br />

rys. 3a widać także czujnik miernika odkształcenia, znajdujący<br />

się w dolnej części stanowiska.<br />

Układ wymuszający<br />

Przyrost temperatury przewodnika powyżej temperatury otoczenia,<br />

zgodnie z prawem Joule’a – Lenza, jest w przybliżeniu<br />

proporcjonalny do kwadratu prądu płynącego w tym przewodniku.<br />

Tak więc, aby drut wykonany z SMA miał określoną stałą<br />

temperaturę, należy zapewnić stałą wartość natężenia prądu,<br />

przepływającego przez dany przewodnik. W tym celu − jako<br />

źródło prądowe − został wybrany zasilacz PSH-3620A o wydajności<br />

36 V i 20 A. Zasilacz ten umożliwia pracę w trybie<br />

stabilizacji prądu lub napięcia.<br />

Pomiar wielkości elektrycznych<br />

Podczas badania charakterystyk statycznych, do jednoczesnego<br />

pomiaru wielkości elektrycznych takich jak: prąd płynący<br />

przez aktuator, napięcie na aktuatorze oraz napięcie wyjściowe<br />

z optycznego przetwornika odkształcenia, użyto trzech<br />

Rys. 3. Widok stanowiska pomiarowego: a) widok rzeczywisty stanowiska, b) model<br />

konstrukcyjny<br />

Fig. 3. View of mechanical construction of laboratory stand: a) Photo of real view,<br />

b) Model of laboratory stand<br />

72<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


multimetrów laboratoryjnych Rigol DM3052. Do akwizycji danych<br />

i generacji sygnałów sterujących użyto komputera PC.<br />

Elementy układu pomiarowego połączone są poprzez sieć<br />

GPIB w standardzie (IEEE 488.2).<br />

Akwizycja danych pomiarowych w badaniach dynamicznych<br />

jest realizowana za pomocą oscyloskopu cyfrowego<br />

Tektronix MSO 2024, posiadającego cztery odseparowane<br />

kanały pomiarowe oraz wbudowany filtr cyfrowy.<br />

Pomiar temperatury<br />

Rys. 4. Obraz cięgna SMA z kamery termowizyjnej<br />

Fig. 4. Thermovision view of SMA wire<br />

miaru temperatury cięgien o niewielkiej średnicy (50…500<br />

µm) zastosowano soczewkę makroskopową closeup 1x<br />

o rozdzielczości powierzchniowej 25 µm. Sygnał pomiarowy<br />

jest przesyłany do komputera za pomocą łącza sieciowego<br />

Do pomiaru temperatury użyto kamery termowizyjnej Flir A325<br />

o częstotliwości rejestracji obrazów do 60 Hz. Rozdzielczość<br />

kamery wynosi 320 × 240 punktów. W celu umożliwienia poo<br />

szybkości transmisji do 1 GB/s (giga Ethernet). Dodatkowo,<br />

realizowany jest ciągły pomiar temperatury otoczenia za<br />

pomocą scalonego termometru cyfrowego DS18B20. Wynik<br />

pomiaru wyświetlany jest na wyświetlaczu. Nie jest on jednak<br />

przekazywany do programu sterującego. Widok okna pomiarowego<br />

z kamery termowizyjnej z soczewką makroskopową,<br />

przedstawiający obraz rozgrzanego cięgna SMA o średnicy<br />

d = 308 µm, pokazano na rys. 4.<br />

Pomiar odkształcenia<br />

Tor pomiarowy miernika odkształcenia składa się z: 1 – czujnika<br />

optycznego (przekształcającego ruch tłoka na ciąg impulsów),<br />

2 – licznika impulsów, 3 – generatora sygnału PWM (wypełnienie<br />

sygnału PWM jest proporcjonalne do liczby zliczonych impulsów),<br />

4 – filtru dolnoprzepustowego oraz 5 – bufora kondycjonującego<br />

sygnał wyjściowy. Amplituda napięcia wyjściowego<br />

przetwornika mierzona jest za pomocą multimetru. Do pomiaru<br />

ruchu służy scalony optyczny czujnik H9720 oraz taśma pomiarowa<br />

z podziałką o wyznaczonej rozdzielczości, wynoszącej<br />

0,085 mm. Czujnik posiada dwa tory optyczne przesunięte<br />

geometrycznie względem siebie. W wyniku tego, sygnały na<br />

dwóch wyjściach (A i B) są względem siebie przesunięte w fazie.<br />

Detekcję kierunku ruchu uzyskuje się na podstawie pomiaru<br />

różnicy faz tych sygnałów. Rolę licznika impulsów oraz<br />

generatora sygnału PWM spełnia ośmiobitowy mikrokontroler<br />

ATmega88, należący do rodziny ośmiobitowych mikrokontrolerów<br />

AVR. Mikrokontroler zlicza impulsy tylko z jednego wejścia<br />

− drugie wejście służy do ustalenia kierunku ruchu. Zliczenie<br />

impulsu odbywa się poprzez detekcję (dowolnego) zbocza,<br />

za pomocą przerwania zewnętrznego INT. Po wykryciu zbocza<br />

sprawdzany jest stan obydwu wejść, jeżeli są one w fazie,<br />

następuje dekrementacja licznika, w przeciwnym przypadku<br />

następuje jego inkrementacja. Schemat blokowy, prezentujący<br />

zasadę działania przetwornika „odkształcenie-napięcie” jest<br />

przedstawiony na rys. 5, natomiast ideę zliczania impulsów<br />

z uwzględnieniem kierunku ruchu przedstawia rys. 6.<br />

a)<br />

b)<br />

RESISTOR<br />

LENS<br />

LED<br />

EMITER SECTION CODE<br />

WHEEL<br />

PHOTO<br />

DIODES<br />

COM<br />

PARA<br />

A TORS<br />

+<br />

A -<br />

B<br />

+<br />

B -<br />

SIGNAL<br />

PROCESSING<br />

CIRCUITRY<br />

V CC<br />

CH A<br />

2<br />

CH B<br />

4<br />

1<br />

DETECTOR SECTION<br />

3<br />

GND<br />

S<br />

In<br />

Dir<br />

WyB Dir<br />

WyA In<br />

L<br />

G<br />

In<br />

In<br />

B<br />

A<br />

Strobe B<br />

L<br />

A<br />

Strobe<br />

K<br />

Mikrokontroler<br />

F<br />

Wy<br />

S - czujnik<br />

L - licznik<br />

B - bufor<br />

G - generator<br />

K - komparator<br />

F - Filtr<br />

B<br />

Strobe<br />

In - wejście zliczające licznika<br />

oraz danych bufora<br />

Dir - wejście kierunku zliczania<br />

Strobe - taktowanie bufora<br />

Rys. 5. Schemat blokowy czujnika Q9720 [1] (a) oraz schemat blokowy przetwornika odkształcenie-napięcie (b)<br />

Fig. 5. Block diagram of: Q9720 sensor (a) and displacement-voltage converter (b)<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 73


In<br />

Dir<br />

Wy<br />

Rys. 6. Zasada działania dwukierunkowego licznika impulsów<br />

Fig. 6. Principle of operation of bidirectional impulse counter<br />

W celu wyznaczenia podziałki na tasiemce pomiarowej<br />

dokonano serii pomiarów przemieszczeń o znanych wartościach.<br />

Na podstawie zliczonych impulsów wyliczono podziałkę.<br />

Zasada działania układu pomiaru odkształcenia<br />

Zliczona liczba impulsów (L S<br />

) na wyjściu licznika (L), znajdującego<br />

się w torze pomiarowym, jest porównywana<br />

w komparatorze (K) z liczbą impulsów (L G<br />

) zliczoną przez<br />

licznik (L), znajdujący się za generatorem impulsów (G).<br />

Licznik impulsów generatora pracuje tylko w jednym kierunku<br />

– przyrostowym. Czas potrzebny do przepełnienia<br />

tego licznika jest równy okresowi generowanego sygnału<br />

PWM. Częstotliwość sygnału PWM wynosi 3,5 kHz. Komparator<br />

porównuje obydwie wielkości. Dopóki: L G<br />

< L S<br />

, na<br />

wyjściu komparatora występuje stan wysoki, przez pozostałą<br />

część okresu − na wyjściu występuje stan niski. Czujnik<br />

oraz przerwanie zewnętrzne mikrokontrolera działają<br />

asynchronicznie. Wyjście licznika impulsów (L S<br />

) jest buforowane<br />

(B), aby w trakcie generowania sygnału PWM, nie<br />

dochodziło do błędów, wynikających z nieprzewidzianych<br />

zmian wartości zliczonej. Zmiana wartości liczby impulsów<br />

w rejestrze buforującym jest możliwa tylko w momencie<br />

przepełnienia licznika zliczającego impulsy z generatora.<br />

Sygnałem sterującym tą operacją jest sygnał Strobe. Jest<br />

to rozwiązanie zaimplementowane sprzętowo w timerach<br />

wbudowanych jako urządzenie peryferyjne w strukturze<br />

mikrokontrolera AVR. Zadaniem filtru dolnoprzepustowego<br />

RC na wyjściu generatora PWM jest przekształcenie<br />

cyfrowej postaci sygnału PWM do postaci analogowej<br />

– uśrednionej. Takie rozwiązanie jest popularne w aplikacjach<br />

czujników z wyjściem PWM, gdy sygnał ten jest mierzony<br />

przyrządem z wejściem analogowym. Bufor wyjściowy<br />

w postaci wzmacniacza operacyjnego o wzmocnieniu<br />

k = 1 kondycjonuje sygnał oraz zapewnia stałą, wysoką<br />

impedancję układu obciążającego filtr. Częstotliwość graniczna<br />

filtru wynosi f g<br />

= 384 Hz.<br />

Podsumowanie<br />

t<br />

Przedstawiony układ pomiaru przemieszczenia mierzy odkształcenie<br />

cięgna SMA w jednostkach bezwzględnych.<br />

W celu wyznaczenia względnego odkształcenia, wartości<br />

zmierzone należy odnieść do długości badanego aktuatora<br />

w fazie austenitu przy minimalnym obciążeniu mechanicznym,<br />

gdyż w takich warunkach osiąga on swój kształt<br />

niezdeformowany. Deformacja (zmiana długości) może<br />

nastąpić w wyniku obciążenia mechanicznego. Duża<br />

siła deformacji może doprowadzić do zajścia przemiany<br />

martenzytycznej w stałej temperaturze. W takim przypadku<br />

zostanie wydzielone ukryte ciepło przemiany [2].<br />

Zjawisko to nosi nazwę superelastyczności lub pseudosprężystości.<br />

Wyniki otrzymywane na stanowisku pomiarowym są<br />

w dużej mierze zgodne z charakterystykami prezentowanymi<br />

w literaturze. Można więc uznać, że stanowisko<br />

do wyznaczania charakterystyk elektro-termo-mechanicznych<br />

oraz optoelektroniczny układ do pomiaru odkształcenia<br />

zostały zaprojektowane i wykonane poprawnie.<br />

Badania przeprowadzone na prezentowanym stanowisku<br />

posłużą do opracowania modelu matematycznego aktuatora<br />

SMA, opisującego odkształcenie w funkcji prądu nagrzewającego.<br />

Opracowywany model powinien mieć możliwość<br />

implementacji w prostych układach sterowania o niewielkiej<br />

mocy obliczeniowej, co wymaga maksymalnego uproszczenia.<br />

Planowany model uwzględniać będzie jedynie najważniejsze<br />

właściwości elementu SMA, przy zachowaniu maksymalnej<br />

zgodności otrzymanych symulacji z przebiegiem<br />

rzeczywistych zjawisk. Taki uproszczony model pozwoli na<br />

upowszechnienie tego typu napędów m.in. w zastosowaniach<br />

w robotyce.<br />

Praca naukowa finansowana ze środków na naukę w latach<br />

2009–<strong>2010</strong> jako projekt badawczy nr N N510 353036.<br />

Literatura<br />

[1] Nota katalogowa czujnika H9720 http://www.avagotech.com/<br />

docs/AV02-0511EN.<br />

[2] Pieczyska E.: Termomechaniczne aspekty przemiany fazowej<br />

w stopie NiTi z pamięcią kształtu indukowanej naprężeniem.<br />

Pomiary Automatyka Kontrola, PAK vol. 55, nr 11/2009,<br />

ss. 958–961.<br />

[3] Kciuk M., Kłapyta G.: Koncepcja stanowiska pomiarowego do<br />

wyznaczania charakterystyk elektro-termo-mechanicznych<br />

stopów z pamięcią kształtu (SMA). Materiały XIII Sympozjum<br />

„Podstawowe Problemy Energoelektroniki, Elektromechaniki<br />

i Mechatroniki” PPEEm 2009 pod patronatem Komitetu Elektrotechniki<br />

PAN, Wisła 14-17.12.2009, ss. 242–247.<br />

[4] Selden B., Cho K.-J., Assada H.: Segmented binary control of<br />

SMA actuator systems using the Peltier effect. IEEE Proceedings<br />

2004. International Conference on Robotics & Automation, New<br />

Orleans, LA.<br />

[5] Cho K.-J., Assada H. H.: Multi-axis SMA actuator array for driving<br />

anthropomorphic robot hand. International Conference on Robotics<br />

& Automation, Spain, April, 2005.<br />

[6] Yam Y., Lei K.-F., Baranyi P.: Control of a SMA actuated artificial<br />

face via Neuro-fuzzy techniques. IEEE International Fuzzy Systems<br />

Conference, 2001.<br />

[7] Kciuk M.: Stanowisko pomiarowe do wyznaczania charakterystyk<br />

elektro-termo-mechanicznych aktuatorów SMA. XI<br />

Międzynarodowe Warsztaty Doktoranckie OWD’2009, Wisła<br />

17–20 października 2009, ss. 251–254.<br />

74<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Poprawa zależności poziomu listków bocznych<br />

od współczynnika kompresji dla sygnałów złożonych<br />

z małą bazą<br />

mgr inż. MARIUSZ ŁUSZCZYK, Przemysłowy <strong>Instytut</strong> Telekomunikacji S.A., Warszawa<br />

W radiolokacji stosowane są połączone techniki syntezy sygnałów<br />

złożonych oraz kompresji impulsów, które pozwalają<br />

na osiągnięcie wysokiej rozdzielczości odległościowej, bez<br />

konieczności skracania czasu trwania impulsu i tym samym<br />

redukcji jego potencjału energetycznego. Struktura czasowoczęstotliwościowa<br />

sygnału sondującego, która jest pochodną<br />

zastosowanej modulacji wewnątrzimpulsowej, określa podstawowe<br />

parametry i właściwości radaru, takie jak potencjalną<br />

rozdzielczość odległościową, odporność na zakłócenia bierne<br />

[1] oraz wrażliwość sygnału echa na efekt Dopplera [6].<br />

Podstawowym rodzajem modulacji stosowanym w złożonych<br />

sygnałach radarowych jest liniowa modulacja częstotliwości<br />

LFM. Właściwości sygnałów złożonych z modulacją typu LFM<br />

są opisane w [2, 5, 6].<br />

Najczęściej stosowaną metodą redukcji poziomu listków<br />

bocznych jest przetwarzanie sygnału z wykorzystaniem okna<br />

wagowego [1, 2, 6]. Sposób doboru funkcji oraz ich wpływ na<br />

postać sygnału po kompresji opisany jest w [1, 6]. Zastosowanie<br />

metody redukcji poziomów listków bocznych za pomocą<br />

filtru wagowego wprowadza niedopasowanie odpowiedzi<br />

impulsowej filtru do sygnału echa. Zastąpienie filtru dopasowanego,<br />

który maksymalizuje stosunek sygnału do szumu<br />

SNR (ang. – Signal to Nosie Ratio) filtrem kompresji skutkuje<br />

pogorszeniem współczynnika kompresji K s<br />

, który definiowany<br />

jest jako iloraz maksymalnej wartości chwilowego stosunku<br />

mocy sygnału do szumu na wyjściu filtru do stosunku mocy<br />

sygnału do szumu na wejściu [3]:<br />

( )<br />

o( )<br />

( ) i<br />

max SNR t0<br />

Ks<br />

=<br />

(1)<br />

SNR<br />

gdzie: t 0<br />

– czas, kiedy amplituda sygnału na wyjściu filtru dopasowanego<br />

osiąga wartość maksymalną.<br />

Zastosowanie obróbki wagowej powoduje pogorszenie<br />

stosunku sygnału do szumu – SNR o<br />

na wyjściu filtru kompresji.<br />

Wielkość straty L m<br />

uzależniona jest od funkcji wagowej<br />

i można ją wyznaczyć z zależności [5]:<br />

i<br />

⎡ ⎤<br />

⎢ ( )<br />

( )<br />

∫w<br />

t dt ⎥<br />

SNR<br />

ow<br />

⎢⎣<br />

0 ⎥⎦<br />

Lm<br />

= =<br />

t<br />

(2)<br />

i<br />

( SNR)<br />

om<br />

2<br />

t w t dt<br />

gdzie: SNR ow<br />

– stosunek sygnału do szumu na wyjściu filtru<br />

kompresji z oknem wagowym, SNR om<br />

– stosunek sygnału do<br />

szumu na wyjściu filtru dopasowanego, w (t) – funkcja wagowa.<br />

t<br />

∫<br />

i<br />

0<br />

( )<br />

2<br />

Pogorszenie wyjściowego stosunku sygnału do szumu<br />

SNR ow<br />

powodowane zastosowaniem funkcji wagowej Hamminga<br />

wynosi 1,36 dB.<br />

Zastosowanie sygnałów z nieliniową modulacją częstotliwości<br />

(ang. – Non-Linear Frequency Modulation) pozwala na<br />

obniżenie czasowych listków bocznych bez konieczności stosowania<br />

filtru wagowego, a tym samym pogorszenia współczynnika<br />

kompresji.<br />

W artykule wykazano cechę sygnałów z nieliniową modulacją<br />

częstotliwości, które w porównaniu do sygnałów LFM wykazują<br />

lepsze właściwości w przypadku kompresji sygnałów<br />

o małych bazach (mniejszych od 100). Poprawa parametrów<br />

dotyczy zarówno rozdzielczości odległościowej jak i poziomu<br />

czasowych listków bocznych.<br />

Algorytm syntezy sygnału NLFM<br />

Funkcja autokorelacji sygnału radarowego stanowi odwrotną<br />

transformatę Fouriera funkcji widma mocy tego sygnału. Dlatego<br />

też postać funkcji autokorelacji (np. w aspekcie poziomu czasowych<br />

listków bocznych) może być kształtowana poprzez postać<br />

widma mocy sygnału. Jedną z metod kształtowania widma<br />

sygnału jest zmiana amplitudy sygnału poprzez zastosowanie<br />

okna wagowego. Kształtowanie widma sygnału w procesie ważenia<br />

realizowanego w dziadzinie czasu. Ze względów energetycznych<br />

(związanych z sygnałem nadawanym) przetwarzanie<br />

z wykorzystaniem okna wagowego możliwe jest tylko po stronie<br />

odbiorczej, co prowadzi do niedopasowania w filtrze kompresji<br />

i straty stosunku sygnału do szumu. Alternatywnym sposobem<br />

kształtowania widma sygnału pozwalającym na unikniecie strat<br />

SNR wynikających niedopasowania jest wprowadzenie nieliniowości<br />

w charakterystyce modulacyjnej sygnału radarowego.<br />

Algorytm syntezy sygnału NLFM z wykorzystaniem metody stacjonarnej<br />

fazy przedstawiony jest poniżej.<br />

Zespoloną obwiednię sygnału złożonego przedstawia zależność:<br />

s t = a t exp jϕ<br />

t<br />

(3)<br />

( ) ( ) [ ( )]<br />

gdzie: a(t) – funkcja modulacji amplitudy, φ(t) – funkcja fazowa<br />

sygnału złożonego.<br />

W rozważanym przypadku sygnał występuje w postaci impulsu<br />

prostokątnego o czasie trwania T :<br />

⎧ t 1<br />

⎪1<br />

≤<br />

⎛ t ⎞ T 2<br />

rect ⎜ ⎟ = ⎨<br />

(4)<br />

⎝T⎠<br />

⎪<br />

t 1<br />

0 ><br />

⎪⎩<br />

T 2<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 75


Transformata Fouriera sygnału s(t) przyjmuje postać:<br />

1<br />

S( f ) = S( f ) exp[ jΦ<br />

( f )] = a( t) exp[ j{ − ωt<br />

+ ϕ( t)<br />

}]<br />

dt (5)<br />

2π<br />

Prezentowana metoda syntezy bazuje na przybliżonym obliczeniu<br />

całek szybkooscylujących funkcji, tzw. metodzie<br />

stacjonarnej fazy. Zakłada się, że funkcje a(t) oraz φ(t) są<br />

funkcjami wolnozmiennymi, zaś funkcja exp(.) jest funkcją<br />

szybkooscylującą. Dla czasu t k<br />

, który jest punktem stacjonarnej<br />

fazy prawdziwa jest zależność:<br />

f = 1<br />

ϕ<br />

(6)<br />

2π ′<br />

Charakter funkcji a(t) oraz φ(t) w pobliżu punktu stacjonarnego<br />

prowadzi do następującej zależności:<br />

2<br />

2 a<br />

( )<br />

( t )<br />

U f<br />

(7)<br />

k<br />

≈ 2π ϕ ′<br />

t<br />

W przypadku sygnału NLFM funkcja a 2 (t) pozostaje stała,<br />

zaś widmo kształtowane jest poprzez funkcję fazową │φ˝(t k<br />

)│.<br />

Funkcja widma U(f ) przybliżona jest funkcją okna V(f ), która<br />

jest definiowana na przedziale:<br />

− B 2 ≤ f ≤ B 2<br />

(8)<br />

gdzie: B – widmo projektowanego sygnału.<br />

W celu wyznaczenia charakterystyki fazowej należy z zalezności<br />

(7) wynaczyć widmo fazowe projektowanego sygnału<br />

jednocześnie wykorzystując funkcję aproksymującą V(f ) :<br />

2<br />

V f<br />

Φ ′′ ( f ) = 2π<br />

(9)<br />

2<br />

a ( t)<br />

Pierwsza pochodna widma fazowego wyznaczana jest z całki<br />

w przedziale (8):<br />

Φ′<br />

f = Φ′′<br />

ζ d<br />

(10)<br />

Opóźnienie grupowe poszukiwanej funkcji wyznaczane jest<br />

z zależności:<br />

T( f ) = − 1<br />

Φ ( f )<br />

(11)<br />

2π ′<br />

Funkcja modulacyjna jest odwrotnością opóźnienia grupowego<br />

T(f):<br />

−1<br />

f t = T f , (12)<br />

zaś funkcja fazowa projektowanego sygnału wyznaczana jest<br />

z całki funkcji modulacyjne:<br />

ϕ t = 2π<br />

f ζ d<br />

(13)<br />

Przedstawiony sposób wyznaczania postaci czasowej sygnału<br />

o zadanej postaci widma amplitudowego można usystematyzować<br />

w postaci algorytmu:<br />

76<br />

∞<br />

∫<br />

−∞<br />

( )<br />

k<br />

t k<br />

f<br />

( )<br />

k<br />

k<br />

( )<br />

( ) ∫ ( ) ζ<br />

−B<br />

2<br />

( ) ( )<br />

( ) ∫ ( ) ζ<br />

t<br />

0<br />

− określenie czasu trwania sygnału impulsowego T oraz dewiacji<br />

sygnału B ,<br />

− wybór funkcji okna V(f ), której kształ przybliża postać widma<br />

amplitudowego projektowanego sygnału w zakresie<br />

częstotliwości 〈–B /2, B /2〉,<br />

− wybór wartości parametru funkcji okna V(f ) (w przypadku<br />

wyboru okna parametrycznego),<br />

− wyznaczenie pierwszej pochodnej funkcji widma fazowego<br />

według zależności (10),<br />

− wyznaczenie opóźnienia grupowego T(f ) według zależności<br />

(11),<br />

− wyznaczenie funkcji odwrotnej do opóźnienia grupowego<br />

(funkcja modulacyjna – f(t )),<br />

− wyznaczenie funkcji fazowej projektowanego sygnału według<br />

zależności (10),<br />

− ostatnik krokiem algorytmu jest synteza sygnału zespolonego<br />

s(t ) według zależności (3).<br />

Przedstawiony algorytm pozwala syntetyzować sygnał<br />

NLFM o zadanych parametrach czasowych i częstotliwościowych.<br />

Wyniki symulacji<br />

Do syntezy sygnału impulsowego NLFM o obwiedni prostokątnej<br />

i postaci widma przybliżonej kwadratem funkcji okna<br />

wagowego zostało wykorzystane okno Kaisera opisywane<br />

zależnością:<br />

⎧ ⎛<br />

2 ⎞<br />

⎪ ⎜ ⎛ 2n<br />

⎞<br />

I<br />

⎟<br />

0<br />

πα 1−<br />

⎪ ⎜<br />

⎜ −1⎟<br />

⎟<br />

w = ⎨ ⎝<br />

⎝ M ⎠<br />

Kn ⎠<br />

, (14)<br />

0 ≤ n ≤ M<br />

⎪ I0( πα)<br />

⎪<br />

⎩ 0<br />

n > M<br />

gdzie: I 0<br />

– zmodyfikowana funkcja Bessela pierwszego rodzaju,<br />

zerowego rzędu, α – współczynnik określający poziom<br />

listków bocznych w dziedzinie częstotliwości, K = α<br />

– parametr okna Kaisera, M – liczba całkowita określająca<br />

długość okna.<br />

Wykorzystując algorytm syntezy sygnału NLFM przedstawiony<br />

w drugiej części niniejszego artykułu dokonano symulacji<br />

dla zadanych parametrów sygnałów radarowych:<br />

− czas trwaniania impulsu T = 20 µs,<br />

− szerokość widma sygnału z modulacją częstotliwości<br />

B = 0,75 MHz.<br />

Przyjęte wartości czasu trwania impulsu oraz dewiacji sygnału<br />

z wewnątrzimpulsową modulacja pozwalają na zakwalifikowanie<br />

sygnału złożonego jako sygnału o małej bazie. Widmo<br />

projektowanego sygnału aproksymowane jest kwadratem<br />

modułu funkcji okna Kaisera z arbitralnie dobranym parametrem<br />

K = 1 [7].<br />

Przy założeniu prostokątnej obwiedni sygnału impulsowego<br />

drugą pochodną widma fazowego syntetyzowanego<br />

sygnału Φ˝(f ) wyznacza się według zależności (9) z wykorzystaniem<br />

funkcji Kaisera │V(f )│. Wykorzystując zależność (10)<br />

wyznacza się wartość pierwszej pochodnej widma fazowego<br />

syntetyzowanego sygnału Φ´(f ).<br />

Obliczenie całki realizowane jest z wykorzystaniem metod<br />

numerycznych dostępnych w oprogramowaniu Matlab.<br />

Funkcja modulacyjna f (t ) wyznaczona jest jako funkcja od-<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Rys. 1. Aproksymacja widma syntetyzowanego sygnału za pomocą<br />

funkcji Kaisera<br />

Fig. 1. Amplitude spectrum of NLFM synthesised signal with Kaiser<br />

window<br />

wrotna opóźniania grupowego (12). Charakterystyki sygnału<br />

NLFM – modulacyjna oraz fazowa – przedstawione są na<br />

rys. 3a i 3b.<br />

Podstawowym narzędziem oceny jakości sygnału złożonego<br />

jest radiolokacyjna funkcja nieoznaczoności χ(τ, f D<br />

).<br />

Funkcję tę interpretuje się jako dwuwymiarową funkcję autokorelacji,<br />

gdzie zmiennymi są opóźnienie sygnału τ oraz<br />

przesunięcie Dopplerowskie f D<br />

. Funkcja autokorelacji jest<br />

szczególnym przypadkiem funkcji nieoznaczoności zachodzącym,<br />

gdy składowa Dopplerowska nie występuje. Funkcje<br />

autokorelacji sygnału LFM i NLFM są przedstawione na<br />

rys. 4a i 4b. Cechą wspólną obu sygnałów jest mała wartość<br />

bazy, która wynosi 15. Przedstawione charakterystyki pozwalają<br />

na wyznaczenie współczynnika kompresji K s<br />

, szerokość<br />

listka głównego na poziomie -4 dB (rozróżnialność odległościowa)<br />

oraz poziom czasowych pierwszych i dalszych listków<br />

bocznych.<br />

Badania symulacyjne zostały przeprowadzone dla szeregu<br />

sygnałów LFM i NLFM. Ocenie podlegały dwa podstawowe<br />

parametry, tj. współczynnik kompresji K s<br />

oraz poziom pierwszych<br />

czasowych listków bocznych PSL (tab.).<br />

Parametry sygnałów po kompresji (LFM i NLFM)<br />

Main characteristics of compressed signal (LFM and NLFM)<br />

Parametry sygnału<br />

LFM<br />

NLFM<br />

K s<br />

PSL [dB] K s<br />

PSL [dB]<br />

T = 5 µs 2,5 -13 13 -29<br />

B = 0,75 MHz<br />

T = 10 µs 5 -20 20 -28<br />

Rys. 2. Widmo amplitudowe sygnału NLFM o parametrach<br />

(T = 20 µs; B = 0,75 MHz)<br />

Fig. 2. Amplitude spectrum of NLFM signal (pulse dutation T = 20<br />

µs; bandwidth B = 0,75 MHz)<br />

B = 6 MHz<br />

T = 20 µs 10 -26 26 -37<br />

T = 1 µs 4 -18 10 -41<br />

T = 5 µs 20 -32 100 -56<br />

T = 10 µs 40 -38 110 -61<br />

T = 20 µs 87 -43 200 -63<br />

Rys. 3. Charakterystyki sygnału NLFM (T = 20 µs, B = 0,75 MHz): a) charakterystyka modulacyjna f (t ), b) charakterystyka fazowa<br />

φ(t )<br />

Fig. 3. Main characteristics of NLFM signal (pulse duration T = 20 µs; bandwidth B = 0,75 MHz): a) modulation function f (t ), b) phase<br />

function φ(t )<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 77


Rys. 4. Funkcja autokorelacji sygnałów złożonych o czasie trwania T = 20 µs oraz dewiacji B = 0,75 MHz: a) sygnał LFM, b) sygnał<br />

NLFM<br />

Fig. 4. Autocorrelation functions of complex radar signals with pulse duration T = 20 µs and bandwidth B = 0,75 MHz: a) LFM signal,<br />

b) NLFM signal<br />

Przykładowe charakterystyki sygnałów LFM i NLFM po<br />

kompresji przedstawione są na rys. 4a i rys. 4b. Sygnały mają<br />

taką samą bazę o wartości 15. Zastosowanie nieliniowej modulacji<br />

częstotliwości poprawia właściwości sygnałów złożonych<br />

charakteryzujących się małą bazą w zakresie poziomu<br />

czasowych listków bocznych oraz rozróżnialności odległościowej<br />

(określanej na podstawie szerokości listka na umownym<br />

poziomie -4 dB). Sygnał LFM ma listki boczne na poziomie -<br />

26 dB, a kształt jego funkcji autokorelacji znacznie odbiega od<br />

ogólnie znanej postaci występującej w przypadku kompresji<br />

sygnałów o bazie większej od 100. W prezentowanym przykładzie<br />

sygnał NLFM charakteryzuje się znacznie niższym<br />

poziomem listków bocznych (-26 dB) oraz lepszą rozróżnialnością<br />

odległościową (tab.).<br />

Podsumowanie<br />

W artykule zaprezentowano algorytm syntezy sygnałów z nieliniową<br />

modulacją częstotliwości. Sygnały takie charakteryzują<br />

się właściwościami pożądanymi w radiolokacji. Ich cechy uwidaczniają<br />

się zwłaszcza w przypadku konieczności redukcji<br />

strefy martwej wykrywania radaru. Dotychczas stosowane<br />

w urządzeniach radarowych sygnały LFM o małej bazie nie<br />

zapewniają wystarczającego współczynnika kompresji (rozróżnialność<br />

odległościowa) oraz poziomu listków bocznych.<br />

Sygnały NLFM o analogicznych parametrach czasowo-częstotliwościowych<br />

zapewniają współczynnik kompresji oraz poziom<br />

listków bocznych lepszy w porównaniu do sygnałów LFM.<br />

Praca wykonana jest w ramach projektu &#8222 – „Rodzina radarów<br />

półprzewodnikowych” Projekt ten współfinansowany jest<br />

przez Narodowe Centrum Badań i Rozwoju w ramach Przedsięwzięcia<br />

IniTech.<br />

Literatura<br />

[1] Endres T. J., Hall R. B., Lopez A. M.: Design and analysis methods<br />

of a DDS-based synthesizer for military spaceborne applications.<br />

Frequency Control Symposium, 1994. 48th. Proceedings of the<br />

1994 IEEE International, 624–632, June 1–3, 1994.<br />

[2] Liu J.: Spectrum characteristic analysis of DDS-based RF digital<br />

modulation. Electrical and Electronic Technology, 2001. TEN-<br />

CON. Proceedings of IEEE Region 10 International Conference,<br />

vol. 2, 588-591, August 19–22, 2001.<br />

[3] Kroszczyński J.: Kompresja impulsu w radiolokacji. Prace PIT<br />

nr 58, 1967.<br />

[4] Levanon N., Mozeson E.: Radar Signals. John Wiley&Sons, Inc.,<br />

Hoboken, New Jersey, 2004.<br />

[5] Cook Ch.E., Bernfeld M.: Radar Signals. An Introduction to Theory<br />

and Application. Artech House, Inc., Boston 1993.<br />

[6] Leśnik C.: Nonlinear Frequency Modulated Signal Design. Acta<br />

Physica Polonica A, vol. 116, 2009.<br />

[7] Łuszczyk M., Łabudziński A.: Synteza sygnałów radarowych<br />

o obniżonym poziomie czasowych listków bocznych. IX Krajowa<br />

Konferencja Elektroniki. Darłówko Wschodnie, 30.05–02.06,<br />

<strong>2010</strong>.<br />

Wpłata w <strong>2010</strong> roku – GwarancjĄ niŻszej ceny prenumeraty o vat!<br />

78<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Między Web 2.0 i 3.0: Mobilne systemy informacyjne<br />

z rozszerzoną rzeczywistością<br />

dr inż. JOLANTA BRZOSTEK-PAWŁOWSKA, <strong>Instytut</strong> Maszyn Matematycznych, Warszawa<br />

Perspektywy rozwoju rynku urządzeń i usług mobilnych są<br />

ogromne. Analitycy np. z ABI Research [1] oceniają, że przede<br />

wszystkim rozwinie się rynek spersonalizowanych usług<br />

świadczonych poprzez mobilny Internet, z których przychody<br />

operatorów sieci komórkowych do 2014 r. wzrosną kilkakrotnie<br />

do ponad 17 mld USD (rys. 1). Z jednej strony na perspektywy<br />

takiego rozwoju wpływa obserwowana dynamika<br />

upowszechniania się urządzeń mobilnych, zwłaszcza smartfonów,<br />

których sprzedaż w 2009 r. wzrosła o 20% oraz rozwój<br />

sklepów z mobilnymi aplikacjami (app stores), których<br />

obecnie ściąga się ok. 2,3 mld rocznie (2009 r.), a do 2014 r.<br />

liczba ta osiągnie 5 mld. Z drugiej strony rosną – wraz z powiększającym<br />

się szumem informacyjnym – potrzeby i oczekiwania<br />

użytkowników na dostarczanie selektywnej informacji,<br />

tu i teraz, szybko i efektywnie. Ponad 3 mld obecnie<br />

użytkowników telefonów mobilnych różnego typu, których<br />

możliwości nie są w pełni wykorzystywane zarówno z powodu<br />

niewiedzy użytkowników, jak i niewystarczającej oferty<br />

rynkowej, stanowi potężne środowisko „ssące” wszelkie<br />

innowacyjne rozwiązania dostarczające w sposób inteligentny<br />

informacji, dokładnie tej która jest użytkownikowi w danej<br />

chwili potrzebna i niewymagające od niego nadmiernej<br />

operatorskiej fatygi. Indywidualizacja obsługi informacyjnej<br />

użytkownika wymaga wiedzy o nim, którą można pozyskać<br />

na przykład z historii jego bilingów, z danych jakie pobierał,<br />

lokalizacji użytkownika i jego dostępności. Powstaje problem<br />

szybkiego dostępu do tych danych oraz ich zintegrowania<br />

w jedną bazę, którego rozwiązanie jest obecnie wyzwaniem<br />

stawianym operatorom sieci komórkowych. Z jednej strony<br />

chodzi o pozyskiwanie wiedzy o użytkowniku, z drugiej strony<br />

o zapewnianie adekwatności (do potrzeb) informacji, którą<br />

to adekwatność można uzyskiwać w wyniku klasyfikacji<br />

i opisu informacji.<br />

Web 2.0 dostarczył portali społecznościowych i kontentu<br />

tworzonego przez użytkowników – klasyfikowanego, opisywanego<br />

i ocenianego, obecnie zaś w ramach Web 3.0 powstają<br />

rozwiązania, które inteligentnie (rozumnie) dostarczają potrzebną<br />

użytkownikowi wiedzę, m.in. pozyskując ją z otwartych<br />

zasobów Web 2.0. Przykładem takiego rozwiązania, które<br />

może budzić nawet niepokój jest w Stanach Zjednoczonych<br />

system billboardów z urządzeniami identyfikującymi użytkownika<br />

na postawie odczytanych numerów rejestracyjnych samochodu<br />

i dostosowujący wyświetlaną reklamę do szybko<br />

uzyskanego z Sieci i zanalizowanego profilu użytkownika.<br />

Ostatnio, od 2009 r. bardzo intensyfikuje się rozwój<br />

mobilnych systemów dostarczających kontekstowej informacji<br />

o otoczeniu, w którym jest lokalizowany użytkownik,<br />

z uwzględnieniem jego preferencji, przekazanych przez niego<br />

systemowi lub uzyskanych przez system na podstawie<br />

historii zachowań użytkownika. Systemy te czerpią informacje<br />

o obiektach znajdujących w najbliższym otoczeniu użytkownika<br />

(o budynkach, produktach, usługach, ludziach) m.in.<br />

z portali społecznościowych (Panoramica, Twitter, Facebook,<br />

Wiki, Flickr, Yelp) oraz z innych źródeł – otwartych bądź komercyjnych<br />

baz danych. A ponieważ, jak wspomniano, ma<br />

być łatwo i przyjemnie dla użytkownika, do przekazywania<br />

tych informacji zastosowano technologie rozszerzonej rzeczywistości<br />

– Augmented Reality (AR), które umożliwiają<br />

nanoszenie wirtualnych obiektów, zawierających informacje<br />

na rzeczywisty obraz zdejmowany kamerą lub strumieniowo<br />

przesyłany na żywo, lub mapę satelitarną. Stąd systemy te<br />

często zwane są przeglądarkami AR lub inaczej przeglądarkami<br />

otoczenia/świata (reality browsers). Systemy tego typu<br />

mogą pełnić rolę przewodników turystycznych i kulturalnych,<br />

edukacyjną, wspierania na stanowisku pracy (performance<br />

suport), marketingową i reklamową, rozrywkową (gry w rzeczywistej<br />

scenerii). Pierwsze mobilne aplikacje AR pojawiły<br />

się w 2008 r., więc bardzo niedawno i ich rozwój, wraz<br />

z rozwojem smartfonów jest błyskawiczny. Zastosowanie<br />

technologii AR stało się technicznie możliwe dzięki rozwojowi<br />

urządzeń mobilnych i ich wydajności, a jest uzasadnione jej<br />

atrakcyjnością w odbiorze i nagromadzonymi zasobami informacyjnymi<br />

w Sieci (jest z czego czerpać). Systemy reality<br />

browsers takie jak najbardziej dojrzały Layar [7], Wikitude [8],<br />

Junaio [9], SekaiCamera [10], Acrossair [11] tworzą nową<br />

jakość na rynku systemów informacyjnych, wynikającą z ich<br />

podręcznej obecności w „komórkach” i ich kontekstowości<br />

informacyjnej oraz immersyjności, pobudzającej użytkownika<br />

do interaktywności i skupiającej jego uwagę na przekazywanych<br />

treściach. Są świetnym przykładem wykorzystania<br />

osiągnięć Web 2.0 i realizacji idei Web 3.0.<br />

Dynamika ich rozwoju, dosłownie z miesiąca na miesiąc<br />

oraz dotychczasowy nieśmiały polski udział w międzynarodowych<br />

rozwiązaniach, przy braku polskich propozycji tego<br />

typu mobilnych systemów, określa cel tego artykułu – ukazuje<br />

trend technologiczny, pokazuje szczegóły rozwiązań i, być<br />

może zachęci do inwestycji i projektów w tym obszarze.<br />

Technologie rozszerzonej rzeczywistości<br />

(Augmented Reality)<br />

Technologie rozszerzonej rzeczywistości są starsze niż termin<br />

Augmented Reality, sformułowany w 1990 r. przez Toma<br />

Caudella, naukowca z Boeinga. Pierwsze aplikacje AR pojawiły<br />

się na przełomie lat 60. i 70., zaś pierwsze mobilne,<br />

jak wcześniej wspomniano, w 2008 r. technologie AR tworzą<br />

świat iluzji, połączenie świata rzeczywistego z wirtualnym<br />

w jedną rzeczywistość – rozszerzoną, która oprócz<br />

fascynujących doznań, umożliwia przekazywanie informacji<br />

i wiedzy dodatkowej, w stosunku do otrzymywanej z obrazu<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 79


Rys. 1. Prezentacja z wykorzystaniem AR, projektu zapory wodnej<br />

koncernu Alstom (Źródło: [12])<br />

Fig. 1. Example presentation using AR of the water dam project<br />

of the Alstom group<br />

rzeczywistości. AR umożliwia również tworzenie symulacji,<br />

służących kształceniu, instruktażowi i rozrywce. Stwierdzenie<br />

„Każda wystarczająco zaawansowana technologia jest<br />

nieodróżnialna od magii” Arthura C. Clarke’a, autora „2001:<br />

Odyseji Kosmicznej”, świetnie pasuje do aktualnego poziomu<br />

rozwoju AR.<br />

Rysunek 1 przedstawia przykład iluzorycznej prezentacji<br />

projektu elektrowni wodnej na wewnętrznym konwencie koncernu<br />

energetycznego Alstom.<br />

Rozwiązania z zastosowaniem AR przyciągają uwagę i angażują<br />

interaktywnie odbiorcę, pod względem siły przekazu<br />

informacji oraz oddziaływania na zmysły i emocje odbiorcy<br />

wydają się obecnie nie mieć konkurencji.<br />

Technologie AR umożliwiają nanoszenie wirtualnych<br />

obiektów różnego typu – tekstów, ikonek, obiektów 2D i 3D<br />

statycznych lub animowanych, obiektów dźwiękowych, a nawet<br />

filmików. Generowane i nanoszone mogą być na obraz<br />

zdejmowany kamerą, lub na fragment mapy satelitarnej<br />

obszaru, w którym znajduje się użytkownik. Generowane<br />

obiekty mogą być wyświetlane, wraz z obrazem mapy lub<br />

obrazem zdejmowanym kamerą, na ekranie lub tablicy projekcyjnej,<br />

tablicy interaktywnej, ekranie smartfona lub palmtopa,<br />

mogą też być wyświetlane na specjalnych okularach,<br />

a nawet soczewkach kontaktowych przez które obserwowany<br />

jest obraz rzeczywisty.<br />

Rys. 2. Przykład matrycy markerów AR – ARTagów (Źródło: [2])<br />

Fig. 2. Example matrix AR markers (ARTags)<br />

80<br />

Obiekty wirtualne mogą być generowane z zastosowaniem<br />

opisanych technik:<br />

1. Poprzez wskazanie przez użytkownika miejsca wygenerowania:<br />

wskazanie może się odbywać za pośrednictwem<br />

urządzeń pomocniczych wyposażonych w sensory dotyku/nacisku<br />

(np. specjalne rękawice) lub w diody LED, które dają<br />

rozpoznawalne punkty świetlne w zdejmowanym obrazie (np.<br />

LED mogą znajdować się na ubraniu, które użytkownik zakłada<br />

lub na przedmiocie trzymanym w ręku). Są to techniki<br />

używane przede wszystkim do gier symulacyjnych.<br />

2. W wyniku rozpoznania odpowiedniego kształtu w analizowanym<br />

obrazie: rozpoznawanym kształtem np. może być<br />

napis, kod kreskowy, etykietka produktu, specjalny marker<br />

(rys. 2), sylwetka człowieka lub jej fragmenty (głowa, ręce).<br />

Obiekt generowany jest w miejscu znajdowania się rozpoznanego<br />

kształtu, przy czym możliwe jest śledzenie ruchu rozpoznanego<br />

kształtu i wraz z nim odpowiednie, zgodne z zaplanowanym<br />

scenariuszem, generowanie obiektu (tego samego<br />

lub przekształconego lub innego, lub usuwanie obiektu albo<br />

jego odpowiednie animowanie). Techniki te stosowane są<br />

głównie w marketingu i reklamie, edukacji i grach symulacyjnych,<br />

zaczynają być stosowane w e-handlu.<br />

Czyli elementami, które mogą być szukane i rozpoznawane<br />

w analizowanym obrazie są:<br />

• różnego rodzaju drukowane markery (leżące na stole, drukowane<br />

w książce lub na planszy, trzymane w rękach lub<br />

umocowane na ubraniu użytkownika);<br />

• punkty świetlne z diod LED, w które są wyposażone specjalne<br />

urządzenia, trzymane przez użytkownika lub<br />

• określone kształty znajdujące się na obrazie np. róg okna,<br />

noga krzesła, ręce, głowa użytkownika, napis, kod kreskowy.<br />

Są to wskaźniki dla oprogramowania, których rozpoznany<br />

kształt i położenie decyduje o rodzaju wygenerowanego<br />

obiektu i jego umiejscowieniu. Wirtualne obiekty mogą zmieniać<br />

swoje położenie wraz ze zmianą położenia wskaźników,<br />

co jest odpowiedzią na zachowania użytkownika (np. poruszanie<br />

markerami, diodami, rękami).<br />

3. Zidentyfikowania geograficznego położenia użytkownika,<br />

jego ruchu i kierunku poruszania się (lub kierunku ruchu trzymanego<br />

np. smartfona lub palmtopa): obiekty wirtualne wyświetlane<br />

są na obrazie zdejmowanym kamerą urządzenia mobilnego<br />

lub wyświetlanej mapie, w miejscach, którym przyporządkowano<br />

takie obiekty informacyjne. Czyli – gdy występuje zgodność ich<br />

współrzędnych geograficznych i które znajdują się (mają współrzędne<br />

geograficzne) w najbliższym w stosunku do lokalizacji<br />

użytkownika obszarze, zakreślonym maksymalnym promieniem,<br />

domniemanym przez system lub zdefiniowanym przez użytkownika.<br />

Technika ta wykorzystuje wbudowane w mobilne urządzenia<br />

GPS, żyroskopy, kompasy i czujniki bezwładnościowe, przekazujące<br />

parametry położenia i ruchu; jest wykorzystywana w mobilnych<br />

systemach informacyjnych omawianych w tym artykule.<br />

Osobnym problemem, obecnie rozwiązywanym przez zespoły<br />

badawcze, jest opracowanie technologii identyfikacji położenia<br />

geograficznego w miejscach bez dostępu do GPS np.<br />

w dużych centrach handlowych, w podziemiach. Dla zapewnienia<br />

działania mobilnego systemu informacyjnego w takich<br />

miejscach muszą być stosowane techniki mieszane np. wykorzystujące<br />

istniejące hot spoty WiFi jako stacje referencyjne lub<br />

oznakowanie terenu markerami, również radiowymi RFID (np.<br />

iPHONE 4 ma już czytnik RFID), jak również techniki rozpo-<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Rys. 3. Przykłady różnych rodzajów obiektów<br />

AR (Źródło: [7, 8, 10])<br />

Fig. 3. Examples of different types of AR<br />

objects<br />

znawania kształtów (innych niż markery). Dla celów mobilnych<br />

systemów informacyjnych obecnie silnie jest rozwijana technologia<br />

Kooaba [13] rozpoznawania nie tyle kształtów co treści<br />

obrazów (np. plakatów, gazet, etykietek produktów, reklam)<br />

i automatycznego kojarzenia rozpoznanych treści z istniejącymi<br />

treściami w Sieci, związanymi tematycznie. W ten sposób użytkownik<br />

otrzymuje (jeśli się godzi na to) „całokształt” informacji.<br />

Na rysunku 3 są przedstawione przykłady kliku rodzajów<br />

wirtualnych obiektów nanoszonych na obraz rzeczywistości.<br />

Oprogramowanie wykorzystujące technologie AR może<br />

być w formie aplikacji internetowych dostępnych przez przeglądarkę<br />

WWW, aplikacji internetowych typu sieciowego, z instalowaną<br />

aplikacją kliencką na komputerze lub urządzeniu<br />

mobilnym użytkownika, która komunikuje się z aplikacją serwerową<br />

lub w formie samodzielnych aplikacji na urządzenia<br />

mobilne. Może też być w formie systemów bardziej złożonych,<br />

opartych na architekturze SOA usług sieciowych, taką architekturę<br />

mają omawiane systemy reality browser, która zostanie<br />

bliżej przedstawiona w dalszej części.<br />

Aplikacje wykorzystujące technologie AR tworzone są<br />

w specjalizowanych środowiskach programistycznych. Dobór<br />

środowiska, a nawet kilku środowisk dostosowanych do specyfiki<br />

i koncepcji tworzonych rozwiązań musi wynikać z dogłębnej<br />

analizy. Obecnie na rynku polskim praktycznie nie ma<br />

dostawców tego typu produktów.<br />

Dostępne są specjalizowane biblioteki AR open source, stanowiące<br />

uzupełnienia popularnych środowisk programistycznych,<br />

np. środowiska programistycznego Adobe Action Scripting<br />

3 (AS3) z bibliotekami open source AR, jak również komercyjne<br />

środowiska programistyczne, z których ARToolkit Professional<br />

firmy ARToolWorks, Inc. wydaje się być dobrze rozwiniętym dla<br />

potrzeb profesjonalnych projektów, a przede wszystkim dostępnym,<br />

do kupienia na jasno określonych warunkach licencji.<br />

Przykłady środowisk programistycznych dla aplikacji AR:<br />

• wspomniany wyżej ActionScript 3.0 – obiektowy język programowania,<br />

bazujący na standardzie języka skryptowego<br />

ECMAScript, dla runtime’u Flash Player 9, z maszyną wirtualną<br />

AVM2 (ActionScript Virtual Machine 2);<br />

• NyARToolkit – biblioteka klas do renderowania obiektów<br />

3D i nanoszenia na obrazy wideo, na platformy Java/Android/C#/ActionScript3;<br />

• FLARToolKit, biblioteki analizy obrazu w technologii Flash<br />

Action Script3 (Adobe Flex/Flash Builder) do wyszukiwania<br />

wzorców w obrazie wideo; inne biblioteki AS3 do<br />

rozpoznawania np. ruchów twarzy, rąk to Papervision3D,<br />

Away3D, Sandy3D, Alternativa3D;<br />

• iPhone ARKit biblioteka do języka C – Objective iPhone’a;<br />

• D-Fusion Studio (Total Immersion) – zintegrowane środowisko<br />

do tworzenia aplikacji Web, Windows i mobilnych,<br />

bardzo rozwinięte, praktycznie zakup jego w Polsce jest<br />

niemożliwy ze względu na usługowe, a nie sprzedażowe,<br />

nastawienie jedynego przedstawiciela w Polsce;<br />

• wspomniany ARToolkit Professional – zintegrowane środowisko<br />

tworzenia aplikacji Web, Windows i mobilnych.<br />

Obszary zastosowań Augmented Reality<br />

AR ma wiele obszarów zastosowań:<br />

• marketing interaktywny (widowiska multimedialne na żywo,<br />

stoiska wystawowe, ulotki i broszury, prezentacje produktów<br />

w webcastach); szczególne zastosowanie AR ma miejsce:<br />

– dla dokładnego oglądania wirtualnie prezentowanych<br />

istniejących produktów, np. o dużych gabarytach lub<br />

bardzo skomplikowanych i drogich, których fizyczne<br />

prezentowanie w stoiskach na targach, wystawach jest<br />

technicznie niemożliwe lub zbyt kosztowne (np. samoloty,<br />

obiekty mieszkalne),<br />

– dla prezentowania planowanych (nieistniejących) projektów<br />

i produktów;<br />

• e-handel/e-usługi (internetowe zakupy z przymierzaniem,<br />

wizualizacja menu, szczegóły produktów znajdujących się<br />

na półkach i dodatkowe informacje o nich);<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 81


• edukacja immersyjna: książki, podręczniki, encyklopedie,<br />

glosariusze, z wizualizacją 3D;<br />

• gry edukacyjne;<br />

• imersyjne wydawnictwa – czasopisma, okładki DVD, które<br />

mogą być wzbogacane o wirtualne obiekty (np. prezentujące<br />

towary, nagrania) za pomocą aplikacji serwerowych<br />

dostępnych z portali wydawnictw lub aplikacji mobilnych;<br />

• inteligentny instruktaż online dla serwisantów, operatorów,<br />

chirurgów, w formie dodatkowych wirtualnych obiektów nakładanych<br />

na widziany rzeczywisty obraz, np. podpowiadających<br />

jak wykonać daną czynność;<br />

• systemy informacyjne typu „reality browsers” informujące<br />

o otoczeniu (omawiane w artykule);<br />

• muzea (rzeczywiste i wirtualne):<br />

– fotografie przestrzenne wnętrz muzealnych znajdujące<br />

się na portalu WWW wzbogacane o obiekty AR dokładnie<br />

prezentujące wskazane przez użytkownika dzieła<br />

i informacje o nich;<br />

– obiekty wirtualne nanoszone na oglądane w rzeczywistości<br />

dzieła muzealne okiem kamery telefonu komórkowego;<br />

• inżynierskie – projektowanie wnętrz i wizualizacja 3D;<br />

• zastosowania rozrywkowe – gry indywidualne i zbiorowe<br />

(w domu, w wesołych miasteczkach);<br />

• i inne.<br />

Najintensywniej rozwijającymi się zastosowaniami są systemy<br />

reality browser, głównie na systemy operacyjne Android oraz<br />

OS iPhone’a, a to ze względu na dostępność bibliotek i dobrze<br />

udokumentowany interfejs API (Application Programming Interface)<br />

dla programistów oraz ze względu na ogromne zainteresowanie<br />

użytkowników tymi systemami, mierzone liczbą ściągnięć<br />

aplikacji AR, będących aplikacjami klienckimi tych systemów.<br />

Poziom rozwoju Augmented Reality<br />

Poziom rozwoju AR, mierzony fazami cyklu (używanymi np.<br />

w raportach Gartner Group, [3]): „wystrzał technologii” „szczyt<br />

rozdmuchanych oczekiwań” „koryto(dno) rozczarowania”<br />

„stok oświecenia” „płaskowyż wydajności” (rys. 4 ) i wynikający<br />

stąd poziom istniejących doświadczeń oraz obszar zastosowań,<br />

można przedstawić, jak poniżej.<br />

Oczekiwania<br />

Rys. 4. Rozwój AR na tle faz cyklu rozwoju technologii wg Gartner<br />

[3] (Źródło: własne)<br />

Fig. 4. Development of AR on the background technology development<br />

cycle according to Gartner [3]<br />

82<br />

Rozpoznawanie<br />

lokalizacji i ruchu<br />

Rozpoznawanie<br />

obrazu<br />

Elastyczne<br />

obiekty 3D,<br />

soczewki AR<br />

„wystrzał<br />

technologii”<br />

Markery<br />

2D<br />

„szczyt<br />

rozdmuchanych<br />

oczekiwań”<br />

UPC (1D)<br />

QR (2D)<br />

” koryto<br />

rozczarowania”<br />

Czas<br />

„płaskowyż<br />

wydajności”<br />

„stok<br />

oświecenia”<br />

Poziom 0 – dobrze opanowany, najstarszy, oparty o kody<br />

kreskowe Universal Product Code – UPC (1D) i kody Quick<br />

Response – QR (2D), które odczytywane kamerą są linkami<br />

do rzeczywistych produktów (np. w magazynach) lub danych<br />

(np. o firmie), jest w fazie” stok oświecenia” zmierzającej<br />

w kierunku „płaskowyżu wydajności”;<br />

Poziom 1 – markery 2D (np. takie jak na rys. 2) są w fazie<br />

„szczytu rozdmuchanych oczekiwań”;<br />

Poziom 2 – bezmarkerowe mobilne rozwiązania oparte<br />

o GPS, kompas lub akcelerometr oraz bezmarkerowe rozwiązania<br />

oparte o rozpoznawanie obrazu (markerem jest rozpoznany<br />

kształt w obrazie wideo) są w fazie „wystrzału technologii”<br />

blisko fazy „szczytu rozdmuchanych oczekiwań”;<br />

Poziom 3 – rozwiązania niedostępne, w fazie badawczolaboratoryjnej,<br />

idące w kierunku uwolnienia użytkownika od<br />

monitora i wyświetlaczy, np. cyfrowe kontaktowe soczewki<br />

(rozwijane na Uniwersytecie Waszyngtona w Seattle) oraz<br />

idące w kierunku elastyczności wirtualnych obiektów, dostosowujących<br />

swój kształt i perspektywę ich oglądania do<br />

bieżącego kształtu i ruchu obiektu rzeczywistego zrelacjonowanego<br />

z obiektem wirtualnym (np. prace Juliena Pileta,<br />

w tym dyplomowa z 2008 r., w których pokonał ograniczenia<br />

markerów polegające na sztywności, niezmienianiu proporcji<br />

i perspektywy), obecnie w fazie „wystrzał technologii”.<br />

W mobilnych systemach informacyjnych mają zastosowania<br />

rozwiązania AR poziomu 0, 1 i 2. Warto zauważyć, że<br />

nanoszone obiekty 3D w rozwiązaniach poziomu 2 mogą podlegać<br />

transformacjom w zależności od odległości użytkownika<br />

i kierunku jego ustawienia w stosunku do POI. Transformacje<br />

te na razie są stosunkowo proste (skalowanie, obrót),<br />

ale wraz z dojrzewaniem innowacji z obecnego poziomu 3,<br />

można oczekiwać ich szybkiej implementacji w mobilnych systemach<br />

AR.<br />

Przykłady rozwiązań i zastosowań<br />

Augmented Reality<br />

Obecnie największe koncerny i firmy stosują AR do swoich<br />

produktów lub prezentacji produktów (General Electric, Alstom,<br />

Nokia, Apple, Intel, Seat, Citroen, Peugeot, Renault,<br />

Infiniti/Nissan, SNECMA-Saffran Group i inne), przykłady:<br />

• prezentacje modeli samochodów Citroen, popularyzacja<br />

elektrowni wiatrowych General Electric,<br />

• e-kioski wystawowe prezentujące m.in. silniki lotnicze<br />

SNECMA-Saffran Group;<br />

• przymierzalnia ubrań w e-sklepach oferujących odzież np.<br />

e-sklep agencji Zugara;<br />

• aplikacja AR, opracowana przez Alcatel-Lucent, rozpoznająca<br />

okładki płyt CD/DVD i nanosząca na ich obraz zdjęty<br />

kamerą wirtualne obiekty, w tym dźwiękowe, informujące<br />

o zawartości płyty;<br />

• Microsoft rozwija system Surface – komputer z oprogramowaniem<br />

aplikacyjnym oraz oprogramowaniem narzędziowym<br />

SDK, z ekranem dotykowym poziomym, jak stolik<br />

(w opracowaniu jest komputer z ekranem wertykalnym,<br />

ściennym), korzystać z niego może wielu użytkowników<br />

jednocześnie, daje informacje o obiekcie wskazanym<br />

palcem/dłonią w postaci wyświetlanych obiektów AR, pozwala<br />

„zoomować” obiekt przez dołączone do komputera<br />

„szkiełko” i pobierać informacje do telefonu komórkowe-<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


go; ekran rozpoznaje kształty położonych na nim przedmiotów<br />

(np. kształt „komórki”) oraz rozpoznaje ruchy rąk;<br />

ruchy rąk przesuwają obiekty lub je „zoomują”; MS Surface<br />

może służyć np. jako kiosk informacyjny w hotelach i na<br />

dworcach, występuje już w tej roli w sieci hoteli Sheraton<br />

(Starwood Hotels);<br />

• grudniowy z 2009 r. numer czasopisma InStyle zawiera<br />

zastosowanie AR do pokazywania zawartości wydrukowanych<br />

w numerze świątecznych pudełek z reklamowanymi<br />

prezentami za pośrednictwem aplikacji serwerowej<br />

dostępnej na portalu wydawnictwa dla posiadaczy zakupionego<br />

numeru;<br />

• AR została po raz pierwszy w Polsce użyta w prasowej<br />

reklamie marki Lech; na ostatniej stronie wakacyjnego dodatku<br />

„Więcej z lata” w „Przekroju” nr 25/2009 pojawił się<br />

specjalny piktogram (tag), który uruchamia wyjątkową animację<br />

3D ukazującą czekającą nagrodę konkursową.<br />

Przewiduje się, że największe zastosowania AR, w tym<br />

mobilne, będą w sektorze marketingu i reklamy, w który to<br />

zauważalnie szybko rozwijający się trend wpisują się mobilne<br />

systemy informacyjne AR. Ich przykłady zostały podane we<br />

Wprowadzeniu. Mogą one służyć reklamie i promocji, mogą<br />

również służyć celom edukacyjnym, turystycznym, kulturalnym.<br />

Mogą obejmować wielorakie zastosowania, w zależności<br />

od kategorii POI, które zgromadziły i obsługują, mogą mieć<br />

specjalizowane zastosowanie, jak system iTacitus odtwarzający<br />

historię na tle współczesności, opisany w dalszej części<br />

Prognozy i fakty.<br />

Zasady działania i architektura mobilnych<br />

systemów AR<br />

Mobilne systemy AR dostarczają informacji, czerpanej z Internetu,<br />

o postrzeganym przez kamerę urządzenia mobilnego<br />

otoczeniu. Informacja jest dobierana do kontekstu otoczenia<br />

na podstawie współrzędnych geograficznych, może też być<br />

dobierana na podstawie rozpoznanego obiektu na obrazie<br />

(np. po twarzy, kształcie, po tekście), rozpoznanej treści zawartej<br />

w obrazie (np. po kodzie lub tagu). Ponieważ masowo<br />

obecnie rozwijają się systemy informacyjne wykorzystujące<br />

dane lokalizacyjne, w dalszej części uwaga zostanie poświęcona<br />

tego typu systemom. Trzeba zaznaczyć, że wymagają<br />

one w urządzeniu mobilnym oprócz kamery i mobilnego Internetu,<br />

GPS, kompasu i akcelerometru oraz właściwego systemu<br />

operacyjnego. Mobilne systemy AR przede wszystkim<br />

są rozwijane na system Android (od wersji 1.5) lub iPhone OS<br />

(od wersji 3.1), zaczynają też być anonsowane rozszerzenia<br />

na inne systemy operacyjne np. Samsung Bada i Nokia Symbian<br />

(open source).<br />

Informacja może być nanoszona na obraz w postaci różnego<br />

rodzaju wirtualnych obiektów. Podnosząc smartfon na<br />

poziom wzroku można zobaczyć na ekranie obraz z kamery<br />

(np. ulice lub krajobraz) z zaznaczonymi za pomocą ikonek<br />

punktami (rys. 3b i 3c), które nas interesują lub wszystkimi<br />

punktami, o których system ma w swojej bazie, lub sieci<br />

współpracujących baz, adnotacje, a które to punkty znajdują<br />

się w domniemanym lub określonym przez użytkownika promieniu<br />

obserwacji, od miejsca znajdowania się użytkownika.<br />

Punkty te, mające swoje geograficzne współrzędne zapisane<br />

w bazie wraz z innymi danymi, nazywają się Point of Interest<br />

(punkt zainteresowania, użyteczne miejsce), przyjętym<br />

skrótem jest POI. Punktem POI może być budynek, jego „zawartość”,<br />

sklep, restauracja, pomnik, człowiek (który np. na<br />

Twitterze podał swoją lokalizację), słup reklamowy, drzewo,<br />

bądź miejsce realizacji określonego projektu. Punkty POI są<br />

kategoryzowane w systemie wg słownika. Przykładowy fragment<br />

takiego hierarchicznego słownika kategorii lub inaczej<br />

warstw (informacyjnych) np.:<br />

Architektura i budynki: informacje o architekturze, nazwy<br />

budynków itp.<br />

Dobroczynność i społeczeństwo: organizacje charytatywne,<br />

wolontariaty, projekty społecznościowe, …<br />

Fotografie i wideo: Flickr, YouTube, z otoczenia<br />

Jedzenie i napoje: restauracje, puby, przewodnik Michelin<br />

itp.<br />

Nieruchomości: budynki i mieszkania na sprzedaż/do wynajęcia,<br />

nieruchomości komercyjne itp.<br />

Noclegi: hotele, campingi, itp.<br />

Rozrywka: kasyna, bilety, kina, itp.<br />

Sztuka: wystawy, wirtualne sztuki, …<br />

Uroda i zdrowi: spa, salony fryzjerskie, kluby fitness itp.<br />

Uczelnie i badania: mapa ośrodków uniwersyteckich, projekty<br />

badawcze itp<br />

WiFi hot spoty: wycieczki po mieście; wycieczki do A, do B,<br />

do C itd.<br />

Wydarzenia: Yahoo – przegląd nadchodzących wydarzeń,<br />

kongresy itp.<br />

Zatrudnienie: oferty pracy, biura, agencje pracy itp.<br />

Warto zwrócić uwagę, że kategoria informacji może mieć<br />

zdefiniowane jako źródło danych o POI bazę danych internetowego<br />

serwisu społecznościowego (Flickr, Yahoo, Wiki,…)<br />

lub istniejącego internetowego serwisu komercyjnego (np.<br />

Funda – serwis z ofertami nieruchomości), do których to baz<br />

(w chmurze) sięga aplikacja kliencka systemu informacyjnego<br />

za pośrednictwem głównego serwera systemu i ewentualnie<br />

serwisu pośredniczącego dostawcy kontentu POI. Oprócz<br />

baz danych istniejących serwisów używanych dla potrzeb systemu<br />

informacyjnego, które publikują swój API (i do niego<br />

odbiorca danych musi się dostosować), źródłami informacji<br />

mogą być specjalnie tworzone bazy danych o POI. Na ogół<br />

są one wypełniane społecznościowo, zaś udostępniane komercyjnie<br />

w ramach współpracy z jednym, lub więcej niż<br />

jednym, systemem informacyjnym. Odbywa się to zgodnie<br />

z założonym modelem biznesowym, by powstawało jak najwięcej<br />

satelickich komercyjnych serwisów z kontentem o POI<br />

udostępnianym poprzez usługi sieciowe. Czyli twórcy kategorii<br />

informacji o POI (jednej lub wielu) tworzą komercyjne serwisy<br />

internetowe, licznie obecnie powstające, rozliczające się<br />

finansowo z właścicielami danego systemu informacyjnego,<br />

oferując usługi sieciowe udostępniania kontentu poprzez dialog<br />

żądanie-odpowiedź (Get i Response w protokole HTTP<br />

1.1). Dialog jest obsługiwany przez główny serwer systemu<br />

informacyjnego, zaś inicjowany jest przez użytkownika za<br />

pośrednictwem aplikacji klienckiej systemu współpracującej<br />

z głównym serwerem. Aplikacja nanosi na obraz zdejmowany<br />

kamerą oznaczenia miejsc POI (np. standardowo ikonki<br />

w postaci kółek kolorowych), zaś klient wybiera POI, o którym/<br />

ych chce wiedzieć więcej. Aplikacja kliencka po uruchomieniu<br />

może oczekiwać na wybór przez użytkownika kategorii POI<br />

do wyświetlenia, znajdujących się w domniemanej lub okre-<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 83


Rys. 5. Przykład fragmentu formularza zbierającego dane o POI na stronie WWW dostawcy kontentu POI (Źródło: [17])<br />

Fig. 5. A part of form for collecting POI data on the site of the POI content provider<br />

ślonej przez użytkownika odległości, może też sama inicjować<br />

transmisje, z serwera systemu, informacji o POI na podstawie<br />

profilu zarejestrowanego użytkownika, który w nim podał preferowane<br />

kategorie. Może też przesyłać podstawowe informacje<br />

o wszystkich POI znajdujących się w pobliżu, bez względu<br />

na kategorię. W tym ostatnim przypadku sprawność przesyłania<br />

zależy od parametrów sprzętu i sieci. Sama aplikacja<br />

kliencka, w wersjach odpowiednich dla danego typu smartfona<br />

(a właściwie sytemu operacyjnego), udostępniana jest do<br />

ściągnięcia na głównym serwerze systemu lub w app store<br />

producenta smartfonów lub jest preinstalowana w smartfonie,<br />

tak jak np. aplikacja kliencka systemu Layar preinstalowana<br />

przez Samsung na smartfonach wyposażonych w system Android<br />

i nowy system Bada, albo przez chińskiego producenta<br />

TCT Mobile smartfonów Alcatel’a.<br />

Podsumowując, struktura mobilnego systemu informacyjnego<br />

AR, której elementy uwidacznia rys. 6, składa się z:<br />

aplikacji klienta – w zależności od rozwiązania wynikającego<br />

z możliwości urządzeń, na jakie jest dedykowana,<br />

może brać na siebie ciężar przetwarzania danych lub być<br />

aplikacją typu proxy-browser korzystającą z mocy obliczeniowych<br />

serwera; typ proxy-browser będzie nadal dominował<br />

w rozwiązaniach, wg analityków do 2013 r., kiedy parametry<br />

wydajnościowe sprzętu pozwolą działać bardziej<br />

samodzielnym aplikacjom;<br />

głównego serwera systemu – komunikującego się z aplikacją<br />

kliencką, z internetową aplikacją zarządzania kategoriami<br />

informacyjnymi oraz z serwisami (poprzez usługi<br />

sieciowe) dostawców kontentu informacyjnego o miejscach<br />

POI;<br />

<br />

<br />

<br />

internetowej aplikacji zarządzania kategoriami informacji<br />

o POI oraz rozliczeń z twórcami kategorii i dostawcami<br />

kontentu – umożliwia tworzenie i aktualizacje kategorii<br />

przez ich dostawców oraz prowadzenie i rozliczanie ich<br />

kont;<br />

serwisów z usługami sieciowymi dostarczającymi zawartość<br />

informacyjną o POI – które mogą opierać się na własnych<br />

bazach danych, będących źródłami informacji o POI,<br />

jak również na bazach serwisów społecznościowych i innych<br />

serwisów dostępnych w Internecie publikujących swój<br />

interfejs API dostępu do zasobów; dostawcy tworzą własne<br />

bazy danych POI zgodnie z narzuconą przez system<br />

informacyjny strukturą danych o POI i formatami zapisu<br />

plików multimedialnych, natomiast dane o kategoriach,<br />

które tworzą, wprowadzają poprzez aplikację zarządzania<br />

kategoriami, do bazy wewnętrznej systemu informacyjnego;<br />

rysunek 5 przedstawia przykład fragmentu formularza<br />

zbierającego dane o POI, wspomagającego się mapą satelitarną<br />

dla określenia koordynat geograficznych opisywanego<br />

POI;<br />

opublikowanego interfejsu API systemu informacyjnego<br />

– umożliwia współpracę systemu z dostawcami informacji<br />

o POI (twórcami i właścicielami kategorii); interfejs<br />

określa sposób przesyłania komunikatów, i ich parametry,<br />

do/z usług sieciowych dostawców kontentu POI, usług raczej<br />

typu REST – opartych o protokół przesyłania HTTP<br />

i format przesyłanych danych bazujący na czystym formatowaniu<br />

danych w XML, a ostatnio coraz częściej na<br />

formacie JSON (JavaScript Object Notation), a nie na zakodowanych<br />

komunikatach SOAP (wydaje się to tenden-<br />

84<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


cją ostatnich lat); usługi są realizowane przez serwlety,<br />

na ogół pisane w PHP (języku nie w pełni przystosowanym<br />

do programowania usług sieciowych) i korzystające<br />

z bazy MySQL zawierającej dane o POI.<br />

API definiuje:<br />

– żądanie (komenda GET w HTTP), definiujące URI<br />

usługi, składające się z URL usługi i parametrów dla<br />

niej: http://?; adres URL aplikacja<br />

kliencka pobiera z danych o kategorii (wybranej<br />

przez użytkownika z listy lub poprzez podanie słów<br />

kluczowych wyszukiwania), dane te dostarcza serwer<br />

główny systemu; parametry w URI dotyczą m.in. identyfikacji<br />

POI, identyfikacji urządzenia i wersji aplikacji<br />

klienta;<br />

– odpowiedź (RESPONSE w HTTP), zawierająca<br />

w swej początkowej części status wykonania żądania<br />

oraz nagłówek informujący o formacie przesyłanych<br />

w odpowiedzi danych (np. w JSON); żądane dane<br />

o POI, przesyłane po tej początkowej części, mają<br />

strukturę zgodną z publikowaną dla potrzeb dostawców<br />

kontentu POI, która dalej w skrócie zostanie<br />

omówiona.<br />

Rysunek 6 przedstawia przykładową architekturę systemu<br />

informacyjnego AR, korzystającego z przetwarzania „w chmurze”,<br />

opartego na usługach sieciowych „trzecich” dostawców<br />

kontentu POI.<br />

Struktura danych opisujących kategorię<br />

POI<br />

Kategorie POI są opisywane przez m.in.:<br />

dane ogólne – nazwa i opis kategorii, typy obiektów AR<br />

zawartych w POI danej kategorii, kraj i obszar geograficzny,<br />

w których POI tej kategorii będą wyświetlane, słowa<br />

kluczowe, nazwa dostawcy, ikonka charakterystyczna dla<br />

kategorii;<br />

dane związane z autoryzacją użytkownika i dostawcy kategorii,<br />

dane dotyczące parametrów wizualizacji kategorii<br />

(rys. 7), ikonki POI w różnej skali (używane zależnie od odległości<br />

POI od użytkownika), wymagania w stosunku do<br />

działania aplikacji klienckiej (np. specyfikacja potrzebnych<br />

kontrolek interfejsu graficznego aplikacji klienta), zakres odległości<br />

(od użytkownika), w jakiej mogą być wyświetlane<br />

POI tej kategorii.<br />

Przetwarzanie „w chmurze”<br />

Dane o kategoriach<br />

i dostawcach<br />

Baza danych POI<br />

dostawcy kategorii<br />

GPS<br />

Aplikacja „klient”<br />

Aplikacja serwerowa<br />

obsługująca żądania<br />

klienta<br />

Get kategoria,<br />

Get POI<br />

Response<br />

Serwer systemu<br />

API systemu<br />

Get POI<br />

Response<br />

Aplikacja serwerowa<br />

zarządzania<br />

kategoriami POI,<br />

obsługująca<br />

dostawców POI<br />

definiujących kategorie<br />

Baza danych POI<br />

dostawcy kategorii<br />

Baza danych POI<br />

dostawcy kategorii<br />

API Twittera<br />

Get POI/<br />

Response<br />

API FB<br />

Get POI/<br />

Response<br />

API FLICKR<br />

Get POI/Response<br />

API NK (?)<br />

(początki)<br />

NK-Nasza<br />

Klasa<br />

Twitter<br />

API Panoramica<br />

Get POI/Response<br />

Facebook<br />

Użytkownicy końcowi<br />

Dostawca kategorii POI<br />

Flickr<br />

Panoramica<br />

Dostęp do informacji w portalach<br />

społecznosciowych poprzez linki dostarczone<br />

(Get POI/Response) przez system<br />

Rys. 6. Przykład architektury SOA mobilnego systemu informacyjnego typu reality browser (Źródło: własne)<br />

Fig. 6. Example SOA mobile information system that is „reality browser”<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 85


Rys. 7. Przykład danych dotyczących wizualizacji kategorii POI (Źródło: [7])<br />

Fig. 7. Example of data visualization POI category<br />

Struktura danych opisujących POI<br />

POI ma przypisaną kategorię, a oprócz niej elementami struktury<br />

danych POI są m.in.:<br />

co najmniej nazwa, współrzędne geograficzne i maksymalna<br />

odległość od użytkownika, w jakiej ma być POI wyświetlany,<br />

potrzebna do decyzji, czy serwer ma przesyłać<br />

do aplikacji klienckiej wstępne dane o tym POI, np. ikonkę<br />

i parametry lokalizacyjne, dla domniemanego lub określonego<br />

przez użytkownika zakresu otoczenia;<br />

może mieć podany adres pocztowy, tekst informacji wyświetlanej,<br />

rodzaj obiektu reprezentującego POI (ikonka, obiekt 2D<br />

lub 3D), adres WWW pliku z obiektem graficznym wyświetlanym<br />

dla POI oraz inne dane związane np. z transformacjami<br />

obiektów 2D/3D (obroty, skalowanie w zależności od pozycji<br />

i odległości użytkownika od rzeczywistego obiektu);<br />

zdefiniowaną akcję, jaka będzie automatycznie uruchamiana,<br />

gdy użytkownik wykona dodatkową czynność np.<br />

naciśnie przycisk more info (przekierowanie na określoną<br />

stronę WWW, wyświetlenie formularza maila, odtworzenie<br />

zapisu dźwiękowego i inne akcje);<br />

określony rejon/obszar, w którym informacja o POI będzie<br />

się pojawiała (np. z kategorii Biura ARiMR – POI będące<br />

Biurami Agencji Restrukturyzacji i Modernizacji Rolnictwa<br />

z przypisanym obszarem województwa mazowieckiego nie<br />

będą wyświetlane dla położenia użytkownika poza tym województwem,<br />

bez względu na odległość ich od użytkownika),<br />

określony termin informacji (np. obiekt informujący o dacie<br />

zakończenia przyjmowania wniosków o dopłaty pojawiający<br />

się w POI będącymi biurami ARiMR w województwie<br />

mazowieckim, nie będzie wyświetlany po tej dacie).<br />

Warto pamiętać, że istnieją różnice<br />

w strukturach danych, zarówno kategorii<br />

jak i POI, pomiędzy różnymi systemami<br />

informacyjnych.<br />

Prognozy i fakty dotyczące<br />

rozwiązań z Augmented<br />

Reality<br />

Odbiór społeczny technologii AR,<br />

mierzony w 2009 r. przez analityków<br />

nagłówkami postów na portalach społecznościowych<br />

(np. Twitter, Facebook)<br />

jest ogromny i m.in. na tej podstawie<br />

analitycy formułują opinię, że technologia<br />

AR jest technologią przyszłości<br />

(a nawet, że była to „technologia roku”<br />

2009). Takie zainteresowanie i fascynacja<br />

AR rokuje uzasadnione nadzieje<br />

przełożenia tego zainteresowania na<br />

coraz większe zamówienia ze strony<br />

odbiorców biznesowych.<br />

Ostatni (z 2009 r.) raport Gartner<br />

Group z cyklu Hype Cycle Book dotyczący<br />

technologii HCI (Human-Computer-Interaction)<br />

[3] przewiduje:<br />

• że do 2012 r. AR będzie jedną z 10<br />

najbardziej „destrukcyjnych” technologii,<br />

w sensie powodowania największych<br />

zmian, najsilniejszego wpływu na dotychczasowy<br />

stan w zakresie modeli biznesowych, procesów, głównych<br />

źródeł dochodów, dynamikę przemysłu.<br />

Juniper Research Ltd. [4] przewiduje że:<br />

• w zakresie mobilnych rozwiązań AR największy przyrost<br />

dochodów w okresie najbliższych 5 lat (do 2014 r.) będzie<br />

pochodził z aplikacji reklamujących oraz z działań „przyrostowych”<br />

(aktualizacje kodu, baz danych, konsultacje,<br />

serwis i inne).<br />

ABI Reaserch [1], prognozuje:<br />

• wzrost przychodów z mobilnych aplikacji AR z 2 mln USD<br />

w <strong>2010</strong> r. do 350 mln USD w 2014 (wg Juniper Research<br />

Ltd. może on być jeszcze większy).<br />

Zgodnie z raportem „<strong>2010</strong> Digital Marketing Outlook”, opublikowanym<br />

przez SoDA (Societies of Digital Agencies) [5],<br />

• marketing cyfrowy w roku bieżącym stanie się jedną z popularniejszych<br />

form rozprowadzania informacji biznesowej<br />

i pozyskiwania klienta, a autorzy tego raportu przytaczają<br />

prognozy i tendencje, które według nich określą ramy<br />

marketingu internetowego w <strong>2010</strong> roku – jedną z nich jest<br />

„rola technologii Augmented Reality w budowaniu zaufania<br />

i znajomości produktów oraz branży”.<br />

W „The Horizon Report. <strong>2010</strong> Edition” [6], corocznym<br />

raporcie wydawanym przez The New Media Consortium<br />

i The EDUCAUSE Learning Initiative, poświęconym kluczowym<br />

trendom technologicznym w krótkim (12 miesięcy),<br />

średnim (2–3 lata) i dłuższym (4–5 lat) horyzoncie czasowym,<br />

• rozpowszechnienie rozwiązań (adopcja) AR przewidywane<br />

jest w średnim horyzoncie czasowym tzn. za 2–3 lata,<br />

gdy w prostej formie (np. mobilnych systemów AR) dotrze<br />

86<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Rys. 8. Przykład informacji AR z Sieci (linki graficzne do portali<br />

społecznościowych) o osobie, której twarz na obrazie z kamery<br />

smartfona została rozpoznana przez system (Źródło: [15])<br />

Fig. 8. Example of AR information from the Web as the graphic<br />

links to social networking sites about the person whose face was<br />

recognized by the system on the picture with the camera smartphone<br />

do masowego odbiorcy, w dalszym zaś horyzoncie plasowana<br />

jest technologia, wspomagająca AR, rozpoznawania<br />

gestów.<br />

W ramach funduszy unijnych i prywatnych prowadzone<br />

były i są prace nad zaawansowanymi aplikacjami AR:<br />

projekt iTacitus (Intelligent Tourism and Cultural Information<br />

through Ubiquitous Services) [14], finansowany z VI<br />

Programu Ramowego, ukończony w 2009 r., który wzbogaca<br />

panoramę zabytków rzymskich o wirtualne obiekty<br />

odtwarzające technikami AR historyczne scenerie i sceny,<br />

np. walki gladiatorów i reakcje publiczności, nanoszone na<br />

zdejmowany kamerą przez turystę obraz Coloseum (sceny<br />

wirtualne dopasowane do fragmentu zabytku obserwowanego<br />

kamerą);<br />

aplikacja TAT Augmented ID, rozwijana przez szwedzką<br />

firmę TAT (The Astonishing Tribe) AB [15], automatycznie<br />

rozpoznaje osobę na podstawie obrazu twarzy uzyskanego<br />

z kamery smartfona i nanosi na ten obraz informacje,<br />

w różnej formie, o osobie, po które sięga do Sieci<br />

(rys. 8);<br />

aplikacja SREngine [16], w opracowaniu przez japońską<br />

firmę Kousei, Inc., jeden z bardziej intrygujących projektów<br />

AR, podobnie jak mniej zaawansowane wcześniejsze<br />

projekty Mobilizy's Wikitude i Tonchidot's Sekai Camera<br />

oraz jak obecnie rozwijana technologia Kooaba, rozpoznaje<br />

kształty, ale w sposób bardziej zaawansowany, ponieważ<br />

potrafi rozpoznać miejsca, przedmioty, statyczne<br />

sceny widziane kamerą smartfona i nanosi o nich informacje<br />

np. porównawcze ceny produktu w centrum handlowym<br />

lub nazwy drzew czy roślin napotkane w przyrodzie.<br />

Te dane, wydaje się są w stanie przekonać, że warto poświęcić<br />

co najmniej uwagę na mobilne zastosowania AR.<br />

Zakończenie<br />

Są już pierwsze polskie podmioty gospodarcze, które dostrzegły<br />

szybko rozwijający się światowy (a właściwie międzynarodowy<br />

– ze względu na przetwarzanie „w chmurze”) trend<br />

rozwoju mobilnych zastosowań AR i włączają się do niego,<br />

na razie jako dostawcy kontentu POI (wprowadzanego przez<br />

społeczność internetową, nieodpłatnie – warto zauważyć). Nie<br />

ma obecnie (połowa <strong>2010</strong> r.) w pełni polskich rozwiązań, jak<br />

i nie ma w pełni lokalizowanych na warunki polskie rozwiązań<br />

zagranicznych. Bo jeśli można spotkać aplikacje, w których<br />

możliwa jest kastomizacja interfejsu z użytkownikiem, to trzeba<br />

jeszcze poczekać, by informacje dostarczane z różnych baz<br />

dostawców usług były w języku polskim. Są też niekiedy kłopoty<br />

w uwzględnianiu Polski w systemach rozliczeń płatności<br />

internetowych w tych mobilnych systemach AR. Wszelkie dane<br />

jednak wskazują, że w krótkim czasie po informacje będziemy<br />

sięgać przede wszystkim do swojego smartfona, a nie komputera<br />

stacjonarnego lub laptopa, co potwierdzają ostatnie doniesienia<br />

z rynku usług mobilnych. Oto wraz z dyfuzją technologii<br />

3G spodziewany jest dalszy wykładniczy wzrost pobieranych<br />

danych poprzez mobilny Internet, zwłaszcza w krajach Zachodniej<br />

Europy i Ameryki Północnej. Przykładowo, Amerykanin<br />

średnio w ciągu roku od 2009 r. do <strong>2010</strong> r. zwiększył mobilny<br />

pobór danych o przeszło połowę, ze 100 do 159 megabajtów.<br />

Problemem operatorów sieci mobilnych jest niezwiększanie się<br />

proporcjonalne ich przychodów ze zwiększonego przesyłu danych.<br />

Oby środki zaradcze, jakie na pewno szybko znajdą, nie<br />

powstrzymały trendu i nie obaliły prognoz. Bo one napawają<br />

nadzieją, że również krajowy rynek usług mobilnych stanie się<br />

międzynarodowym partnerem w zakresie mobilnych systemów<br />

AR, partnerem równym liderom takim jak np. holenderski Layar<br />

czy austriacki Wikitude, a nie tylko tzw. „trzecim partnerem”, co<br />

ma miejsce obecnie i to w bardzo początkującym stadium. Niezagospodarowanych<br />

obszarów przez „przeglądarki AR”, poza<br />

zagospodarowanymi takimi jak systemy operacyjne iPhone OS<br />

i Andoird Google’a, jest wiele, choćby systemy operacyjne Nokii<br />

– dotychczasowy Symbian i wschodząca gwiazda (?) MeeGo.<br />

Warto pamiętać, że mimo obecnej przewagi konkurencyjnej<br />

Apple’a, do Nokii nadal należy 30% udziału w rynku urządzeń<br />

mobilnych…<br />

Literatura<br />

[1] Mobile Applications Market Data, The Mobile Consumer Research<br />

Service Report, ABI Research, June <strong>2010</strong>.<br />

[2] Cawood S., Fiala M.,: Augmented Reality. A Practical Guide, Editor<br />

The Pragmatic Programmers, LLL., 2008.<br />

[3] Hype Cycle for Human-Computer Interaction Report. July 2009,<br />

Gartner Group<br />

[4] Embedded Mobile&M2 M Strategies. Healthcare, Telematics,<br />

Metering and Connected Buildings 2009-2014. Juniper Research,<br />

2009.<br />

[5] Taylor R.: Reality, But Not As We Know It, Raport “Two Thousand<br />

and Ten Digital Marketing Outlook”, wyd. Society of Digital Agencies,<br />

<strong>2010</strong>.<br />

[6] The Horizon Report <strong>2010</strong> Edition by The New Media Consortium<br />

& The EDUCAUSE Learning Initiative, <strong>2010</strong>.<br />

Netografia<br />

[7] www.layar.com<br />

[8] www.wikitude.org<br />

[9] www.junaio.com<br />

[10] http://sekaicamera.com<br />

[11] www.acrossair.com<br />

[12] http://www.t-immersion.com<br />

[13] http://www.kooaba.com<br />

[14] www.itacitus.org<br />

[15] www.tat.se<br />

[16] www.srengine.com 17. www.maperia.pl<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 87


Using multi-frequency coherent signal to measurement<br />

of frequency response of narrow-band LF and HF circuits<br />

(Zastosowanie wieloczęstotliwościowego sygnału koherentnego<br />

do pomiaru charakterystyk częstotliwościowych układów<br />

wąskopasmowych m.cz. oraz w.cz.)<br />

dr JAN DUCHIEWICZ, dr ANDRZEJ S. SOWA, dr JERZY S. WITKOWSKI,<br />

dr hab. ANDRZEJ FRANCIK, dr ANDRZEJ L. DOBRUCKI, mgr BARTOSZ IDŹKOWSKi<br />

Politechnika Wrocławska, <strong>Instytut</strong> Telekomunikacji, Teleinformatyki i Akustyki<br />

mgr TOMASZ DUCHIEWICZ, ELOKON Polska, Sp. z. o. o. Warszawa<br />

Coherent (synchronous) detection is applied first of all in cases<br />

of very weak signal reception. The presence of a so-called<br />

reference signal, synchronous with respect to the received<br />

signal carrier is necessary to make it possible to apply coherent<br />

detection. Coherent detection can be realized by means<br />

of an analogue or digital set-up – the method of realization<br />

depends mainly on the frequency range of received signals.<br />

If a signal is in the form of a HF carrier, amplitude modulated<br />

with a periodic LF signal, a two-stage coherent detection<br />

is applied for the reception of such a signal. The first coherent<br />

detection (in the HF area) is usually analogue and is performed,<br />

because of the unknown phase shift of the HF channel,<br />

by means of an I&Q demodulator. Output signals from the I&Q<br />

demodulator are subjected to the second coherent detection,<br />

also using, because of the unknown phase shift of the LF<br />

channel, the quadrature detection. This second detection is<br />

currently most often realized in a digital way.<br />

The need to use a HF measuring signal modulated simultaneously<br />

with many LF signals sometimes occurs in the<br />

measuring procedure. The application in such a case of synchronous<br />

detection of all modulating LF signals allows high<br />

sensitivity of the measuring set-up to be obtained – similar to<br />

that obtained in the case of a single modulating signal. Such<br />

measuring signals have many advantages.<br />

In general the application of synchronous detection for M of<br />

modulating signals requires applying M of modulating generators<br />

and M of synchronous detectors, utilizing 2xM of amplifiers (I and<br />

Q channels) and 2xM of analog-to-digital converters (ADC).<br />

The authors proposed the principle of simultaneous homodyne<br />

detection of many modulating signals [1, 2, 4 – 9] and designed<br />

the multi-channel homodyne receiver, making it possible<br />

to measure M signals by means of one M-output generator and<br />

one synchronous detector utilizing two amplifiers and two ADC.<br />

The multi-channel homodyne receiver is described in the<br />

next part of the paper. An example is presented of using the<br />

multi-channel homodyne receiver to measure frequency characteristics<br />

of the selective object.<br />

multi-channel coherent demodulator. It is obvious that such<br />

a demodulator can be realized only in the digital technology.<br />

A simplified block diagram of a multi-channel coherent receiver<br />

designed by the authors is shown in Fig. 2 and 3.<br />

The receiver contains two independent measuring channels<br />

for complex signals I and Q, present at the outputs of the I&Q<br />

demodulator. A programmed amplifier, a band-pass filter and<br />

a 16 bit ADC are applied in each channel. A modulating generator<br />

implemented in CPLD circuit produces 16 signals synchronous<br />

with respect to the reference frequency. The reference<br />

frequency is also the processing frequency of both ADC.<br />

Fig. 1. HF Measuring setup with a multi-channel coherent receiver<br />

Rys. 1. Układ pomiarowy w.cz. z wielokanałowym, koherentnym<br />

odbiornikiem<br />

Simultaneous coherent reception of many<br />

signals<br />

A HF measuring setup proposed by the authors, equipped<br />

with a multi-channel coherent receiver is shown in Fig.1. The<br />

simultaneous, coherent reception of many signals requires<br />

that each of these signals possess a different frequency, and<br />

all these frequencies are synchronous with respect to the<br />

common reference frequency. The synchronism of all these<br />

signals with respect to the common reference frequency<br />

enables to separate each of them by means of a common,<br />

88<br />

Fig. 2. Measuring setup with the authors’ multi-channel coherent<br />

receiver<br />

Rys. 2. Wielokanałowy, koherentny układ odbiorczy – zasada<br />

(opracowany przez autorów)<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


The module of the demodulated lines can be determined<br />

directly from (4) on the basis of dependence:<br />

<br />

Re<br />

2<br />

Im<br />

2<br />

I Sm = ⎡<br />

I S ⎤<br />

m + ⎡<br />

I S ⎤<br />

m<br />

⎣ ⎦ ⎣ ⎦<br />

(5a)<br />

<br />

2 2<br />

Re<br />

Im<br />

Q<br />

Sm = ⎡<br />

Q<br />

S ⎤<br />

m + ⎡<br />

Q<br />

S ⎤<br />

m<br />

⎣ ⎦ ⎣ ⎦<br />

(5b)<br />

<br />

2<br />

m = ⎡ ⎡<br />

I m ⎤ +<br />

Q m<br />

S ⎣ S ⎦ S ⎤<br />

⎣ ⎦<br />

2<br />

(5c)<br />

Fig. 3. Simplified block diagram of the authors’ multi-channel coherent<br />

receiver<br />

Rys. 3. Wielokanałowy, koherentny układ odbiorczy – uproszczony<br />

schemat blokowy<br />

The digital coherent demodulation consists of multiplying<br />

of the train of registered signal samples by the digital reference<br />

function (sinusoidal or rectangular), and adding up the results<br />

of successive multiplications. Process of the digital coherent<br />

demodulation for both I and Q channels can be described in<br />

the following way:<br />

<br />

<br />

(1a)<br />

(1b)<br />

where S(n) – signal at the output of the ADC, R m<br />

(n) – digital<br />

reference function for m signal.<br />

Signals I and Q at AD converters outputs can be described<br />

as:<br />

<br />

<br />

I<br />

I<br />

S ( m) = S ( n) ⊗ R ( n)<br />

Q<br />

Q<br />

S ( m) = S ( n) ⊗ R ( n)<br />

M −1<br />

k⋅(2⋅M −1)<br />

⋅K<br />

I<br />

I<br />

S ( n) U ( n)<br />

= ∑ ∑<br />

m= 0 n=<br />

0<br />

M −1<br />

k⋅(2⋅M −1)<br />

⋅K<br />

Q<br />

Q<br />

S ( n) U ( n)<br />

= ∑ ∑<br />

m= 0 n=<br />

0<br />

m<br />

m<br />

(2a)<br />

(2b)<br />

The length of the train of registered samples is of the basic<br />

importance – the signal separation and the signal to noise ratio<br />

is improving with the increase in the length of this train. The<br />

designed receiver makes possible registration of practically<br />

non limited number of samples. From 1000 to 8000000 samples<br />

were registered during experiments. The sampling frequency<br />

was 180 kHz, frequencies of modulating signals were<br />

situated in the range of (3 – 6) kHz. The time of registration<br />

of 100000 samples was about 0.4 sec.<br />

Simultaneous reception of many coherent<br />

signals modulating HF carrier<br />

If an applied signal is of the form of a HF carrier, amplitude modulated<br />

with a group of sinusoidal coherent LF signals, each<br />

line of the spectrum belonging to the lower side band and to<br />

the upper side band should be, generally, determined separately<br />

in the demodulation process. The solution is an application<br />

of the technique known in the field of radio receivers as<br />

the „zero-IF” (homodyne, direct-conversion) technique [e.g. 3].<br />

It requires an extra shift of the Q signal by 90 o (it corresponds<br />

to multiplying by j in the complex number domain). Knowing<br />

I and Q signals one can determine the lower side band components<br />

and the upper side band components. The total of I and<br />

jQ signals allows to determine components of the lower side<br />

band of the amplitude modulated signal and the difference of<br />

I and jQ signals allows to determine components of its upper<br />

side band. Denoting the lower side band signal by A and the<br />

upper side band signal by B it can be written as [3]:<br />

where U(n) means the succeeding numeric values of the signal<br />

(1), K – total number of samples. Sinusoidal reference<br />

function:<br />

<br />

<br />

A = I + j ⋅Q<br />

B = I − j ⋅Q<br />

(6a)<br />

(6b)<br />

s<br />

⎡ n ⎤ ⎡ n ⎤<br />

Rm<br />

( n) = ∑ cos ⎢2⋅π<br />

⋅ ⎥ + j ⋅ sin 2 π (3)<br />

n= 0 N( m) ∑ ⎢ ⋅ ⋅ ⎥<br />

⎣ ⎦ n=<br />

0 ⎣ N( m)<br />

⎦<br />

where N(m) = k · (M + m) – number of samples corresponding<br />

M −<br />

I<br />

Q<br />

I<br />

A( n) = S ( n) + j ⋅ S ( n) = U ( n)<br />

+<br />

to the single period of m signal, k – proportional ratio (k = 2<br />

was engaged), l = 0,1,2, L(m), L(m) – number of periods for <br />

(7a)<br />

M −1 k⋅(2⋅M −1) ⋅K M −1<br />

k⋅(2⋅M −1)<br />

⋅K<br />

m signal.<br />

I<br />

Q<br />

I<br />

Q<br />

A( n) = S ( n) + j ⋅ S ( n) =<br />

In general the phase of the demodulated signals is unknown<br />

which causes the need to apply a quadrature detection,<br />

∑ ∑ U ( n) + ∑ ∑ U ( n)<br />

m= 0 n= 0 m= 0 n= k⋅ ( M + m)/2<br />

making it possible to determine both amplitudes and phases<br />

of the received signals.<br />

M −<br />

I<br />

Q<br />

I<br />

Dependence (1) with applying the reference function (3), <br />

B( n) = S ( n) − j ⋅ S ( n) = U ((7b)<br />

n)<br />

−<br />

allows determining real and imagining components of each<br />

M −1 k⋅(2⋅M −1) ⋅K M −1<br />

k⋅(2⋅M −1)<br />

⋅K<br />

signal:<br />

I<br />

Q<br />

I<br />

Q<br />

B( n) = S ( n) − j ⋅ S ( n) = U ( n) −<br />

U ( n)<br />

<br />

<br />

N ( m) N ( m)<br />

= S + j⋅<br />

S<br />

Re Im<br />

I S m I m I m<br />

= S + j⋅<br />

S<br />

Re Im<br />

Q S m Q m Q m<br />

(4a)<br />

(4b)<br />

Executing operations (6) on trains (2) one can obtain 2xM<br />

of numeric trains, representing separated sums of upper side<br />

band lines A(n) and lower side band lines B(n). It can be written<br />

as:<br />

∑ ∑ ∑ ∑<br />

m= 0 n= 0 m= 0 n= k⋅ ( M + m)/2<br />

1k⋅(2⋅M −1) ⋅K M −1<br />

k⋅(2⋅M<br />

−1<br />

∑ ∑ ∑ ∑<br />

m= 0 n= 0 m= 0 n= k⋅ ( M + m<br />

1k⋅(2⋅M −1) ⋅K M −1<br />

k⋅(2⋅M<br />

−1)<br />

∑ ∑ ∑ ∑<br />

m= 0 n= 0 m= 0 n= k⋅ ( M + m<br />

In expressions (7) the fact was utilized, that the shift<br />

of the periodic signal by 90 o is corresponding to the shift<br />

of the numeric samples train representing it by ¼ of the num-<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 89


er of samples falling into one period of the signal. Carrying<br />

out a quadrature detection circumscribed by expression (1) for<br />

each component of signals A and B circumscribed by expressions<br />

(7) complex components of each line of both lower side<br />

band and upper side band are determined:<br />

Re Im<br />

Re Im<br />

A = A + j⋅<br />

A B = B + j ⋅ B<br />

(8)<br />

m m m<br />

m m m<br />

where m = 0,1,2... M-1 – number of the succeeding line.<br />

From dependence (7) it is possible to determine both the<br />

amplitude and the phase of each line:<br />

<br />

<br />

Re<br />

2<br />

Im<br />

2<br />

A A<br />

Am = ⎡A ⎤ ⎡<br />

m A ⎤<br />

⎣ ⎦ + ⎣ m ⎦ , ϕm<br />

= arctg( ) A<br />

Re<br />

m<br />

Im<br />

m<br />

Re<br />

2<br />

Im<br />

2<br />

B B<br />

Bm = ⎡B ⎤ ⎡<br />

m B ⎤<br />

⎣ ⎦ + ⎣ m ⎦ , ϕm<br />

= arctg( ) B<br />

Re<br />

m<br />

Im<br />

m<br />

(9a)<br />

(9b)<br />

Fig. 5. Block diagram of transmission setup for the measurement<br />

of frequency characteristics of HF object<br />

Rys. 5. Schemat blokowy transmisyjnego układu pomiarowego<br />

charakterystyki częstotliwościowej układu w.cz.<br />

If amplitude characteristics of the tested HF object are flat<br />

in the range of +/- f mod max<br />

where f mod max<br />

is the maximum modulating<br />

frequency, the lines of the lower side band and of the<br />

upper side band of the demodulated HF signal have the same<br />

amplitudes and it is usually enough assigning of the module<br />

of the demodulated lines in accordance to dependence:<br />

S = A + B = 2⋅ A = 2⋅B<br />

(10)<br />

Obviously, this module can be determined directly from (4)<br />

on the basis of dependence (5).<br />

Example<br />

m m m m m<br />

Simultaneous, multipoint measurement of frequency<br />

transmittance of a narrow-band quartz filter<br />

Simultaneous, multipoint measurement of transmittance<br />

of a narrow-band circuit is possible by means of many HF<br />

generators tuned to many little differing frequencies. Also the<br />

other solution is possible – it relies on the usage of a HF signal,<br />

amplitude modulated with several LF signals. Simplified<br />

amplitude spectrum of a HF signal modulated with the multifrequency<br />

LF signal described above is shown in Fig. 4.<br />

The amplitude spectrum of measured signal is expected<br />

to be identical in both methods. The transmittance determined<br />

for each line of the spectrum in the latter way is identical<br />

to the one determined in case of the independent generator<br />

with a suitable frequency equal to the frequency of the line.<br />

Fig. 6. Block diagram of reflection setup for the measurement of<br />

frequency characteristics of HF object<br />

Rys. 6. Schemat blokowy odbiciowego układu pomiarowego<br />

charakterystyki częstotliwościowej układu w.cz.<br />

The possibility of applying a two-stage homodyne detection<br />

that uses a multi-channel coherent receiver is an obvious<br />

advantage of the solution with the amplitude modulated signal.<br />

The homodyne detection allows to obtain high sensitivity<br />

of the setup. It also allows the simultaneous, multi-signal measurement<br />

for some types of non-stationary circuits, impossible<br />

to carry out with the sweep generator method. Application<br />

of the described earlier<br />

multi-channel coherent<br />

receiver allows the simul-<br />

Fig. 4. Simplified amplitude spectrum of multi-frequency LF signal (a) and HF signal modulated with<br />

a multi-frequency LF signal (b). H(f ) – example frequency characteristics of the object under test<br />

Rys. 4. Uproszczone widmo częstotliwości: (a) wieloczęstotliwościowego sygnału m.cz., (b) sygnału<br />

w.cz. zmodulowanego amplitudowo sygnałem (a). H(f ) – przykładowa charakterystyka częstotliwościowa<br />

badanego układu<br />

taneous measurement<br />

of the amplitudes and<br />

phases for 32 frequencies<br />

in both the transmission<br />

(Fig. 5) and reflection<br />

(Fig. 6) setup.<br />

If the need of measurement<br />

in the broader<br />

frequency band arises,<br />

the HF generator should<br />

be sequentially tuned by<br />

an adequate frequency<br />

interval to cover the required<br />

frequency spectrum.<br />

90<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


The amplitude modulated measuring signal from the HF<br />

generator (e.g. microwave) is applied to the input of a measured<br />

HF object. The signal from the output of this object is directed<br />

to the RF input of the I&Q demodulator. The not-modulated<br />

signal is applied to the LO input of the I&Q demodulator<br />

from the same HF generator. I and Q signals being the result<br />

of the synchronous HF demodulation drive I and Q inputs of<br />

the multi-channel coherent receiver. As a result the values of<br />

the amplitudes and phases of all components of the upper and<br />

lower sidebands are determined which allows to assign the<br />

transmittance of the object under test.<br />

In the reflection setup, in similar way as in the transmission<br />

setup, the amplitude modulated measuring signal from the HF<br />

generator (e.g. microwave) is applied to the input of a measured<br />

object by a directional coupler. Signal reflected from<br />

the object appears on the decoupling output of the directional<br />

coupler and is directed to the RF input of the I&Q demodulator.<br />

The not-modulated signal is applied to the LO input of the<br />

I&Q demodulator from the same HF generator. As a result of<br />

the measurement the values of the amplitudes and phases of<br />

all components of the upper and lower sideband are stated.<br />

Finally the measurement allows to determine the reflection coefficient<br />

of the object.<br />

Results of measurements<br />

The above described measuring method can be applied<br />

for measurements in a wide range of frequencies. As an<br />

example a narrow-band quartz filter was measured in the<br />

transmission setup. The LF generator produced 16 signals<br />

(3 to 6 kHz) modulating the HF signal. Taking into account<br />

both side bands, the simultaneous measurement was made<br />

for 32 frequencies. The measurement was repeated for the<br />

next two frequencies of the HF generator, with a frequency<br />

step equal to 3 kHz It allowed to cover the total range<br />

of 18 kHz (96 frequency points). Continuation of the measurement<br />

required a change of the frequency of the HF<br />

generator by 12 kHz and a repetition of measurements in<br />

the way described earlier. This procedure can be, certainly,<br />

multiplied. The result of the measurement of the filter transmittance<br />

is shown in the Fig.7.<br />

The results obtained demonstrate that the usage of multichannel<br />

synchronous detection allows the measurement for<br />

many (here 32) frequencies at the same time with good accuracy<br />

and high dynamics.<br />

Fig. 7. Transmittance of the narrow-band quartz filter<br />

Rys. 7. Transmitancja wąskopasmowego filtru kwarcowego<br />

Conclusions<br />

The measurement method proposed by the authors and applied<br />

to the receiver of their design allows simultaneous measurement<br />

by means of many periodic signals. Frequencies<br />

of the signals were chosen in such a way that the executed<br />

measurement is the coherent measurement for all used signals.<br />

Due to coherent detection, high sensitivity of the receiver<br />

and high dynamic range of the measuring set-up were obtained.<br />

Doubling of the measurement channel of the receiver<br />

allows its effective application in measurements of HF signals<br />

demodulated by means of the I&Q demodulator. In this case<br />

two-stage coherent detection allows high sensitivity and high<br />

dynamic range of the measurement set-up for modulated HF<br />

signals to be obtained. The measurement method described<br />

ensures the simultaneous measurement of many signals in<br />

the similar period of time as in the case of one signal in a single-channel<br />

receiver and for the same sensitivity of the receiver.<br />

The proposed method allows simultaneous multi-point measurement<br />

of frequency characteristics of some non-stationary<br />

selective narrow-band circuits or systems.<br />

The example presented in the paper demonstrate the<br />

practical usefulness of the technology of simultaneous coherent<br />

measurement of many signals in multi-point measurement<br />

of selective, particularly narrow-band circuits. Other<br />

examples of the application are given in [1, 7–9].<br />

The designed coherent receiver can be applied wherever<br />

the simultaneous measurement of many signals is necessary<br />

or recommended, both HF as well as LF. In case of HF signals<br />

the set-up shown in Fig. 2 has to be equipped with a proper<br />

HF measurement channel designed for the frequency of interest.<br />

References<br />

[1] Duchiewicz J., Sowa A., Witkowski J., Duchiewicz T.: Simultaneous<br />

coherent measurement of many HF signals. Proc. of the XVI<br />

Intern. Conf. on Microwaves, Radar and Wireless Communications,<br />

Poland, Krakow, May 22–26, 2006, Vol.1.<br />

[2] Sowa A., Witkowski J., Duchiewicz J.: Idea and Design Of Modulating<br />

Generator and Synchronous Receiver for Multi-Channel<br />

MST System. Proc.of the 17th Int. Wroclaw Symp. and Exhibition<br />

on EMC, Wroclaw, Poland, June 29–July 1, 2004.<br />

[3] Abidi A.: Direct-Conversion Radio Transceivers for Digital Communications.<br />

IEEE J. Solid State Circuits, vol. 30, Dec. 1995.<br />

[4] Duchiewicz J., Sowa A., Witkowski J., Duchiewicz T., The Use of<br />

Multi-Frequency Coherent Signal to Measurement of Frequency<br />

Response of Narrow-Band Circuits, IEEE International Conference<br />

on „Computer as a tool” EUROCON 2007, Warsaw, Poland,<br />

September 9–12, 2007.<br />

[5] Duchiewicz J., T. Duchiewicz, A. E. Sowa, J. S. Witkowski:<br />

Method of Physical Phenomena Measurement and a Setup to<br />

Physical Phenomena Measurement. Patent Pending, P-378477,<br />

2005, Poland (in Polish).<br />

[6] Duchiewicz J., T. Duchiewicz T., A.E. Sowa, J.S.Witkowski:<br />

Method of Physical Phenomena Measurement with a Multi-Channel<br />

Homodyne Detection and a Setup for Physical Phenomena<br />

Measurement with a Multi-channel Homodyne Detection. Patent<br />

Pending, P-377185, 2005, Poland (in Polish).<br />

[7] Duchiewicz J., Sowa A., Witkowski J., Duchiewicz T.: Synchronous<br />

measurement of inter-modulation products using a multifrequency<br />

coherent signal. International Conference on Signals<br />

and Electronic Systems ICSES’08, September 14–17, 2008,<br />

Kraków, Poland.<br />

[8] Duchiewicz J., Sowa A., Witkowski J., Duchiewicz T.: Multi-Frequency<br />

Coherent Signal in Measurement Technology. Proc. 19th<br />

Int. Wroclaw Symp. and Exhibition on EMC, Wroclaw, Poland,<br />

June 11–13, 2008.<br />

[9] Duchiewicz J., Sowa A., Witkowski J., Duchiewicz T., Francik A.,<br />

Dobrucki A., Idźkowski B., Sadowski A.: Multi-frequency coherent<br />

signal – properties and applications in measuring technique.<br />

Electronics – Constructions, Technologies, Applications, July<br />

<strong>2010</strong> (in Polish).<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 91


Radiokomunikacyjne aspekty planu implementacji<br />

strategii e-nawigacji<br />

dr inż. KAROL KORCZ<br />

Akademia Morska w Gdyni, Katedra Telekomunikacji Morskiej<br />

Szybki rozwój technologiczny w zakresie elektroniki, radiokomunikacji<br />

i informatyki inspiruje pojawianie się coraz to nowych<br />

propozycji zmian urządzeń i systemów stosowanych<br />

na statkach morskich. Zasady i zakres wyposażania statków<br />

morskich, związane z zapewnieniem ich bezpieczeństwa, są<br />

ściśle regulowane przez Międzynarodową Organizację Morską<br />

– IMO (International Maritime Organization). W 2006 roku<br />

kilka państw zgłosiło na forum Komitetu bezpieczeństwa na<br />

morzu IMO – MSC (Maritime Safety Committee), propozycję<br />

przygotowania szerokiej strategii włączenia nowych technologii<br />

w sposób strukturalny, z zapewnieniem ich zgodności<br />

z już istniejącymi różnymi technologiami nawigacyjnymi i komunikacyjnymi<br />

oraz usługami. Nadrzędnym celem tej strategii<br />

miałaby być poprawa efektywności, bezpieczeństwa i zmniejszenie<br />

kosztów całego systemu, zapewniającego globalne<br />

pokrycie oraz mającego zastosowanie dla wszystkich typów<br />

statków morskich [1].<br />

W odpowiedzi na tą propozycję MSC podjęło decyzję<br />

o rozpoczęciu prac nad projektem „Przygotowanie strategii<br />

e-nawigacji”, zlecając prowadzenie go przez dwa podkomitety<br />

techniczne IMO: Podkomitet ds. bezpieczeństwa żeglugi<br />

– NAV (Sub-Committee on Safety of Nawigation) oraz Podkomitet<br />

ds. radiokomunikacji, poszukiwań i ratownictwa – COM-<br />

SAR (Sub-Committee on Radiocommunications, Search and<br />

Rescue). Koordynatorem projektu został wyznaczony Podkomitet<br />

NAV [1]. Do prac nad tym projektem zaproszone zostały<br />

również: Międzynarodowe Stowarzyszenie Władz Latarni<br />

Morskich – IALA (International Association of Marine Aids to<br />

Navigation and Lighthouse Authorities), Międzynarodowa Organizacja<br />

Hydrograficzna – IHO (International Hydrographic<br />

Organization) i Międzynarodowa Federacja Stowarzyszeń<br />

Kapitańskich – IFSMA (The International Federation of Shipmasters’<br />

Associations).<br />

„Strategia e-nawigacji” została przyjęta na 85. sesji Komitetu<br />

bezpieczeństwa na morzu IMO (MSC85), która miała<br />

miejsce w grudniu 2008 r. w Londynie. Konsekwencją tej<br />

decyzji było wprowadzenie do programu pracy podkomitetów<br />

IMO NAV, COMSAR i dodatkowo Podkomitetu ds. szkolenia<br />

zawodowego i obowiązków wachtowych – STW (Sub-Committee<br />

on Standards of Training and watchkeeping), na lata<br />

2009 – 2012 nowego tematu „Przygotowanie planu implementacji<br />

strategii e-nawigacji”.<br />

Ogólne założenia strategii e-nawigacji<br />

W pracach nad strategią e-nawigacji przyjęto, jako podstawowe<br />

założenie, iż potencjalny system powinien być rozwijany<br />

w funkcji oczekiwań jego użytkowników (na statku i na lądzie),<br />

a nie możliwości technicznych aktualnie dostępnych technologii<br />

informacyjnych i systemów radiokomunikacyjnych [2, 3].<br />

W odniesieniu do technicznych rozwiązań komunikacyjnych,<br />

we wstępnej fazie rozważań nad strategią e-nawigacji,<br />

przyjęła następujące założenia [2, 3]:<br />

• priorytetem jest transmisja danych; telefonia będzie mogła<br />

być częścią e-nawigacji ale na obecnym etapie nacisk powinien<br />

być położony na transmisję danych,<br />

92<br />

• mogą być różne wymagania odnośnie dostępności danych<br />

w zależności od nadawanej informacji (priorytety),<br />

• statek może odbierać bardzo dużo informacji, istotne dla<br />

załogi będzie zatem właściwe zarządzanie danymi,<br />

• e-nawigacja nie powinna być ograniczona jedynie do funkcji<br />

związanych z bezpieczeństwem i ochroną na morzu<br />

oraz środowiska naturalnego, ale powinna również dawać<br />

wyraźne korzyści dla statków i załóg,<br />

• powinna być wykorzystana transmisja danych drogą satelitarną<br />

jak i z użyciem zakresów naziemnych MF, HF i VHF.<br />

Po dyskusjach przyjęto następującą definicję koncepcji<br />

e-nawigacji [4]: E-nawigacja jest zharmonizowanym zbieraniem,<br />

integracją, wymianą, prezentacją i analizą morskich<br />

informacji na statkach i lądzie, za pomocą środków<br />

elektronicznych, poprawiających nawigację od portu do<br />

portu i powiązane serwisy bezpieczeństwa oraz ochronę<br />

na morzu, a także ochronę środowiska naturalnego.<br />

Zgodnie z tą definicją, zadaniem e-nawigacji ma być spełnienie<br />

obecnych i przyszłych potrzeb użytkowników, poprzez<br />

harmonijne współdziałanie morskich systemów nawigacyjnych<br />

oraz wspierających je serwisów lądowych. Natomiast<br />

celem nadrzędnym jest poprawa bezpieczeństwa nawigacji<br />

i redukcja szeroko rozumianych błędów, w tym powodowanych<br />

przez człowieka.<br />

Główne cele e-nawigacji<br />

Aby spełnić oczekiwania użytkowników na statkach i na lądzie,<br />

do głównych celów e-nawigacji powinno należeć [4, 5]:<br />

• ułatwienie bezpiecznej nawigacji statków mając wzgląd na<br />

informację hydrograficzną, meteorologiczną i nawigacyjną<br />

oraz ryzyko,<br />

• ułatwienie obserwacji ruchu statków i zarządzanie nim korzystając<br />

z dostępu do stosownej brzegowej/lądowej bazy<br />

danych – jeśli istnieje,<br />

• ułatwienie łączności, włącznie z wymianą danych w relacji:<br />

statek-statek, statek-stacja brzegowa, stacja brzegowastatek,<br />

stacja brzegowa-stacja brzegowa i pomiędzy innymi<br />

użytkownikami,<br />

• zapewnienie możliwości zwiększenia efektywności transportu<br />

i logistyki,<br />

• wsparcie efektywnych działań w sytuacji kryzysowej oraz<br />

podczas akcji poszukiwania i ratowania – akcji SAR (Search<br />

And Rescue),<br />

• zaprezentowanie określonego poziomu dokładności, integracji<br />

i ciągłości, odpowiedniego do systemu decydującego<br />

o bezpieczeństwie,<br />

• integracja i prezentacja informacji na statku i na lądzie poprzez<br />

interfejs użytkownika zapewniający maksymalizację<br />

korzyści dotyczących bezpieczeństwa żeglugi i minimalizację<br />

jakichkolwiek zagrożeń wynikających z dezorientacji<br />

i błędnej interpretacji ze strony użytkownika,<br />

• integracja i prezentacja informacji na statku i na lądzie<br />

umożliwiająca zarządzanie obciążeniem pracy użytkowników,<br />

a także motywowania i wspierania ich podczas podejmowania<br />

decyzji,<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


• włączenie wymagań szkoleniowych użytkowników poprzez<br />

przygotowane i wdrożone procesy,<br />

• ułatwienie globalnego pokrycia, wprowadzenie jednolitych<br />

standardów i ustaleń, wzajemnej kompatybilności i współdziałania<br />

sprzętu, systemów, symboliki i operacyjnych<br />

procedur tak, aby uniknąć potencjalnych różnic pomiędzy<br />

działaniami użytkowników,<br />

• wsparcie integracji tak, aby ułatwić wykorzystanie e-nawigacji<br />

przez wszystkich potencjalnych morskich użytkowników.<br />

Priorytetowe potrzeby użytkowników<br />

Priorytetowe potrzeby użytkowników powinny być określone<br />

z uwzględnieniem elementów składowych definicji e-nawigacji<br />

i odpowiednich analiz oraz czynnika ludzkiego, odnoszących<br />

się do wszystkich użytkowników. W ustalaniu tych potrzeb<br />

uczestniczyły państwa członkowskie IMO oraz różne organizacje<br />

powiązane z gospodarką morską. Przyjęte na dzień dzisiejszy<br />

potrzeby odnoszą się głównie do statków podlegających<br />

Międzynarodowej konwencji o bezpieczeństwie życia na<br />

morzu – Konwencja SOLAS (International Convention for the<br />

Safety of Life at Sea) i oczekiwań władz morskich. Bardziej<br />

szczegółowe potrzeby będą zidentyfikowane później w ramach<br />

prac nad planem wdrożenia e-nawigacji. W wyniku tego<br />

metodycznego podejścia przyjęto następujące, priorytetowe,<br />

wstępne potrzeby użytkowników:<br />

• ujednolicone morskie informacje/struktura danych;<br />

na statku potrzebne są informacje zgodne z obowiązującymi<br />

regulacjami, odnoszące się do planowania i realizacji podróży<br />

i oceny ryzyka nawigacyjnego; informacje te powinny<br />

być dostępne z jednego zintegrowanego systemu; użytkownicy<br />

lądowi potrzebują informacje odnoszące się do ich morskiej<br />

dziedziny, włączając statyczne i dynamiczne informacje<br />

o statkach i ich podróżach,<br />

• zautomatyzowane i standaryzowane funkcje raportowania;<br />

e-nawigacja powinna zapewniać zautomatyzowane i standaryzowane<br />

funkcje raportowania dla optymalnej komunikacji ze<br />

statkiem i pozyskiwania informacji o podróży; obejmuje to nadawanie<br />

na ląd informacji dotyczących bezpieczeństwa, nadawanie<br />

informacji z lądu na statek i informacji odnoszących<br />

się do ochrony statku i portu oraz ochrony środowiska naturalnego,<br />

a przeznaczonych do wszystkich użytkowników,<br />

• efektywna i dobrej jakości komunikacja;<br />

użytkownicy lądowi potrzebują efektywnych środków do komunikacji<br />

ze statkami aby zapewnić bezpieczeństwo, ochronę<br />

statków i portu oraz ochronę środowiska, a także aby dostarczać<br />

informacji operacyjnych; aby komunikacja z i między<br />

statkami była efektywna, powinna wykorzystywać środki audio-wizualne<br />

oraz standardowe określenia dla minimalizacji<br />

problemów językowych, sprawiających kłopot operatorom,<br />

• prezentacja przyjazna użytkownikowi;<br />

wyświetlacze nawigacyjne powinny być projektowane tak, aby<br />

jasno wskazywały ryzyko i optymalnie wspierały podejmowanie<br />

decyzji; wszystkie wyświetlacze powinny być projektowane<br />

tak, aby ograniczały możliwość pomyłki i błędnej interpretacji<br />

w przypadku większej liczby informacji dotyczącej bezpieczeństwa;<br />

system e-nawigacji powinien być zaprojektowany<br />

tak, aby odpowiednio zajmował i motywował użytkownika,<br />

• interfejs przyjazny użytkownikowi;<br />

ponieważ urządzenia elektroniczne/informacyjne odgrywają<br />

coraz większą rolę, muszą być budowane tak, aby obejmowały<br />

i prezentowały informacje w sposób wizualny, z uwzględnieniem<br />

wiedzy i doświadczenia użytkowników; prezentacja<br />

informacji dla wszystkich użytkowników powinna być zaprojektowana<br />

tak, aby zredukować możliwość błędów jednej<br />

osoby a uwydatniać pracę zespołową; jest wyraźna potrzeba<br />

rozwiązań zgodnych z zasadami ergonomii zarówno co do fizycznej<br />

konstrukcji urządzeń jak i zastosowania oświetlenia,<br />

kolorów, symboliki i języka,<br />

• integralność danych i całego systemu;<br />

system e-nawigacji powinien być elastyczny i brać pod uwagę<br />

sprawy wiarygodności danych i ich integralności tak, aby był<br />

niezawodny; należy rozważyć sprawę redundancji danych,<br />

w szczególności w odniesieniu do systemów określających<br />

pozycję statków,<br />

• analiza pracy systemu;<br />

system e-nawigacji powinien wspierać podejmowanie dobrych<br />

decyzji, poprawiać pracę i zapobiegać błędom jednoosobowym;<br />

aby to osiągnąć systemy statkowe powinny zawierać<br />

funkcje analizy, które wspierałyby użytkownika w stosowaniu<br />

przepisów, planowaniu podróży, ocenie ryzyka, unikaniu kolizji<br />

i wejścia na mieliznę, włączając obliczenia zanurzenia statku;<br />

systemy lądowe powinny wspierać analizę uwarunkowań<br />

środowiskowych, planowanie ruchu statków – w przyszłość,<br />

ocenę zagrożenia/ryzyka, wskaźniki raportowania i zapobieganie<br />

wypadkom,<br />

• kwestie implemantacji systemu;<br />

przed wdrożeniem technicznym należy zapewnić efektywne<br />

praktyki, szkolenia i zapoznawanie się z odpowiednimi aspektami<br />

e-nawigacji; do przeprowadzania szkoleń i ich oceny zalecane<br />

jest stosowanie symulatorów; e-nawigacja powinna,<br />

tak daleko jak to możliwe, potrzeć w przód ale i do tyłu oraz<br />

zapewnić zgodność z wymaganymi przepisami międzynarodowymi<br />

i krajowymi w zakresie urządzeń i systemów a także<br />

standardów eksploatacyjnych; gdzie to możliwe należy szukać<br />

interoperacyjności między e-nawigacją a systemami zewnętrznymi.<br />

Kluczowe elementy strategii e-nawigacji<br />

Do kluczowych elementów strategii e-nawigacji, wynikających<br />

z priorytetowych potrzeb użytkowników, należy zaliczyć: architekturę,<br />

czynnik ludzki, konwencje i standardy, określanie<br />

pozycji, technologie komunikacyjne i systemy informacyjne,<br />

elektroniczne mapy nawigacyjne, standaryzacje urządzeń<br />

i systemów oraz ich skalowalność [4, 5]. Podstawowe zalecenia<br />

dotyczące realizacji tych kluczowych elementów można<br />

ująć następująco:<br />

• ogólna funkcjonalna i techniczna koncepcja e-nawigacji<br />

powinna być przygotowana w szczególności pod względem<br />

opisu procesu, struktury danych, systemów informacyjnych,<br />

technologii komunikacyjnych i przepisów;<br />

• niezmiernie ważne są odpowiednie szkolenia, kompetencje,<br />

umiejętności językowe, obciążenie i motywacja<br />

użytkowników; sprawami istotnymi są również rozważne<br />

zarządzanie, przeciążenie informacyjne i ergonomia; powyższe<br />

aspekty e-nawigacji powinny być wzięte pod uwagę<br />

z uwzględnieniem stosownych zaleceń IMO w tym zakresie<br />

(np. IMO’s Human Element work),<br />

• w pracach nad e-nawigacją należy uwzględnić międzynarodowe<br />

konwencje, regulaminy i przewodniki, a także narodowe<br />

przepisy i standardy,<br />

• systemy określania pozycji powinny spełniać wymagania<br />

użytkowników w zakresie dokładności, integracji, niezawodności<br />

i redundancji systemów, z uwzględnieniem odpowiedniego<br />

poziomu ryzyka i wielkości ruchu statków,<br />

• technologie komunikacyjne i systemy informacyjne muszą<br />

spełniać wymagania użytkowników; powyższe może<br />

pociągać za sobą ulepszenie istniejących systemów lub<br />

opracowanie nowych; powinien być rozważony każdy<br />

wpływ (pozytywny/negatywny) istniejących systemów, pod<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 93


kątem standardów i protokołów technicznych, na strukturę<br />

danych, a także zajętość pasma i niezbędny przydział częstotliwości,<br />

• bardzo ważna jest ogólnoświatowa dostępność elektronicznych<br />

map nawigacyjnych – ENC (Electronic Navigational<br />

Chart), dlatego IHO oraz państwa członkowskie<br />

powinny kontynuować wysiłek w tym kierunku; perspektywicznie,<br />

e-nawigacja z pewnością zwiększy funkcjonalność<br />

przyszłych rozwiązań ENC,<br />

• standaryzacja urządzeń i ich skalowalność będą rozwijane<br />

wraz z postępem prac nad standardami eksploatacyjnymi,<br />

jako wynik współpracy między użytkownikami i producentami;<br />

państwa członkowskie IMO są odpowiedzialne<br />

za bezpieczeństwo wszystkich klas statków, stąd bardzo<br />

ważna jest możliwość tworzenia e-nawigacji w różnej skali,<br />

z przeznaczeniem dla różnych użytkowników; rozszerzenie<br />

koncepcji e-nawigacji na statki nie podlegające Konwencji<br />

SOLAS (np. jachty, łodzie motorowe) jest ważnym<br />

zadaniem, które powinno być realizowane przy ścisłej<br />

współpracy z użytkownikami (żeglarzami),<br />

• zarządzanie projektem e-nawigacji powinno należeć do<br />

jednej instytucji, mającej odpowiednie kompetencje techniczne,<br />

operacyjne i prawne, niezbędne do zdefiniowania<br />

i zrealizowania najważniejszych, obejmujących całą<br />

strukturę, elementów wdrożenia, działania i wprowadzenia<br />

w życie systemu, występujących na poziomie globalnym,<br />

regionalnym, narodowym i lokalnym; nie oznacza to, iż<br />

zarządzająca organizacja musi przeprowadzić wszystkie<br />

cele sama, może bowiem niektóre z nich przekazać innej<br />

kompetentnej instytucji; jedyną organizacją zdolną do spełnienia<br />

tych wymagań, mogącą wziąć odpowiedzialność za<br />

system i jego kontrolę jest IMO.<br />

Uwzględniając powyższe, implementacja strategii e-nawigacji<br />

powinna być zmiennym, interaktywnym procesem, składającym<br />

się z następujących elementów (rys. 1) [4, 5]:<br />

Implementacja<br />

Analiza<br />

braków<br />

Implemantacja<br />

Identyfikacja źródeł<br />

finansowani<br />

Planowanie przejścia<br />

Analiza braków<br />

technicznych<br />

Identyfikacja istniejących<br />

systemów<br />

Analiza braków<br />

operacyjnych<br />

Analiza braków<br />

regulacyjnych<br />

Przegląd<br />

zdobytego<br />

doświadczenia<br />

Przegląd<br />

zdobytego<br />

doświadczenia<br />

Potrzeby<br />

użytkownika<br />

Identyfikacja<br />

użytkowników<br />

Definicja podstawowych<br />

potrzeb użytkowników<br />

Identyfikacja<br />

funkcji i usług<br />

Definicja architektury<br />

Definicja koncepcji<br />

działania<br />

Korzyści finansowe<br />

i analiza ryzyka<br />

Analiza potrzeb szkoleniowych<br />

Analiza instytucjonalna<br />

i pod względem przepisów<br />

Rys. 1. Etapy implementacji strategii e-nawigacji<br />

Fig. 1. Stages of an e-navigation strategy implementation<br />

Architektura<br />

i analizy<br />

• potrzeby użytkownika,<br />

• architektura i analizy,<br />

• analiza braków,<br />

• implementacja,<br />

• przegląd zdobytego doświadczenia.<br />

Wstępny plan implementacji strategii<br />

e-nawigacji<br />

Jednym z trzech podstawowych elementów strategii e-nawigacji,<br />

obok elektronicznej mapy nawigacyjnej – ENC (Electronic<br />

Navigation Chart) oraz elektronicznego systemu określania<br />

pozycji, jest radiokomunikacja. Uwzględniając wytyczne<br />

MSC85, z punktu widzenia wspólnego planu prac podkomitetów<br />

COMSAR, NAV i STW, za najważniejsze należy uznać<br />

prace nad określeniem architektury systemu oraz dokonaniem<br />

wstępnej analizy braków a także kosztów/zysków i ryzyka całego<br />

projektu [6].<br />

W odniesieniu do architektury systemu, za priorytetowe<br />

należy uznać zidentyfikowanie istniejących i nowych technologii<br />

komunikacyjnych niezbędnych dla zaspokojenia potrzeb<br />

użytkownika w systemie e-nawigacji oraz zdefiniowanie warunków<br />

wyboru sprzętu i oprogramowania dla potrzeb e- nawigacji.<br />

W kwestii analizy braków powinno skupić się na identyfikacji<br />

istniejących systemów, które mogłyby zostać integralną<br />

częścią struktury e-nawigacji oraz określeniu luk w obszarze<br />

technicznym, regulacyjnym oraz operacyjnym.<br />

Jeśli chodzi o analizę kosztów/zysków i ryzyka należy<br />

przeprowadzić wstępną analizę w zakresie odnoszącym się<br />

jedynie do systemów komunikacyjnych oraz służby poszukiwania<br />

i ratowania – SAR (Serach And Rescue).<br />

Na obecnym etapie prac nad architekturą systemu, przyjęto<br />

jedynie koncepcyjną architekturę e-nawigacji przedstawioną<br />

na rys. 2, uznając, iż jest ona w ogólności poprawna i może<br />

stanowić dobry punkt wyjścia do dalszych prac<br />

nad ostateczną, bardziej szczegółową wersją tej<br />

architektury [6].<br />

Z perspektywy systemu e-nawigacji kluczowymi<br />

elementami środowiska statkowego są: stacja<br />

nadawczo-odbiorcza (transceiver station), sensory<br />

i aplikacje połączone z „transceiverem”, zintegrowany<br />

system nawigacyjny INS (Integrated<br />

Navigation System) oraz system zintegrowanego<br />

mostka nawigacyjnego IBS (Integrated Bridge System)<br />

(rys. 2).<br />

Stacja nadawczo-odbiorcza (w rzeczywistości<br />

może być ich więcej) komunikuje się ze stosownymi<br />

serwisami technicznymi e-nawigacji na lądzie<br />

za pomocą łączy fizycznych (radiowych, optycznych).<br />

Łączność ze statkiem od strony lądu<br />

zapewniają specjalne serwisy techniczne, odpowiedzialne<br />

za to. Operatorzy lądowi np.: operatorzy<br />

służby kontroli ruchu statków morskich VTS<br />

(Vessel Traffic Service), operatorzy stacji pilotowych,<br />

czy operatorzy firm obsługujących statki,<br />

osiągają swoje cele przez współpracę z daną<br />

aplikacją statkową. Z punktu widzenia operatorów<br />

lądowych, połączenia ze statkiem mają<br />

charakter połączeń (linii) funkcjonalnych pomiędzy<br />

aplikacjami użytkowników lądowych a odpowiednimi<br />

aplikacjami statkowymi. Podobne interakcyjne<br />

związki występują także w pozostałych<br />

relacjach z jakimi mamy do czynienia w radiokomunikacji<br />

morskiej tj. relacji statek-statek oraz<br />

ląd-statek.<br />

94<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Ogólnoświatowy system radionawigacyjny (WWRNS) włączając satelitarne<br />

i naziemne systemy nawigacyjne<br />

Inne<br />

statki<br />

Oficer<br />

na<br />

statku<br />

IBS<br />

Środowisko<br />

statku<br />

INS<br />

Sensory<br />

statkowe<br />

Stacja<br />

nadawczo-odbiorcza<br />

Aplikacje<br />

statkowe<br />

Łącze<br />

fizyczne<br />

(np. łącze<br />

radiowe)<br />

Inne<br />

statki<br />

Aplikacja- aplikacja<br />

(połączenie funkcjonalne)<br />

Lądowe<br />

serwisy<br />

techniczne<br />

e-nawigacji<br />

Użytkownicy<br />

ladowi<br />

(np.<br />

operatorzy<br />

VTS)<br />

Środowisko<br />

lądowe<br />

Rys. 2. Koncepcyjna architektura e-nawigacji Fig. 2. Conceptual e-navigation architecture<br />

Na rysunku 2 pokazano także rolę Światowego systemu<br />

radionawigacyjnego WWRNS (World-Wide Radionavigation<br />

System), zapewniającego w koncepcji e-nawigacji m.in. informację<br />

o pozycji statku, skorelowanej z czasem jej określenia.<br />

Uznając bardzo ważną rolę urządzeń/systemów oraz odpowiedniego<br />

dla nich oprogramowania, ustalono wstępną listę<br />

ogólnych kryteriów dotyczących wyboru sprzętu i oprogramowania<br />

na potrzeby e-nawigacji takich jak np.: systemy i aplikacje<br />

powinny odpowiadać potrzebom użytkownika, być łatwe w utrzymaniu,<br />

być niezawodne, elastyczne, wykorzystywać sprawdzone<br />

i zweryfikowane technologie czy być przyjazne w obsłudze<br />

zarówno dla użytkowników na statku jak i na lądzie.<br />

W celu przeprowadzenia oceny, dokonano przeglądu istniejących<br />

systemów, nowych technologii komunikacyjnych,<br />

regulacji międzynarodowych i standardów wraz z wymogami<br />

technicznymi oraz zapotrzebowaniem na pasmo i nowe częstotliwości<br />

dla potrzeb e-nawigacji. Stworzono rozległe, tabelaryczne<br />

zestawienie wszystkich istniejących systemów radiokomunikacyjnych,<br />

systemów obecnie opracowywanych oraz<br />

systemów planowanych wraz z przywołaniem odpowiednich<br />

regulacji Konwencji SOLAS, Regulaminu Radiokomunikacyjnego<br />

Międzynarodowego Związku Telekomunikacyjnego – ITU<br />

(International Telecommunication Union) oraz innych dokumentów<br />

IMO. W tabeli tej określono także funkcje realizowane przez<br />

poszczególne systemy. Zdecydowano, że zestawienie to powinno<br />

być otwarte na umieszczanie w nim ciągle nowych systemów<br />

przydatnych dla e-nawigacji. Ponadto ustalono, że dokument<br />

ten powinien jeszcze być rozbudowany o dodatkowe kolumny<br />

określające potrzeby użytkownika poszczególnych systemów<br />

zarówno na lądzie jak i na morzu oraz, że przy pracach nad e-<br />

nawigacją bardzo przydatnym byłoby uwzględnienie wymogów<br />

IMO dotyczących zasad budowy nowoczesnych mostków nawigacyjnych.<br />

Stwierdzono również, iż system e-nawigacji będzie<br />

wymagał zmian w gospodarce widmem ale na obecnym etapie<br />

prac trudno jest sformułować szczegółowe zapotrzebowanie na<br />

pasma i częstotliwości dla systemów komunikacyjnych, potencjalnie<br />

funkcjonujących w ramach e-nawigacji.<br />

Rozważając kwestię, czy Światowy system radionawigacyjny<br />

(WWRNS) powinien być centralnym elementem systemu<br />

e-nawigacji zgodzono się, że WWRNS, uwzględniając<br />

jego znaczenie dla bezpieczeństwa żeglugi, powinien być kluczowym<br />

elementem systemu. Ponadto uznano, że ze względu<br />

na bezpieczeństwo i ochronę żeglugi konieczna jest budowa<br />

globalnego, naziemnego systemu radionawigacyjnego, który<br />

funkcjonowałby również jako system awaryjny (backup) dla<br />

systemów nawigacji satelitarnej.<br />

W odniesieniu do problemu wspólnej struktury danych stosowanej<br />

przy wymianie informacji uznano, że jest to problem<br />

bardzo istotny, wymagający nowego podejścia do zagadnienia<br />

ale na obecnym etapie prac nad systemem e-nawigacji,<br />

wymagający jeszcze dalszych prac. Uzgodniono, że prace te<br />

powinny być prowadzone przy współudziale IALA i IHO (mających<br />

już doświadczenie w tym temacie) a także, iż powinny<br />

one dotyczyć formatów zbierania jak również nadawania danych<br />

oraz standardów interfejsów.<br />

W dyskusji nad wstępną analizą braków uznano, że rzadko<br />

dotychczas poruszany był bardzo ważny i aktualny obecnie<br />

problem ochrony (security) użytkowników systemu e-nawigacji.<br />

Stwierdzono również, że wiele aplikacji typu security wymaga<br />

transmisji szerokopasmowych, co w przyszłości może<br />

wpływać na zapotrzebowanie na nowe częstotliwości (pasma)<br />

dla e-nawigacji. Poruszono także kwestie niezawodności systemu<br />

e-nawigacji.<br />

Stwierdzono, że przeprowadzenie szczegółowych analiz<br />

kosztów i zysków jest na obecnym etapie prac niecelowe.<br />

Analiza taka wymagałaby ustalenia wielu współczynników<br />

oraz wyceny wielu elementów niematerialnych. Dopiero gdy<br />

dokona się tych ustaleń, możliwe i celowe będzie wyliczenie<br />

konkretnych sum pieniędzy oraz porównanie kosztów<br />

i zysków. Stwierdzono, że analiza ryzyka związana z pracami<br />

prowadzonymi w ramach IMO została już zdefiniowana<br />

w dokumencie „IMO Formal Safety Assessment Guidelines”<br />

(FSA). Dokument ten jest spójny z rezolucjami Zgromadzenia<br />

Ogólnego IMO, dotyczącymi metod pracy organizacji i stanowi<br />

bardzo dobrą podstawę dla analizy ryzyka. Zgadzając się<br />

co do potrzeby dalszych prac nad analizą kosztów/zysków<br />

i ryzyka, zaaprobowano metodykę pracy nad tymi zagadnieniami<br />

zalecaną we wspomnianym dokumencie IMO.<br />

Odnośnie ostatniego z rozważanych zagadnień dotyczących<br />

planu implementacji strategii e-nawigacji zgodzono<br />

się, że potrzeby instytucji SAR jako znaczącego użytkow-<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 95


nika systemu, powinny być brane pod uwagę nie tylko przy<br />

pracach nad rozwojem e-nawigacji, ale także przy pracach<br />

nad nowymi elementami i procedurami w GMDSS. Najistotniejsze<br />

potrzeby instytucji SAR jako użytkownika e-nawigacji<br />

zostały zebrane w formie stabelaryzowanej, z krótkim uzasadnieniem<br />

każdej potrzeby oraz odniesieniem do przyjętej<br />

strategii IMO.<br />

Podsumowanie<br />

Z zaproponowanego przez MSC85 harmonogramu czasowego<br />

dla przygotowania „Planu implementacji strategii e-nawigacji”,<br />

zrealizowane zostały (w pełni lub wstępnie) następujące<br />

elementy:<br />

• 2009 r. – zidentyfikowano potrzeby użytkowników, dokonano<br />

ich przegląd i ustalono ich priorytetowość; przeprowadzono<br />

skoordynowany przegląd architektury systemu,<br />

włączając sprawy sprzętowe, danych, informatyczne,<br />

technologii komunikacyjnej i oprogramowania, spełniającej<br />

potrzeby użytkowników,<br />

• <strong>2010</strong> r. – przyjęto, biorącą pod uwagę także element ludzki,<br />

wstępną analizę braków uwzględniającą aspekty techniczne,<br />

regulacyjne, operacyjne i szkoleniowe; zakończono<br />

wstępną identyfikację potrzeb użytkowników.<br />

Zgodnie z powyższym planem czasowym, w najbliższych<br />

dwóch latach, pozostały do zrealizowania następujące zagadnienia:<br />

• 2011 r. – ukończenie analizy ryzyka i korzyści finansowych,<br />

które powinny wspierać podejmowanie strategicznych decyzji<br />

jak i kiedy pewne funkcje będą możliwe do zastosowania;<br />

analizy powinny uwzględniać zarówno aspekty<br />

finansowe i ekonomiczne tych funkcji, jak i ocenę wpływu<br />

na bezpieczeństwo, ochronę i środowisko;<br />

• 2012 r. – rozpoczęcie wdrożenia „Planu implementacji<br />

strategii e-nawigacji”, który powinien zawierać wskazanie<br />

zakresu odpowiedzialności organizacji/stron wdrażających<br />

go, plan dla okresu przejściowego oraz harmonogram faz<br />

wdrożenia wraz z możliwą mapą drogową, wspomagającą<br />

powszechne zrozumienie potrzeb implementacji.<br />

Przedstawiony wyżej plan czasowy pokazuje złożoność<br />

i rozmiar prac związanych z implementacją koncepcji morskiej<br />

e-nawigacji, w której rolę szczególną mają odegrać nowoczesne<br />

systemy informacyjne i radiokomunikacyjne. Należy<br />

zauważyć, iż wiele z obecnych propozycji ma charakter tematów<br />

zamkniętych, ale wiele ma wciąż charakter wstępny i jest<br />

traktowanych jako podstawa do dalszych dyskusji.<br />

W kontekście zaprezentowanych radiokomunikacyjnych<br />

aspektów planu implementacji strategii e-nawigacji, otwartą<br />

staje się dyskusja dotycząca roli w tej koncepcji, stosowanego<br />

obecnie w radiokomunikacji morskiej systemu GMDSS,<br />

którego dziesięciolecie pełnej implementacji obchodzono<br />

w ubiegłym roku. Uznając potrzebę modernizacji systemu<br />

GMDSS należy podkreślić, iż zmiany elementów i procedur<br />

tego systemu powinny być dokonywane w zgodzie z rozwojem<br />

koncepcji e-nawigacji. Należy przy tym pamiętając, iż tak<br />

jak koncepcja e-nawigacji, system GMDSS przede wszystkim<br />

ma zapewniać bezpieczeństwo żeglugi, a podstawowym kryterium<br />

stosowanym w procesie jego modernizacji, powinny<br />

być rzeczywiste potrzeby użytkowników systemów GMDSS<br />

i e-nawigacji, a nie dostępne na rynku nowości technologiczne.<br />

Tylko takie podejście zapewni bowiem, wspomniane już<br />

bezpieczeństwo żeglugi statków morskich.<br />

Literatura<br />

[1] Raport z obrad 81 sesji Komitetu IMO ds. bezpieczeństwa<br />

na morzu – MSC (Maritime Safety Committee). IMO, Londyn,<br />

2006.<br />

[2] Raport z obrad 11 sesji Podkomitetu IMO ds. radiokomunikacji,<br />

poszukiwań i ratownictwa – COMSAR (Sub-Committee on Radiocommunications,<br />

Search and Rescue). IMO, Londyn, 2007.<br />

[3] Korcz K.: GMDSS as a Data Communication Network for E-Navigation.<br />

7th International Symposium on Marine Navigation and<br />

Safety of Sea Transportation, TransNav’2007. Gdynia Maritime<br />

University, Gdynia, 2007.<br />

[4] Raport z obrad 12 sesji Podkomitetu IMO ds. radiokomunikacji,<br />

poszukiwań i ratownictwa – COMSAR (Sub-Committee on<br />

Radiocommunications, Search and Rescue). IMO, Londyn,<br />

2008.<br />

[5] Korcz K.: Strategia e-nawigacji w żegludze morskiej. Przegląd<br />

Komunikacyjny i Wiadomości Telekomunikacyjne nr 5, 2009.<br />

[6] Raport z obrad 14 sesji Podkomitetu IMO ds. radiokomunikacji,<br />

poszukiwań i ratownictwa – COMSAR (Sub-Committee on<br />

Radiocommunications, Search and Rescue). IMO, Londyn,<br />

<strong>2010</strong>.<br />

Przypominamy o prenumeracie miesięcznika <strong>Elektronika</strong> na 2011 r.<br />

96<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Urządzenie do generacji silnych i ultrasilnych pól<br />

magnetycznych oraz sprzężony z nim układ pomiarowy<br />

mgr DARIUSZ PŁOCH, mgr PAWEŁ ŚLIŻ, prof. dr hab. EUGENIUSZ SZEREGIJ<br />

Uniwersytet Rzeszowski, Zakład Elektroniki Fizycznej, <strong>Instytut</strong> Fizyki<br />

Współczesna fizyka to fizyka w zakresie „ultra”: ultra silnych pól<br />

magnetycznych, ultra niskich temperatur. Ażeby we współczesnej<br />

fizyce doświadczalnej doprowadzić do znaczących odkryć<br />

naukowych potrzeba ekstremalnych warunków badań, przykładem<br />

tutaj może być Kwantowy Efekt Halla (temp. 0,1K).<br />

W Zakładzie Elektroniki Fizycznej Uniwersytetu Rzeszowskiego<br />

od 1998 roku pracuje Instalacja Silnych Impulsowych<br />

Pól Magnetycznych. W ciągu 12 lat eksploatacji instalacja ta<br />

przechodziła przez kolejne etapy udoskonaleń i dostosowywania<br />

do współczesnych możliwości stwarzanych przez elektronikę,<br />

informatykę, materiałoznawstwo. Ewaluacja instalacji<br />

dotyczyła zarówno wytwarzania cewki wielozwojowej [1], jak<br />

i akwizycji danych pomiarowych. Te kolejne etapy pozwalały<br />

na uzyskanie interesujących wyników badań [2–4].<br />

W artykule opisana została ostateczna wersja aparatury,<br />

pozwalającej na osiągnięcie dostatecznej czułości i dokładności<br />

pomiarów Rezonansu Magnetofononowego dla różnych<br />

rodzajów struktur, w tym podwójnych sprężonych studni<br />

kwantowych.<br />

Schemat instalacji: Badanie Efektów<br />

Galwanomagnetycznych w Silnych<br />

i Super-silnych Polach Magnetycznych<br />

Urządzenie zostało zaprojektowane na potrzeby pomiarów<br />

galwanometrycznych (efekt Halla, magnetoopór itp.) w silnych<br />

polach magnetycznych (SPM).<br />

Mówiąc silne pola magnetyczne mamy na myśli pola większe<br />

od 30 Tesli. Takie wartości indukcji pola magnetycznego są<br />

możliwe do osiągnięcia zazwyczaj metodami impulsowymi.<br />

Powyższa aparatura została oparta o ideę równoległego<br />

obwodu LC.<br />

Schemat blokowy i opis poszczególnych układów<br />

• Układ załadowania baterii kondensatorów i rozładowania.<br />

Na rysunku 1. Przedstawiony został schemat blokowy<br />

głównych bloków urządzenia. Można tutaj wyróżnić:<br />

Rys. 1. Schemat blokowy aparatury do generowania ultra-silnych impulsowych pól magnetycznych<br />

Fig. 1. Block diagram of the system for generation ultra-strong magnetics fields<br />

1. Blok ładowania baterii kondensatorów w przedziale<br />

0…3000 V<br />

2. Baterię kondensatorów wraz z modułem tyrystorowym<br />

o pojemności 10 mF, podzieloną na cztery sekcje po<br />

2,5 mF<br />

3. Główna cewkę roboczą – wytwarzającą pole magnetyczne<br />

4. Zestaw kart pomiarowo-sterujących wraz z komputerem<br />

klasy PC<br />

5. Wzmacniacze pomiarowe formujące mierzone sygnały.<br />

Szczegółowy opis każdego z tych bloków razem z schematami<br />

ideowymi przedstawiono niżej.<br />

• Układ załadowania baterii kondensatorów i rozładowania.<br />

Na rys. 2 pokazany został schemat zasadniczy baterii<br />

kondensatorów gromadzących energię potrzebna dla generacji<br />

super-silnych pól magnetycznych, razem z blokiem<br />

załadowania i rozładowania kondensatorów. Kondensatory<br />

o pojemności 100 µF każdy o maksymalnym napięciu 3 kV<br />

połączono równoległe (sumaryczna pojemność 10 mF przy<br />

tym samym napięciu), co pozwała na zgromadzenie maksymalnej<br />

energii 45 kJ.<br />

• Sterowanie fazowe tyrystorów. Proces ładowania baterii<br />

kondensatorów odbywa się poprzez transformator podwyższający<br />

napięcie do wartości 3000 V zasilany z sieci napięcia<br />

230 V. W celu uzyskania zadanej wartości napięcia na<br />

baterii po stronie pierwotnej zastosowano układ tyrystorowy<br />

sterowany fazowo. Sercem modułu jest znany układ TCA<br />

785 firmy Infineon [5]. Takie rozwiązanie umożliwia płynny<br />

start procesu naładowania kondensatorów bez ryzyka przeciążenia<br />

obwodu zasilającego.<br />

• Pomiar napięcia na baterii kondensatorów. Podczas<br />

ładowania baterii kondensatorów bardzo ważną rolę<br />

odgrywa prawidłowy odczyt napięcia. W tym przypadku<br />

zastosowano wzmacniacz izolacyjny AD 210 [6], spełniający<br />

bardzo ważną role – galwaniczne odseparowanie<br />

badanego modułu od pozostałych bloków pomiarowych.<br />

Pomiar napięcia odbywa się<br />

w czasie rzeczywistym a cała<br />

regulacja zadanego napięcia<br />

pracuje w pętli sprzężenia<br />

zwrotnego.<br />

• Rozładowanie – generacja<br />

pola magnetycznego. Rozładowanie,<br />

a co za tym idzie<br />

wygenerowanie pola magnetycznego<br />

odbywa się poprzez<br />

przyłączenie wcześniej naładowanej<br />

baterii kondensatorów<br />

do cewki roboczej.<br />

Realizowane jest to poprzez<br />

zespół tyrystorów mocy sterowanych<br />

z kart pomiarowych. W celu<br />

zwiększenia wydajności prądowej<br />

zastosowano 4 tyrystory połączone<br />

równolegle.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 97


Rys. 2. Schemat układu załadowania baterii kondensatorów Fig. 2. Diagram of the system for charge capacitors battery<br />

Rys. 3. Schemat układu rozładowania baterii kondensatorów i rozładowania<br />

Fig. 3. Diagram of the system for recharge capacitors battery<br />

98<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Rys. 4. Cewka robocza widok schematyczny<br />

Fig. 4. Magnetics coil – schematics view<br />

Rys. 5. Uchwyt pomiarowy (1 – złącze pomiarowe; 2 – przewody<br />

sygnałowe, 3 – ekran; 4 – cewka pomiarowa pola magnetycznego;<br />

5 – czujnik temperatury (PT-100); 6 – badana próbka)<br />

Fig. 5. The sample holder<br />

• Cewka robocza. Cewka robocza została zaprojektowana<br />

i wykonana w Zakładzie Elektroniki Fizycznej (rys. 4).<br />

Umożliwia wielokrotną generację pól magnetycznych do<br />

35 T, maksymalne wartości osiągają 50 T (w takich przypadka<br />

następuje destrukcja uzwojenia). Uzwojenie zostało<br />

wykonane z drutu miedzianego (płaskownika) o powierzchni<br />

przekroju 3×1 mm izolowanego oplotem bawełnianym odpornym<br />

na wysokie napięcia przebicia.<br />

Badane próbki umieszczane są centralnie w otworze cewki<br />

o średnicy ∅11 mm przy długości l = 90 mm. W rezultacie pry<br />

indukcyjności cewki L = 0,4 mH czas trwania impulsu stanowił<br />

14 ms. Inny wariant cewki tzw. optymalnej [1, 4] ma parametry:<br />

∅20 mm, l = 20 mm, L = 0,1 mH. Maksymalne uzyskane<br />

pole magnetyczne przy tych parametrach 55 T przy długości<br />

impulsu 6 ms. Jednorodność pola optymalnej cewki wyznaczona<br />

eksperymentalnie [4] okazała się zadawalająca.<br />

Dla uzyskania jak najlepszej dobroci cewki podczas<br />

pracy całość chłodzona jest ciekłym azotem. Jednorodność<br />

pola magnetycznego w centrum cewki jest lepsza niż<br />

0,2%/mm 3 (zwykle grubość badanej próbki nie przekracza<br />

1 mm).<br />

W przypadku pierwszej wersji cewki, to wytrzymała ona<br />

ponad 2000 impulsów (w tym około 500 o maksymalnym polu)<br />

w ciągu 8 lat. Druga wersja (cewka optymalna) nie jest tak<br />

wytrzymała na pola maksymalne.<br />

Uchwyt próbki<br />

• Schemat blokowy. Dokonywanie pomiarów w wysokich impulsowych<br />

polach magnetycznych możliwe jest po umieszczeniu<br />

badanej próbki we wnęce roboczej cewki. Ze względu<br />

na bardzo małe rozmiary próbki oraz konstrukcję cewki<br />

konieczne jest zastosowanie specjalnego uchwytu (rys. 5),<br />

pozwalającego na precyzyjne łatwe umieszczenie badanej<br />

próbki we wnęce roboczej cewki z jednoczesnym łatwym<br />

dostępem do sygnałów pomiarowych.<br />

Uchwyt wyposażony jest w czujniki temperatury (PT-100)<br />

oraz cewkę pomiaru pola magnetycznego. Szczegółowy opis<br />

każdego z nazwanych elementów uchwytu podano niżej.<br />

• Pomiar indukcji pola magnetycznego B. Jak wcześniej<br />

zostało zaznaczone pomiar pola magnetycznego realizowany<br />

jest poprzez cewkę umieszczoną na uchwycie bezpośrednio<br />

przy badanej próbce.<br />

Istotnym elementem badań w impulsowych polach magnetycznych<br />

jest problem dokładnego pomiaru indukcji pola<br />

magnetycznego, a co za tym idzie wykalibrowania cewki pomiarowej.<br />

Kalibracja cewki pomiarowej polega na określeniu łącznej<br />

powierzchni wszystkich jej zwojów poprzez porównanie<br />

otrzymywanych w niej indukowanych w zewnętrznym polu<br />

magnetycznych sygnałów z sygnałami uzyskiwanymi w tych<br />

samych warunkach w cewce wzorcowej (o znanej powierzchni<br />

zwojów). W celu kalibracji cewki pomiarowej umieszcza<br />

się ją wewnątrz cewki kalibracyjnej. Następnie z generatora<br />

podaje się sygnały o sinusoidalnych przebiegach na cewkę<br />

pierwotną, cewka ta wytwarza pole magnetyczne. W cewkach<br />

wzorcowych oraz w cewce pomiarowej indukuje się prąd<br />

i odczytuje się ich amplitudy. Porównując wielkości amplitudy<br />

sygnałów z poszczególnych cewek wzorcowych i z cewki pomiarowej<br />

określa się tę cewkę wzorcową, z której sygnał jest<br />

najbardziej zbliżony wielkością do sygnału cewki pomiarowej.<br />

Dalej badając już tylko sygnał z tej wybranej cewki wzorcowej<br />

i z cewki pomiarowej, ze stosunku otrzymanych z nich sygnałów<br />

oblicza się szukaną sumaryczną powierzchnię zwojów<br />

cewki pomiarowej:<br />

<br />

A<br />

S<br />

pom<br />

=<br />

A<br />

pom<br />

wz<br />

Wiadomo, że sygnał z cewki pomiarowej w czasie trwania<br />

impulsu zmienia się według zależności:<br />

S<br />

wz<br />

(1)<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 99


dϕ<br />

(2)<br />

E = −<br />

dt<br />

gdzie: strumień magnetyczny Φ jest równy:<br />

Φ = B sin ωt<br />

(3)<br />

S pom<br />

gdzie: S pom<br />

– pole przekroju cewki pomiarowej,<br />

ω – częstość drgań własnych obwodu SPM,<br />

B o<br />

– maksymalna wartość indukcji magnetycznej<br />

pola cewki wytwarzającej pole magnetyczne. Zatem:<br />

E = ωSB<br />

cosωt<br />

(4)<br />

0<br />

Sygnał ten podany jest na pierwszy kanał karty komputerowej.<br />

Po całkowaniu numerycznym tego sygnału otrzymujemy<br />

napięcie<br />

U = kB S sin ωt<br />

(5)<br />

cal<br />

0<br />

pom<br />

dzięki któremu odtwarzany jest impuls indukcji pola magnetycznego<br />

w cewce.<br />

Wygenerowane napięcie w czasie impulsu ma charakter<br />

dB/dt, dlatego wymagana jest operacja całkowania tegoż sygnału,<br />

pozwalająca na uzyskanie wartości właściwej B.<br />

Wiemy, że sygnał z cewki umieszczonej w solenoidzie<br />

zmienia się wg wzoru:<br />

dΦ<br />

U<br />

we<br />

= = ωB0S<br />

pom<br />

cosωt<br />

(6)<br />

dt<br />

gdzie: ω – częstotliwość, S pom<br />

– pole przekroju cewki pomiarowej,<br />

B o<br />

– maksymalna wartość indukcji magnetycznej pola<br />

cewki.<br />

Sygnał ten podany jest na układ:<br />

U wej<br />

R<br />

0<br />

C<br />

U wyj<br />

Wiemy, że U wyj<br />

= q/c oraz:<br />

1<br />

U wy<br />

= ∫ idt<br />

(7)<br />

C<br />

U<br />

wej<br />

gdzie: i = otrzymujemy zatem wyrażenie:<br />

R<br />

U<br />

we 1<br />

(8)<br />

U<br />

wy<br />

= ∫ dt = ∫ω0B0S<br />

pom<br />

cosωtdt<br />

RC RC<br />

Wykonując całkowanie wyrażenie powyższe przyjmie postać:<br />

B0S<br />

pom<br />

U<br />

wyj<br />

= sin ωt<br />

(9)<br />

RC<br />

Jako że pomiar pola magnetycznego stanowi tutaj bardzo<br />

ważny parametr determinujący wiarygodność i jakość badań<br />

naukowych, zastosowano niezależnie drugi tor pomiaru pola<br />

magnetycznego z wykorzystaniem miniaturowego hallotronu<br />

punktowego. Należy zaznaczyć że zdublowane pomiary nie<br />

obiegają od siebie więcej niż 0,8…1%, co może świadczyć<br />

o dokładnej realizacji pomiaru.<br />

• Zasilanie badanej próbki. Na końcu uchwytu pomiarowego<br />

montowana jest próbka badanego materiału. W zasadzie<br />

wszystkie pomiary galwanomagnetyczne odbywają się przy<br />

udziale przepływającego prądu o stałej wartości natężenia<br />

I(t). W tej konstrukcji próbka zasilana ze źródła stałoprądowego<br />

zbudowanego z wykorzystaniem przetwornika 16-bitowego<br />

C/A AD 420 firmy Analog Devices (rys. 6). Układ ten<br />

został wykorzystany w standardowej aplikacji producenta.<br />

Zadawanie wartości prądu realizowane jest przez wyjścia<br />

cyfrowe (CLOCK1, DATA1) jednej z kart pomiarowych (NI<br />

PXI 6259). Zakres ustawianego prądu zawiera się w przedziale<br />

0…24 mA, z rozdzielczości 300 nA, przy maksymalnym<br />

napięciu wymuszającym 32 V.<br />

• Wzmacniacze pomiarowe badanych sygnałów. W celu<br />

dopasowania sygnałów pomiarowych zastosowano blok<br />

wzmacniaczy pomiarowych. W tej aplikacji wykorzystano<br />

specjalizowane układy PGA 207 oraz PGA 103 firmy Burr-<br />

Brown [7].<br />

Zastosowanie tych układów podyktowane jest tym, że posiadają<br />

niezależne wejścia różnicowe a wybór wzmocnienia<br />

odbywa się poprzez wejścia cyfrowe A0, A1, A2, A3, dzięki<br />

temu możliwe jest ustalenie wzmocnienia w zakresie 1…1000<br />

V/V. Tak szeroki zakres wzmocnienia umożliwia pomiary sygnałów<br />

o dużej rozpiętości amplitud.<br />

• Kontrola temperatury. Ważnym parametrem jest pomiar<br />

temperatury w jakiej mierzone są zjawiska galwanomagnetyczne.<br />

W systemie zastosowano rezystor PT-100 zasilany<br />

ze źródła prądu stałego o wartości 1 mA zbudowanego na<br />

układach REF 200 [8]<br />

Pomiar temperatury odbywa się w wyniku pomiaru spadku<br />

napięcia na PT-100. Realizuje to karta NI PXI 6259 wykorzystując<br />

jedno z wejść analogowo-cyfrowych.<br />

Rys. 6. Schemat źródła prądowego zasilającego badaną próbkę Fig. 6. Diagram of the current sources<br />

100<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Rys. 7. Schemat wzmacniaczy pomiarowych<br />

Fig. 7. Diagram of the instrumentations amplifiers<br />

Rys. 9. Karty analogowo-cyfrowe<br />

Fig. 9. The Multifunction Data Acquisition (DAQ)<br />

Rys. 8. Schemat źródła prądowego zasilającego czujnik temperatury<br />

Fig. 8. Diagram of the current sources<br />

Akwizycja sygnałów pomiarowych<br />

Sercem układu pomiarowego jest tandem dwóch kart pomiarowych<br />

NI PXI 6120 oraz NI PXI 6259, umieszczonych w zewnętrznym<br />

chassis PXI 1002 (rys. 9) firmy National Instruments.<br />

Cały zestaw połączony jest z komputerem klasy PC z pomocą<br />

interfejsu o przepustowości 1,3 GS/s, sprawiającego,<br />

że karty pomiarowe pracują z maksymalnymi parametrami<br />

(nie następują straty w transferze danych).<br />

Wykorzystanie zewnętrznego chassis<br />

pozwoliło na skrócenie przewodów w torze<br />

pomiarowo-sterującym i jednocześnie<br />

zostały odseparowane galwanicznie moduły<br />

pomiarowe od pulpitu sterującego<br />

(komputer PC).<br />

Szybka karta pomiarowa sygnałów<br />

od próbki i od pola<br />

W przypadku pomiarów w impulsowych<br />

polach magnetycznych najważniejszym<br />

parametrem jest prędkość próbkowania<br />

przy maksymalnie dużej rozdzielczości.<br />

Zadanie to spełnia pierwsza z zastosowanych<br />

kart, NI PXI 6120. W jej strukturze<br />

znajdują się 4 równolegle niezależnie<br />

próbkujące 16-bitowe przetworniki A/C<br />

pracujące z częstotliwością 1 MS/s. Dodatkowo<br />

karta posiada dwa wyjścia C/A<br />

2,5 MS/s oraz 8 wejść typu I/O wykorzystywanych<br />

do włączania lub wyłączania<br />

poszczególnych funkcji aparatury.<br />

Ze względu na parametry, karta ta stanowi trzon układu<br />

akwizycji danych pomiarowych. Pierwszy kanał został zaaplikowany<br />

do pomiarów pola magnetycznego, natomiast<br />

pozostałe trzy kanały w zależności od konfiguracji służą do<br />

akwizycji trzech sygnałów szybkozmiennych (napięcie Halla<br />

oraz napięcie proporcjonalne do magnetooporu ρ (B)). Napięcie<br />

wygenerowane poprzez wyjście C/A jest wykorzystywane<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 101


do sterowania procesem ładowania baterii kondensatorów<br />

(współpraca z wejściem sterującym układu TCA 785). Pozostałe<br />

wejścia cyfrowe tzw. I/O wykorzystywane są do sterowania<br />

funkcjami typu włącz/wyłącz (wszelakie układy zabezpieczające,<br />

wyzwalające itp.)<br />

Multipleksowana karta pomiarowa sygnałów<br />

wolnozmiennych<br />

Dopełnieniem systemu pomiarowo sterującego jest druga<br />

karta 16-kanałowa z multipleksowanym próbkowaniem i częstotliwością<br />

1,25 MS/s. Ze względu dużą ilość wejść karta ta<br />

obsługuje system w zakresie sygnałów wolnozmiennych, między<br />

innymi pomiar temperatury, napięcia baterii kondensatorów,<br />

pomiar natężenia prądu przez próbkę, docelowo pomiar<br />

poziomu azotu w dewarze itp.<br />

Oprogramowanie układu pomiarowego<br />

Główna aplikacja pomiarowa<br />

Oprogramowanie sytemu stanowi dopełnienie całości systemu<br />

pomiarowego. Program sterujący napisany został w środowisku<br />

Lab View 8.0, niezaprzeczalnie lidera w tej klasie<br />

zastosowań. Wybór takiego środowiska podyktowany był również<br />

pełną kompatybilnością z częścią sprzętową – ten sam<br />

producent, National Instruments [10, 11].<br />

Struktura programu składa się z kilku podprogramów wywoływanych<br />

z poziomu głównego menu.<br />

Pierwszym jest okno typu Konfiguracja, poziomu której<br />

ustawia się wstępne parametry kart pomiarowych, między<br />

innymi wzmocnienie poszczególnych kanałów, częstotliwość<br />

próbkowania itp.<br />

Ustawienie żądanej wartości prądu płynącego przez badaną<br />

strukturę. Na tym etapie wybiera się współczynnik całkowania<br />

cewki pomiarowej pola magnetycznego. Dodatkowo<br />

deklaruje się miejsce zapisu<br />

Rys. 11. Podmenu Konfiguracja<br />

Fig. 11. Configuration menu<br />

Podprogram „Diagnostyka”<br />

Podmenu Diagnostyka umożliwia szybkie skontrolowanie<br />

poprawności połączeń oraz wstępne wykonanie pomiarów<br />

bez pola magnetycznego (rys. 12).<br />

Na tym etapie przeprowadza się kalibrację nowego uchwytu<br />

pomiarowego z użyciem cewki wzorcowej. Jest to aplikacja<br />

szczególnie przydatna na etapie stosowania nowego uchwytu<br />

nie w pełni skalibrowanego.<br />

Rys. 12. Podmenu Diagnostyka<br />

Fig. 12. Diagnostics menu<br />

Rys. 10. Menu główne programu<br />

Fig. 10. Programs menu<br />

Podprogram załadownia i wyzwolenia<br />

Następnym etapem jest główny podprogram tzw. „STRZAŁ”<br />

z poziomu którego następuje pełna kontrola procesem pomiarowym.<br />

Jest to aplikacja otwierająca drogę do przeprowadzenia<br />

pomiarów w polu magnetycznym.<br />

Na rys. 13 możemy wskazać kontrolki do ustawiania maksymalnego<br />

napięcia załadowania w danym cyklu pomiarowym,<br />

przelicznik pozwalający na orientacyjne oszacowanie<br />

pola magnetycznego odpowiadającego aktualnemu napięciu<br />

na baterii kondensatorów. Występują tutaj włączniki procesu<br />

ładowania baterii kondensatorów, awaryjnego rozładowania<br />

wcześniej naładowanych kondensatorów w przypadku awarii<br />

102<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Rys. 13. Podmenu wyzwalania „Strzał”<br />

Fig. 13. „Strzał” menu<br />

systemu. Dodatkowo można regulować tzw. fazą włączenia<br />

tyrystorów w obwodzie ładowania pojemności, co ma wpływ<br />

na szybkość i liniowość procesu. W prawej części okna znajdują<br />

się wskaźniki informujące o aktualnym stanie całego<br />

systemu pomiarowego. Można tutaj odczytać aktualną temperaturę<br />

na uchwycie pomiarowym, temperaturę otoczenia,<br />

wskazanie natężenia prądu zasilającego badaną próbkę, docelowo<br />

przewidziany jest pomiar poziomu azotu oraz pomiar<br />

rezystancji uzwojenia cewki roboczej – wskazanie poziomu<br />

schłodzenia uzwojenia.<br />

Aplikacja analizy danych<br />

Czwartym etapem, wieńczącym pomiar jest analiza numeryczna<br />

rys 14. otrzymanych wyników. W tym celu wybieramy<br />

z głównego okna dialogowego podmenu „Analiza Danych”.<br />

Jest to aplikacja pozwalająca na dokonanie szeregu czynności<br />

mających na celu wyeliminowanie składowych pasożytniczych<br />

dających negatywny wkład w końcowy efekt.<br />

Występuje tutaj szereg matematycznych reguł, przybliżeń<br />

wielomianowych, uśrednień zarówno krzywej przedstawiającej<br />

pomiar pola magnetycznego jak również właściwych sygnałów<br />

pomiarowych pochodzących od badanej próbki.<br />

Na tym etapie następuje wyeliminowanie zależności czasowych<br />

a wszystkie sygnały pomiarowe mają odniesienie<br />

w stosunku do pola magnetycznego. Efektem końcowym tych<br />

operacji są krzywe przedstawiające oscylacje Rezonansu<br />

Magnetofononowego [12] dla podwójnych studni kwantowych<br />

(rys. 15).<br />

Rys. 15. Krzywe eksperymentalna dla podwójnej studni kwantowej<br />

(DQW) – dwie wartości załadowania baterii kondensatorów<br />

(2000 i 3000 V) [12]<br />

Fig. 15. The experimental curves for Double Quantum Wells<br />

(DQW) – two value charging capacitors (2000 and 3000 V)<br />

Analiza Fourierowska<br />

Jako piąty pomocniczy poziom „Pomoc- Analiza FFT”. występuje<br />

aplikacja do analizy furierowskiej sygnałów mierzonych.<br />

Wykorzystywana jest jako narzędzie do weryfikacji składowych<br />

sygnału szczególnie wpływu szumów na widmo<br />

sygnałów zmierzonych oraz wpływu zakłóceń zewnętrznych<br />

na proces pomiarowy. Na rys. 16 przedstawiono przykładowy<br />

rozkład fourierowski szumów toru pomiarowego kanału I.<br />

Można zauważyć widmo szumów składające się z przynajmniej<br />

7 linii reprezentujących składowe harmoniczne zakłóceń<br />

pochodzących od sieci 50 Hz.<br />

Rys. 16. Podmenu Pomoc- analiza furierowska<br />

Fig. 16. Fourier analyze menu<br />

Rys. 14. Podmenu Analiza Danych<br />

Fig. 14. Data analyze menu<br />

Wnioski<br />

Celem pracy było stworzenie współczesnej instalacji do wytwarzania<br />

silnych impulsowych pól magnetycznych wraz z systemem<br />

akwizycji danych pomiarowych. Do realizacji tego zadania zastosowano<br />

nowoczesne podzespoły elektroniczne, a całe urządzenie<br />

oprogramowano z wykorzystaniem środowiska Lab View.<br />

Efektem tego uzyskano narzędzie do przeprowadzania<br />

szybkich badań galwanomagnetycznych kryształów i struktur<br />

(głównie półprzewodników) w super silnych polach magnetycznych<br />

z wystarczającą czułością napięciową (na poziome<br />

nV). W wyniku przeprowadzonych badań dla podwójnych<br />

sprzężonych studni kwantowych uzyskano subtelną strukturę<br />

rezonansu magnetofononowego [12].<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 103


Literatura<br />

[1] Kąkol T., Rybak R.,. Sheregii E.M.: Optymalna konstrukcja cewki<br />

do wytwarzania silnych impulsowych pól magnetycznych. <strong>Elektronika</strong>,<br />

nr 2, 13–17, 2000.<br />

[2] Tomaka G., Cebulski J., Sheregii E.M., Sciuk W., Strupiński W.,<br />

Dobrzanski I.: Magnetophonon Resonance as method of controlling<br />

of the thermal stress in the multiple quantum wells. Material<br />

Science and Engineering, A288, 138–141, 2000.<br />

[3] Tomaka G., Sheregii E.M., Kakol T.: Application of Magnetophonon<br />

Resonance to control of the thermal stress in multiple quantum<br />

wells. Material Science and Engineering, B80, 173–177, 2001.<br />

[4] Kakol T.,. Sheregii E.M.: The optimal coil build-up aimed to generate<br />

high pulsed magnetic fields. Physica B, 298, 594–598,<br />

2001.<br />

[5] Nota aplikacyjna układu scalonego TCA 785.<br />

[6] Nota aplikacyjna układu scalonego AD 210.<br />

[7] Nota aplikacyjna układu scalonego PGA 207, PGA 103.<br />

[8] Nota aplikacyjna układu scalonego REF 200.<br />

[9] Nota specyfikacyjna kart pomiarowych NI 6120 oraz NI 6259.<br />

[10] Stadler A.W.: Systemy akwizycji i przesyłania danych. Rzeszów<br />

2002.<br />

[11] Tłaczała W.: Środowisko LabView w eksperymencie wspomaganym<br />

komputerowo. WNT 2002.<br />

[12] Płoch D.,. Sheregii E.M,. Marchewka M, Woźny M., Tomaka G.:<br />

Magnetophonon Resonance in Double Quantum Wells. Phys.<br />

Rev. B 79,195434 (2009).<br />

Aspekty algorytmiczne organizacji jednostki<br />

procesorowej do mnożenia liczb Cayleya<br />

dr hab. inż. profesor ZUT ALEXANDR ŢARIOV, dr GALINA ŢARIOVA<br />

Zachodniopomorski Uniwersytet Technologiczny w Szczecinie, Wydział Informatyki<br />

Dynamiczny rozwój technik przetwarzania danych oraz potrzeba<br />

realizacji coraz to bardziej skomplikowanych zadań<br />

i aplikacji praktycznych wymagają zastosowania w tym celu<br />

coraz bardziej zaawansowanych i skomplikowanych metod<br />

oraz formalizmów matematycznych. Obecnie na coraz szerszą<br />

skalę do syntezy, opisu i realizacji wysoce efektywnych<br />

algorytmów CPS wykorzystuje się aparat matematyczny algebry<br />

liczb hiperzespolonych [1–6]. Dotyczy to głównie zadań<br />

związanych z uogólnieniem dyskretnej transformaty Fouriera<br />

oraz innych transformacji ortogonalnych, filtracji sygnałów<br />

jedno- i wielowymiarowych, wielorozdzielczej reprezentacji<br />

danych [7–11].<br />

Przy realizacji wspomnianych zadań w przestrzeniach hiperzespolonych<br />

jedną z najbardziej pracochłonnych operacji<br />

(a raczej „makrooperacji”) jest makrooperacja mnożenia<br />

dwóch oktonionów, gdyż wymaga ona wykonania kilkudziesięciu<br />

konwencjonalnych operacji mnożenia oraz dodawania<br />

liczb rzeczywistych [13]. Z przyczyn naturalnych intencją projektantów<br />

wysoce efektywnych algorytmów CPS od zawsze<br />

było poszukiwanie sposobów redukcji liczby operacji arytmetycznych<br />

w algorytmach obliczeniowych. Dotyczy to zwłaszcza<br />

minimalizacji liczby operacji mnożenia będących od samego<br />

początku rozwoju komputeryzacji najbardziej czasochłonnymi<br />

operacjami w całym zbiorze operacji przetwarzania danych.<br />

W tym celu w pracy [14] został zaproponowany sposób obliczania<br />

iloczynu dwóch kwaternionów wymagający, względem<br />

pierwowzoru, dwukrotnie mniejszej liczby operacji mnożenia<br />

kosztem więcej niż potrójnego zwiększenia liczby operacji<br />

algebraicznego dodawania liczb rzeczywistych. W [15] zaproponowano<br />

algorytmiczne rozwiązanie, pozwalające przy<br />

mnożeniu kwaternionów istotnie zredukować liczbę dodawań<br />

przy zachowaniu takiej samej liczby mnożeń. Zastosowane<br />

w cytowanym artykule rozwiązania można rozszerzyć na<br />

mnożenie liczb Cayleya. Celem niniejszej pracy jest przedstawienie<br />

oryginalnych wyników syntezowanego przez nas<br />

zracjonalizowanego algorytmu obliczenia iloczynu dwóch oktonionów,<br />

wymagającego wykonania mniejszej liczby operacji<br />

mnożenia liczb rzeczywistych w stosunku do bezpośredniego<br />

naiwnego sposobu obliczania kosztem pewnego zwiększenia<br />

liczby operacji dodawania.<br />

Synteza zracjonalizowanej procedury<br />

mnożenia liczb Cayleya<br />

Rozważmy problem mnożenia dwóch liczb Cayleya (oktonionów)<br />

gdzie:<br />

d = ab, (1)<br />

a = ( a0 + ia1<br />

+ ja2<br />

+ ka3<br />

+ Ea4<br />

+ Ia5<br />

+ Ja6<br />

+ Ka7<br />

) ,<br />

b = ( b0 + ib1<br />

+ jb2<br />

+ kb3<br />

+ Eb4<br />

+ Ib5<br />

+ Jb6<br />

+ Kb7<br />

) ,<br />

d = ( d<br />

0<br />

+ id1<br />

+ jd<br />

2<br />

+ kd3<br />

+ Ed<br />

4<br />

+ Id5<br />

+ Jd<br />

6<br />

+ Kd<br />

7<br />

) ,<br />

a i<br />

∈ R, i = 0,7, zaś i, j, k, E, I, J, K – są jednostkami urojonymi.<br />

Operacja mnożenia oktonionów może być przedstawiona<br />

w postaci iloczynu wektorowo-macierzowego:<br />

, Y = X<br />

8 × 1 8 8×<br />

1<br />

(2)<br />

gdzie:<br />

Τ<br />

Y<br />

8× 1<br />

= [ y0<br />

, y1,<br />

y2<br />

, y3,<br />

y4<br />

, y5,<br />

y6<br />

, y7<br />

] , y = d , i = 0, 7 ,<br />

Τ<br />

X<br />

8× 1<br />

= [ x0<br />

, x1,<br />

x2<br />

, x3,<br />

x4<br />

, x5,<br />

x6<br />

, x7<br />

] , xi = ai<br />

, i = 0, 7 ,<br />

⎡b0<br />

− b1<br />

− b2<br />

− b3<br />

− b4<br />

− b5<br />

− b6<br />

− b7<br />

⎤<br />

⎢<br />

⎥<br />

⎢<br />

b1<br />

b0<br />

b3<br />

− b2<br />

b5<br />

− b4<br />

− b7<br />

b6<br />

⎥<br />

⎢b<br />

−<br />

− − ⎥<br />

2<br />

b3<br />

b0<br />

b1<br />

b6<br />

b7<br />

b4<br />

b5<br />

⎢<br />

⎥<br />

⎢b3<br />

b2<br />

− b1<br />

b0<br />

b7<br />

− b6<br />

b5<br />

− b4<br />

Q =<br />

⎥<br />

8<br />

.<br />

⎢b<br />

− − −<br />

⎥<br />

4<br />

b5<br />

b6<br />

b7<br />

b0<br />

b1<br />

b2<br />

b3<br />

⎢<br />

⎥<br />

⎢b5<br />

b4<br />

− b7<br />

b6<br />

− b1<br />

b0<br />

− b3<br />

b2<br />

⎥<br />

⎢b<br />

− −<br />

− ⎥<br />

6<br />

b7<br />

b4<br />

b5<br />

b2<br />

b3<br />

b0<br />

b1<br />

⎢<br />

⎥<br />

⎢⎣<br />

b7<br />

− b6<br />

b5<br />

b4<br />

− b3<br />

− b2<br />

b1<br />

b0<br />

⎥⎦<br />

i<br />

i<br />

104<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Można zapisać:<br />

~<br />

(1) (2)<br />

(1)<br />

(2)<br />

Y = B X = ( B + B X = B X B X , (3)<br />

gdzie:<br />

8 × 1 8 8×<br />

1 8 8<br />

)<br />

8×<br />

1 8 8×<br />

1<br />

+<br />

~<br />

Τ<br />

Y<br />

8× 1<br />

= [ −y0<br />

, y1,<br />

y2<br />

, y3,<br />

y4<br />

, y5,<br />

y6<br />

, y7<br />

] ,<br />

⎡−<br />

b0<br />

b1<br />

b2<br />

b3<br />

b4<br />

b5<br />

b6<br />

b7<br />

⎤<br />

⎢<br />

⎥<br />

⎢<br />

b1<br />

b0<br />

b3<br />

− b2<br />

b5<br />

− b4<br />

− b7<br />

b6<br />

⎥<br />

⎢ b −<br />

− − ⎥<br />

2<br />

b3<br />

b0<br />

b1<br />

b6<br />

b7<br />

b4<br />

b5<br />

⎢<br />

⎥<br />

⎢ b3<br />

b2<br />

− b1<br />

b0<br />

b7<br />

− b6<br />

b5<br />

− b4<br />

B =<br />

⎥<br />

8<br />

, (4)<br />

⎢ b − − −<br />

⎥<br />

4<br />

b5<br />

b6<br />

b7<br />

b0<br />

b1<br />

b2<br />

b3<br />

⎢<br />

⎥<br />

⎢ b5<br />

b4<br />

− b7<br />

b6<br />

− b1<br />

b0<br />

− b3<br />

b2<br />

⎥<br />

⎢ b<br />

− −<br />

− ⎥<br />

6<br />

b7<br />

b4<br />

b5<br />

b2<br />

b3<br />

b0<br />

b1<br />

⎢<br />

⎥<br />

⎢⎣<br />

b7<br />

− b6<br />

b5<br />

b4<br />

− b3<br />

− b2<br />

b1<br />

b0<br />

⎥⎦<br />

⎡b0<br />

b1<br />

b2<br />

b3<br />

b4<br />

b5<br />

b6<br />

b7<br />

⎤<br />

⎢<br />

⎥<br />

⎢<br />

b1<br />

b0<br />

b3<br />

b2<br />

b5<br />

b4<br />

b7<br />

b6<br />

⎥<br />

⎢b<br />

⎥<br />

2<br />

b3<br />

b0<br />

b1<br />

b6<br />

b7<br />

b4<br />

b5<br />

⎢<br />

⎥<br />

= ⎢b3<br />

b2<br />

b1<br />

b0<br />

b7<br />

b6<br />

b5<br />

b<br />

(1)<br />

4<br />

B ⎥<br />

8<br />

, (5)<br />

⎢b<br />

⎥<br />

4<br />

b5<br />

b6<br />

b7<br />

b0<br />

b1<br />

b2<br />

b3<br />

⎢<br />

⎥<br />

⎢b5<br />

b4<br />

b7<br />

b6<br />

b1<br />

b0<br />

b3<br />

b2<br />

⎥<br />

⎢b<br />

⎥<br />

6<br />

b7<br />

b4<br />

b5<br />

b2<br />

b3<br />

b0<br />

b1<br />

⎢<br />

⎥<br />

⎢⎣<br />

b7<br />

b6<br />

b5<br />

b4<br />

b3<br />

b2<br />

b1<br />

b0<br />

⎥⎦<br />

⎡−<br />

b0<br />

0 0 0 0 0 0 0 ⎤<br />

⎢<br />

⎥<br />

⎢<br />

0 0 0 − b2<br />

0 − b4<br />

− b7<br />

0<br />

⎥<br />

⎢ 0 − b<br />

⎥<br />

3<br />

0 0 0 0 − b4<br />

− b5<br />

⎢<br />

⎥<br />

<br />

⎢ 0 0 − b1<br />

0 0 − b6<br />

0 − b<br />

(2)<br />

4<br />

B =<br />

⎥<br />

8<br />

2<br />

. (6)<br />

⎢ 0 − b − − 0 0 0 0 ⎥<br />

5<br />

b6<br />

b7<br />

⎢<br />

⎥<br />

⎢ 0 0 − b7<br />

0 − b1<br />

0 − b3<br />

0 ⎥<br />

⎢ 0 0 0 − b − 0 0 − ⎥<br />

5<br />

b2<br />

b1<br />

⎢<br />

⎥<br />

⎢⎣<br />

0 − b6<br />

0 0 − b3<br />

− b2<br />

0 0 ⎥⎦<br />

Realizacja operacji (1) wymaga wykonania 64 operacji<br />

mnożenia oraz 56 operacji dodawania algebraicznego liczb<br />

rzeczywistych. Wykorzystując sposób minimalizacji złożoności<br />

obliczeniowej zastosowany w pracy [15] do mnożenia<br />

kwaternionów, rozszerzając go do wyznaczenia iloczynu<br />

oktonionów oraz wykorzystując metodę [16] do dalszej racjonalizacji<br />

obliczeń można syntetyzować efektywny algorytm<br />

realizacji rozpatrywanej operacji ze zminimalizowaną<br />

liczbą operacji mnożenia i dodawania liczb rzeczywistych.<br />

A ponieważ bloki mnożenia wymają przy implementacji<br />

znacznie większych nakładów sprzętowych niż sumatory<br />

i zajmują lwią część płaszczyzny kryształu układu scalonego<br />

lub zużywają dużą liczbę elementów operacyjnych FPGA, to<br />

minimalizacja operacji mnożenia w istotny sposób oszczędza<br />

resurs operacyjny jednostki procesorowej. Dlatego też<br />

redukcja bloków mnożących niezbędnych do implementacji<br />

jednostki dedykowanej do wyznaczania iloczynu liczb Cayleya<br />

jest aktualna.<br />

W celu syntezy efektywnej wektorowo-macierzowej procedury<br />

obliczeniowej wyznaczenia poszukiwanego iloczynu<br />

zdefiniujmy kilka macierzy:<br />

8<br />

8×<br />

1<br />

A<br />

30×<br />

16<br />

⎡1<br />

1<br />

⎤<br />

⎢<br />

⎥<br />

⎢<br />

1 1<br />

0<br />

⎥<br />

4<br />

⎢1<br />

−1<br />

⎥<br />

⎢<br />

⎥<br />

⎢ 1 −1<br />

⎥<br />

(1)<br />

08<br />

× 22<br />

A<br />

30<br />

= ⎢<br />

1 1<br />

⎥ ,<br />

⎢<br />

1 1 ⎥<br />

⎢ 0<br />

⎥<br />

4<br />

⎢<br />

1 −1<br />

⎥<br />

⎢<br />

⎥<br />

⎢<br />

1 −1<br />

⎥<br />

⎣<br />

022<br />

× 8<br />

I<br />

22<br />

⎦<br />

⎡1<br />

1<br />

⎤<br />

⎢ 02<br />

⎥<br />

⎢<br />

1 −1<br />

0<br />

⎥<br />

4<br />

⎢ 1 1<br />

0<br />

⎥<br />

⎢<br />

2<br />

1 −1<br />

⎥<br />

(2) ⎢<br />

08<br />

× 22<br />

A =<br />

⎥<br />

30<br />

1 1<br />

.<br />

⎢<br />

0 ⎥<br />

2<br />

⎢<br />

1 −1<br />

⎥<br />

⎢ 04<br />

1 1 ⎥<br />

⎢<br />

02<br />

⎥<br />

⎢<br />

1 −1<br />

⎥<br />

⎢⎣<br />

022<br />

× 8<br />

I<br />

22<br />

⎥⎦<br />

⎡1<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢1<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

= ⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎣<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

−1<br />

0<br />

1<br />

22×<br />

8<br />

1<br />

−1<br />

1<br />

−1<br />

1<br />

−1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

0<br />

8<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

⎤<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

1<br />

⎥<br />

⎥<br />

1⎥<br />

⎥<br />

1⎦<br />

,<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 105


P 16 × 8<br />

⎡1<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

= ⎢<br />

1<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎣<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

⎤<br />

⎡ 1<br />

⎤<br />

⎥<br />

⎢<br />

⎥<br />

⎢<br />

1<br />

⎥<br />

⎢ 1<br />

⎥<br />

⎢<br />

⎥<br />

⎢<br />

1<br />

⎥<br />

⎢<br />

1<br />

⎥<br />

⎢<br />

⎥<br />

⎢<br />

1<br />

⎥<br />

⎢<br />

1<br />

⎥<br />

⎢<br />

1⎥<br />

Τ ⎢<br />

1<br />

⎥<br />

[ A 8 × 16<br />

] = ⎢<br />

−1<br />

⎥<br />

⎢<br />

⎥<br />

⎢ −1<br />

⎥<br />

⎢<br />

⎥<br />

⎢<br />

−1<br />

⎥<br />

⎢<br />

−1<br />

⎥<br />

⎢<br />

⎥<br />

⎢<br />

−1<br />

⎥<br />

⎢<br />

−1<br />

⎥<br />

⎢<br />

⎥<br />

⎢<br />

−1<br />

1⎥<br />

, ⎢<br />

⎥ ⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥ ,<br />

⎦<br />

⎣<br />

−1⎦<br />

[ A<br />

]<br />

Τ<br />

16×<br />

30<br />

⎡1<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢1<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

= ⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎣<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

−1<br />

−1<br />

0<br />

22×<br />

8<br />

1<br />

1<br />

−1<br />

−1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

0<br />

8<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

⎤<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

1<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

1<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

1⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎦ .<br />

~<br />

A<br />

30×<br />

16<br />

⎡1<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢1<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

= ⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎢<br />

⎣<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

−1<br />

−1<br />

0<br />

22×<br />

8<br />

1<br />

1<br />

−1<br />

−1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

0<br />

8<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

1<br />

⎤<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

1⎥<br />

⎥<br />

⎥<br />

1⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

1⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎥<br />

⎦ .<br />

Uwzględniając wprowadzone konstrukcje macierzowe<br />

zracjonalizowaną procedurę mnożenia dwóch liczb Cayleya<br />

można przedstawić następująco:<br />

~<br />

(1) (2) (2) (1)<br />

Y<br />

8× 1<br />

= I8A<br />

8×<br />

16A16×<br />

30A<br />

30<br />

A<br />

30<br />

D30A<br />

30<br />

A<br />

30<br />

A<br />

30×<br />

16P16<br />

× 8X8×<br />

1<br />

, (7)<br />

gdzie I ~ = diag (–1,1,1,1,1,1,1,1), D = diag (c , c ,..., c ), zaś<br />

8 30 0 1 29<br />

elementy {c i<br />

}, i = 0,29 mogą zostać wyznaczone za pomocą<br />

następującego wyrażenia:<br />

106<br />

(2) (1) ~<br />

C<br />

30× 1<br />

= Ν<br />

30A<br />

30<br />

A<br />

30<br />

A<br />

30×<br />

16P16<br />

× 8B8×<br />

1<br />

,<br />

(8)<br />

gdzie:<br />

Τ<br />

B<br />

8× 1<br />

= [ b0<br />

, b1<br />

, b2<br />

, b3<br />

, b4<br />

, b5<br />

, b6<br />

, b7<br />

] ,<br />

Ν<br />

30<br />

= diag(1 8,1 8,1 8,1 8,1 8,1 8,1 8,1 8 ,<br />

2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2) ,<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


c 0<br />

H 2<br />

c 1<br />

H 2<br />

H 2<br />

H 2<br />

c 2<br />

c 3<br />

H 2<br />

H 2<br />

x 0<br />

x 1<br />

c 4<br />

H 2<br />

H<br />

H c 2<br />

2<br />

5<br />

H<br />

2<br />

−y 0<br />

y 1<br />

x 2<br />

x 3<br />

H 2<br />

H 2<br />

c 6<br />

c 7<br />

H 2<br />

H 2<br />

y 2<br />

y 3<br />

x 4<br />

x 5<br />

c 8<br />

c 9<br />

y 4<br />

y 5<br />

x 6<br />

x 7<br />

c 10<br />

c 11<br />

y 6<br />

y 7<br />

c 12<br />

c 13<br />

c 14<br />

c 15<br />

c 16<br />

c 17<br />

Rys. 1. Model grafostrukturalny organizacji układu do wyznaczania<br />

iloczynu dwóch liczb Cayleya zgodnie z procedurą (7)<br />

Fig. 1. The graph-structural model of organization of processing<br />

unit for multiplication of two Cayley numbers corresponding to (7)<br />

c 18<br />

c 19<br />

c 20<br />

c 21<br />

c 22<br />

c 23<br />

c 24<br />

c 25<br />

c 26<br />

c 27<br />

c 28<br />

c 29<br />

Na rysunku 1 została pokazana struktura algorytmiczna<br />

układu procesorowego, realizującego obliczanie iloczynu<br />

liczb Cayleya według zracjonalizowanego algorytmu zgodnie<br />

z procedurą (7). Kółkami na rysunku oznaczone są operacje<br />

mnożenia liczb rzeczywistych (bloki mnożące) przez wartości<br />

wpisane w poszczególne kółka. Linie proste symbolizują operacje<br />

(kanały) transferu danych. W przypadku skupienia tych<br />

linii w odpowiednich punktach struktury mamy do czynienia<br />

z operacją sumowania nadchodzących danych realizowanymi<br />

za pomocą zwykłych sumatorów. Prostokątami w tym przypadku<br />

oznaczone są operacje (bloki) mnożenia dwuelementowych<br />

podwektorów przez macierze Hadamarda drugiego<br />

⎛1<br />

1 ⎞<br />

rzędu H 2<br />

= ⎜ ⎟ . Linie przerywane tutaj symbolizują ope-<br />

⎝1<br />

−1⎠<br />

racje zmiany znaku odpowiedniego operandu na przeciwny.<br />

Z kolei na rysunku 2 przedstawiona została struktura algorytmiczna<br />

układu pomocniczego, służącego do wyznaczenia<br />

elementów zgodnie z procedurą (8).<br />

b 0<br />

b 1<br />

b 2<br />

b 3<br />

b 4<br />

b 5<br />

b 6<br />

b 7<br />

H 2<br />

Rys. 2. Struktura algorytmiczna bloku do obliczania elementów<br />

zgodnie z procedurą (8)<br />

Fig. 2. The algorithmic structure of processor unit for calculating<br />

of corresponding to (8)<br />

H 2<br />

H 2<br />

H 2<br />

H 2<br />

H 2<br />

H 2<br />

⅛<br />

⅛<br />

⅛<br />

⅛<br />

⅛<br />

⅛<br />

⅛<br />

⅛<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

2<br />

c 0<br />

c 1<br />

c 2<br />

c 3<br />

c 4<br />

c 5<br />

c 6<br />

c 7<br />

c 8<br />

c 9<br />

c 10<br />

c 11<br />

c 12<br />

c 13<br />

c 14<br />

c 15<br />

c 16<br />

c 17<br />

c 18<br />

c 19<br />

c 20<br />

c 21<br />

c 22<br />

c 23<br />

c 24<br />

c 25<br />

c 26<br />

c 27<br />

c 28<br />

c 29<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 107


Dyskusja aspektów realizacyjnych<br />

Omawiana w artykule koncepcja organizacji układu do obliczania<br />

iloczynu dwóch oktonionów stanowi próbę racjonalizacji<br />

jego struktury pod kątem minimalizacji liczby układów<br />

mnożących niezbędnych do całkowicie równoległej sprzętowej<br />

realizacji obliczeń. W tabeli pokazano liczby bloków<br />

mnożących oraz sumatorów niezbędnych do realizacji dwóch<br />

różnych sposobów organizacji obliczeń, określonych odpowiednio<br />

wzorem (1) oraz za pomocą zaproponowanego algorytmu.<br />

Jak widać liczba operacji mnożenia (bloków mnożących)<br />

w przypadku proponowanego rozwiązania jest ponad<br />

dwukrotnie mniejsza od metody naiwnej, natomiast liczba<br />

dodawań (bloków dodawania) się zwiększyła.<br />

Oszacowanie liczby bloków operacyjnych przy realizacji dwóch różnych<br />

metod obliczania iloczynu dwóch oktonionów<br />

Number of operational units for two various methods of quaternion<br />

multiplier realization<br />

Metoda<br />

Liczba bloków<br />

mnożących<br />

Liczba sumatorów<br />

Na podstawie wzoru (1) 64 56<br />

Proponowane rozwiązanie 30 94<br />

Należy nadmienić, że w większości zadań cyfrowego przetwarzania<br />

sygnałów jeden z przemnażanych oktonionów jest<br />

tak zwanym „stałym”, co oznacza, że jego współczynniki { b i<br />

}<br />

są z góry znanymi stałymi liczbami rzeczywistymi. Oznacza<br />

to, że elementy macierzy D 30<br />

mogą być jeden tylko raz z góry<br />

wyliczone oraz przechowywane w pamięci stałej układu obliczeniowego.<br />

Wtedy proponowane w artykule rozwiązanie<br />

staje się jeszcze bardziej efektywne, ponieważ wymaga tyle<br />

samo bloków operacyjnych jak w przypadku realizacji metody<br />

naiwnej, wśród których tylko 30 bloków (dwukrotnie mniej) są<br />

blokami mnożącymi, zaś tylko 70 – sumatorami. Sumaryczna<br />

liczba bloków operacyjnych niezbędnych do implementacji<br />

układu procesorowego do wyznaczania pozyskiwanego<br />

iloczynu w tym przypadku wynosi 100 (120 w przypadku<br />

implementacji metody naiwnej). W ten sposób proponowane<br />

w artykule rozwiązanie staje się najlepszym pod kątem sumarycznej<br />

liczby operacji arytmetycznych lub bloków operacyjnych<br />

(w przypadku realizacji sprzętowej) niezbędnych do<br />

obliczenia iloczynu oktonionów.<br />

Podsumowanie<br />

W artykule poruszono aspekty algorytmiczne organizacji specjalizowanego<br />

układu obliczeniowego do wyznaczenia iloczynu<br />

dwóch oktonionów stanowiących szczególny przypadek<br />

liczb hiperzespolonych. Można stwierdzić, że pod kątem złożoności<br />

realizacji sprzętowej proponowane rozwiązanie jest<br />

najlepszym ze znanych, ponieważ wymaga przy implementacji<br />

tyle samo bloków mnożących, co w przypadku implementacji<br />

sposobu [15], lecz potrzebuje znacznie mniej sumatorów.<br />

Oznacza to, że proponowane w artykule rozwiązanie algorytmiczne<br />

pozwala zminimalizować zapotrzebowanie na zasoby<br />

sprzętowe, co z kolei stwarza dodatkowe korzyści przy realizacji<br />

układu do wyznaczania iloczynu oktonionów w środowisku<br />

reprogramowalnym.<br />

Literatura<br />

[1] Kantor I., Solodovnikov A.: Hypercomplex numbers. Springer-<br />

Verlag, New York, 1989.<br />

[2] Bülow T., Sommer G.: Hypercomplex signals – a novel extension<br />

of the analytic signal to the multidimensional case. IEEE Trans.<br />

Sign. Proc., vol. SP-49, no. 11, pp. 2844–2852, Nov. 2001.<br />

[3] Schütte H.-D., Wenzel J.: Hypercomplex numbers in digital signal<br />

processing. In Proc. ISCAS ’90, New Orleans, 1990, pp.<br />

1557–1560.<br />

[4] Alfsmann D.: On families of 2N-dimensional hypercomplex algebras<br />

suitable for digital signal processing. In Proc. European Signal<br />

Processing Conf. (EUSIPCO 2006), Florence, Italy, 2006.<br />

[5] Alfsmann D., Göckler H. G., Sangwine S. J. and Ell T. A.: Hypercomplex<br />

Algebras in Digital Signal Processing: Benefits and<br />

Drawbacks (Tutorial). Proc. EURASIP 15th European Signal<br />

Processing Conference (EUSIPCO 2007), Poznań, Poland,<br />

2007, pp. 1322–1326.<br />

[6] Sangwine S. J., Bihan N. Le: Hypercomplex analytic signals: extension<br />

of the analytic signal concept to complex signals. Proc.<br />

EURASIP 15th European Signal Processing Conference (EU-<br />

SIPCO 2007), Poznań, Poland, 2007, Poznań, pp. 621–624.<br />

[7] Sangwine S. J.: Fourier transforms of color images using quaternion<br />

or hypercomplex, numbers. In Electronics Letters, 10<br />

Oct. 1996, vol. 32, pp. 1979–1980.<br />

[8] Sangwine S. J., Ell T. A.: Hypercomplex auto- and cross-correlation<br />

of color images. In Proc. ICIP, 1999, pp. 319–323.<br />

[9] Moxey C. E., Sangwine S. J., Ell T. A.: Hypercomplex correlation<br />

techniques for vector images. IEEE Trans. Signal Processing,<br />

vol. 51, pp. 1941–1953, July 2003.<br />

[10] Ueda K., S Takahashi. I.: Digital filters with hypercomplex coefficients.<br />

In Proc. IEEE Int. Symp. Circuits Syst., May 1993, vol. 1,<br />

pp. 479–482.<br />

[11] Bayro-Corrochano E.: Multi-resolution image analysis using the<br />

quaternion wavelet transform. Numerical Algorithms, vol. 39, no<br />

1-3, July, 2005, pp. 35–55.<br />

[12] Conway J.H., Smith D.: On quaternions and octonions. A.K. Peters,<br />

2003.<br />

[13] Doukhnitch E.: Octonion CORDIC Algorithms for DSP.In Proc.<br />

of the 6th Symp. on Signal Processing, DSPCS’2002, Sydney,<br />

Australia, Jan. 2002., pp. 158–163.<br />

[14] Makarov О.М.: An algorithm for the multiplication of two quaternion.<br />

Computational Mathematics and Mathematical Physics.<br />

1977, т. 17, no. 6 pp. 1574–1575. (in Russian).<br />

[15] Ţariova G., Ţariov A.: Aspekty algorytmiczne redukcji liczby<br />

bloków mnożących w układzie do obliczania iloczynu dwóch<br />

kwaternionów. Pomiary, Automatyka, Kontrola, nr 7, <strong>2010</strong>,<br />

ss. 668–690.<br />

[16] Ţariov A.: Strategie racjonalizacji obliczeń przy wyznaczaniu<br />

iloczynów macierzowo-wektorowych. Metody Informatyki Stosowanej,<br />

nr 1, 2008, ss. 147–158.<br />

Wpłata w <strong>2010</strong> roku – GwarancjĄ niŻszej ceny prenumeraty o vat!<br />

108<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Technika próżni i technologie próżniowe<br />

Biuletyn Polskiego Towarzystwa Próżniowego 1 (50) <strong>2010</strong><br />

pod redakcją dr hab. inż. Ryszarda Korbutowicza<br />

e-mail: Ryszard.Korbutowicz@pwr.wroc.pl<br />

http://www.ptp.pwr.wroc.pl<br />

Biuletyn dofinansowany przez Ministerstwo Nauki i Szkolnictwa Wyższego<br />

XVIII Walne Zebranie Członków Polskiego Towarzystwa Próżniowego<br />

Zebranie sprawozdawczo-wyborcze<br />

Wrocław 24.09.<strong>2010</strong><br />

Odbyło się podczas połączonych konferencji ELTE <strong>2010</strong><br />

oraz IMAPS-CMPT <strong>2010</strong>, zwołane przez Zarząd Polskiego<br />

Towarzystwa Próżniowego na podstawie uchwały Zarządu<br />

PTP nr 3/<strong>2010</strong>:<br />

Uchwała Zarządu PTP nr 3/<strong>2010</strong><br />

Na podstawie §25 pkt. 4 statutu PTP Zarząd zwołuje Walne<br />

Zebranie Sprawozdawczo-Wyborcze Członków Towarzystwa<br />

we Wrocławiu, w budynku Politechniki Wrocławskiej<br />

w dniu 24.09.<strong>2010</strong> – w trakcie Konferencji ELTE<br />

<strong>2010</strong><br />

na godz. 19 00 w I terminie,<br />

na godz. 19 15 w II terminie<br />

z następującym porządkiem obrad:<br />

1. Otwarcie zebrania<br />

2. Wybory Przewodniczącego i Sekretarza Walnego Zebrania<br />

3. Przyjęcie porządku obrad<br />

4. Wybory Komisji Skrutacyjnej<br />

5. Wybory Komisji Wnioskowej<br />

6. Sprawozdanie finansowe Zarządu za 2009 r.<br />

7. Sprawozdanie z działalności Zarządu i Sekcji Tematycznych<br />

PTP w okresie VI kadencji (2007-<strong>2010</strong>)<br />

8. Sprawozdanie z działalności Komisji Rewizyjnej PTP<br />

i przedstawienie wniosku w sprawie absolutorium dla<br />

ustępującego Zarządu PTP<br />

9. Dyskusja nad sprawozdaniami Zarządu i Komisji Rewizyjnej<br />

10. Głosowanie w sprawie udzielenia absolutorium ustępującemu<br />

Zarządowi PTP<br />

11. Przekazanie prof. dr hab. J. Szuberowi, dotychczasowemu<br />

Przewodniczącemu-Elektowi, obowiązków<br />

Przewodniczącego PTP<br />

12. Przedstawienie planu działania PTP na nową kadencję<br />

oraz budżetu PTP na rok <strong>2010</strong><br />

13. Dyskusja nad planem działania i budżetem<br />

14. Wybory Przewodniczącego-Elekta<br />

15. Wybory Przewodniczącego Komisji Rewizyjnej<br />

16. Ustalenie składu Zarządu nowej kadencji i wybory<br />

członków Zarządu<br />

17. Ustalenie składu Komisji Rewizyjnej nowej kadencji<br />

i wybory członków Komisji<br />

18. Wolne wnioski<br />

19. Sprawozdanie Komisji Wnioskowej<br />

20. Zakończenie Walnego Zebrania<br />

Zarząd zobowiązuje sekretarza organizacyjnego PTP<br />

do zawiadomienia wszystkich Członków Towarzystwa<br />

o terminie, miejscu i porządku Walnego Zebrania.<br />

Warszawa, 29 stycznia <strong>2010</strong> r.<br />

za Zarząd<br />

dr inż. Katarzyna Olszewska<br />

Sekretarz Organizacyjny<br />

Polskiego Towarzystwa Próżniowego<br />

Miejsce zebrania: budynek Wydziału Elektroniki Mikrosystemów<br />

i Fotoniki Politechniki Wrocławskiej C2, Wrocław,<br />

ul. Janiszewskiego 11/17<br />

Termin I godz. 19 00 – zebranie nie odbyło się z powodu<br />

braku kworum.<br />

Aktualna liczba członków zwyczajnych PTP – 136 osób<br />

– obecnych 25 członków<br />

Aktualna liczba członków wspierających PTP – 13 – obecnych<br />

5 członków<br />

Załącznik 1 – Lista obecności członków PTP na Walnym<br />

Zebraniu w dniu 24.09.<strong>2010</strong> r.<br />

Zgodnie z §20 pkt 1 Statutu zebranie rozpoczęło się w terminie<br />

II o godz. 19 15 przy obecnych 25 członkach zwyczajnych<br />

i 5 członkach wspierających miał przebieg zgodny<br />

z wyżej zamieszczonym Porządkiem Obrad.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 109


Przebieg zebrania<br />

Walne Zebranie otworzył przewodniczący PTP prof. St. Hałas<br />

witając zebranych i zaproponował wybranie Przewodniczącego<br />

i Sekretarza Zebrania. Walne Zebranie w jawnym<br />

głosowaniu jednomyślnie wybrało: na przewodniczącego<br />

Walnego Zebrania prof. M. Godlewskiego, na sekretarza<br />

Walnego Zebrania dr K. Olszewską. Przewodniczący zebrania<br />

poddał pod głosowanie przyjęcie zaproponowanego Porządku<br />

Obrad. Walne Zebranie jednogłośnie w głosowaniu<br />

jawnym wyraziło zgodę. Następnie wybrano jednomyślnie<br />

Komisję Skrutacyjną w składzie: dr hab. Z. Stępień, dr D. Radziewicz,<br />

prof. A. Ciszewski. Komisja wybrała na przewodniczącego<br />

dr D. Radziewicza i stwierdziła liczbę członków PTP<br />

biorących udział w Walnym Zebraniu. W jawnym głosowaniu<br />

wybrano też jednomyślnie Komisję Wnioskową Zebrania<br />

w składzie: dr hab. E. Czerwosz i dr K. Marszałek.<br />

Kolejnym punktem obrad było sprawozdanie finansowe<br />

Zarządu za 2009 rok oraz realizacja budżetu za ten<br />

okres, które złożył skarbnik PTP dr M. Kozłowski. Następnie<br />

Przewodniczący PTP prof. St. Hałas przedstawił sprawozdanie<br />

merytoryczne z działalności Zarządu i Sekcji<br />

PTP w okresie kadencji (2007–<strong>2010</strong>).<br />

Sprawozdanie Zarządu Polskiego<br />

Towarzystwa Próżniowego z działalności<br />

w okresie od 26 kwietnia 2007 r.<br />

do 24 września <strong>2010</strong> r. w kadencji 2007–<strong>2010</strong><br />

W okresie sprawozdawczym Zarząd PTP pracował<br />

w składzie wybranym na Walnym Zebraniu Sprawozdawczo-Wyborczym<br />

we Wrocławiu w dniu 26 kwietnia 2007<br />

i uzupełnionym na Walnym Zebraniu w Janowie Lubelskim<br />

22 września 2008:<br />

Przewodniczący – do dnia 26 kwietnia 2008 – prof. Leszek<br />

MICHALAK, od 22 września 2008 – prof. Stanisław<br />

HAŁAS<br />

Przewodniczący poprzedniej kadencji – prof. Marek<br />

SZYMONSKI<br />

Przewodniczący-Elekt – prof. Jacek SZUBER<br />

Sekretarz Naukowy – dr hab. inż. Marek TŁACZAŁA<br />

Sekretarz organizacyjny – dr inż. Katarzyna OLSZEWSKA<br />

Skarbnik – dr Mirosław KOZŁOWSKI<br />

Przewodniczący Sekcji:<br />

prof. Witold PRECHT, od 22 września 2008 dr Ryszard<br />

MANIA – Sekcja Plazmowej Inżynierii Powierzchni<br />

dr hab. Zdzisław STĘPIEŃ – Sekcja Nauki o Powierzchni<br />

prof. Elżbieta CZERWOSZ – Sekcja Techniki Próżni<br />

dr hab. inż. Regina PASZKIEWICZ – Sekcja Cienkich<br />

Warstw<br />

Pozostali Członkowie Zarządu to: dr Janusz BUDZIOCH,<br />

prof. Antoni CISZEWSKI, dr hab. inż. Ryszard KORBUTO-<br />

WICZ, dr hab. Leszek MARKOWSKI.<br />

Zgodnie z planem pracy przyjętym dla całej kadencji<br />

2007–<strong>2010</strong> w okresie sprawozdawczym prace Zarządu<br />

PTP koncentrowały się na następujących zagadnieniach:<br />

1. działalności naukowej,<br />

2. działalności szkoleniowej,<br />

3. działalności wydawniczej (biuletyn, witryna internetowa<br />

PTP: http://ww.ptp.pwr.wroc.pl, oraz wydawanie<br />

materiałów szkoleniowych i edukacyjnych)<br />

4. współpracy z zagranicznymi organizacjami próżniowymi<br />

oraz IUVSTA<br />

5. sprawach organizacyjno-członkowskich<br />

Materiały źródłowe:<br />

1. Plan działania PTP na lata 2007–<strong>2010</strong><br />

2. Sprawozdanie Zarządu Polskiego Towarzystwa Próżniowego<br />

z działalności w okresie od 26 kwietnia 2007<br />

do 22 września 2008 w kadencji 2007–<strong>2010</strong><br />

3. Sprawozdanie Zarządu Polskiego Towarzystwa Próżniowego<br />

działalności w okresie od 22 września 2008<br />

do 20 maja 2009 w kadencji 2007–<strong>2010</strong><br />

4. Biuletyny PTP: Nr 44–45 (2007), Nr 46–47 (2008) i 48<br />

(2009) opublikowane w Elektronice i rozesłane do<br />

wszystkich członków PTP.<br />

5. Protokoły z posiedzeń Zarządu PTP:<br />

• nr 1/3/2007 z dn.26.04.2007 r. we Wrocławiu,<br />

• nr 1/2008 z dn. 11.01.2008 w Warszawie<br />

• nr 2/2008 z dn. 13.06.2008 w Warszawie<br />

• nr 3/2008 z dn. 22.09.2008 w Janowie Lubelskim<br />

• nr 1/2009 z dn. 30.01.2009 w Warszawie<br />

• nr 2/2009 z dn. 19.05.2009 w Szklarskiej Porębie<br />

• nr 3/2009 z dn. 06.11.2009 w Warszawie<br />

• nr 1/<strong>2010</strong> z dn. 29.01.<strong>2010</strong> w Warszawie<br />

6. Witryna internetowa prowadzona na serwerze Politechniki<br />

Wrocławskiej: www.ptp.pwr.wroc.pl<br />

I. Działalność naukowa<br />

Konkurs im. J. Groszkowskiego<br />

Odpowiedzialnym za organizację konkursu jest prof. dr<br />

hab. inż. Marek Tłaczała, przewodniczący komisji konkursowej.<br />

Komisja Konkursowa działa w składzie: prof.<br />

dr hab. Jacek Szuber, doc. dr hab. inż. Elżbieta Czerwosz,<br />

prof. dr hab. Zdzisław Stępień oraz dr hab. inż. Regina<br />

Paszkiewicz, prof. PWr.<br />

Komisja po rozpatrzeniu zgłoszonych do konkursu prac<br />

oraz na podstawie opinii wystawianej przez każdorazowo<br />

powoływanego recenzenta, przyznała w czasie obecnej<br />

kadencji następujące nagrody:<br />

• Rok 2007 – dr inż. Monika Kwoka w kategorii najlepsza<br />

praca doktorska za rozprawę pt. Studies of surface<br />

properties of L-CVD SnO 2<br />

thin films wykonaną<br />

w Instytucie Fizyki Politechniki Śląskiej w Gliwicach<br />

pod kierunkiem prof. Jacka Szubera. Laureatka nagrodzonej<br />

pracy doktorskiej odebrała nagrodę (dyplom)<br />

i wygłosiła komunikat na IV Kongresie PTP w Janowie<br />

Lubelskim. W kategorii prace magisterskie w tej edycji<br />

konkursu nie została zgłoszona żadna praca.<br />

• Rok 2008 – mgr inż. Michał Krupiński w kategorii najlepsza<br />

praca magisterska za pracę pt. Badanie wpływu<br />

surfaktantów na wzrost cienkich warstw w układzie<br />

cienkowarstwowym wykonaną na Wydziale<br />

Fizyki i Informatyki Stosowanej, Akademii Górniczo-<br />

Hutniczej pod kierunkiem doc. dr hab. Marty Wolny-<br />

Marszałek. W kategorii prace doktorskie w tej edycji<br />

nie została zgłoszona żadna praca<br />

• Rok 2009 – dr inż. Mateusz Wośko w kategorii najlepsza<br />

praca doktorska za rozprawę pt. Opracowanie<br />

konstrukcji i technologii fotodetektorów z zastosowaniem<br />

nanostruktur półprzewodników AIIIBV o ciągłej<br />

zmianie składu wykonaną na Wydziale Elektroniki<br />

Mikrosystemów i Fotoniki Politechniki Wrocławskiej<br />

pod kierunkiem dr hab. inż. Reginy Paszkiewicz,<br />

prof. PWr. W kategorii prace magisterskie w tej edycji<br />

nie została zgłoszona żadna praca.<br />

110<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Na początku września br. prof. Tłaczała ogłosił kolejny<br />

konkurs na najlepsze prace doktorskie i magisterskie prace<br />

obronione w <strong>2010</strong> r. Termin składania propozycji prac<br />

mija 31 grudnia.<br />

Konferencje zorganizowane pod patronatem PTP<br />

Organizacja konferencji naukowych stanowi jedną z najważniejszych<br />

merytorycznie form działania PTP z powodzeniem<br />

realizowaną przez poszczególne sekcje tematyczne<br />

Towarzystwa. W okresie sprawozdawczym<br />

zorganizowano pod patronatem PTP 6 konferencji, 4 warsztatów<br />

naukowych o zasięgu międzynarodowym oraz 2<br />

Szkoły Techniki Próżniowej.<br />

Rok 2008<br />

VI International Workshop on Semiconductor Gas<br />

Sensors – SGS2008, Zakopane, 14-19 września 2008.<br />

Organizatorami byli: Sekcja Nauki o Powierzchni i Sekcja<br />

Cienkich Warstw PTP oraz Europejskie Centrum Doskonałości<br />

CESIS i Krajowe Centrum Doskonałości NANOMET<br />

przy Zakładzie Technologii Elektronowej Politechniki Śląskiej<br />

w Gliwicach. Konferencja otrzymała dofinansowanie<br />

z Ministerstwa Nauki i Szkolnictwa Wyższego z wniosku<br />

PTP. Dyrektorem Konferencji był prof. Jacek Szuber, Politechnika<br />

Śląska w Gliwicach. W warsztatach SGS 2008<br />

wzięło udział 53 uczestników, w tym 40 z zagranicy. Podczas<br />

konferencji zostały wygłoszone referaty plenarne<br />

i komunikaty, oraz prezentacje na sesji producentów. Materiały<br />

pokonferencyjne ukazały się w czasopiśmie Thin<br />

Solid Films 517 (2009), 6147-6216 (Elsevier).<br />

IV Kongres PTP i VIII Krajowa Konferencja Techniki<br />

Próżni, Janów Lubelski, 21–24 września 2008. Organizatorzy<br />

to Sekcja Techniki Próżni, <strong>Instytut</strong> Tele- i Radiotechniczny,<br />

<strong>Instytut</strong> Fizyki UMCS. Otrzymano dofinansowanie<br />

z Ministerstwa Nauki i Szkolnictwa Wyższego z wniosku<br />

PTP. Dyrektorem Konferencji była prof. Elżbieta Czerwosz,<br />

<strong>Instytut</strong> Tele-Radiotechniczny, Warszawa. W czasie konferencji<br />

zostały wygłoszone referaty plenarne i komunikaty,<br />

oraz prezentacje na sesji producentów. Materiały pokonferencyjne<br />

ukazały się w miesięczniku <strong>Elektronika</strong> nr 1/2009.<br />

IV Workshop on Hybrid Nanostructured Materials, Advanced<br />

Nanomaterials, Their Preparation and Analysis,<br />

Wrocław, 24–27 października 2008.<br />

Organizatorzy: Sekcja Cienkich Warstw PTP, Wydział<br />

Elektroniki Mikrosystemów i Fotoniki Politechniki Wrocławskiej.<br />

W trakcie konferencji zostały wygłoszone referaty<br />

plenarne i komunikaty.<br />

Rok 2009<br />

XI Seminarium Powierzchnia i Struktury Cienkowarstwowe,<br />

Szklarska Poręba, 19–22 maja 2009. Organizatorzy:<br />

Sekcja Nauki o Powierzchni i Sekcja Struktur Cienkowarstwowych<br />

PTP, Wydział Elektroniki Mikrosystemów<br />

i Fotoniki Politechniki Wrocławskiej. Seminarium otrzymało<br />

dofinansowanie z Ministerstwa Nauki i Szkolnictwa Wyższego<br />

z wniosku PTP. W konferencji uczestniczyło blisko 70<br />

osób. Program konferencji obejmował referaty zaproszone<br />

i krótkie komunikaty z prac własnych, praz prezentacje plakatowe.<br />

Materiały pokonferencyjne ukazały się w czasopismach<br />

Optica Applicata oraz Materials Science Poland.<br />

III Krajowa Konferencja Nanotechnologii, Warszawa,<br />

22–26 czerwca 2009. Organizatorami były: Polskie<br />

Towarzystwo Próżniowe, <strong>Instytut</strong> Technologii Elektronowej,<br />

<strong>Instytut</strong> Fizyki PAN, Uniwersytet Warszawski. Konferencja<br />

otrzymała dofinansowanie z Ministerstwa Nauki<br />

i Szkolnictwa Wyższego z wniosku PTP. Dyrektorem<br />

Konferencji był prof. Maciej Bugajski, <strong>Instytut</strong> Technologii<br />

Elektronowej, Warszawa. Program konferencji składał się<br />

z serii zaproszonych referatów plenarnych, oraz z referatów<br />

zaproszonych jak i krótkich komunikatów ustnych<br />

w ramach równoległych sesji tematycznych. Konferencji<br />

towarzyszyła duża wystawa sprzętu pomiarowego.<br />

Materiały pokonferencyjne ukazały się w czasopiśmie<br />

Acta Physica Polonica A.<br />

IV Symposium on Vacuum based Science and Technology,<br />

Koszalin, 21–23 września 2009. Organizatorzy<br />

to: Polskie Towarzystwo Próżniowe, Niemieckie Towarzystwo<br />

Próżniowe, Politechnika Koszalińska. Dyrektor<br />

Konferencji: prof. Witold Gulbiński, Politechnika Koszalińska.<br />

Program konferencji składał się z serii zaproszonych<br />

referatów plenarnych oraz komunikatów. Odbyła<br />

się również sesja producentów aparatury próżniowej<br />

i pomiarowej.<br />

VI International Workshop on Semiconductor Surface<br />

Passivation, Zakopane, 13–18 września 2009. Organizatorzy<br />

Workshopu: Sekcja Nauki o Powierzchni<br />

i Sekcja Cienkich Warstw PTP oraz Europejskie Centrum<br />

Doskonałości CESIS i Krajowe Centrum Doskonałości<br />

NANOMET w Instytucie Elektroniki Politechniki Śląskiej<br />

w Gliwicach. Otrzymano dofinansowanie z Ministerstwa<br />

Nauki i Szkolnictwa Wyższego z wniosku PTP. Dyrektor<br />

Konferencji: prof. Jacek Szuber, Politechnika Śląska<br />

w Gliwicach. W warsztatach SSP 2009 wzięło udział 48<br />

uczestników, w tym 40 z zagranicy. W czasie konferencji<br />

zostały wygłoszone referaty plenarne i komunikaty, oraz<br />

prezentacje plakatowe. Materiały pokonferencyjne ukazały<br />

się w czasopiśmie Applied Surface Science 256 (<strong>2010</strong>)<br />

5697-5794 (Elsevier), <strong>2010</strong> r.<br />

IV Krajowa Konferencja Nanotechnologii, Poznań,<br />

22–26 czerwca <strong>2010</strong>. Organizatorzy: Polskie Towarzystwo<br />

Próżniowe, <strong>Instytut</strong> Technologii Elektronowej, <strong>Instytut</strong><br />

Fizyki PAN, Uniwersytet Warszawski. Uzyskano dofinansowanie<br />

z Ministerstwa Nauki i Szkolnictwa Wyższego<br />

z wniosku PTP. Dyrektor Konferencji: prof. Ryszard Czajka,<br />

Politechnika Poznańska. Program konferencji składał<br />

się z serii zaproszonych referatów plenarnych, oraz z referatów<br />

zaproszonych jak i krótkich komunikatów ustnych<br />

w ramach równoległych sesji tematycznych. Konferencji<br />

towarzyszyła duża wystawa sprzętu pomiarowego. Materiały<br />

pokonferencyjne ukażą się w czasopiśmie Acta<br />

Physica Polonica A.<br />

XXIV International Conference on Atomic Collisions<br />

in Solids, Kraków, 18–23 lipca <strong>2010</strong>. Organizatorzy to:<br />

Polskie Towarzystwo Próżniowe, <strong>Instytut</strong> Fizyki – Centrum<br />

NANOSAM Uniwersytetu Jagiellońskiego. Otrzymano<br />

dofinansowanie z Ministerstwa Nauki i Szkolnictwa Wyższego<br />

na wniosek PTP. Dyrektor Konferencji: prof. Marek<br />

Szymoński, Uniwersytet Jagielloński, Kraków.<br />

VII International Workshop on Semiconductor Gas<br />

Sensors – SGS<strong>2010</strong>, Kraków, 12–-16 września <strong>2010</strong>.<br />

Organizatorzy: Sekcja Nauki o Powierzchni i Sekcja Cienkich<br />

Warstw PTP, oraz Europejskie Centrum Doskonałości<br />

CESIS i Krajowe Centrum Doskonałości NANOMET<br />

w Instytucie Elektroniki Politechniki Śląskiej w Gliwicach.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 111


Uzyskano dofinansowanie z Ministerstwa Nauki i Szkolnictwa<br />

Wyższego z wniosku PTP. Dyrektor Konferencji: prof.<br />

Jacek Szuber, Politechnika Śląska w Gliwicach. W warsztatach<br />

SGS <strong>2010</strong> wzięło udział 58 uczestników, w tym<br />

45 z zagranicy. W czasie konferencji zostały wygłoszone<br />

referaty plenarne i komunikaty ustne, oraz prezentacje plakatowe.<br />

Podsumowaniem konferencji była Dyskusja Okrągłego<br />

Stołu (RTD). Materiały pokonferencyjne ukażą się<br />

w roku 2011 w czasopiśmie Thin Solid Films (Elsevier).<br />

Aktualnie prowadzone są już prace nad przygotowaniem<br />

najbliższych konferencji naukowych organizowanych<br />

przez lub pod auspicjami PTP.<br />

II. Działalność szkoleniowa<br />

Działalność szkoleniowa jest jednym z najważniejszych<br />

zadań statutowych Towarzystwa. Działalność ta jest ściśle<br />

związana z zadaniami statutowymi Towarzystwa, jest<br />

niezwykle potrzebna i będzie kontynuowana zgodnie<br />

z wcześniejszymi ustaleniami, tj. kolejne cykle szkolenia<br />

będą odbywać się co 2 lata.<br />

W dniach 25–28 czerwca 2008 odbyła się III Szkoła<br />

Podstaw Techniki Próżni, którą zorganizowali: Centrum<br />

Badan Układów Nanoskopowych i Zaawansowanych Materiałów,<br />

NANOSAM, <strong>Instytut</strong> Fizyki, Uniwersytet Jagielloński<br />

w Krakowie. Dyrektorem Szkoły był dr Janusz Budzioch<br />

z Uniwersytetu Jagiellońskiego w Krakowie.<br />

Szkołę zorganizowano jako imprezę towarzyszącą<br />

II Krajowej Konferencji Nanotechnologii w Krakowie.<br />

Obejmowała dwa zakresy: podstawowy i zaawansowany.<br />

W kursie podstawowym wzięło udział 47 osób,<br />

a w zaawansowanym – 42 osoby, przy czym część osób<br />

uczestniczyła w obu kursach. 32 osoby były pracownikami<br />

instytucji związanych z przemysłem, a 21 osób<br />

z instytucjami naukowymi. Każdy z uczestników otrzymał<br />

podręcznik A. Hałasa i P. Szwemina, który stanowił<br />

podstawowy materiał wykładowy. Szkolenie zostało<br />

zakończone testem kwalifikacyjnym. Sprawozdanie<br />

ze Szkoły dostępne jest na stronie internetowej PTP<br />

http://www.ptp.pwr.wroc.pl.<br />

IV Szkoła Techniki Próżniowej została zorganizowana<br />

w dniach 19–23.07.<strong>2010</strong> r. w Krakowie równolegle do<br />

międzynarodowych konferencji: 24th International Conference<br />

on Atomic Collisions in Solids (ICACS-24) oraz<br />

10th International Conference on Computer Simulation<br />

of Radiation Effects in Solids (COSIRES-10). Połączenie<br />

Szkoły Techniki Próżniowej z w/w konferencjami ma na<br />

celu obniżenie kosztów organizacyjnych STP, jak również<br />

ma umożliwić udział w Szkole wielu krajowym i międzynarodowym<br />

wystawcom aparatury badawczej i próżniowej.<br />

Uczestnicy STP zapoznali się z szeroką gamą produktów<br />

oraz z dostawcami i producentami sprzętu w trakcie trwania<br />

wystawy i sesji producentów.<br />

W szkoleniu wzięło udział 41 osób. Kursanci reprezentowali<br />

zarówno instytuty naukowo-badawcze, uczelnie<br />

jak i zakłady przemysłowe w bardzo podobnym stopniu.<br />

W szkoleniu wzięło udział również 7 pań. Wykładowcami<br />

byli: prof. dr hab. inż. Andrzej Hałas, prof. dr hab. inż.<br />

Witold Precht, dr inż. Konstanty Marszałek, dr Jacek<br />

Marzec, dr Artur Kajoch, dr Janusz Budzioch i mgr inż.<br />

Witold Skrzypulec. Uroczystego powitania uczestników<br />

oraz zamknięcia szkolenia dokonał prof. dr hab. Marek<br />

Szymoński. Wiele informacji na temat tej szkoły można<br />

znaleźć na stronie internetowej: http://www.if.uj.edu.pl/<br />

NANOSAM/PTP<strong>2010</strong>/<br />

112<br />

III. Działalność wydawnicza<br />

Działalność wydawnicza jest również jednym z najważniejszych<br />

zadań statutowych Towarzystwa. W ramach tej działalności<br />

w okresie sprawozdawczym ukazały się Biuletyny<br />

PTP: 1–2/2007 – wraz z numerem 12/2007 czasopisma<br />

<strong>Elektronika</strong>, 1–2/2008 – wraz z numerem 12/2008 czasopisma<br />

<strong>Elektronika</strong>, 1/2009 – wraz z numerem 11/2009 czasopisma<br />

<strong>Elektronika</strong>.<br />

Sprawnie działa witryna internetowa PTP http://www.<br />

ptp.pwr.wroc.pl i jest na bieżąco aktualizowana, co zawdzięczamy<br />

dr hab. inż. Ryszardowi Korbutowiczowi. Posiada<br />

ona linki aktualności, informacji, konferencji, Konkursu<br />

im. J. Groszkowskiego, biuletynów PTP oraz ma też<br />

ma rubrykę wspomnień (znajdujemy w niej wspomnienie<br />

o doc. Januszu Sobańskim, jednym z założycieli Polskiego<br />

Towarzystwa Próżniowego), prof. Piotrze Szweminie,<br />

prof. Leszku Michalaku oraz o prof. Edwardzie Lei. Są<br />

tam także linki do stron internetowych organizacji i towarzystw<br />

próżniowych. Wśród ciekawostek znajdujemy<br />

zredagowany przez prof. Andrzeja Hałasa opis pierwszego<br />

eksperymentu próżniowego przeprowadzonego<br />

na dworze króla Władysława IV. Dzięki uprzejmości prof.<br />

J. Zdanowskiego zainteresowani znajdą również szeroki<br />

przegląd osiągnięć polskiej próżni w materiałach Polska<br />

<strong>Elektronika</strong> Próżniowa Wczoraj i Dziś wydanych przez<br />

oficynę Wydawniczą Politechniki Wrocławskiej.<br />

IV. Współpraca z zagranicznymi<br />

towarzystwami próżniowymi i IUVSTA<br />

W tym zakresie prowadzono rozmowy z Niemieckim Towarzystwem<br />

Próżniowym, Szwajcarskim Towarzystwem<br />

Próżniowym i Szwedzkim Towarzystwem Próżniowym na<br />

temat zorganizowanej we wrześniu 2009 w Koszalinie kolejnej<br />

wspólnej konferencji naukowej p.t. Vacuum Based<br />

Science and Technology oraz wystawy sprzętu próżniowego.<br />

Konferencja ta jest kontynuacją wcześniejszych<br />

konferencji z tego cyklu zorganizowanych w Krakowie<br />

(2005) i Darmstadt (2006). Kolejna konferencja z tego<br />

cyklu odbędzie się w Kaiserslautern (Niemcy) w dnach<br />

27–29 września <strong>2010</strong>.<br />

Współpraca z IUVSTA<br />

Polskie Towarzystwo Próżniowe bardzo aktywnie współpracuje<br />

z IUVSTA. Wyrazem naszej pozycji w tej organizacji<br />

było zaproszenie Prezesa Polskiego Towarzystwa<br />

Próżniowego do objęcia funkcji Przewodniczącego Komitetu<br />

Reprezentantów Europejskich Towarzystw Próżniowych,<br />

będącego kolegium Prezesów tych Towarzystw,<br />

lub ich Zastępców, którego zadaniem jest zaktywizowanie<br />

na płaszczyźnie europejskiej współpracy naukowej<br />

i edukacyjnej w zakresie nauki o próżni i dziedzinach<br />

wykorzystujących próżnię oraz technologii próżniowych.<br />

Istotnym materialnym wymiarem polskiego uczestnictwa<br />

w pracach tego gremium było uzyskanie aż trzykrotnie<br />

dofinansowania IUVSTA do imprez szkoleniowych i naukowych<br />

organizowanych wcześniej w Polsce pod patronatem<br />

PTP, na łączną kwotę 17 tysięcy franków szwajcarskich.<br />

Jednocześnie w wyniku dotychczasowych prac<br />

Komitet pod przewodnictwem przedstawiciela Polski<br />

przedstawił swoje ustalenia Radzie Wykonawczej IUV-<br />

STA z rekomendacją, aby nie podejmować kroków w kierunku<br />

utworzenia nowych struktur organizacyjnych typu<br />

Europejskie Towarzystwo Próżniowe. W okresie bieżącej<br />

kadencji funkcję tę pełni prof. Marek Szymoński.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


V. Sprawy organizacyjno-członkowskie<br />

W okresie sprawozdawczym PTP poniosło bolesne straty,<br />

zmarli niezwykle zasłużeni dla Towarzystwa następujący<br />

członkowie:<br />

Prof. Leszek Michalak – UMCS Lublin, wieloletni członek<br />

Zarządu PTP, Sekretarz Naukowy, Przewodniczący PTP<br />

w latach 2007 – 2008, zmarł 26 kwietnia 2008 r.<br />

Prof. Piotr Szwemin – Politechnika Warszawska, wieloletni<br />

członek Zarządu PTP oraz organizator Konkursu im. J.<br />

Groszkowskiego, zmarł 9 stycznia 2008 r.<br />

Prof. Edward Leja – AGH Kraków, „najlepszy malarz wśród<br />

fizyków”, zmarł 5 października 2009 r.<br />

prof. Jerzy Haber – członek rzeczywisty PAN, <strong>Instytut</strong> Katalizy<br />

i Fizykochemii Powierzchni PAN, zmarł 1 stycznia<br />

<strong>2010</strong> r.<br />

mgr inż. Maria Tymieniecka, firma COMEF, zmarła 11 marca<br />

<strong>2010</strong> r.<br />

Wspomnienia o pierwszych trzech wymienionych zmarłych<br />

zostały opublikowane w biuletynach PTP.<br />

W okresie sprawozdawczym przyjęto 28 nowych<br />

członków indywidualnych PTP. Ponadto przyjęto również<br />

1 członka wspierającego – firmę Kurt J. Lesker. Obecnie<br />

PTP ma 136 członków indywidualnych, w tym 3 honorowych<br />

i 13 członków wspierających.<br />

VI. Realizacja planu działalności Zarządu<br />

PTP na okres kadencji 2007–<strong>2010</strong><br />

przyjętego na Walnym Zebraniu<br />

Sprawozdawczo-Wyborczym<br />

26 kwietnia 2007 r. we Wrocławiu:<br />

Zgodnie z planem działalności Zarządu PTP Towarzystwo<br />

miało tak jak do tej pory realizować wszystkie cele sformułowane<br />

w statucie PTP. Szczególna aktywność Zarządu,<br />

poszczególnych Sekcji i Członków miała być skupiona na<br />

następujących celach:<br />

1. Zwiększeniu liczebności PTP przez zainteresowanie<br />

pracami Towarzystwa szczególnie tzw. młodych ludzi<br />

nauki i techniki związanych z szeroko rozumianą tematyką<br />

próżniową. Działania te powinny być prowadzone<br />

między innymi przez:<br />

• Kontynuację organizowania konferencji naukowych,<br />

a także szkół i kursów dla użytkowników<br />

urządzeń próżniowych,<br />

• Kontynuację aktywności Towarzystwa w relacjach<br />

z IUVSTA, a szczególnie pod kątem zdobywania<br />

środków IUVSTA na działalność szkoleniową,<br />

• Szeroką promocję osiągnięć Towarzystwa jak<br />

i poszczególnych członków,<br />

• Lepszą i szerszą promocję nagród PTP. Działaniu<br />

temu powinna towarzyszyć praca nad doskonaleniem<br />

regulaminów nagród.<br />

2. Poprawie sytuacji finansowej Towarzystwa, która<br />

umożliwi jeszcze lepszą Jego promocję; pozwoli<br />

rozwinąć działalność popularyzatorską; zapewni<br />

płynność w przyznawaniu ustanowionych nagród.<br />

Działania te powinny być prowadzone między innymi<br />

przez:<br />

• poszukiwanie sponsorów nagród,<br />

• zwiększenie liczby członków.<br />

3. Rozwijaniu kontaktów i współpracy z zagranicznymi<br />

Towarzystwami Próżniowymi oraz krajowymi towarzystwami<br />

o zbliżonej tematyce działalności<br />

Poniżej przedstawiono krótki komentarz do przyjętego<br />

planu działania ilustrujący zakres i stopień realizacji planowanych<br />

działań.<br />

1. Zwiększenie liczy Członków Towarzystwa PTP, odnotowano<br />

stały wzrost przyjęć; przyjęto 28 nowych<br />

członków indywidualnych oraz 1 członka wspierającego.<br />

Nastąpiła pewna konsolidacja działalności firm<br />

próżniowych w Polsce w korelacji z działalnością Towarzystwa.<br />

Udało się zaktywizować niektórych członków<br />

wspierających uczestniczących w imprezach organizowanych<br />

przez PTP (szkolenia i wystawy sprzętu<br />

próżniowego przy okazji konferencji organizowanych<br />

pod auspicjami PTP).<br />

Udało się natomiast przeprowadzić, mimo niespodziewanej<br />

śmierci prof. P. Szwemina – przewodniczącego<br />

Komisji Konkursowej – kolejne edycje Konkursu im. J.<br />

Groszkowskiego za lata 2007- 2009. Przewodniczący<br />

Komisji Konkursowej ogłosił kolejną edycję konkursu<br />

– za prace obronione w <strong>2010</strong> r.<br />

2. Poprawa sytuacji finansowej Towarzystwa – jest stabilna<br />

i zapewnia pokrycie wszystkich planowanych<br />

wydatków, w tym nagród w Konkursie im J. Groszkowskiego.<br />

Zarząd pozyskał zgodnie z planem środki<br />

z Ministerstwa Nauki i Szkolnictwa Wyższego na imprezy<br />

organizowane przez lub pod auspicjami PTP.<br />

3. Rozwijanie współpracy. Jak wspomniano przy opisie<br />

działalności naukowej, były prowadzone rozmowy<br />

z Niemieckim Towarzystwem Próżniowym na<br />

temat organizacji w 2009 r. w Koszalinie i wspólnej<br />

cyklicznej konferencji naukowej pt. Vacuum Based<br />

Science and Technology oraz wystawy sprzętu<br />

próżniowego.<br />

Iberyjskie Towarzystwo próżniowe nawiązało kontakt<br />

z PTP w związku z organizacją konferencji w Salamance<br />

we wrześniu <strong>2010</strong> r. Do komitetu naukowego tej konferencji<br />

wszedł prof. Jacek Szuber, a do komitetu wystawienniczego<br />

– doc. Czesław Kiliszek.<br />

Prezes Pakistańskiego Towarzystwa Próżniowego<br />

zwrócił się z prośbą do Przewodniczącego Towarzystwa<br />

o wydelegowanie przedstawiciela PTP na organizowana<br />

przez nich konferencje w Islamabadzie w listopadzie<br />

<strong>2010</strong> r. Niestety nikogo nie udało się namówić do wyjazdu<br />

na koszt organizatorów ze względu na niebezpieczną sytuację<br />

w regionie.<br />

Obecnie przedstawicielami PTP do IUVSTA na lata<br />

<strong>2010</strong>–2013 są:<br />

Nominacje na GM 17 oraz Triennium <strong>2010</strong>–2013<br />

Szefem delegacji PTP jest prof. dr hab. Marek Szymoński<br />

(UJ), delegatem jest dr hab. Leszek Markowski (UWr).<br />

Profesor Marek Szymoński został także członkiem Rady<br />

Wykonawczej, a jego zmiennikiem jest profesor Jacek<br />

Szuber (PŚl.).<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 113


Przedstawicielami PTP w Sekcjach Tematycznych IUVSTA są:<br />

Sekcja Imię i nazwisko Adres kontaktowy<br />

Applied Surface Science Leszek Markowski <strong>Instytut</strong> Fizyki Doświadczalnej, Uniwersytet Wrocławski,<br />

Pl. Maksa Borna 9, 50-204 Wrocław<br />

Electronic Materials and Processing Marek Godlewski <strong>Instytut</strong> Fizyki PAN, al. Lotników 32/46, 02-668 Warszawa<br />

Nanometer Structures Marek Szymoński Uniwersytet Jagielloński, ul. Reymonta 4, 30-059 Kraków<br />

Plasma Science and Technique Marek Tłaczała Politechnika Wrocławska, Wydział Elektroniki Mikrosystemów<br />

i Fotoniki, ul. Janiszewskiego 11/17 50-372 Wrocław<br />

Surface Engineering Witold Precht Politechnika Koszalińska, Wydział Mechaniczny,<br />

ul. Racławicka 15-17, 75-620 Koszalin<br />

Surface Science Antoni Ciszewski <strong>Instytut</strong> Fizyki Doświadczalnej Uniwersytetu Wrocł.<br />

pl. Maksa Borna 9, 50-204 Wrocław<br />

Thin Films Regina Paszkiewicz Politechnika Wrocławska, Wydział Elektroniki Mikrosystemów<br />

i Fotoniki, ul. Janiszewskiego 11/17 50-372 Wrocław<br />

Vacuum Science and Technology Elżbieta Czerwosz <strong>Instytut</strong> Tele- i Radiotechniczny, ul. Długa 44/50,<br />

00-241 Warszawa<br />

Sprawozdanie z działalności Komisji Rewizyjnej PTP<br />

przedstawił przewodniczący tej Komisji doc. Cz. Kiliszek.<br />

Na zakończenie sprawozdania doc. Cz. Kiliszek w imieniu<br />

Komisji Rewizyjnej postawił wnioski o przyjęcie sprawozdań:<br />

Skarbnika i Przewodniczącego i udzielenie absolutorium<br />

ustępującemu zarządowi PTP.<br />

Ponieważ zebrani nie mieli żadnych uwag do sprawozdań,<br />

Przewodniczący Walnego Zebrania prof. M. Godlewski<br />

zaproponował podjęcie trzech uchwał w sprawie<br />

przedstawionych sprawozdań oraz udzielenia absolutorium<br />

Zarządowi i skwitowania Komisji Rewizyjnej. Zebrani<br />

w głosowaniu jawnym podjęli jednomyślnie Uchwałę nr 1<br />

w sprawie sprawozdania Zarządu i sprawozdania finansowego<br />

PTP za 2009 r. Walne Zebranie w głosowaniu jawnym<br />

podjęło jednomyślnie także Uchwałę nr 2 w sprawie<br />

udzielenia Zarządowi absolutorium oraz Uchwałę nr 3<br />

w sprawie skwitowania Komisji Rewizyjnej.<br />

Po udzieleniu absolutorium Zarządowi PTP i skwitowaniu<br />

Komisji Rewizyjnej prof. St. Hałas przekazał dotychczasowemu<br />

Przewodniczącemu-Elektowi prof. J. Szuberowi<br />

obowiązki Przewodniczącego PTP. Przewodniczący<br />

PTP prof. J. Szuber przedstawił plan działania PTP na<br />

nową kadencję, a dr M. Kozłowski przedstawił preliminarz<br />

budżetu PTP na rok <strong>2010</strong>.<br />

Plan działania<br />

Zarządu Polskiego Towarzystwa Próżniowego w kadencji<br />

<strong>2010</strong>–2013 przyjęty na Zebraniu Sprawozdawczo-<br />

Wyborczym w dniu 24 września <strong>2010</strong> r. we Wrocławiu.<br />

114<br />

Towarzystwo będzie tak jak do tej pory, realizować przede<br />

wszystkim główne cele sformułowane w statucie PTP.<br />

Szczególna aktywność Zarządu, poszczególnych Sekcji<br />

Tematycznych, a także samych Członków, również Wspierających,<br />

powinna być skupiona na 4 głównych celach:<br />

1. Promocji tematyki próżniowej, w tym zainteresowaniu<br />

nią szczególnie młodych ludzi nauki i techniki nie tylko<br />

bezpośrednio związanych z tą tematyką, co powinno<br />

przyczynić się do zwiększenia liczebności PTP i większe<br />

zainteresowanie pracami samego Towarzystwa.<br />

Działania te powinny być prowadzone w następujących<br />

obszarach:<br />

• Dalszym organizowaniu nie tylko konferencji naukowych,<br />

ale i warsztatów naukowych, zwłaszcza<br />

pod kątem udziału w nich młodych ludzi<br />

nauki i techniki związanych z tematyką sekcji<br />

naukowych PTP<br />

• Dalszym organizowaniu szkół i kursów, nie tylko<br />

dla użytkowników urządzeń próżniowych, ale<br />

także dla użytkowników aparatury badawczej<br />

i technologicznej, zgodnie z zakresem działania<br />

sekcji tematycznych PTP<br />

• Szerokiej promocji osiągnięć PTP jak i poszczególnych<br />

jego członków poprzez m.in. uruchomienie<br />

linków do stron własnych zespołów badawczych<br />

• Szerokiej promocji nagród PTP, połączonej<br />

z pracą nad doskonaleniem regulaminów tych<br />

nagród.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


2. Poprawie sytuacji finansowej Towarzystwa m.in.<br />

przez:<br />

• poszukiwanie sponsorów nagród,<br />

• zwiększenie liczby członków, co ułatwi jego promocję,<br />

oraz zwiększy możliwości m.in. działalności<br />

popularyzatorskiej, a także zapewni np.<br />

płynność w przyznawaniu ustanowionych nagród.<br />

3. Dalszym rozwijaniu kontaktów i współpracy z zagranicznymi<br />

Towarzystwami Próżniowymi m.in. poprzez<br />

organizowanie wspólnych konferencji i warsztatów<br />

naukowych oraz krajowymi towarzystwami o zbliżonej<br />

tematyce działalności (PTF, PTCh).<br />

4. Kontynuację aktywności Towarzystwa w relacjach<br />

z IUVSTA, także pod kątem zdobywania środków<br />

IUVSTA na działalność szkoleniową.<br />

Zebrani dyskutowali następnie nad problemem istnienia<br />

w strukturach PTP Sekcji Nanostruktur i Nanotechnologii,<br />

zastanawiając się czy z uwagi na brak jej formalnego działania<br />

przy rozwijaniu tej tematyki w innych sekcjach (np.<br />

Sekcja Cienkich Warstw) nie zlikwidować tej sekcji. Analizowano<br />

również słuszność założonej wielkości wpływów<br />

ze składek członków wspierających. Uznano, że nie należy<br />

zmieniać tej kwoty. Więcej uwag nie było. W tej sytuacji<br />

przewodniczący Walnego Zebrania zaproponował podjęcie<br />

Uchwały nr 4 w sprawie planu działalności i budżetu<br />

na <strong>2010</strong> r. Walne Zebranie w głosowaniu jawnym podjęło<br />

ją jednomyślnie.<br />

Prof. Godlewski, przechodząc do następnego punktu<br />

Obrad, poprosił o zgłaszanie kandydatów do pełnienia<br />

funkcji Przewodniczącego-Elekta PTP. Prof. Ciszewski zaproponował<br />

dr hab. Leszka Markowskiego, który wyraził<br />

zgodę na kandydowanie. Nie zgłoszono więcej kandydatur.<br />

Walne Zebranie w głosowaniu tajnym dokonało wyboru<br />

nowego Przewodniczącego-Elekta PTP. Przewodniczący<br />

Walnego Zebrania stwierdził, że zgodnie ze statutem PTP<br />

§20 na Przewodniczącego-Elekta został wybrany dr hab.<br />

L. Markowski.<br />

Kolejne wybory dotyczyły Przewodniczącego Komisji<br />

Rewizyjnej Towarzystwa. Zgłoszono jednego kandydata<br />

na to stanowisko – dr M. Kozłowskiego, który wyraził zgodę<br />

na kandydowanie. Przy jednym głosie wstrzymującym<br />

się dr Kozłowski został wybrany i pełni funkcję Przewodniczącego<br />

Komisji Rewizyjnej.<br />

Walne Zebranie zobowiązane jest do ustalenia ilości<br />

członków Zarządu PTP. W związku z tym wywiązała się<br />

dyskusja ponownie dotycząca Sekcji Nanostruktur i Nanotechnologii.<br />

Biorąc pod uwagę, że przewodniczący<br />

sekcji wchodzą automatycznie w skład Zarządu, w celu<br />

uniknięcia powiększania Zarządu wakatami, prof. A. Hałas<br />

zaproponował anulowanie Uchwały Nr 3 z dnia 12<br />

września 2003 w sprawie utworzenia Sekcji Nanostruktur<br />

i Nanotechnologii. Zebrani jednogłośnie podjęli taką<br />

decyzję. Zaproponowano, aby Zarząd nowej kadencji<br />

składał się z 12 osób. W związku z tym została poddana<br />

pod głosowanie Uchwała nr 5 w sprawie ustalenia liczby<br />

członków Zarządu ustalająca liczbę członków Zarządu na<br />

12 osób. Walne Zebranie w głosowaniu jawnym podjęło<br />

ją jednomyślnie.<br />

Zgodnie ze Statutem §26, §31 w skład Zarządu wchodzą<br />

wcześniej wybrani:<br />

– Przewodniczący prof. J. Szuber (Walne Zebranie PTP<br />

26.04.2007 r.)<br />

– Przewodniczący-Elekt dr hab. L. Markowski (wybrany<br />

24.09.<strong>2010</strong> r.)<br />

– Przewodniczący poprzedniej kadencji prof. St. Hałas<br />

– Przewodniczący Sekcji:<br />

dr R. Mania – Sekcja Plazmowej Inżynierii Powierzchni<br />

dr hab. Z. Stępień – Sekcja Nauki Powierzchni<br />

dr hab. E. Czerwosz – Sekcja Techniki Próżni<br />

dr hab. inż. R. Paszkiewicz – Sekcja Cienkich Warstw<br />

Walne Zebranie w głosowaniu tajnym wybrało pozostałych<br />

5 członków Zarządu PTP, którymi zostali zgodnie Protokołem<br />

Komisji Skrutacyjnej:<br />

1. prof. M. Tłaczała (WEMiF PWr) – 30 głosów<br />

2. dr hab. inż. R. Korbutowicz (WEMiF PWr) – 29 głosów<br />

3. dr inż. K. Olszewska (ITR) – 30 głosów<br />

4. A. Zawada (ITR) – 29 głosów<br />

5. dr K. Marszałek (AGH) – 28 głosów<br />

W głosowaniu wzięło udział 30. Członków Towarzystwa.<br />

Przewodniczący Walnego Zebrania stwierdził, że zgodnie<br />

z §20 statutu PTP wymienieni członkowie Towarzystwa<br />

zostali wybrani do Zarządu PTP.<br />

Następnie Walne Zebranie ustaliło, zgodnie z §36 statutu,<br />

podejmując jednomyślnie w głosowaniu jawnym Uchwałę<br />

nr 6, że Komisja Rewizyjna będzie liczyć 3 osoby.<br />

Walne Zebranie w głosowaniu tajnym wybrało pozostałych<br />

2 członków Komisji Rewizyjnej, którymi zostali zgodnie<br />

Protokołem Komisji Skrutacyjnej:<br />

1. Cz. Kiliszek – 29 głosów<br />

2. J. Marks – 30 głosów<br />

W głosowaniu wzięło udział 30. Członków Towarzystwa.<br />

Uwzględniając smutne okoliczności z ubiegłego roku,<br />

prof. A. Hałas zgłosił wniosek, aby Zarząd na następne<br />

walne zebranie przygotował zmiany w statucie PTP określające<br />

jednoznacznie sposób postępowania na wypadek<br />

śmierci Przewodniczącego lub innych przypadkach<br />

losowych uniemożliwiających mu dalsze wypełnianie<br />

obowiązków.<br />

Na podstawie problemów przedstawionych w wygłoszonych<br />

sprawozdaniach i planach działania Towarzystwa<br />

w przyszłości oraz poruszonych w dyskusji i zgłoszonych<br />

w wolnych wnioskach, Komisja Wnioskowa Zebrania sformułowała<br />

wnioski z Walnego Zebrania, W głosowaniu jawnym<br />

sprawozdanie Komisji Wnioskowej zostało przyjęte<br />

jednomyślnie.<br />

O godz. 21 50 Walne Zebranie zostało zamknięte.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 115


Dr inż. Mateusz Wośko – Laureat Konkursu<br />

im. J. Groszkowskiego<br />

Wzorem lat ubiegłych ogłoszono i rozstrzygnięto<br />

Konkurs na najlepszą pracę doktorską i magisterską<br />

z dziedzin interesujących Polskie Towarzystwo Próżniowe<br />

(www.ptp.pwr.wroc.pl). Do konkursu w 2009<br />

roku zgłoszono tylko jedną pracę – pracę doktorską.<br />

Przedstawiamy protokół komisji konkursowej oraz<br />

krótki opis nagrodzonej pracy.<br />

Wrocław 9.09.<strong>2010</strong> r.<br />

Protokół z posiedzenia komisji konkursowej<br />

Polskiego Towarzystwa Próżniowego o nagrodę<br />

im. Janusza Groszkowskiego<br />

Komisja Konkursowa w składzie:<br />

Prof. dr hab. inż. Marek Tłaczała<br />

Wydział Elektroniki Mikrosystemów i Fotoniki,<br />

Politechnika Wrocławska<br />

Prof. dr hab. Jacek Szuber<br />

<strong>Instytut</strong> Elektroniki Politechnika Śląska w Gliwicach<br />

<strong>Instytut</strong> Tele- i Radiotechniczny, Warszawa<br />

Doc. dr hab. inż. Elżbieta Czerwosz<br />

Prof. dr hab. Zdzisław Stępień<br />

<strong>Instytut</strong> Fizyki, Akademia im. Jana Długosza, Częstochowa<br />

po zapoznaniu się z pracą przedstawioną do konkursu<br />

oraz recenzją zatwierdzonego recenzenta postanowiła<br />

przyznać w kategorii prac doktorskich dr inż. Mateuszowi<br />

Wośko I Nagrodę im. JANUSZA GROSZKOWSKIEGO za<br />

pracę „Opracowanie konstrukcji i technologii fotodetektorów<br />

z zastosowaniem nanostruktur półprzewodników<br />

AIIIBV o ciągłej zmianie składu” wykonaną na<br />

Wydziale Elektroniki Mikrosystemów i Fotoniki Politechniki<br />

Wrocławskiej. Promotorem rozprawy jest dr hab. inż. Regina<br />

Paszkiewicz, prof. PWr.<br />

Zaprezentowane w pracy rezultaty stanowią wkład<br />

w badania nad funkcjonalnymi materiałami gradientowymi,<br />

w szczególności nad półprzewodnikowymi strukturami<br />

fotoczułymi z warstwami o ciągłej zmianie składu. Celem<br />

pracy było przedstawienie problematyki związanej z projektowaniem,<br />

wytwarzaniem i charakteryzacją fotodetektorów<br />

na bazie półprzewodnikowych struktur gradientowych.<br />

W pracy zawarto wyniki analizy teoretycznej i badań doświadczalnych<br />

nad półprzewodnikowymi materiałami gradientowymi<br />

AIIIBV, na przykładzie układu materiałowego<br />

GaAs/Al x<br />

Ga 1-x<br />

As, oraz ich wykorzystaniem do wytwarzania<br />

fotodetektorów, w celu kształtowania ich charakterystyk<br />

spektralnych. Rezultatem przeprowadzonych badań<br />

było opracowanie konstrukcji fotodetektorów z obszarami<br />

czynnymi z ciągłą zmianie składu, w szczególności analiza<br />

numeryczna wpływu warstwy gradientowej GaAs/Al x<br />

Ga 1-<br />

As na parametry heterozłączowego fotodetektora; opracowanie<br />

metodologii wytwarzania epitaksjalnych struktur<br />

x<br />

gradientowych GaAs/Al x<br />

Ga 1-x<br />

As, w której kluczowym elementem<br />

jest uwzględnienie dynamiki procesu krystalizacji<br />

i interakcji między poszczególnymi fazami w reaktorze;<br />

charakteryzacja warstw o ciągłej zmianie składu oraz wytworzenie<br />

struktur przyrządowych fotodetektorów.<br />

Wzorem lat ubiegłych, Laureat poza nagroda pieniężną,<br />

otrzymał możliwość wygłoszenia wykładu podczas<br />

konferencji. Tym razem była to konferencja „Technologia<br />

Elektronowa” ELTE<strong>2010</strong>, organizowana przez Wydział<br />

Elektroniki Mikrosystemów i Fotoniki Politechniki Wrocławskiej.<br />

Wykład poświęcony był prezentacji tez zawartych<br />

w nagrodzonej rozprawie doktorskiej. Laureat przedstawił<br />

główne wyniki prac badawczych przeprowadzonych w ramach<br />

realizacji pracy. Szerzej omówił problemy związane<br />

z krystalizacją warstw o ciągłej zmianie składu metodą<br />

epitaksji ze związków metaloorganicznych oraz przedstawił<br />

założenia oraz wyniki opracowanej metody określania<br />

składu warstw gradientowych na podstawie widm odbicia<br />

światła.<br />

za komisję konkursową<br />

Prof. dr hab. inż. Marek Tłaczała<br />

Nota o laureacie:<br />

Mateusz Wośko, ur. 25.10.1980 r. w Bolesławcu. W 2004 r. obronił z wyróżnieniem pracę<br />

magisterską „Opracowanie technologii osadzania warstw GaN na podłożach krzemowych”.<br />

W latach 2004–2009 uczestnik Studium Doktoranckiego na Wydziale Elektroniki Mikrosystemów<br />

i Fotoniki PWr. Pracę doktorską „Opracowanie konstrukcji i technologii fotodetektorów<br />

z zastosowaniem nanostruktur półprzewodników AIIIBV o ciągłej zmianie składu”<br />

obronił z wyróżnieniem. Laureat m.in. konkursu o Nagrodę Naukową Polskiego Towarzystwa<br />

Wzrostu Kryształów.<br />

116<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Spektrometr Elektronowego Rezonansu<br />

Paramagnetycznego (EPR), do pomiaru liczby<br />

spinów w badanym materiale<br />

dr JAN DUCHIEWICZ, dr ANDRZEJ DOBRUCKI, dr hab. ANDRZEJ FRANCIK<br />

Politechnika Wrocławska, <strong>Instytut</strong> Telekomunikacji, Teleinformatyki i Akustyki<br />

dr WACŁAW STACHOWICZ, <strong>Instytut</strong> Chemii i Techniki Jądrowej, Warszawa<br />

inż. TADEUSZ OLEŚ, Uniwersytet Jagielloński, Wydział Biochemii, Biofizyki i Biotechnologii, Kraków<br />

mgr TOMASZ DUCHIEWICZ, ELOKON Polska, Warszawa<br />

Spektrometria Elektronowego Rezonansu Paramagnetycznego<br />

(EPR) należy do podstawowych metod badania materii.<br />

Wyróżnić można dwa kierunki prowadzenia pomiarów tą metodą:<br />

• Rejestracja sygnałów EPR, umożliwiająca wykrycie<br />

w badanej substancji lub w mieszaninie zawierającej różne<br />

substancje, związków charakteryzujących się obecnością<br />

niesparowanych elektronów (spinów) w cząsteczce, takich<br />

jak centra paramagnetyczne, wolne rodniki oraz jony paramagnetyczne<br />

metali. Każdemu z tych związków można<br />

przypisać charakterystyczne widmo EPR. Analiza kształtu<br />

zarejestrowanego widma umożliwia stwierdzenie obecności<br />

danego związku paramagnetycznego w badanej próbce<br />

oraz w niektórych przypadkach także określenie jego<br />

budowy chemicznej.<br />

• Rejestracja widma sygnału EPR, umożliwiająca nie tylko<br />

wykrycie obecności w badanej substancji składników zawierających<br />

niesparowane elektrony, ale również ilościowe<br />

określenie ich zawartości wyrażone liczbą spinów. Jest<br />

to niezbędne np. w badaniach dozymetrycznych oraz do<br />

określania wydajności rodnikowych procesów radiacyjnych<br />

i fotochemicznych.<br />

Ilościowe pomiary liczby spinów są zatem niezbędne w różnych<br />

dziedzinach nauki i techniki, ponieważ pozwalają określić<br />

nie tylko sam fakt obecności danego związku paramagnetycznego<br />

w badanej próbce, ale również jego procentową<br />

zawartość. W badaniach często zachodzi potrzeba określenia<br />

ich składu ilościowego, czyli wprost określenia liczby spinów,<br />

produktów paramagnetycznych biorących udział w badanym<br />

procesie. Potrzeba takich pomiarów ilościowych występuje<br />

np. w pracach związanych z ochroną środowiska (większość<br />

procesów degradacyjnych jest związana z obecnością wolnych<br />

rodników), w gospodarce żywnościowej (wolne rodniki<br />

w żywności powstają np. wskutek działania promieniowania<br />

jonizującego), w przemyśle chemicznym (produkty pośrednie),<br />

paliwowym (surowa ropa naftowa jest najczęściej zanieczyszczona<br />

związkami wanadu z porfirynami oraz związkami<br />

węgla, doskonale wykrywalnymi metodą EPR).<br />

Pomiary ilościowe są, jak wspomniano, absolutnie niezbędne<br />

w radiacyjnej dozymetrii EPR, ponieważ umożliwiają<br />

określenie dawki promieniowania za pomocą kalibrowanego<br />

dozymetru np. alaninowego dołączanego do dowolnego napromieniowanego<br />

materiału, jak również oszacowanie dawki<br />

promieniowania, jaką zastosowano podczas radiacyjnej pasteryzacji<br />

danego produktu żywnościowego. Jest to szczególnie<br />

istotne z uwagi na ustalenia Kodeksu Żywnościowego<br />

WHO/FAO dopuszczającego napromieniowanie żywności jedynie<br />

w określonych granicach dawek określonych dla każdej<br />

grupy produktów spożywczych, nie przekraczających 10 kGy.<br />

Teoretycznie, pomiary ilościowe mogą być dokonywane<br />

niemal na każdym spektrometrem EPR. Jednak w praktyce<br />

jest to bardzo trudne, a często wręcz niemożliwe, ponieważ<br />

wymaga stosowania odpowiednich, nie zawsze dających<br />

się odczytać w widmie EPR wzorców odniesienia. Ponadto,<br />

związek pomiędzy poziomem zarejestrowanego sygnału EPR<br />

a liczbą spinów w próbce zależy od wielu, często trudnych do<br />

określenia czynników. Poniżej przedstawiono trzy wpływające<br />

na ilościowe pomiary liczby spinów czynniki, przy czym dwa<br />

pierwsze są bardzo trudne do określenia:<br />

1. Wartość składowej magnetycznej mikrofalowego pola<br />

elektromagnetycznego wewnątrz badanej próbki. Wartość<br />

ta zależy od właściwości samej próbki, od poziomu mocy<br />

mikrofalowej oraz od konstrukcji i dobroci rezonatora pomiarowego;<br />

2. Stopień sprzężenia badanej próbki z rezonatorem. Zależy<br />

on, między innymi od stosunku objętości samej próbki do<br />

objętości rezonatora;<br />

3. Wzmocnienie skuteczne toru odbiorczego oraz poziom<br />

modulacji pomocniczej – parametry te są stosunkowo najłatwiejsze<br />

do określenia.<br />

W jaki sposób są więc dokonywane<br />

pomiary ilościowe w praktyce?<br />

Ponieważ bezpośrednie określenie liczby spinów jest bardzo<br />

trudne, najczęściej do pomiarów ilościowych jest stosowana<br />

metoda porównawcza, polegająca na porównaniu zarejestrowanego<br />

sygnału EPR badanej próbki z zarejestrowanym sygnałem<br />

EPR próbki wzorcowej o znanej liczbie spinów. Metoda<br />

porównawcza może być realizowana na kilka sposobów:<br />

1) W możliwie tych samych warunkach są dokonywane kolejno<br />

dwie, niezależne rejestracje sygnału EPR: dla próbki<br />

wzorcowej oraz dla próbki badanej. Jest to metoda najprostsza,<br />

ale najmniej pewna – w praktyce jest niezwykle<br />

trudno zapewnić takie same warunki w obydwu pomiarach,<br />

szczególnie wtedy, gdy różnice gęstości fizycznej i elektronowej<br />

oraz liczby spinów pomiędzy próbką wzorcową<br />

i badaną są znaczne.<br />

2) Jest dokonywana jednoczesna rejestracja sygnału EPR<br />

próbki badanej i wzorcowej, umieszczonej wewnątrz<br />

próbki badanej. Jest to metoda znacznie pewniejsza od<br />

poprzedniej – oba sygnały są rejestrowane w tych samych<br />

warunkach. Jednak w praktyce jest na ogół trudno<br />

wprowadzić próbkę wzorcową do próbki badanej oraz, co<br />

ważniejsze, występuje bardzo często nakładanie się widm<br />

próbki i wzorca uniemożliwiające praktycznie pomiar.<br />

3) Jest stosowany rezonator podwójny, zawierający dwie<br />

identyczne wnęki pomiarowe o możliwie podobnym rozkła-<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 117


dzie mikrofalowego pola elektromagnetycznego. W jednej<br />

z nich jest umieszczana próbka badana, w drugiej wzorcowa.<br />

Zapewnia to badanie obu próbek w niemal takich samych<br />

warunkach. Rezonator podwójny jest konstruowany<br />

np. poprzez odpowiednie złożenie dwóch identycznych rezonatorów<br />

prostokątnych o modzie TE102. Powstaje wówczas<br />

podwójny rezonator prostokątny o modzie TE104.<br />

Rezonator podwójny dla pasma X oferuje np. firma Bruker<br />

(model ER 4105DR) z informacją, że nadaje się on szczególnie<br />

do pomiarów ilościowych. Zastosowanie rezonatora<br />

podwójnego zapewnia większą dokładność określenia<br />

liczby spinów w porównaniu z dwoma, powyżej opisanymi<br />

sposobami pod warunkiem, że oba sygnały próbki i wzorca<br />

zawierają zbliżone liczby spinów. Jest tak dlatego, ponieważ<br />

oba sygnały EPR są rejestrowane jednocześnie tym<br />

samym torem odbiorczym, więc aby uniknąć maskowania<br />

sygnału słabego przez sygnał silny, poziomy obu sygnałów<br />

muszą do siebie zbliżone. Należy więc zapewnić, aby liczby<br />

spinów w obu próbkach były podobne. Jest to najczęściej<br />

niemożliwe i staje się konieczne stosowanie wielu pośrednich<br />

próbek wzorcowych lub zestawu próbek znacznie<br />

różniących się liczbą spinów. W rezultacie dokładność pomiaru<br />

spada i nie jest lepsza jak w przypadku stosowania<br />

wzorca zewnętrznego (p.1).<br />

4) Jest stosowany rezonator podwójny (podobnie jak w wyżej<br />

opisanej metodzie) oraz dwa niezależne tory odbiorcze sygnału<br />

EPR: jeden dla sygnału próbki badanej, drugi dla sygnału<br />

próbki wzorcowej. W celu uniknięcia wzajemnych interferencji,<br />

częstotliwości pracy obu torów powinny znacznie się<br />

różnić pomiędzy sobą; optymalnym wydaje się tutaj wybór<br />

częstotliwości 100 kHz dla toru próbki badanej oraz 1 kHz<br />

dla toru próbki wzorcowej. Ponieważ parametry (amplituda<br />

modulacji pomocniczej, wzmocnienie skuteczne, szerokość<br />

pasma) obu torów odbiorczych mogą być dobierane niezależnie,<br />

więc intensywności (amplitudy) sygnałów EPR obu<br />

próbek mogą się różnić między sobą nawet o kilka rzędów<br />

wielkości. Przy takiej konfiguracji rezonatora jedna próbka<br />

wzorcowa może być stosowana do pomiarów wielu próbek<br />

o znacznie różniących się liczbach spinów. Jest oczywiście<br />

konieczny specjalny system komputerowy, umożliwiający<br />

jednoczesną rejestrację sygnałów z obu torów.<br />

Dwukanałowy spektrometr EPR<br />

Idea spektrometru dwukanałowego jest znana i opisana w podręcznikach<br />

dotyczących zagadnień technicznych spektrometrii<br />

EPR [1, 2] – jest jednak bardzo rzadko wykorzystywana<br />

w praktyce. Przyczyną tego jest fakt, że system taki składa się<br />

niemal z dwóch spektrometrów, współpracujących jednocześnie<br />

z jednym, podwójnym rezonatorem. Według posiadanych<br />

informacji, żadna z firm komercyjnych nie oferuje dziś takiego<br />

spektrometru. W roku 1982 [3, 4] przebudowano spektrometr<br />

EPR na pasmo X firmy Varian: wyposażono go w dodatkowy tor<br />

odbiorczy o małej częstotliwości (od spektrometru na pasmo<br />

Q) na dwukanałowy spektrometr, umożliwiający jednoczesną<br />

rejestrację sygnału dwóch próbek w paśmie X. Jako rezonator<br />

podwójny zastosowano dwa rezonatory prostokątne o modzie<br />

TE102, złączone ze sobą węższymi ściankami, tworząc w ten<br />

sposób rezonator 2-komorowy o modzie TE104. Rozwiązanie<br />

takie umożliwiło przeprowadzanie pomiarów ilościowych oraz<br />

pomiary współczynnika g z dużą dokładnością – nie osiągalną<br />

w spektrometrze klasycznym. Uproszczony schemat blokowy<br />

spektrometru dwukanałowego jest przedstawiony na rys. 1.<br />

Rezonator podwójny<br />

Najistotniejszym elementem dwukanałowego spektrometru<br />

EPR jest rezonator podwójny, zapewniający możliwie podobne<br />

warunki rejestracji sygnałów obu próbek. Wyróżnić tutaj<br />

można dwa rozwiązania:<br />

a. rezonator podwójny, składający się z dwóch rezonatorów<br />

tworzących w sumie jeden rezonator 2-wnękowy,<br />

b. dwa niezależne, identyczne rezonatory, włączone w taki<br />

sposób, aby nawzajem na siebie nie wpływały, natomiast<br />

aby ich sygnały wyjściowe się dodawały.<br />

Rys. 1. Schemat blokowy dwukanałowego spektrometru EPR Fig. 1. Block diagram of the two-channel EPR spectrometer<br />

118<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Rezonator prostokątny 2-wnękowy<br />

Rezonator 2-wnękowy prostokątny najprościej można utworzyć<br />

przez odpowiednie złożenie dwóch identycznych rezonatorów<br />

prostokątnych o modzie TE102 – zostanie wówczas utworzony<br />

rezonator TE104, posiadający dwie, ściśle sprzężone ze sobą<br />

wnęki. Przykład takiego rezonatora na pasmo X jest przedstawiony<br />

na rys. 2. Ograniczeniem są tutaj rozmiary rezonatora prostokątnego,<br />

które są odwrotnie proporcjonalne do częstotliwości<br />

pracy generatora mikrofal. Stąd też opisane rozwiązanie jest<br />

w zasadzie stosowane tylko dla pasma X oraz Q. Schemat poglądowy<br />

dwukanałowego spektrometru na pasmo X jest przedstawiony<br />

na rys. 3. Zastosowano w nim 2-wnękowy rezonator<br />

przedstawiony na rys. 2.<br />

Rezonator 2-wnękowy typu Loop-Gap<br />

Klasyczna konstrukcja rezonatora typu Loop-Gap, posiadającego<br />

dwie wnęki sprzężone ze sobą szczeliną jest przedstawiona<br />

na rys. 4 [5].<br />

W ramach projektu rozwojowego NCBiR nr N R01 0018 04<br />

pt. „Prototyp spektrometru EPR do badań dozymetrycznych<br />

i identyfikacji napromieniowanej żywności” opracowano<br />

i wykonano kompletny spektrometr EPR na pasmo<br />

L, w którym zastosowano rezonator typu Loop-Gap, którego<br />

widok jest przedstawiony na rys. 5 [6–10].<br />

Na rysunku 6 jest przedstawiony widok rezonatora typu<br />

Loop-Gap przewidzianego do zastosowania w dwukanałowym<br />

spektrometrze na pasmo L.<br />

Ze względu na stosunkowo małą odległość pomiędzy obiema<br />

wnękami pomiarowymi rezonatora, średnica cewek modulacyjnych<br />

powinna być na tyle mała, aby wytwarzane przez<br />

nie pola magnetyczne nie wpływały wzajemnie na siebie.<br />

Dwa rezonatory niezależne<br />

Rys. 2. Podwójny rezonator 2-wnękowy o modzie TE104 na pasmo X<br />

Fig. 2. X-Band double resonator with mode TE104<br />

Rezonator podwójny można również utworzyć z dwóch identycznych<br />

rezonatorów włączonych do systemu tak, aby nawzajem<br />

na siebie nie oddziaływały, natomiast aby sygnały<br />

pochodzące od nich dodawały się. Przykładowy sposób podłączenia<br />

dwóch rezonatorów w bloku mikrofalowym przedstawiono<br />

na rys. 7.<br />

Rys. 3. Schemat poglądowy dwukanałowego spektrometru EPR na pasmo X z rezonatorem o modzie TE104<br />

Fig. 3. Illustrative schema of the two-channel X-Band EPR spectrometer with the TE104 mode double resonator<br />

Rys. 4. Rezonator typu Loop-Gap z dwiema wnękami, sprzężonymi szczeliną: A- konstrukcja stosowana dla niższych pasm (UHF,<br />

L, S), B – konstrukcja stosowana dla wyższych pasm (X, Q). Oznaczenia: h, r – odpowiednio długość oraz promień pojedynczej wnęki,<br />

W, t – odpowiednio szerokość oraz długość szczeliny<br />

Fig. 4. Loop-Gap resonator with two cylindrical cavity coupled with a gap.: A- construction used for lower frequency bands (UHF,<br />

L, S), B – construction used for higher frequency bands (X, Q). where h, r – length and radius of the cavity respectively, W, t – width<br />

and length of the gap respectively<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 119


Rys. 5. Rezonator typu Loop-Gap zastosowany w opracowanym<br />

spektrometrze EPR na pasm L<br />

Fig. 5. Loop-Gap resonator used in the designed L-Band EPR<br />

spectrometer<br />

Rys. 6. Rezonator typu Loop-Gap do dwukanałowego spektrometru<br />

EPR na pasm L<br />

Fig. 6. Loop-Gap resonator for two-channel L-Band EPR spectrometer<br />

Rys. 7. Sposób włączenia dwóch rezonatorów<br />

Fig. 7. Manner of proper connection of two resonators in microwave<br />

unit<br />

Program sterujący<br />

Opracowano specjalny program komputerowy EPR System<br />

do sterowania poszczególnymi blokami spektrometru (stabilizator<br />

pola magnetycznego, odbiorniki cyfrowe 1 i 100 kHz<br />

oraz blok mikrofalowy) poprzez interfejs USB, zapewniający<br />

jednoczesną rejestrację obu sygnałów EPR. Dodatkowo,<br />

program umożliwia zawansowaną obróbkę cyfrową zarejestrowanych<br />

sygnałów (wygładzanie, sumowanie, całkowanie,<br />

wzajemne porównywanie, różniczkowanie itp.). Program<br />

może też współpracować z innymi, dodatkowymi przyrządami<br />

pomiarowymi (sterowanymi w standardzie USB lub RS232),<br />

którymi mogą być np. jądrowy miernik pola magnetycznego,<br />

miernik częstotliwości i mocy mikrofalowej oraz regulator temperatury<br />

badanej próbki i inne. Okno główne programu EPR<br />

System przedstawiono na rysunku 8. Z uwagi na zastosowanie<br />

interfejsu USB, program EPR System może być zainstalowany<br />

zarówno w komputerze stacjonarnym (desktop) jak<br />

w komputerze przenośnym (laptop).<br />

Rys. 8. Okno główne programu EPR System. Fig. 8. Main window of the program EPR System<br />

120<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Rys. 9. Dwukanałowe spektrometry EPR na pasmo X na Akademii Medycznej w Bydgoszczy oraz na Uniwersytecie w Białymstoku<br />

Fig. 9. X-Band two-channel EPR spectrometers in the Medical Academy in Bydgoszcz and in the University in Bialystok<br />

Przykłady realizacji<br />

Jak już wcześniej zaznaczono, żadna firma komercyjna nie<br />

oferuje dwukanałowego spektrometru EPR, nadającego się<br />

do względnych pomiarów intensywności sygnałów EPR.<br />

Spektrometr taki można np. uzyskać w ramach modernizacji<br />

starszego modelu spektrometru. W ostatnich latach grupa<br />

pracowników Politechniki Wrocławskiej zbudowała trzy dwukanałowe<br />

spektrometry EPR na pasmo X, które zostały przedstawione<br />

poniżej.<br />

a. Akademia Medyczna w Bydgoszczy oraz Uniwersytet<br />

w Białymstoku.<br />

Dwukanałowe spektrometry EPR na pasmo X zestawiono na<br />

bazie spektrometrów EPR wykonanych niegdyś przez nieistniejącą<br />

już firmę Radioman w Poznaniu.<br />

b. <strong>Instytut</strong> Fizyki Polskiej Akademii Nauk w Warszawie<br />

Dwukanałowy spektrometr EPR na pasmo X zestawiono na<br />

bazie spektrometru EPR, wykonanego przez Politechnikę<br />

Wrocławską.<br />

Podsumowanie<br />

Z przedstawionych informacji i przykładów konkretnych rozwiązań<br />

wynika, że przy stosunkowo niewielkim nakładzie<br />

kosztów można uzyskać zupełnie nowe narzędzie badawcze,<br />

dotychczas nie oferowane przez światowy przemysł<br />

aparaturowy. Możliwość dokładnych pomiarów intensywności<br />

sygnałów EPR ma duże znaczenie niemal we wszystkich<br />

dziedzinach nauki i techniki, w których wykorzystuje się spektrometrię<br />

EPR.<br />

Do budowy takiego spektrometru szczególnie nadaje się<br />

opracowany przez nasz zespół spektrometr na pasmo L.<br />

Uzupełniając ten spektrometr o dodatkowy, cyfrowy odbiornik<br />

sygnału EPR częstotliwości 1 kHz oraz nieznacznie zmieniając<br />

konstrukcję zastosowanego rezonatora Loop-Gap, uzyskuje<br />

się dwukanałowy spektrometr EPR na pasmo L nowej<br />

generacji.<br />

Spektrometr taki nadaje się szczególnie do pomiarów<br />

dozymetrycznych, m.in. napromieniowanej żywności różnego<br />

rodzaju. Ponieważ tłumienie pola elektromagnetycznego<br />

wnoszone przez wodę w paśmie L jest znacznie mniejsze niż<br />

w przypadku wyższych pasm częstotliwości (X, K, Q), więc<br />

spektrometr na pasmo L jest szczególnie przydatny w badaniach<br />

substancji o dużej zawartości wody (tj. próbek z natury<br />

mokrych), co jest typową cechą większości żywności oraz<br />

materiałów biologicznych. W przeciwieństwie do pasma L,<br />

próbki lite o dużej zawartości wody są trudno mierzalne lub<br />

też w ogóle niemierzalne w wyższych pasmach częstotliwości<br />

(X, K, Q).<br />

Można przyjąć, że spektrometr taki znalazłby zastosowanie<br />

nie tylko w stacjach sanitarno-epidemiologicznych, lecz<br />

również w jednostkach badawczych i firmach zajmujących<br />

się zarówno badaniem jak też przetwarzaniem i dystrybucją<br />

materiałów biologicznych oraz wszelkiego rodzaju żywności<br />

krajowej, jak też importowanej.<br />

Literatura<br />

Rys. 10. Dwukanałowy spektrometr EPR na pasmo X w Instytucie<br />

Fizyki Polskiej Akademii Nauk w Warszawie<br />

Fig. 10. X-Band two-channel EPR spectrometer in the Institute<br />

of Physics of the Polish Academy of Science in Warszawa<br />

[1] Poole Jr. C. P., Charles P.: Electron Spin Resonance. Interscience<br />

Publication, Y. York, London, Sydney 1967.<br />

[2] Wertz J. E., Bolton J. R.: Electron Spin Resonance. Elementary Theory<br />

and Practical Applications. McGraw Hill Co., New York 1972.<br />

[3] Lange J.P., Gutsze A., Karge H.G.: Coke Formation through the<br />

Reaction of Olefins over Hydrogen Modernite. Journal of Catalysis,<br />

114, 136 (1988).<br />

[4] Witzel F., Karge H.G., Gutsze A.: ESR Measurements for the<br />

Characterization of Acidic Lewis Sites in Zeolites. Proc of 9th<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 121


International Zeolite Conference, Montreal 1992, Bulterworth-<br />

Heinemann 1993, 283.<br />

[5] Ferenc S., Ferenc M. ESR spectrometer with a loop-gap resonator<br />

for CW and time resolved studies in a superconducting magnet,<br />

Journal of Magnetic Resonance 173 (2005).<br />

[6] Duchiewicz J., Dobrucki A., Francik A., Duchiewicz T., Sadowski<br />

A., Idźkowski B., Kutynia A.: Cyfrowy odbiornik z układem FPGA<br />

do spektrometru EPR. <strong>Elektronika</strong> – Konstrukcje, Technologie,<br />

Zastosowania, nr 3/<strong>2010</strong>.<br />

[7] Duchiewicz J., Dobrucki A., Francik A., Duchiewicz T., Sadowski<br />

A., Idźkowski B., Kutynia A,, Błaszczyk J.: Źródło pola magnetycznego<br />

do spektrometru Elektronowego Rezonansu Paramagnetycznego<br />

(EPR) na pasmo L. <strong>Elektronika</strong> – Konstrukcje, Technologie,<br />

Zastosowania, nr 4/<strong>2010</strong><br />

[8] Duchiewicz J., Dobrucki A., Francik A., Duchiewicz T., Sadowski<br />

A., Idźkowski B., Kutynia A,, Błaszczyk J.: Moduł mikrofalowy do<br />

Spektrometru Elektronowego Rezonansu Paramagnetycznego<br />

(EPR) na pasmo L. <strong>Elektronika</strong> – Konstrukcje, Technologie, Zastosowania,<br />

nr 5/<strong>2010</strong>.<br />

[9] Duchiewicz J., Dobrucki A., Francik A., Duchiewicz T., Sadowski<br />

A., Idźkowski B., Kutynia A,, Błaszczyk J.: Spektrometr<br />

Elektronowego Rezonansu Paramagnetycznego (EPR) na pasmo<br />

L. <strong>Elektronika</strong> – Konstrukcje, Technologie, Zastosowania,<br />

nr 6/<strong>2010</strong>.<br />

[10] Duchiewicz J., Dobrucki A., Francik A., Duchiewicz T., Sadowski<br />

A., Idźkowski B., Kutynia A., Błaszczyk J.: L-Band Electron Paramagnetic<br />

Resonance (EPR) Spectrometer. 4th Microwave and<br />

Radar Week MIKON <strong>2010</strong>, 14–18 June, <strong>2010</strong> in Vilnius (Litwa).<br />

Moduły lokalizacji oraz testy chipów GNSS<br />

dla elementów systemu PROTEUS<br />

mgr inż. RADOSLAV DARAKCHIEV, mgr inż. PIOTR SITEK, mgr inż. KAROL BRZOSTOWSKI,<br />

mgr inż. ANNA FOKS-RYZNAR, dr inż. MACIEJ KALARUS, mgr inż. RYSZARD ZDUNEK<br />

Centrum Badań Kosmicznych PAN, Warszawa<br />

Celem projektu PROTEUS – „Zintegrowanego mobilnego<br />

systemu wspomagającego działania antyterrorystyczne i antykryzysowe”<br />

– jest wykorzystanie innowacyjnych technologii<br />

w stworzeniu demonstratora nowoczesnego mobilnego systemu<br />

dla Straży Pożarnej, antyterrorystów, Policji oraz Centrów<br />

Zarządzania Kryzysowego.<br />

Jednym z elementów systemu są roboty mobilne, których<br />

zadaniem będzie wykonywanie czynności niebezpiecznych<br />

dla ludzi. Roboty będą w stanie pracować w ekstremalnych<br />

warunkach środowiskowych, dlatego będą mogły poruszać<br />

się w rejonach niebezpiecznych i zbierać próbki gleby, cieczy<br />

i powietrza. Dzięki wyposażeniu w kamery wizyjne, kamery<br />

termowizyjne oraz inne czujniki, będą umożliwiać podgląd<br />

miejsca akcji.<br />

Całość działań będzie kierowana z Mobilnego Centrum<br />

Dowodzenia. Będzie ono wyposażone w sieć komputerową,<br />

poprzez którą będą dostępne m.in. aktualne dane o pogodzie,<br />

mapy cyfrowe, zdjęcia satelitarne zagrożonego rejonu,<br />

co umożliwi bardziej efektywne dowodzenia akcją ratowniczą.<br />

Będą w nim przetwarzane oraz wizualizowane dane ze<br />

wszystkich elementów systemu. Narzędzia multimedialne<br />

oraz mnogość danych o sytuacji zapewnią bardzo dobre warunki<br />

do dowodzenia akcją ratowniczą.<br />

Bezzałogowy Statek Latający będzie użyty do obserwacji<br />

strefy objętej sytuacją kryzysową. Będzie w stanie pozostawać<br />

w powietrzu przez minimum 8 godzin. Będzie mógł być<br />

kierowany przez osobę nie posiadającą umiejętności pilotażu,<br />

w oparciu jedynie o zadane współrzędne geograficzne na<br />

mapie. Kamera wizyjna, kamera termowizyjna oraz ultranowoczesne<br />

czujniki płomieni będą dostarczać do Mobilnego<br />

Centrum Dowodzenia wielowarstwowe dane z zagrożonego<br />

rejonu.<br />

Dodatkową możliwością Systemu jest użycie Przenośnego<br />

Zestawu Czujników. Jego zadaniem będzie zbieranie fizycznych<br />

i chemicznych danych z miejsca, w którym zostanie<br />

umieszczony.<br />

Pełna lista elementów systemu PROTEUS jest następująca:<br />

1. Mobilne Centrum Dowodzenia<br />

2. Mobilne Centrum Operatorów Robotów<br />

122<br />

3. Mały Robot Mobilny<br />

4. Robot Mobilny o Zwiększonej Funkcjonalności<br />

5. Interwencyjny Robot Mobilny<br />

6. Przenośne stanowiska Operatora Robota<br />

7. Bezzałogowy Samolot<br />

8. Przenośny Zestaw Czujników<br />

9. Nasobne Zestawy Czujników<br />

10. Symulator Robota Mobilnego<br />

Elementy systemu wyposażone w moduły lokalizacji konstrukcji<br />

CBK PAN zostały podkreślone.<br />

Moduły lokalizacji<br />

Wymagania dla poszczególnych modułów lokalizacji zostały<br />

określone przez ich użytkowników końcowych. Potrzebują<br />

oni rozwiązań specjalistycznych, niekomercyjnych, dedykowanych<br />

dla ich zastosowań. Oprócz określonych parametrów<br />

lokalizacyjnych (dostępność i dokładność wyznaczanej pozycji)<br />

wymagana jest również łatwość obsługi modułu oraz duża<br />

odporność na warunki środowiskowe.<br />

Pojazdy, czyli Mobilne Centrum Dowodzenia oraz Mobilne<br />

Centrum Operatorów Robotów, wymagają dokładności lokalizacji,<br />

jaką zapewnia Standard Positioning Service (SPS)<br />

systemu GPS (≤9 m w płaszczyźnie horyzontalnej i ≤15 m<br />

w płaszczyźnie wertykalnej). Masa, zużycie energii oraz wielkość<br />

modułów mają mniejsze znaczenie. Jednak wymagana<br />

jest dokładna znajomość kierunku północy oraz kątów inklinacji<br />

(+/- 1 stopień) w celu ustawienia anteny dla łączności satelitarnej.<br />

Należy pamiętać, że północ geograficzna różni się od<br />

północy magnetycznej – jest to tzw. deklinacja magnetyczna.<br />

Przyczyną tej różnicy jest nie tylko różnica położenia bieguna<br />

północnego geograficznego i magnetycznego, lecz także<br />

anomalie pola magnetycznego. W Polsce różnica ta wynosi<br />

3-6 stopni. Północ geograficzna będzie wyznaczana na podstawie<br />

pomiarów z magnetometrów.<br />

Roboty mobilne wymagają najlepszej możliwej dokładności<br />

lokalizacji (do 2 m w niektórych przypadkach). Ich pozycja jest<br />

wykorzystywana w Mobilnym Centrum Dowodzenia w celu ich<br />

lokalizacji oraz wyznaczenia trasy przejazdu przez teren ak-<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Rys. 1. Schemat blokowy modułu lokalizacji<br />

Fig. 1. Block schema of the location module<br />

cji. W przypadku utraty łączności Robot Mobilny o Zwiększonej<br />

Funkcjonalności musi mieć możliwość autonomicznego<br />

powrotu do punktu wyjściowego. Roboty mobilne wymagają<br />

również wyznaczenia kierunku północy oraz kąta inklinacji<br />

z błędem mniejszym od 1 stopnia. W przypadku robotów moduły<br />

lokalizacji muszą także spełnić ściśle określone wymagania<br />

dotyczące masy oraz rozmiarów. Ograniczenia poboru<br />

mocy również występują, jednak moduły lokalizacji zużywają<br />

znacznie mniej energii niż napęd mechaniczny robotów.<br />

Zestawy Nasobne powinny zapewnić dokładność lokalizacji<br />

zgodną ze standardem SPS (Standard Positioning Service)<br />

a ich zasilanie powinno zapewnić 8 godzin nieprzerwanej pracy.<br />

Rozmiar prototypu Zestawu Nasobnego jest ograniczony<br />

do wielkości paczki papierosów (7 × 5 × 2 cm). Rzeczywiste<br />

rozmiary Zestawu Nasobnego mogą być znacznie mniejsze<br />

od wielkości zakładanej. Maksymalna jego masa bez obudowy<br />

i bez akumulatorów to 200 g, jednak masa rzeczywistego<br />

urządzenia prawdopodobnie nie przekroczy 100 g. Kolejnym<br />

wymogiem jest bezpieczeństwo – Zestaw Nasobny musi być<br />

odporny na wibracje, wysoką wilgotność oraz duże gradienty<br />

temperatury.<br />

Pojedynczy moduł lokalizacji składa się z następujących<br />

bloków:<br />

• zasilanie,<br />

• housekeeping,<br />

• blok sterowania i komunikacji,<br />

• blok lokalizacji.<br />

Schemat blokowy modułu lokalizacji pokazany jest na rys. 1.<br />

Różnicą pomiędzy Nasobnym Zestawem Czujników a modułami<br />

lokalizacji dla robotów mobilnych oraz pojazdów jest to, że<br />

Nasobny Zestaw Czujników posiada niezależny układ bateryjny,<br />

natomiast inne moduły lokalizacji są zasilane ze źródła zasilania<br />

robota lub pojazdu. Blok housekeepingu 1 może składać<br />

się z czujnika temperatury, czujnika wilgotności, czujnika wibracji<br />

oraz pomiaru napięcia zasilania. Zbiór czujników housekeepingu<br />

różni się w poszczególnych typach modułów lokalizacji.<br />

Pomiary z czujnika wilgotności pozwolą określić, kiedy należy<br />

wyłączyć urządzenie, jeśli istnieje zagrożenie zwarciem<br />

z powodu wysokiej wilgotności. Dane te mogą również umożliwić<br />

ochronę innych czujników wrażliwych na wilgotność. Wysoka<br />

wilgotność może pojawić się np. gdy robot lub ratownik<br />

wejdą w strefę pożaru gaszonego wodą. Czujnik wibracji<br />

w Nasobnym Zestawie Czujników służy do monitorowania ruchów<br />

ratownika i pozwala określić, czy pozostaje w bezruchu,<br />

czy też doświadcza wstrząsu fizycznego. Bezruch ratownika<br />

może sugerować jego utratę przytomności.<br />

1<br />

housekeeping – monitorowanie parametrów pracy urządzenia<br />

Monitorowanie zasilania w Nasobnym Zestawie Czujników<br />

pozwala na zdalny pomiar napięcia baterii oraz informowanie<br />

o możliwym ich rozładowaniu. Jeśli nie ma zagrożenia dla<br />

powodzenia operacji, ratownik może się wycofać w celu wymiany<br />

baterii. Projektowane urządzenie gwarantuje 8 godzin<br />

nieprzerwanej pracy. Jednakże taki czujnik może być przydatny,<br />

gdy z jakiegoś powodu niemożliwa jest praca w pełnym<br />

wymiarze czasu. Wtedy ratownik może wycofać się z akcji<br />

i wymienić akumulatory.<br />

Czujniki lokalizacji zawierają w podstawowej wersji układ<br />

lokalizacji satelitarnej. W wersji rozbudowanej mogą być one<br />

wyposażone w inklinometr, czujnik pola magnetycznego 3D<br />

oraz żyroskop 3D.<br />

Układy lokalizacji satelitarnej (GNSS) dostarczają informacji<br />

o bezwzględnym położeniu. Żyroskopy 3D, akcelerometry<br />

i magnetometry, tworzące razem Inercyjny System Nawigacji<br />

(INS) po zintegrowaniu z modułem GNSS pozwolą na uzyskanie<br />

informacji o położeniu w trudnych warunkach środowiskowych<br />

(kanion miejski, las) lub nawet przy całkowitej utracie<br />

sygnału GNSS (np. w budynku). Poprawne działanie (wyznaczanie<br />

pozycji z zakładaną dokładnością) tego systemu jest<br />

mocno ograniczone czasowo, ponieważ w INS jako metodzie<br />

względnej zliczeniowej (wyznacza pozycję na podstawie pomiarów<br />

oraz pozycji poprzedniej) występuje zjawisko kumulowania<br />

błędu z biegiem czasu. Inklinometry są niezbędne dla<br />

robotów mobilnych, aby określić, czy istnieje zagrożenie przewrócenia<br />

się robota np. gdy wspina się on na schody lub inną<br />

przeszkodę. Mogą one również wykryć sam fakt przewrócenia<br />

się robota.<br />

Inklinometr w połączeniu z magnetometrem 3D zostanie<br />

użyty w Mobilnym Centrum Dowodzenia do poprawnego ustawienia<br />

anteny dla łączności satelitarnej. Będzie to następować<br />

automatycznie poprzez działanie urządzeń pokładowych.<br />

Interfejsy implementowane w modułach lokalizacji to:<br />

• CAN,<br />

• UART/RS-232,<br />

• I2C,<br />

• Interfejs firmowy Motoroli oparty na USB Host,<br />

• GPRS.<br />

Protokół GPRS zostanie użyty do przesyłania poprawek<br />

GNSS (DGPS lub RTK) w celu uzyskania lepszej dokładności<br />

wyznaczania pozycji robotów mobilnych. Poprawki te pozwalają<br />

wyeliminować błędy wynikające z opóźnienia sygnału w jonosferze<br />

i troposferze, błędu zegara satelity oraz błędów efemeryd.<br />

Taka metoda pomiarowa pozwala wyznaczyć pozycję z dokładnością<br />

2…3 metrów (poniżej 1 metra dla RTK), zamiast błędu<br />

10 metrów, którym jest obarczony pomiar bez użycia poprawek.<br />

Interfejs CAN jest używany do komunikacji pomiędzy modułami<br />

lokalizacji oraz CPU (Central Processing Unit – główny<br />

procesor) robotów mobilnych lub pojazdów. Używany będzie<br />

interfejs CAN wersji 2.0 A oraz 2.0 B. UART/RS-232 jest podstawowym<br />

interfejsem między CPU modułów lokalizacji oraz<br />

modułami GPS/INS. Szyna I2C jest odpowiednia dla czujników<br />

housekeepingu. Interfejs firmowy Motorola pozwala na<br />

komunikację między modułem lokalizacji i radiotelefonem,<br />

wchodzących w skład Nasobnego Zestawu Czujników. Będzie<br />

on użyty do wysłania bieżącej pozycji do Mobilnego Centrum<br />

Dowodzenia.<br />

Rolą CPU (Central Processing Unit – procesora głównego)<br />

jest inicjalizacja wszystkich czujników, zbieranie i przetwarzanie<br />

danych pomiarowych oraz przesyłanie ich do Mobilnego<br />

Centrum Dowodzenia poprzez odpowiedni interfejs. Następujące<br />

rodziny CPU spełniają wymagania projektu:<br />

• ATmega,<br />

• ARM,<br />

• PIC.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 123


Mikrokontrolery ARM mają bardzo dużą moc obliczeniową,<br />

co powoduje duże zużycie energii. Z tego powodu zostały one<br />

wyłączone z projektu. Mikrokontrolery ATmega i PIC spełniają<br />

wszystkie wymagania projektu. Ponieważ producent mikrokontrolerów<br />

ATmega nie był w stanie dostarczyć procesorów w wyznaczonym<br />

terminie, zostały wybrane mikrokontrolery PIC.<br />

Zasilanie<br />

Spośród wszystkich możliwych układów zasilania, wybrano<br />

przetwornice DC/DC. Powodem tego jest ich wysoka sprawność,<br />

do 90% oraz możliwość otrzymania wyjściowego napięcia<br />

z szerokiego zakresu napięć wejściowych, w tym także<br />

wyższych niż napięcie wejściowe. W urządzeniach mobilnych,<br />

takich jak Nasobne Zestawy Czujników oraz roboty mobilne<br />

wymagana jest wysoka sprawność przetwarzania z powodu<br />

ograniczonej pojemności baterii. Zasilanie wydzielające ciepło<br />

może spowodować błędy urządzeń wrażliwych na temperaturę,<br />

np. kamer zakresu podczerwieni. Z tych powodów stabilizatory<br />

liniowe nie są dobrym rozwiązaniem dla tego projektu.<br />

W niniejszym projekcie wymagane jest przetwarzanie napięcia<br />

wejściowego na napięcie wyjściowe 5 V dla następujących<br />

zakresów napięć wejściowych:<br />

0,9…2,5 V dla Nasobnego Zestawu Czujników,<br />

12…48 V dla robotów mobilnych,<br />

18…32 V dla pojazdów.<br />

Zadaniem zespołu CBK PAN w tym zakresie było zaproponowanie<br />

zasilania dla Nasobnych Zestawów Czujników.<br />

W celu wyboru optymalnego źródła zasilania została wykonana<br />

dogłębna analiza. Rozważano zarówno jednorazowe<br />

baterie konwencjonalne, jak i akumulatory. Baterie jednorazowe<br />

zostały odrzucone ze względu na ich szkodliwość dla<br />

środowiska. Natomiast akumulatory NiCd posiadają efekt pamięciowy,<br />

a akumulatory Li-ion są trudne do ładowania (niezbędna<br />

jest specjalna ładowarka o skomplikowanej budowie),<br />

są wrażliwe na zwarcia (mogą eksplodować) i tracą swoją<br />

pojemność szybko po dużej liczbie ładowań i rozładowań.<br />

Wybrane akumulatory NiMH pozbawione są efektu pamięciowego,<br />

a sposób ich ładowania jest prostszy niż akumulatorów<br />

Li-ion.<br />

Dwa akumulatory NiMH o pojemności 4800 mAh zapewniają<br />

8-godzinną pracę Nasobnego Zestawu Czujników.<br />

Czujnik lokalizacji<br />

W projekcie zostaną użyte następujące rozwiązania lokalizacyjne:<br />

1. Global Navigation Satellite System – globalny system nawigacji<br />

satelitarnej (GNSS),<br />

2. Inertial Navigation System – inercyjny system nawigacyjny<br />

(INS).<br />

Największymi zaletami systemów GNSS są: łatwość<br />

w użytkowaniu oraz ich globalny zasięg bez względu na porę<br />

dnia i warunki pogodowe. Współcześnie producenci rozwijają<br />

technologię high-sensitivity, pozwalającą na detekcję sygnału<br />

GNSS słabszego o 30 dB niż sygnał odbierany przez starsze<br />

odbiorniki GNSS. Pozwala to na poprawną pracę odbiorników<br />

w trudnych warunkach środowiskowych (kanion miejski, las,<br />

niektóre budynki).<br />

W celu poprawy dostępności i dokładności rozwiązań lokalizacyjnych<br />

do wybranych modułów włączono czujniki INS.<br />

Pozwalają one lokalizować obiekty, gdy sygnał GNSS zostaje<br />

utracony. Niestety, czujniki INS zmniejszają dokładność wyznaczania<br />

pozycji wraz z upływem czasu. Dzieje się tak dlatego,<br />

iż pomiar pozycji następuje poprzez podwójne całkowanie<br />

przyspieszenia, co powoduje kumulację błędu.<br />

124<br />

Rys. 2. Położenie punktów P1 i P2 względem budynku CBK PAN<br />

Fig. 2. Location of P1 and P2 points in relation to SRC PAS building<br />

W CBK PAN przeprowadzono szereg testów. Ich celem<br />

był wybór najlepszego modułu GNSS dla projektu. Testy składały<br />

się z części statycznej i kinematycznej. Testy statyczne<br />

przeprowadzono w punktach oznaczonych jako P1 i P2. Na<br />

początku przeprowadzono precyzyjne pomiary referencyjne<br />

(geodezyjnymi odbiornikami GPS metodą Fast Static) położenia<br />

punktów P1 i P2. Następnie wyniki testowanych modułów<br />

GNSS porównywano z wcześniej wyznaczoną pozycją<br />

referencyjną. Punkt P1 znajdował się przed budynkiem CBK<br />

PAN, w odległości 30 m od niego. Widoczność satelitów w tym<br />

punkcie była bardzo dobra (8…10 satelitów). Punkt P2 był położony<br />

blisko budynku, który zasłaniał dużą część widnokręgu<br />

(od północnej i zachodniej strony i częściowo od południowej)<br />

przez co ograniczał liczbę obserwowanych satelitów oraz potęgował<br />

efekt wielotorowości sygnału (tzw. Multipath).<br />

Każdy test statyczny trwał godzinę. Odczyt pozycji następował<br />

co sekundę. Lista testowanych odbiorników jest następująca:<br />

• Fastrax (Mediatek MT3318),<br />

• MTK (Mediatek MT3328),<br />

• Modulestek (SirfStar III),<br />

• u-blox LEA 4P,<br />

• u-blox LEA 5H,<br />

• Novatel (OEMV-3G).<br />

Na rysunku 3 pokazano wyniki pomiarów dwóch generacji<br />

chipów u-blox LEA (4P I 5H) w punkcie P1. Mimo bardzo<br />

dobrych warunków dla lokalizacji GNSS u-blox LEA 4P wykazuje<br />

bardzo słabą precyzję pomiaru (duży rozrzut). u-blox 5<br />

zachowuje się wzorcowo, zarówno dokładność jak i precyzja<br />

pomiaru, są na bardzo wysokim poziomie.<br />

Na rysunku 4 pokazano wyniki testu tych samych chipów<br />

na punkcie P2. Ze względu na niekorzystne położenie punktu<br />

pomiarowego zarówno u-blox LEA 4P i u-blox LEA 5H wykorzystują<br />

zaledwie 4…6 satelitów do wyznaczenia pozycji.<br />

Dodatkowym problemem jest bardzo duży wpływ efektu wielotorowości<br />

sygnału (multipath) na jakość wyznaczanej pozycji.<br />

Możemy zauważyć, że chip u-blox LEA 4P zupełnie nie<br />

radzi sobie z tak złymi warunkami pomiarowymi i błąd osiąga<br />

wartości rzędu kilkunastu, a nawet 20–30 metrów. u-blox LEA<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Rys. 3. Wyniki testu u-blox LEA 4P (lewy) oraz 5H (prawy) w punkcie P1<br />

Fig. 3. Results of u-blox LEA 4P (left) and 5H (right) tests at P1 point<br />

Rys. 4. Wyniki testu u-blox LEA 4P (lewy) oraz 5H (prawy) w punkcie P2 Fig. 4. Result of u-blox LEA 4P (left) and 5H (right) tests at P2 point<br />

5H radzi: sobie z tak trudnymi warunkami znacznie lepiej od<br />

swojego poprzednika. Zarówno precyzja, jak i dokładność pomiaru<br />

są na poziomie zadowalającym dla potrzeb modułów<br />

lokalizacji elementów systemu PROTEUS.<br />

Wyniki testów statycznych wykazały, że najlepszym testowanym<br />

chipem GNSS był u-blox LEA 5H, ponieważ posiadał<br />

on najmniejszy błąd oraz bardzo dobrą precyzję. Dodatkowymi<br />

zaletami modułu u-blox 5H są jego małe wymiary<br />

(17 × 22,4 × 3 mm) oraz niewielkie zużycie mocy. Zatem moduł<br />

u-blox LEA 5H spełnia wymagania dla Nasobnego Zestawu<br />

Czujników oraz modułów lokalizacji robotów mobilnych.<br />

Wyniki testów kinematycznych potwierdziły słuszność wyboru<br />

modułu GNSS. Średnia dostępność lokalizacji w terenie<br />

otwartym, w lesie oraz w kanionie miejskim była na poziomie<br />

95%, a dokładność wynosiła 2–3 m (w najgorszych warunkach<br />

do 7…10 m).<br />

Dodatkowo przeprowadzono również testy modułu u-blox<br />

LEA5H w budynku (budynek 3-kondygnacyjny, materiał: beton<br />

+ drewno), które wykazały niespodziewaną dostępność<br />

rozwiązań pozycyjnych na poziomie 70% oraz dokładność<br />

rzędu 10–30 m. Wyniki testów wewnątrz budynków pokazano<br />

na rys. 5. Test polegał na akwizycji pozycji satelitów, wejściu<br />

do budynku i przejściu przez niego. Droga pokazana przez u-<br />

blox LEA 5H była najbliższa rzeczywistej trasie przejścia.<br />

Rys. 5. Wyniki testu wewnątrz budynku u-blox LEA 5H<br />

Fig. 5. Results of u-blox LEA 5H tests inside a building<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 125


• wstrząsy w ciągu 1 godziny,<br />

• upadek z 1 metra na metalową płytę.<br />

We wszystkich przypadkach będzie sprawdzane bezpieczeństwo<br />

urządzeń. Mogą one ulec uszkodzeniu, lecz nie<br />

mogą spowodować uszkodzeń ciała.<br />

Podsumowanie<br />

Rys. 6. Wyniki testów u-blox LEA 5H na parkingu<br />

Fig. 6. Results of u-blox LEA 5H tests in a parking lot<br />

Dodatkowo przeprowadzono test wewnątrz parkingu wielopoziomowego.<br />

Trasa prowadziła przez wszystkie poziomy<br />

parkingu (3 kondygnacje). Wyniki dla u-blox LEA 5H pokazano<br />

na rys. 6. Dostępność sygnału dla całego testu była na<br />

poziomie 70% (brak odbioru sygnału na 1 i 2 kondygnacji<br />

w środkowej części parkingu).<br />

Próby wytrzymałościowe Nasobnego<br />

Zestawu Czujników<br />

Szczególną uwagę należy zwrócić na Nasobne Zestawy Czujników.<br />

Powinny one spełnić wiele kryteriów bezpieczeństwa<br />

zdefiniowanych przez końcowych użytkowników – Policję,<br />

Straż Pożarną i antyterrorystów. Z tego powodu należy przeprowadzić<br />

następujące testy:<br />

• zalanie wodą w temperaturze pokojowej,<br />

• podgrzanie do temperatury 80 stopni Celsjusza,<br />

• schłodzenie do -20 stopni Celsjusza,<br />

W ramach projektu PROTEUS powstają innowacyjne urządzenia,<br />

których lokalizacja będzie opierała się na zastosowaniu<br />

systemów GNSS oraz integracji odbiorników GNSS z modułami<br />

INS. Każdy element systemu PROTEUS powstaje<br />

w ścisłej współpracy z użytkownikami końcowymi, którymi są<br />

przede wszystkim Straż Pożarna, Policja i służby antyterrorystyczne.<br />

To właśnie odbiorcy Proteusa brali udział w opracowywaniu<br />

założeń taktyczno-technicznych systemu i jego<br />

elementów precyzyjnie określając wymagania, których spełnienie<br />

stanowi wyzwanie dla realizujących projekt inżynierów<br />

i narzuca stosowanie innowacyjnych rozwiązań. Opracowywane<br />

w CBK PAN moduły lokalizacji są szczególnym przypadkiem,<br />

ponieważ są przed nimi stawiane zróżnicowane<br />

wymagania środowiskowe i dokładnościowe w zależności od<br />

warunków pracy i funkcji elementów systemu, które mają być<br />

w nie wyposażone. Dlatego zaprojektowanie i opracowanie<br />

modułów dla każdego elementu systemu wymaga indywidualnego<br />

podejścia.<br />

Dotychczas przeprowadzone testy dostępnych komercyjnie<br />

modułów GNSS w trybie statycznym i kinematycznym<br />

wykazały, że najlepszym modułem lokalizacyjnym dla elementów<br />

systemu jest U-blox LEA 5H. Jednak przewidziane<br />

są testy nowych modułów, które powinny wykazać jeszcze<br />

lepszą jakość.<br />

Prace nad połączeniem metod GNSS i INS w Zestawach<br />

Nasobnych pozwolą na lokalizację osób przy słabej lub braku<br />

widoczności satelitów GNSS.<br />

Projekt powstaje we współpracy 7 polskich instytucji naukowych:<br />

Przemysłowy <strong>Instytut</strong> Automatyki i Pomiarów, Centrum<br />

Badań Kosmicznych PAN, Centrum Naukowo-Badawcze Ochrony<br />

Przeciwpożarowej, <strong>Instytut</strong> Technologii Materiałów <strong>Elektronicznych</strong>,<br />

Politechnika Poznańska, Politechnika Warszawska,<br />

Wojskowa Akademia Techniczna.<br />

Projekt jest współfinansowany przez Europejski Fundusz Rozwoju<br />

Regionalnego w ramach Programu Operacyjnego Innowacyjna<br />

Gospodarka na lata 2007–2013.<br />

Pełna informacja o projekcie dostępna jest pod adresem<br />

http://www.projektproteus.pl<br />

Wpłata w <strong>2010</strong> roku – GwarancjĄ niŻszej ceny prenumeraty o vat!<br />

126<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Analiza możliwości zastosowania sensorów<br />

inercjalnych MEMS w projekcie PROTEUS<br />

dr inż. MACIEJ KALARUS, mgr inż. PIOTR SITEK, mgr inż. KAROL BRZOSTOWSKI,<br />

mgr inż. ANNA FOKS-RYZNAR, mgr inż. RADOSLAV DARAKCHIEV<br />

Centrum Badań Kosmicznych PAN, Warszawa<br />

Podstawowy system lokalizacyjny w projekcie PROTEUS<br />

wykorzystuje moduły GNSS, jednak takie rozwiązanie nie zapewnia<br />

ciągłości pozycjonowania podczas utraty widoczności<br />

satelitów (w tunelach, pośród wysokiej zabudowy, wewnątrz<br />

budynków etc.). W tej sytuacji naturalne wydaje się wykorzystanie<br />

alternatywnych możliwości lokalizacji obiektów. Jedną<br />

z nich jest zastosowanie systemów inercjalnych (bezwładnościowych),<br />

które pozwalają na zwiększenie dokładności pozycjonowania<br />

w warunkach częściowej utraty widoczności satelitów,<br />

a także na kontynuowanie (w ograniczonej skali czasu)<br />

nawigacji po całkowitej utracie sygnałów satelitarnych.<br />

W tym kontekście przetestowany został moduł Xsens MTiG<br />

(system zintegrowany INS/GPS) zawierający m.in. trójosiowe<br />

sensory przyspieszenia (akcelerometry) oraz sensory prędkości<br />

kątowej (żyroskopy) wykonane w technologii MEMS. Ponadto,<br />

moduł wyposażony jest również w magnetometr i barometr,<br />

które poprawiają dokładność wyznaczenia odpowiednio<br />

orientacji względem Ziemi oraz względnych zmian wysokości.<br />

W celu oceny praktycznych możliwości pomiarowych układu<br />

MTiG przeprowadzono dwa rodzaje testów. Pierwszy określono<br />

jako test w warunkach statycznych polegający na obserwacji<br />

układu pozostającego w spoczynku, dzięki czemu wyznaczono<br />

głównie parametry szumowe sensorów. Natomiast drugi test<br />

wykonano w warunkach bardziej zbliżonych do rzeczywistych<br />

pokonując trasę wewnątrz budynku w celu oszacowania błędu<br />

pozycji obliczonej na podstawie danych z podsystemu INS.<br />

Pomiary podstawowych parametrów<br />

podsystemu INS układu Xsens MTiG<br />

Wybrane parametry układu zostały wyznaczone z szeregów<br />

czasowych zarejestrowanych podczas dwunastogodzinnego<br />

pomiaru w warunkach statycznych (tab. 1).<br />

Tab. 1. Parametry sensorów inercjalnych<br />

Tabl. 1. Parameters of the inertial sensors<br />

Sensor Akcelerometr [x, y, z] Żyroskop [x, y, z]<br />

wartość średnia -0.311<br />

(1)<br />

0.599 m/s 2<br />

9.794<br />

odchylenie<br />

standardowe<br />

0.0080<br />

0.0086 m/s 2<br />

0.0084<br />

gęstość szumów 0.0017<br />

0.0011 (0.002) m/s 2 / √Hz ) (3) 0.0017<br />

poziom szumów<br />

(błądzenie<br />

przypadkowe)<br />

stabilność<br />

(bias stability)<br />

velocity random walk (VRW)<br />

0.05<br />

0.06 [m/s/ √h ]<br />

0.06<br />

1.02 (72s)<br />

1.54 (14s) [m/h 2 ]<br />

1.32 (83s)<br />

(1)<br />

składowe przyspieszenia siły ciężkości<br />

(2)<br />

bias<br />

(3)<br />

parametr podany przez producenta [4]<br />

0.0120<br />

Wstępnej oceny urządzenia można dokonać np. na podstawie<br />

wartości średniej prędkości kątowej, która powinna<br />

wynosić zero (układ pozostawał w spoczynku). Z pomiarów<br />

wynika jednak, że ω y<br />

= 0.2315°/s, co oznacza jeden fikcyjny<br />

obrót na 26 min. Niestety nie jest to wada układu, a możliwa<br />

wartość tego parametru dla sensorów inercjalnych wykonanych<br />

w technologii MEMS. Szczęśliwie, w tym przypadku<br />

algorytm wyznaczania pozycji może korygować dane pomiarowe<br />

uwzględniając znany błąd sensorów, jednak dość duże<br />

szumy (tab. 1) w dalszym ciągu będą degradować dokładność<br />

rozwiązania (pozycji) w miarę upływu czasu. Dla wyjaśnienia,<br />

parametry szumowe VRW (Velocity Random Walk), ARW<br />

(Angle Random Walk) oraz stabilność (bias stability) odczytać<br />

można bezpośrednio z wykresu przedstawiającego wariancję<br />

Allana [3]. Przykładowo ARW x<br />

= 2.16°/√h oznacza, że po jednej<br />

godzinie odchylenie standardowe błędu orientacji będzie<br />

na poziomie 2.16°, po dwóch godzinach √2 · 2.16° = 3.05° itd.<br />

Natomiast stabilność określana jest jako minimum wariancji<br />

Allana dla czasu uśredniania t. Zatem BS x<br />

= 2.84 · 10 - 4 m/s 2 dla<br />

czasu uśredniania t = 72 s oznacza, że jeśli znany jest bias<br />

B(t), to B(t + 72 s) będzie zmienną losową o wartości oczekiwanej<br />

B(t) i odchyleniu standardowym BS x<br />

.<br />

Dla porównania warto wspomnieć, że układy INS stosowane<br />

w systemach militarnych charakteryzują się parametrami<br />

lepszymi nawet o kilka rzędów wielkości. Z drugiej strony<br />

niewątpliwą zaletą układów MEMS jest o kilka rzędów niższa<br />

cena i niewielka waga, co sprawia, że mogą być stosowane<br />

w niskobudżetowych aplikacjach oraz na większą skalę.<br />

Kolejny test wykonano pokonując pieszo trasę o długości<br />

ok. 250 m wewnątrz budynku. Dane zarejestrowano z częstotliwością<br />

100 Hz otrzymując ok. 30 tys. punktów pomiarowych.<br />

Układ starano się utrzymywać w pozycji horyzontalnej<br />

w taki sposób, aby oś x skierowana była w kierunku ruchu,<br />

a oś z do dołu. Rys. 1 potwierdza, że wartości<br />

średnie składowych x i y oscylują w pobliżu<br />

zera, natomiast składowa z zawiera stały<br />

sygnał odpowiadający przyspieszeniu siły<br />

ciężkości.<br />

Z punktu widzenia analizy dynamiki pieszego<br />

najbardziej interesujące dane pochodzą<br />

z sensorów zorientowanych w kierunku z, tj.<br />

zmiany przyspieszenia w kierunku pionowym<br />

oraz zmiany orientacji horyzontalnej (rys. 2).<br />

W przypadku żyroskopu wyraźnie widoczne są<br />

momenty wykonywania zwrotów (składowa ω z<br />

),<br />

natomiast w składowej przyspieszenia a z<br />

wyodrębnić<br />

można sygnał odpowiadający częstotliwości<br />

stawiania kroków czyli oscylacji góra-dół.<br />

Potwierdza to rys. 3 przedstawiający analizę<br />

spektralną danych z akcelerometrów z zaznaczonym<br />

maksimum dla częstotliwości ok. 1,84<br />

Hz. Ponadto, w okolicach 210. sekundy szeregu<br />

czasowego a z<br />

widoczne są oscylacje o większej<br />

amplitudzie odpowiadające wykonywaniu<br />

skłonów.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 127<br />

(2)<br />

0.2315 °/s<br />

-0.1879<br />

0.3300<br />

0.3723 °/s<br />

0.3053<br />

0.052<br />

0.059 (0.05) °/s/ √Hz ) (3)<br />

0.048<br />

angle random walk (ARW)<br />

2.16<br />

2.44 [°/ √h ]<br />

2.01<br />

23.0 (143s)<br />

21.9 (247s) [°/ h ]<br />

18.7 (493s)


a x<br />

[m/s 2 ]<br />

a y<br />

[m/s 2 ]<br />

a z<br />

[m/s 2 ]<br />

Rys. 1. Dane pomiarowe z akcelerometrów<br />

Fig. 1. Accelerometer measurement data<br />

x [deg/s]<br />

y [deg/s]<br />

z [deg/s]<br />

Rys. 2. Dane pomiarowe z żyroskopów<br />

Fig. 2. Gyro measurement data<br />

PSD: a x<br />

PSD: a y<br />

PSD: a z<br />

Standardowe podejście do nawigacji<br />

inercjalnej<br />

Nawigacja inercjalna polega na pomiarze i całkowaniu przyspieszeń<br />

liniowych a (1) oraz prędkości kątowej ω (2), dzięki<br />

czemu możliwe jest wyznaczenie prędkości v, położenia p<br />

oraz orientacji α obiektu w przestrzeni.<br />

p = ( adtνo<br />

) dt po<br />

(1)<br />

α = ωdt + α<br />

o<br />

∫ (2)<br />

128<br />

2<br />

0<br />

-2<br />

0 50 100 150 200 250 300<br />

2<br />

0<br />

-2<br />

0 50 100 150 200 250 300<br />

15<br />

10<br />

5<br />

0 50 100 150 200 250 300<br />

czas [s]<br />

50<br />

0<br />

-50<br />

0 50 100 150 200 250 300<br />

50<br />

0<br />

-50<br />

0 50 100 150 200 250 300<br />

100<br />

0<br />

-100<br />

0 50 100 150 200 250 300<br />

czas [s]<br />

10 -2<br />

10 -4<br />

10 -2<br />

10 -4<br />

10 -2<br />

10 -4<br />

10 -2 10 -1 10 0 10 1<br />

10 -2 10 -1 10 0 10 1<br />

10 -2 10 -1 10 0 10 1<br />

cz stotliwo [Hz]<br />

Rys. 3. Dane z akcelerometrów – PSD<br />

Fig. 3. PSD of the accelerometer measurement data<br />

∫ ∫<br />

+<br />

Otrzymana pozycja jest pozycją względną i w celu nawiązania<br />

jej np. do globalnego układu współrzędnych należy<br />

dostarczyć odpowiednie stałe całkowania – pozycję p o<br />

i prędkość<br />

v o<br />

wyznaczone przez GPS. Ponadto, z uwagi na szybko<br />

rosnący (z kwadratem czasu) błąd pozycji wyznaczanej tylko<br />

z INS, konieczna jest okresowa kalibracja, czyli dowiązanie<br />

do bardziej wiarygodnego rozwiązania, które standardowo<br />

dostarcza GPS, ale może to być także inny system pozycjonowania,<br />

działający z wykorzystaniem np. radiolatarni lub lokalnych<br />

infrastruktur sieci WiFi.<br />

Powyższe podejście zastosowano do oszacowania błędu<br />

pozycji wyznaczanej przez INS analizując dane pomiarowe<br />

(rys. 1 i 2) zebrane podczas przemieszczania się wewnątrz<br />

budynku bez dowiązywania rozwiązań zewnętrznych<br />

(z GPS). W tym celu scałkowano dane z żyroskopów otrzymując<br />

bieżącą orientację oraz podwójnie całkowano dane<br />

z akcelerometrów otrzymując przemieszczenia. W obliczeniach<br />

uwzględniono macierz transformacji z układu związanego<br />

z sensorami do układu nawigacyjnego oraz wyeliminowano<br />

wpływ przyspieszenia siły ciężkości. Otrzymany<br />

rezultat nie pozostawił złudzeń co do sensowności stosowania<br />

takiego podejścia, gdyż błąd pozycji po ok. 2 min<br />

osiągnął wartość kilkuset metrów. Dalsza analiza wykazała,<br />

że dominującym czynnikiem degradującym był dość duży<br />

szum akcelerometrów, gdyż nawet wizualnie trudno wyodrębnić<br />

ze składowej a x<br />

(rys. 1) (zgodnej z kierunkiem ruchu)<br />

momenty oraz wartości przyspieszeń podczas zmiany prędkości<br />

poruszania się – zatrzymywania się i ruszania. Natomiast<br />

bardzo dobrze udało się wyznaczyć zmiany orientacji<br />

w czasie (rys. 4), gdzie wyraźnie widać zwroty o 90 o , 180 o<br />

oraz obroty.<br />

[deg]<br />

720<br />

630<br />

540<br />

450<br />

360<br />

270<br />

180<br />

90<br />

-90 0<br />

-180<br />

-270<br />

-360<br />

-450<br />

-540<br />

0 50 100 150 200 250 300<br />

czas [s]<br />

Rys. 4 Zmiany orientacji obiektu w czasie<br />

Fig. 4. Attitude in time<br />

Alternatywne podejście do wyznaczania<br />

pozycji<br />

Jak pokazano wcześniej, dane pomiarowe charakteryzują<br />

się dość dużym poziomem szumów dlatego dalsze postępowanie<br />

skupione zostało na przetwarzaniu sygnałów a z<br />

i ω z<br />

,<br />

które z racji sposobu przeprowadzania testu (utrzymywanie<br />

modułu w pozycji horyzontalnej) charakteryzują się największą<br />

wartością SNR. Jako alternatywę dla poprzedniego<br />

podejścia zaproponowano analizę polegającą na wyodrębnieniu<br />

z a z<br />

cech sygnału charakterystycznych dla obiektu<br />

pomiarowego (w tym przypadku pieszego). Taką cechą jest<br />

oczywiście wspomniana wcześniej oscylacja o częstotliwości<br />

f o<br />

= 1,84 Hz. Ze względu na niestacjonarność szeregu<br />

czasowego a z<br />

, do wykrycia przedziałów czasowych w których<br />

występuje oscylacja (marsz lub postój), wykorzystano<br />

transformatę falkową (wavelet) (2), która umożliwia określenie<br />

parametrów zarówno w dziedzinie czasu jak i częstotliwości.<br />

Podstawowy wzór wyraża się następująco:<br />

x<br />

y<br />

z<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


|w|<br />

⎛ t − b ⎞<br />

wˆ<br />

( a,<br />

b)<br />

= z t Ψ ⎜ ⎟dt<br />

∫ +∞ ( )<br />

(3)<br />

−∞<br />

⎝ a ⎠<br />

gdzie a i b nazywane są odpowiednio parametrem dylatacji<br />

(skali) i translacji (przestrzeni), natomiast Ψ jest funkcją analizującą,<br />

która w zależności od zastosowań może przyjmować<br />

różne postaci. W niniejszej analizie wykorzystano zespoloną<br />

falkę Morleta (4), gdzie szerokość pasma f b<br />

dobrano empirycznie<br />

w celu otrzymania optymalnej jakości rozwiązania<br />

w dziedzinie czasu.<br />

<br />

6<br />

4<br />

2<br />

0<br />

0 50 100 150 200 250 300<br />

czas [s]<br />

Rys. 5. Moduł transformaty Falkowej<br />

Fig. 5. Module of the wavelet transform<br />

Następnie wyznaczono moduł transformaty falkowej |ω^ (a, b) (t)|<br />

(rys. 5), przy stałym parametrze skali odpowiadającym częstotliwości<br />

f o<br />

. Ustalono też, że wartość modułu powyżej<br />

1,5 oznacza przemieszczanie się pieszego z prędkością<br />

v p<br />

= 4,5 km/h w kierunku wyznaczonym przez żyroskop<br />

z (rys. 4).<br />

Okazało się, że takie rozwiązanie daje mniejszy błąd<br />

pozycji (nie rosnący z kwadratem czasu) co wynika przede<br />

wszystkim z braku kłopotliwego podwójnego całkowania<br />

przyspieszeń. W teście w zależności od ustalonej prędkości<br />

v p<br />

, błąd pozycji po 2 min, wynosił ok. kilkadziesiąt metrów,<br />

czyli znacznie mniej niż przy zastosowaniu klasycznego<br />

całkowania. Obiektywnie jednak trzeba zauważyć,<br />

że rezultat otrzymywany jest przez pośrednie szacowanie<br />

przy znajomości dynamiki obiektu. Zatem takie rozwiązanie<br />

może nie sprawdzić się podczas pozycjonowania robota,<br />

choć może i wówczas da się wyodrębnić charakterystyczne<br />

drgania wynikające z przemieszczania się z określoną<br />

prędkością.<br />

Podsumowanie<br />

Ψ(<br />

t)<br />

= e<br />

2<br />

−t<br />

f b<br />

−iπt<br />

Przeprowadzone analizy pozwalają zakwalifikować układ<br />

Xsens MTiG jako urządzenie rozpoznające typowe, w tym<br />

przypadku dla człowieka, zachowania wewnątrz budynku tj.<br />

przemieszczanie się po linii prostej, zwroty, przysiady, schylanie<br />

się, etc. Jednak nie trudno wyobrazić sobie, że wyodrębnienie<br />

różnych faz ruchu znacząco zależy od miejsca<br />

umocowania przyrządu, co powinno zostać uwzględnione<br />

w dalszych badaniach.<br />

Wyniki pokazują również dość duży udział szumów,<br />

szczególnie w pomiarach przyspieszeń, choć ich poziom<br />

e<br />

(4)<br />

jest typowy dla miniaturowych sensorów MEMS. Dodając<br />

do tego trudności w przetwarzaniu skomplikowanej dynamiki<br />

pieszego należy stwierdzić, że autonomiczne zastosowanie<br />

klasycznej nawigacji zliczeniowej (dead reckoning)<br />

będzie miało praktyczny sens jedynie przez kilka do kilkunastu<br />

sekund. Warto odnotować, że algorytmy przetwarzania<br />

dla systemów nawigacyjnych typu strapdown, a takim jest<br />

omawiany układ, są zdecydowanie bardziej skomplikowane<br />

i wymagają bardziej precyzyjnych pomiarów niż algorytmy<br />

stosowane w systemach kardanowych. Reasumując, błędy<br />

pomiarowe przyspieszeń są duże i dodatkowo kumulują się<br />

dając błąd pozycji rosnący z kwadratem czasu. Natomiast<br />

niewątpliwie łatwiej przetwarza się informację o prędkości<br />

kątowej, gdyż orientację układu otrzymuje się po jednokrotnym<br />

całkowaniu (błąd rośnie liniowo). Jednak pomimo<br />

fabrycznej kalibracji żyroskopy generują zauważalne błędy<br />

systematyczne, które muszą być korygowane przed dalszymi<br />

obliczeniami.<br />

Testy układu MTiG wykazały, że podstawowym problemem<br />

jest czasowa stabilność podsystemów inercjalnych MEMS,<br />

co powoduje, że najbardziej optymalna integracja INS/GPS<br />

w celu wyznaczenia pozycji odbywa się w skali czasu równej<br />

pojedynczym sekundom. Dla mniejszej skali czasu (


Antena mikropaskowa o poszerzonym paśmie pracy<br />

mgr inż. MARIUSZ PERGOŁ, dr hab. inż. WŁODZIMIERZ ZIENIUTYCZ,<br />

mgr inż. ŁUKASZ SOROKOSZ<br />

Politechnika Gdańska, Wydział ETI, Katedra Inżynierii Mikrofalowej i Antenowej<br />

Anteny mikropaskowe są elementami szeroko stosowanymi<br />

we współczesnej elektronice, w komunikacji bezprzewodowej<br />

(WiFi, WiMAX, GSM) i w systemach radarowych (lotnictwo).<br />

Cechuje je: mała waga, odporność na zmieniające się warunki<br />

atmosferyczne, niski koszt oraz powtarzalność produkcji.<br />

Atrakcyjność tego typu anten sprawia, iż są one obiektem wielu<br />

badań. Prace te są często skupione na poszukiwaniu sposobu<br />

uzyskania stosunkowo szerokiego pasma, którego brak<br />

jest jedną z głównych wad anten mikropaskowych. Problem<br />

ten jest szczególnie istotny w systemach radarowych, w których<br />

standardowo wykorzystuje się pasma komunikacyjne<br />

rzędu 30% przy poziomie współczynnika fali stojącej (WFS)<br />

poniżej 1,5. W bibliografii można znaleźć szereg publikacji na<br />

temat metod poszerzenia pasma, np. poprzez zwiększenie<br />

grubości podłoża i zasilanie przez sprzężenie elektromagnetyczne<br />

[2], zastosowanie dodatkowych elementów reaktancyjnych<br />

w płaszczyźnie umieszczenia łaty lub w łacie [3, 6].<br />

Dodatkowym warunkiem poprawnej pracy anteny jest odpowiednia<br />

charakterystyka promieniowania o szerokości wiązki<br />

głównej równej co najmniej kilkudziesięciu stopniom.<br />

W niniejszym artykule przedstawiono wyniki prac nad anteną<br />

spełniającą powyższe wymagania. Szerokie, 30% pasmo<br />

pracy anten z dopasowaniem na poziomie WFS W).<br />

Aby uniknąć tego efektu można zastosować szczelinę typu<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Wcześniej opisano budowę anteny mikropaskowej oraz<br />

poruszono problem efektywnego jej zasilania. W tej części<br />

przedstawiono wyniki badań, których celem było zaprojektowanie<br />

anteny mikropaskowej zasilanej przez sprzężenie<br />

elektromagnetyczne, posiadającej możliwie jak najszersze<br />

pasmo. Częstotliwość środkowa pracy została określona na<br />

1,1 GHz. Projekt anteny mikropaskowej podzielony został na<br />

3 etapy:<br />

– badania numeryczne anteny mikropaskowej zasilanej<br />

przez szczelinę typu H, bez ekranu (ADS Momentum),<br />

– badania numeryczne anteny mikropaskowej ze zmodyfikowanym<br />

układem zasilania, bez ekranu,<br />

– wykonanie anteny, weryfikacja eksperymentalna, dostrojenie<br />

anteny.<br />

W pierwszych dwóch etapach projektu, ze względu na<br />

oszczędność czasu, prowadzono badania numeryczne an-<br />

a) b)<br />

teny bez ekranu. Na wstępie przyjęto<br />

parametry podłoża, na którym antena<br />

miała być zrealizowana. Wyniosły<br />

one: d 1<br />

= d 3<br />

= 0,76 mm; ε r1<br />

= ε r3<br />

= 3,5;<br />

d 2<br />

= 0,40 mm; ε r2<br />

= 1,05. Wyniki symulacji<br />

pokazały, że wykorzystując pojedynczą<br />

szczelinę typu H można uzyskać około<br />

16-procentowe pasmo pracy. W drugim<br />

etapie projektu przystąpiono do próby<br />

poszerzenia pasma pracy. W tym celu<br />

zaproponowano wprowadzenie do układu<br />

sprzęgającego drugiej szczeliny, tym<br />

razem prostokątnej (rys. 5). W efekcie<br />

uzyskano znaczne poszerzenie pasma<br />

do 29%. Wymiary szczelin (l, w, a, d)<br />

oraz radiatora (L, W) anteny mikropaskowej<br />

z poszerzonym pasmem podano<br />

w tabeli. Wyniki numeryczne dla<br />

Rys. 3. Oznaczenia wymiarów radiatora i szczeliny: a) warstwa radiatora; b) warstwa<br />

masy<br />

zaprojektowanej anteny przedstawiono<br />

na rys. 6. W kolejnym kroku antenę<br />

Fig. 3. Radiator and slot dimension: a) radiator layer; b) groundplane<br />

zrealizowano w Katedrze Inżynierii Mikrofalowej<br />

i Antenowej Politechniki Gdańskiej i zmierzono jej<br />

a<br />

wejściowy współczynnik odbicia. Ze względu na zastosowane<br />

przybliżenie w procesie projektowania (pominięcie warstwy<br />

d<br />

I<br />

ekranu dolnego) kolejnym etapem było eksperymentalne dobranie<br />

odległości między ekranem a anteną (d 4<br />

ef<br />

) oraz między<br />

warstwą układu szczelin i radiatora (d 2<br />

), tak aby uzyskać maksymalne<br />

pasmo pracy.<br />

Rys. 4. Szczelina typu H i jej ekwiwalentny odpowiednik w postaci<br />

szczeliny prostokątnej<br />

Fig. 4. H-type slot with corresponding rectangular slot<br />

d<br />

a<br />

H, której efektywna długość l ef<br />

jest znacznie większa niż jej<br />

faktyczna szerokość a (l ef<br />

≅ a + 4d > a, rys. 4). Dodatkowo<br />

zastosowanie szczeliny typu H pozwala zmniejszyć wymiary<br />

poprzeczne radiatora, co ma duże znaczenie, szczególnie<br />

w przypadku szyków antenowych [4]. Poprzez efektywną długość<br />

szczeliny typu H rozumiemy tutaj długość hipotetycznej<br />

szczeliny prostokątnej, której zastosowanie zapewniłoby zbliżone<br />

pod względem efektywności sprzężenie radiatora z linią<br />

mikropaskową.<br />

Projekt anteny mikropaskowej zasilanej<br />

przez sprzężenie elektromagnetyczne<br />

w<br />

Rys. 5. Układ szczelin zapewniający szersze pasmo anteny mikropaskowej<br />

Fig. 5. Proposed slot configuration resulted in broadband bandwidth<br />

Uzyskane z symulacji komputerowych wymiary szczelin stosowanych<br />

w antenie mikropaskowej o poszerzonym paśmie pracy. L 1,<br />

W 1<br />

– wymiary<br />

zastosowanego w realizacji anteny laminatu (rys. 3)<br />

Dimension of slots configuration obtained from numerical results.<br />

L 1<br />

, W 1<br />

– dimension of dielectric substrate used in experiment (Fig. 3)<br />

l [mm] w [mm] a [mm] d [mm] L [mm] W [mm] L 1<br />

[mm] W 1<br />

[mm]<br />

46,5 4,1 72,3 8,6 89,4 72,3 27,5 13<br />

Ostatecznie ustalone wielkości to d 2<br />

= 35 mm i d 4<br />

= 41 mm.<br />

Wyniki pomiarów charakterystyki dopasowania zrealizowanej<br />

anteny przedstawiono na rys. 6. Antena charakteryzuje się<br />

29% pasmem pracy (0,94…1,26 GHz) przy wartości WFS


Rys. 6. Wyniki numeryczne anteny mikropaskowej z układem<br />

dwóch szczelin sprzęgających, bez ekranu<br />

Fig. 6. Numerical results of unshielded patch antenna with two<br />

slots configuration<br />

Rys. 7. Wyniki eksperymentalne anteny mikropaskowej z układem<br />

dwóch szczelin sprzęgających z obecnością ekranu<br />

Fig. 7. Experimental results shielded patch antenna with two<br />

slots configuration<br />

eksperymentalnych, szczególnie dla dolnych częstotliwości<br />

pasma pracy. Dzięki zastosowaniu drugiej szczeliny sprzęgającej<br />

pasmo pracy zostało poszerzone z 16 do 29%.<br />

Charakterystyka promieniowania anteny<br />

mikropaskowej<br />

Po wykonaniu anteny, okazało się, że na wyższych częstotliwościach<br />

w pomierzonych charakterystykach promieniowania<br />

występują niepożądane minima [5]. Źródła tych zniekształceń<br />

szukano w skończonych wymiarach warstw układu szczelin<br />

i radiatora (L 1<br />

= 27,5 mm i W = 13 mm). W celu weryfikacji postawionej<br />

hipotezy zaprojektowano i wykonano kolejną wersję<br />

1<br />

anteny (rys. 8) ze zmienionym wymiarem L 1<br />

= 14 mm. W rezultacie<br />

uzyskano charakterystyki promieniowania (rys. 9),<br />

które nie wykazywały niepożądanych minimów. Na rys. 10,<br />

celem porównania przedstawiono analogiczne charakterystyki<br />

uzyskane z symulacji numerycznych.<br />

a) b)<br />

Rys. 8. Zdjęcie wykonanej anteny: a) warstwa układu szczelin; b)<br />

warstwa radiatora<br />

Fig. 8. Photo of the antenna: a) slot configuration; b) radiator<br />

layer<br />

Rys. 9. Charakterystyka promieniowania anteny po modyfikacji; a) płaszczyzna E; b) płaszczyzna H. Wyniki eksperymentalne<br />

Fig. 9. Radiation pattern of the modified patch antenna; a) E-plane; b) H-plane. Experimental results<br />

132<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Rys. 10. Charakterystyka promieniowania anteny po modyfikacji; a) płaszczyzna E; b) płaszczyzna H. Wyniki numeryczne<br />

Fig. 10. Radiation pattern of the modified patch antenna; a) E-plane; b) H-plane. Numerical results<br />

Promieniowanie wsteczne, które zaobserwowano w badanej<br />

antenie wynika ze skończonych wymiarów ekranu dolnego<br />

użytego przy pomiarach. Uwzględniając fakt, że antena<br />

będzie docelowo zamontowana na płaszczyźnie przewodzącej<br />

o większej powierzchni, ww. poziom promieniowania<br />

wstecznego nie prowadzi do ograniczenia w zastosowaniu<br />

anteny.<br />

Podsumowanie<br />

W pracy przedstawiono wyniki badań anteny mikropaskowej<br />

zasilanej przez układ dwóch szczelin we wspólnej masie. Zaproponowana<br />

w trakcie symulacji numerycznych modyfikacja<br />

układu sprzęgającego przyniosła efekt w postaci szerszego,<br />

29% pasma pracy. Ten sam efekt uzyskano w badaniach<br />

eksperymentalnych, dostrajając antenę poprzez dobór odpowiedniej<br />

odległości między warstwą linii mikropaskowej<br />

a ekranem. Przeprowadzone badania dwóch wersji anten<br />

o różnych wymiarach płaszczyzny wspólnej masy pokazały,<br />

że zastosowanie odpowiednich wymiarów warstwy wspólnej<br />

masy usuwa zniekształcenia charakterystyki promieniowania<br />

(dodatkowe minima w wiązce głównej) obserwowane w wersji<br />

anteny o większych rozmiarach wspólnej masy.<br />

Praca naukowa finansowana ze środków na naukę w latach<br />

2007-<strong>2010</strong> jako Projekt Badawczy Zamawiany PBZ-MNiSW-DBO-<br />

04/I/2007.<br />

Autorzy pragną podziękować Przemysłowemu <strong>Instytut</strong>owi Telekomunikacji<br />

o. Gdańsk za pomoc przy przeprowadzeniu pomiarów<br />

charakterystyk promieniowania badanych anten.<br />

Literatura<br />

[1] Pozar D.M.: A Microstrip Antenna Aperture-Coupled to a Microstrip<br />

Line. Electronic Letters; vol. 21; 1985; pp. 49–50.<br />

[2] Garg R.i in.: Microstrip Antenna Design Handbook; Artech House<br />

2001.<br />

[3] Kowalczys M., Zieniutycz W.: Prostokątne anteny mikropaskowe<br />

zasilane przez sprzężenie elektromagnetyczne. <strong>Elektronika</strong> –<br />

konstrukcje, technologie, zastosowania; vol. 5; 2008; pp. 74–79.<br />

[4] Zieniutycz W., Mazur M., Pergol M.: IFF/SSR Rectangular Patch<br />

Antenna with Reduced Transversal Size, Microwave and Optical<br />

Technology Letters (przyjęte do druku)<br />

[5] Pergol M., Zieniutycz W., Mazur M.: Broadband Microstrip Patch<br />

Antenna with Reduced Transversal Size. 18th International Conference<br />

on Microwaves, Radar and Wireless Communications,<br />

Wilno, Litwa, <strong>2010</strong>, vol. 1, pp. 86–88, <strong>2010</strong>.<br />

[6] McFeetors G., Weldon M.A.J., Okoniewski M.: Aperture-Fed<br />

Patch Antenna with Planar Reactive Load. IEEE Antennas and<br />

Wireless Propagation Letters; vol. 3; 2004; pp. 182–185.<br />

Przypominamy o prenumeracie miesięcznika <strong>Elektronika</strong> na 2011 r.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 133


Properties of feature contours for audio<br />

classification tasks<br />

(Właściwości konturów cech w zadaniach klasyfikacji<br />

sygnałów akustycznych)<br />

dr inż. TOMASZ MĄKA, Zachodniopomorski Uniwersytet Technologiczny w Szczecinie<br />

Increasing technological advances stimulates development of<br />

fully automated speech/audio interaction systems in everyday<br />

life. Such systems include various modules to handle different<br />

acoustical streams like speech, music, background noises,<br />

events, etc. Additionally, there are systems that use extra information<br />

from video streams or data captured from dedicated sensors<br />

to improve their efficiency in human-machine interaction<br />

and interactive multimodal tasks [1]. However, the final effectiveness<br />

is still highly dependent on audio processing stage.<br />

The processing flow involve two basic steps: segmentation<br />

and classification [2–3]. In the segmentation phase input<br />

signal is divided into regions, where important audio information<br />

from interaction point of view can be extracted. Then, for<br />

every region audio classification is performed. Due to high<br />

sensitivity on the variable external audio conditions, selection<br />

of features with good separability properties for a given<br />

set of classes is crucial [4–5]. Nevertheless in many systems<br />

a feature set is selected arbitrary [6]. Typical audio classification<br />

phase works in the following manner: first, input signal<br />

is divided into short frames, then for each frame a set of audio<br />

features is computed making feature vector. Every vector<br />

is labeled with signal class and then used in learning phase<br />

where database is created. Finally, obtained database is exploited<br />

in the classification process [7]. The main obstacle in<br />

such approach is the impossibility of features modification (for<br />

adaptation to changing conditions) without repeating learning<br />

phase. Therefore, we propose approach based on properties<br />

of feature contours. We decided to use the dynamic properties<br />

of peaks occurrences in feature contours as a basic measure<br />

for audio classification.<br />

Simultaneously, a set of feature contours are composed of<br />

extracted frames. Then peaks are calculated (1) using adaptive<br />

thresholding function (2) – an example of this procedure<br />

is shown in Fig. 2. Threshold is obtained by smoothing the<br />

feature contour in defined time window, where: x (n) – feature<br />

contour values, y (n, p) – generated peaks, k (n, p) – adaptive<br />

threshold function, p – window size [frames], β – scaling<br />

factor, N – number of frames, s a<br />

(n, p) = max (n – p/ 2,0),<br />

s b<br />

(n, p) = min(n + p/ 2, N – 1), n = 0,..., N – 1.<br />

⎧ 1, if x(n) > x(n -1) ∧ x(n) > x(n + 1) ∧ x(n) > k(n, p);<br />

y( n,<br />

p)<br />

= ⎨<br />

(1)<br />

⎩0,<br />

otherwise.<br />

sb<br />

β<br />

k ( n,<br />

p)<br />

=<br />

∑ x(<br />

i)<br />

(2)<br />

s ( n,<br />

p)<br />

− s ( n,<br />

p)<br />

b<br />

i=<br />

sa<br />

In the next step, the set of time offsets for each peak is determined<br />

H = 〈t 1<br />

, t 2<br />

,..., t R–1<br />

〉, where: t i<br />

< t i + 1<br />

, – number of peaks Ĥ ,<br />

and the distances between peaks are evaluated (3). Finally,<br />

the mean, variance, minimum and maximum of Ĥ are calculated<br />

and can be used in the classification stage.<br />

Hˆ<br />

t − t , t − t , K,<br />

t − t , K t t (3)<br />

=<br />

2 1 3 4 i+ 1 i<br />

,<br />

R−1<br />

−<br />

R−2<br />

a<br />

Analysis of feature contours<br />

An architecture of audio classification scheme utilizing feature<br />

contours is depicted in Fig. 1. Input audio signal is divided into<br />

short-time frames which are used in region detection module<br />

utilizing BIC (Beyesian Information Criterion) segmentation<br />

scheme [8].<br />

Fig. 2. Example of peaks detection: feature contour with calculated<br />

threshold (upper figure) and obtained peaks (lower<br />

figure)<br />

Rys. 2. Przykład detekcji pików: kontur cechy wraz z obliczonym<br />

progiem decyzyjnym (górny rysunek) oraz wynikowe<br />

piki (dolny rysunek)<br />

Fig. 1. Audio classification scheme<br />

Rys. 1. System klasyfikacji akustycznej<br />

134<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Experimental evaluation<br />

In order to verify proposed approach we have generated a test<br />

set containing W = 100 audio files, each 30 seconds long,<br />

for three binary classification scenarios: speech/music, speech/background<br />

and music/background. At the feature contours<br />

extraction stage (frame length was 30 ms, frame overlap<br />

50%), we used 220 features for 300 files (with 2000 frames<br />

per file). Then, for obtained contours, peak distances with<br />

β = 1, p = 120 was calculated.<br />

To compare scenarios we used the absolute difference<br />

between variance of peak distances in class 1 (σ 1<br />

) and class<br />

2 (σ 2<br />

) as separability measure (ψ). The obtained results for<br />

each scenario are shown in Tab. 1 (is the number of features<br />

with highest ψ value for all files in test set). Results in Tab. 2<br />

show the first ten of the most prominent features for each case:<br />

AEZR – autocorrelation energy to zero-crossing rate, DFC<br />

– dominant frequency component [6], LAT – log attack time<br />

[5], MFSCx – Mel-frequency spectral coefficients [9], NACFPI<br />

– peak index of the normalized autocorrelation function [10],<br />

PARCORx – partial autocorrelation coefficients [11], SBPx<br />

– spectral band energy, SSLOPE – spectral slope [5], STILT<br />

– spectral tilt [11], TIRR – time domain irregularity [12].<br />

Tabl. 1. Overall characteristic of class separability for the test set<br />

Tab. 1. Ogólna charakterystyka rozdzielności klas dla zbioru testowego<br />

Taking the results from Tabl. 1 and Tabl. 2 into account,<br />

a binary classifier for the task [6] of discriminating between<br />

speech and music (first scenario) was proposed. As can be<br />

seen in Tabl. 2, the feature with highest number of instances<br />

for whole test set is SBP2. This feature denotes an energy in<br />

the “second“ band (Fig. 3, where N = 22050 Hz). Simple classification<br />

by comparison σ 1<br />

and σ 2<br />

values leads to 79% accuracy.<br />

Since this value is not satisfactory, an additional parameter<br />

had to be chosen. In order to minimize the features extraction<br />

overhead, first three SBPx features were selected because utilizing<br />

max (Ĥ (c)<br />

SBP1,2,3<br />

) value improve classification ratio. In fact,<br />

above analysis has to be performed in fully automated manner<br />

involving heuristic algorithm. Based on the preliminary results,<br />

the decision function was defined as sum of comparisons between<br />

two classes for variances and maxima:<br />

Ф = sgn(σ 1<br />

– σ 2<br />

) + sgn(m 1<br />

– m 2<br />

), where: c – class number<br />

(1 – speech, 2 – music), σ c<br />

– variance of Ĥ (c) , m =<br />

SBP 2 c<br />

max (Ĥ (c)<br />

(c)<br />

(c)<br />

) + max (Ĥ ) + max (Ĥ ).<br />

SBP1 SBP2 SBP3<br />

The variances σ 1<br />

(speech), σ 2<br />

(music) and maxima m 1<br />

(speech),<br />

m 2<br />

(music) for all regions are depicted in Fig. 4. The final<br />

decision whether audio region contains speech or music is<br />

made by checking the value of Ф: if Ф ≥ 0 then region contains<br />

speech signal. As the result 98% accuracy was obtained.<br />

Scenario F<br />

W<br />

W<br />

( w)<br />

( w)<br />

min σ1<br />

− σ<br />

( w)<br />

( w)<br />

2 max σ1<br />

− σ 2<br />

w = 1<br />

w = 1<br />

Speech / Music 35 376.11 29257.9<br />

Speech / Background 38 448.14 15799<br />

Music / Background 42 82.434 131698<br />

Tabl. 2. The first ten of the most frequent features for each scenario<br />

Tab. 2. Dziesięć pierwszych z najczęściej występujących cech dla<br />

każdego scenariusza<br />

Speech / Music Speech / Background Music / Background<br />

Feature Instances Feature Instances Feature Instances<br />

SBP2 11 SBP2 9 STILT 6<br />

MFSC38 8 STILT 8 TIRR 6<br />

STILT 8 MFSC9 7 NACFPI 6<br />

MFSC40 7 MFSC39 6 LAT 6<br />

MFSC39 7 MFSC11 6 MFSC12 5<br />

MFSC11 6 MFSC12 5 AEZR 5<br />

MFSC37 5 LAT 5 SSLOPE 4<br />

AEZR 4 MFSC40 4 PARCOR1 4<br />

MFSC10 4 MFSC36 4 SBP2 4<br />

MFSC9 4 MFSC8 4 DFC 3<br />

Fig. 3. Frequency bands defined for SBPx features (N is the half<br />

of sampling frequency)<br />

Rys. 3. Pasma częstotliwości cech SBPx (N jest równe połowie<br />

częstotliwości próbkowania)<br />

Fig. 4. Variances σ 1<br />

, σ 2<br />

(upper figure) and sum of maxima m 1<br />

, m 2<br />

(lower figure) for the whole test set<br />

Rys 4. Wariancje σ 1<br />

, σ 2<br />

(górny rysunek) oraz sumy maksimów m 1<br />

,<br />

m 2<br />

(dolny rysunek) dla całego zbioru testowego<br />

Conclusions<br />

In this paper, we have presented an approach for classification<br />

of audio regions. The distances between peaks calculated<br />

from various feature contours seems to suffice for binary<br />

classification tasks. Properties like mean, variance, minimum,<br />

and maximum of these distances are used to determine the<br />

discrimination function. We have performed exhaustive analysis<br />

of statistical properties for large set of low-level audio<br />

features and comparison of the differences between two<br />

classes in three scenarios. In the result, we found the most<br />

distinctive features for our audio test set. The classification<br />

attributes was obtained by choice and analysis of feature<br />

with the highest separability measure. The final configuration<br />

of classifier utilize feature contours needed to calculate attributes<br />

and determine the discrimination function. We have tested<br />

proposed configuration for speech/music discrimination task<br />

on the set containing audio files with two classes. Exploiting<br />

proposed approach in our experiments 98% accuracy was obtained.<br />

Proposed technique can be used for classifying complex<br />

audio scenes by using more sophisticated parameters<br />

fusions defined in a heuristic way. Despite the need to specify<br />

decision function by analysis of feature contours, presented<br />

approach may leads to low-complexity classifier designs.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong> 135


References<br />

[1] Thiran J., Marques F., Bourlard H.: Multimodal Signal Processing<br />

– Theory and Applications for Human-Computer Interaction.<br />

Academic Press, <strong>2010</strong>.<br />

[2] Hanna P., et. al.: Audio Features for Noisy Sound Segmentation.<br />

ISMIR’04 Conference, Barcelona, Spain, 2004.<br />

[3] Aggarwal C.: On classification and segmentation of massive audio<br />

data streams. Knowledge Information Systems, 20, Springer-<br />

Verlag, 2009, pp. 137–156.<br />

[4] Pachet F., Roy P.: Exploring Billions of Audio Features, CBMI’07,<br />

Bordeaux, 2007.<br />

[5] Kim H., Moreau N., Sikora T.: MPEG-7 Audio and Beyond. John<br />

Wiley & Sons, Ltd., 2005.<br />

[6] Al-Shoshan A.: Speech and Music Classification and Separation:<br />

A Review, J. King Saud Univ., Eng. Sci., vol. 19 (1), Riyadh,<br />

2006, pp. 95–133.<br />

[7] Rabaoui A., et. al.: Using Robust Features with Multi-Class SVMs<br />

to Classify Noisy Sounds. ISCCSP’08, Malta, 2008.<br />

[8] Zibert J., Brodnik A., Michelic F.: An adaptive BIC approach for<br />

robust audio segmentation. Interspeech, Brighton, 2009.<br />

[9] Schutte K., Glass J.: Features and Classifiers for Robust Automatic<br />

Speech Recognition, Research Abstracts, Research<br />

Project. MIT CSAIL Publications and digital archives, 2007.<br />

[10] Tian Y., Wang Z., Lu D.: Nonspeech segment rejection based on<br />

prosodic information for robust speech recognition. IEEE Signal<br />

Processing Letters, vol. 9, issue 11, 2002, pp. 364–367.<br />

[11] Kondoz A.: Digital Speech – Coding for Low Bit Rate Communication<br />

Systems. John Wiley & Sons, Ltd., 2004.<br />

[12] Krimphoff J., McAdams S., Winsberg S.: Characterization of the<br />

timbre of complex sounds. 2. Acoustic analysis and psychophysical<br />

quantification. J. de Physique, 4(C5), 1994, pp. 625–628.<br />

Nagroda Siemensa<br />

W październiku br. minęło 15 lat od zawarcia porozumienia pomiędzy<br />

Politechniką Warszawską a firmą Siemens Sp. z o.o.,<br />

na mocy którego ustanowiona została Nagroda Siemensa dla<br />

polskich naukowców. Porozumienie podpisali ówczesny Rektor<br />

Politechniki Warszawskiej prof. Stanisław Mańkowski oraz<br />

prezes zarządu Siemens Sp. z o.o. Peter Baudrexl.<br />

Porozumienie ustanowiło coroczną Nagrodę, która służy<br />

promowaniu wybitnych osiągnięć w badaniach prowadzonych<br />

przez pracowników polskich instytucji akademickich<br />

i pozaakademickich. Fundatorem Nagrody, w łącznej wysokości<br />

70 tysięcy złotych, jest firma Siemens Sp. z o.o.<br />

Zgodnie z regulaminem ubiegać się o nią mogą autorzy prac<br />

naukowych z dziedziny elektrotechniki i energetyki, elektroniki,<br />

telekomunikacji i informatyki, automatyki, transportu szynowego,<br />

inżynierii biomedycznej, a także z zakresu inżynierii środowiska,<br />

budowy maszyn i inżynierii materiałowej. Decydującym<br />

kryterium przy nagradzaniu zgłaszanych opracowań naukowych<br />

jest możliwość ich praktycznego zastosowania. Nagroda jest<br />

przyznawana raz w roku naukowcowi lub zespołowi z polskich<br />

uczelni lub naukowych instytucji pozaakademickich.<br />

Dla oceny wniosków i wyłonienia laureata Nagrody powołane<br />

zostało jury, złożone z ośmiu przedstawicieli środowisk<br />

politechnicznych i przemysłu oraz przedstawiciela fundatora<br />

Nagrody.<br />

Wśród laureatów dotychczasowych Nagród, we wszystkich<br />

ocenianych kategoriach, znaleźli się przedstawiciele<br />

polskiego świata nauki z wielu ośrodków akademickich na terenie<br />

kraju. (cr)<br />

Polacy zasłużeni dla elektryki. Początki elektrotechnicznego<br />

szkolnictwa wyższego, pionierzy elektryki. Praca<br />

zbiorowa pod redakcją Jerzego Hickiewicza, PTETiS,<br />

Warszawa – Gliwice – Opole 2009<br />

136<br />

Staraniem Polskiego Towarzystwa<br />

Elektrotechniki Teoretycznej<br />

i Stosowanej ukazała się książka<br />

poświęcona zagadnieniom kultury<br />

materialnej Polski w XX w. Jest<br />

to rodzaj leksykonu polskich inżynierów<br />

elektryków od początku<br />

istnienia tej dyscypliny do czasów<br />

najnowszych, który podsumowuje<br />

osiągnięcia polskiej elektryki<br />

w XX w. – wieku elektryczności.<br />

Otwierają go rozdziały poświęcone początkom polskiego<br />

szkolnictwa elektrycznego na pierwszych polskich politechnikach<br />

– Lwowskiej i Warszawskiej. W kolejnych rozdziałach<br />

przedstawiono biogramy 87 najwybitniejszych polskich elektryków<br />

poczynając od Józefa Hermana Osińskiego – zakonnika,<br />

wykładowcy Colegium Nobilium, autora pierwszej polskiej<br />

książki o elektrotechnice „Sposób ubezpieczający życie<br />

i majątki od piorunów”. Przedstawiono też sylwetkę Michała<br />

Doliwo-Dobrowolskiego – jednego z najwybitniejszych elektrotechników<br />

w skali światowej, wynalazcy asynchronicznego<br />

silnika klatkowego, który zapoczątkował w technice epokę<br />

prądu zmiennego.<br />

Ponadto m.in. zawarto sylwetki następujących elektryków<br />

polskich:. Juliana Ochorowicza – prekursora transmisji<br />

obrazów na odległość, Ignacego Mościckiego – autora nowatorskiej<br />

metody pozyskiwania kwasu azotowego z powietrza,<br />

rektora Politechniki Lwowskiej, profesora Politechniki<br />

Warszawskiej, prezydenta Polski w latach 1926–1939, Mariana<br />

Lutosławskiego – projektanta mostu Poniatowskiego,<br />

urządzeń prądotwórczych dla elektrowni na Powiślu i hotelu<br />

Bristol w Warszawie, Alfonsa Hoffmanna – prof. Politechniki<br />

Gdańskiej, opiekuna Związku Harcerstwa Polskiego, Tadeusza<br />

Cholewickiego – pioniera zastosowania teorii macierzy do<br />

analizy obwodów elektrycznych i teorii układów nieliniowych,<br />

Jana Obrąpalskiego – prekursora teorii i praktyki napędów<br />

elektrycznych w górnictwie i hutnictwie. Zwraca uwagę brak<br />

wzmianki na temat tak wybitnych polskich elektryków jak Janusz<br />

Groszkowski, Adam Smoliński.<br />

Każdy z rozdziałów zamyka wykaz publikacji przedstawianych<br />

postaci, uzupełnianych często nazwiskami doktorantów<br />

oraz bibliografia ich dotycząca. Sponsorami wydawnictwa<br />

były następujące firmy: Elektrobudowa SA Katowice, Energotest-Energopomiar<br />

Gliwice, Energotest-Diagnostyka Brzezie<br />

k. Opola, Energoefekt Ruda Śląska.<br />

<strong>Elektronika</strong> 11/<strong>2010</strong>


Zaprenumeruj wiedz fachow<br />

2011<br />

WWW.SIGMA-NOT.PL<br />

Nasze czasopisma według branż<br />

Ceny wersji podstawowej (brutto) prenumeraty rocznej na 2011 r. Prenumerata w pakiecie (z płytą CD)– należy doliczyć 24,40 zł do ceny podstawowej prenumeraty<br />

Przemysł<br />

Spożywczy<br />

Budownictwo<br />

<strong>Elektronika</strong>,<br />

Energetyka,<br />

Elektrotechnika<br />

Hutnictwo,<br />

Górnictwo<br />

Czasopisma<br />

Ogólnotechniczne<br />

Czasopisma<br />

Wielobranżowe<br />

Chłodnictwo<br />

(miesięcznik)<br />

252,00 zł<br />

Ciepłownictwo,<br />

Ogrzewnictwo,<br />

Wentylacja<br />

(miesięcznik)<br />

258,30 zł<br />

<strong>Elektronika</strong> – Konstrukcje,<br />

Technologie,<br />

Zastosowania<br />

(miesięcznik)<br />

302,40 zł<br />

Hutnik<br />

+ Wiadomości Hutnicze<br />

(miesięcznik)<br />

277,20 zł<br />

Atest<br />

– Ochrona Pracy<br />

(miesięcznik)<br />

239,40 zł<br />

Aura<br />

– Ochrona Środowiska<br />

(miesięcznik)<br />

176,40 zł<br />

Gazeta Cukrownicza (dwumiesięcznik)<br />

214,20 zł<br />

Gaz, Woda<br />

i Technika Sanitarna<br />

(miesięcznik)<br />

258,30 zł<br />

Przegląd<br />

Elektrotechniczny<br />

(miesięcznik)<br />

315,00 zł<br />

Inżynieria Materiałowa<br />

(dwumiesięcznik)<br />

270,90 zł<br />

Maszyny, Technologie,<br />

Materiały – Technika<br />

zagraniczna<br />

(dwumiesięcznik)<br />

88,20 zł<br />

Dozór Techniczny (dwumiesięcznik)<br />

154,35 zł<br />

Gospodarka Mięsna<br />

(miesięcznik)<br />

252,00 zł<br />

Przegląd<br />

Gastronomiczny<br />

(miesięcznik)<br />

201,60 zł<br />

Przegląd Piekarski<br />

i Cukierniczy<br />

(miesięcznik)<br />

187,92 zł<br />

Przegląd<br />

Zbożowo-Młynarski<br />

(miesięcznik)<br />

302,40 zł<br />

Przemysł Spożywczy<br />

(miesięcznik)<br />

258,30 zł<br />

Przemysł Fermentacyjny<br />

i Owocowo-Warzywny<br />

(miesięcznik)<br />

258,30 zł<br />

Materiały Budowlane<br />

(miesięcznik)<br />

245,70 zł<br />

Przegląd Geodezyjny<br />

(miesięcznik)<br />

264,60 zł<br />

Szkło i Ceramika<br />

(dwumiesięcznik)<br />

129,15 zł<br />

Wokół Płytek<br />

Ceramicznych<br />

(kwartalnik)<br />

71,40 zł<br />

Przegląd Telekomunikacyjny<br />

+ Wiadomości<br />

Telekomunikacyjne<br />

(miesięcznik)<br />

277,20 zł<br />

Wiadomości<br />

Elektrotechniczne<br />

(miesięcznik)<br />

315,00 zł<br />

Przemysł<br />

Lekki<br />

Przegląd Włókienniczy – Włókno,<br />

Odzież, Skóra (miesięcznik)<br />

315,00 zł<br />

Rudy i Metale<br />

Nieżelazne<br />

(miesięcznik)<br />

277,20 zł<br />

Przemysł<br />

Pozostały<br />

Gospodarka Wodna<br />

(miesięcznik)<br />

315,00 zł<br />

Przegląd Papierniczy<br />

(miesięcznik)<br />

245,70 zł<br />

Przemysł Chemiczny<br />

(miesięcznik)<br />

478,80 zł<br />

Problemy Jakości<br />

(miesięcznik)<br />

308,70 zł<br />

Przegląd Techniczny<br />

(dwutygodnik)<br />

270,27 zł<br />

Zakład Kolportażu Wydawnictwa SIGMA-NOT – ul. Ku Wiśle 7, 00-707 Warszawa,<br />

tel. (22) 840 35 89, 840 30 86, faks (22) 891 13 74, e-mail: kolportaz@sigma-not.pl<br />

Ochrona Przed Korozją<br />

(miesięcznik)<br />

422,10 zł<br />

Opakowanie<br />

(miesięcznik)<br />

226,80 zł<br />

Dla prenumeratorów<br />

– bezpłatny dostęp do publikacji<br />

i możliwość umieszczenia<br />

wizytówki firmy w Portalu<br />

www.sigma-not.pl

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!