RF SWITCHING OPTIONS: - ElectronicsAndBooks
RF SWITCHING OPTIONS: - ElectronicsAndBooks
RF SWITCHING OPTIONS: - ElectronicsAndBooks
Transform your PDFs into Flipbooks and boost your revenue!
Leverage SEO-optimized Flipbooks, powerful backlinks, and multimedia content to professionally showcase your products and significantly increase your reach.
“How can I tell if a power supply is reliable?”There’s an indicator on the front.It says “Agilent.” With a typical MTBF of 40,000 hours, over half-acenturyof experience, and with more than 250 models to choosefrom, Agilent’s power supplies are the ones you can count on. In factthe array of our power supplies is so extensive, it wouldn’t fit on thispage. For clean, low-noise, programmable power to countless DUTs,there’s an Agilent power supply with your name on it. Actually, it’sour name on it, but you know what we mean.Take a brief quiz and save up to 10%.And enter to win an iPod touch.www.agilent.com/find/poweronquizAgilent Authorized Distributor© Agilent Technologies, Inc. 2009.Limited time offer. See Web site for full promotion details.Discount offer applies to 144 power supplies.866-925-7633 www.conres.com/test-equipment
Enter xx at www.edn.com/info
NI LabVIEW.LimitedOnlybyYourImagination.Perform real-time analysisfor thought-based controlCombine .m file-basedtextual math withgraphical programmingQuickly develop userinterfaces to plot anddisplay algorithm resultsEasily create a scalablesystem to acquirereal-world signalsInteractively adjustalgorithm parametersto quickly iterateon designsReal-Time and EmbeddedSignal ProcessingHigh-Performance TestIndustrial ControlPRODUCT PLATFORMNI Advanced SignalProcessing ToolkitNI Motion and Vision ModulesNI LabVIEW MathScriptNI CompactRIO EmbeddedHardware PlatformAmbient cofounders and research students from the Universityof Illinois chose NI LabVIEW software to create the first thoughtcontrolledwheelchair. Using LabVIEW – a graphical systemdesign environment ideal for signal synthesis, frequency analysis,and digital signal processing – they developed sophisticatedalgorithms that translate neurological signals into commands,empowering people with disabilities.>> Explore the thought-controlled interface at ni.com/imagine/sp 866 337 5041© 2007 National Instruments Corporation. All rights reserved. CompactRIO, LabVIEW, National Instruments, NI, and ni.com are trademarks of National Instruments.Other product and company names listed are trademarks or trade names of their respective companies. 2007-8589-821-101
9.17.09contents<strong>RF</strong> switching options:The right fit mightcome with a lossareoffering SOI and30ManufacturersMEMS alternativesto PIN-diode, GaAs, and electromechanicalswitches for a varietyof <strong>RF</strong> applications, but you needto understand <strong>RF</strong>-switch specsbefore you commit to a newtechnology.by Rick Nelson, Editor-in-Chiefpulse13 Osram develops direct-emittinggreen-laser diode13 1-GHz Cortex-A8 maintainscycle-accurate operation14 Versatile audio analyzer easilyquantifies performance ofaudio components, products14 Mentor unveils strategyat DACBreaking down thesensor-signal pathBy understanding the25stages of an analog-signalpath, digital-system developerscan more accurately capturesensor data for a variety of applications.Dilbert 14by Aaron GL Podbelski,Cypress Semiconductor15 Quad 12-bit DAC has internalreference and EEPROM15 Workshop addressesLED challenges16 Voices: Newark/PremierFarnell’s Gary Nevison:scoping out the EnvironmentalDesign of ElectricalEquipment ActIn search of a betterDRAM: evolvingto floating bodiesWidely investigated float-memories appear39ing-bodyto be compelling replacementsfor conventional DRAMs. A newfloating-body memory uses theintrinsic bipolar transistor to storesignificantly greater charge.by Serguei Okhonin, PhD,Innovative SiliconDESIGNIDEASR 61kR 31kCLOCKR 71kIC 1AHC14SIGNAL IND 4BZD23-4V71D 1D1N4148IC 1BHC14R 21k45 Missing pulse detects position or produces a delay47 Emulate SPI signals with a digital-I/O cardCLKRSTQ0Q1Q2IC 2 Q34024Q4Q5Q6232D 2D1N4148D 3D1N4148IC 4AHC08D SET QIC 3A4013QNRST48 Resistive DAC and op amp form hybrid divider49 Connect two buttons with just two wiresSEPTEMBER 17, 2009 | EDN 3
The NewestWireless SolutionsZigBee ®<strong>RF</strong>IDBluetooth GPSWLANCellularISMNew Products from:SG / SR GPS Receiver Modulesand Kitswww.mouser.com/linxsgsrZICM2410P0 MeshConnectModulewww.mouser.com/cel/aSocketModem® GPRSEmbedded Wireless Modemwww.mouser.com/multitechsystems/The ONLY New Catalog Every 90 DaysExperience Mouser’s time-to-marketadvantage with no minimums and same-dayshipping of the newest products from morethan 390 leading suppliers.MiniCore RCM5600W Wi-Fi Modulewww.mouser.com//rabbit_rcm5600/The Newest ProductsFor Your Newest Designswww.mouser.comOver A Million Products Online(800) 346-6873
EDN.COMMENT,,BY RICK NELSON, EDITOR-IN-CHIEFRobots, jobs, and warRobotics was a focus of attention at National Instruments’NIWeek event in Austin, TX, last month, when presentersdiscussed the technical capabilities and ethical considerationssurrounding robot use. Although the technical issuesgot most of the attention, it may be the ethical onesthat prove to be more difficult. Consider mining accidents.Six miners died tragically in the 2007 Crandall Canyon Mine disaster.The loss 10 days later of three would-be rescuers compounded thetragedy. Could the use of robots toperform the rescue reconnaissancehave averted those three deaths?At NIWeek, Thomas Bewley, a professorat the University of California—San Diego, described the challengesrobots can face. Those small enoughto access a collapsed mine tend to betoo small to climb over the debris theyencounter once inside. That problemis one Bewley and his students are addressingby finding ways to have smallrobots climb over large obstacles. Arobot should roll when possible, hesays, but use multifunction mechanisms,including plungers, for example,when necessary.It would clearly be ethical to haverobots search for survivors in collapsedmines, sparing rescue workersthe risk. If such robots can reconnoitercollapsed mines, however, they couldtake over mining itself.Mining isa dangerousjob, but is it betterthan no job at all? Inless dangerous occupations,is it ethicalto substitute robotsfor humans?In a recent article, Gregory Clark,a professor of economics at the Universityof California—Davis, doesn’tdiscuss the ethics of the situationbut rather the consequences of whathe takes to be the inevitable (Reference1). Clark writes that the currentdownturn is a minor blip in technology-driveneconomic growth, and, hecautions, “The economic problems ofthe future will not be about growthbut about … the ineluctable increasein the number of people with no marketableskills and technology’s rolenot as the antidote to social conflict,but as its instigator” as machines displacepeople.In a keynote address at NIWeek, DavidBarrett, PhD, director of SCOPE(Senior Capstone Program in Engineering)at Franklin W Olin Collegeof Engineering (Needham, MA), describedrobots that are or will be takingover human tasks, including mining,industry, construction, and agriculture.It won’t just be unskilled workers whomight have something to fear. Barrettalso described various medical robots,including ones that perform surgery.What should we do about robots’displacement of people? Clark picturesa dystopia: “We could imaginecities where entire neighborhoods arepopulated by people on state support.In France, generous welfare has alreadyproduced huge suburban housingestates, les banlieues, populatedwith a substantially unemployed andimmigrant population, parts of whichhave periodically burst into violentprotest.” To support such populations,he says, “you tax the winners—thosewith the still uniquely human skillsand those owning the capital andland—to provide for the losers.”Perhaps the most difficult problemof ethics centers on robots’ use inthe military. In another recent article(Reference 2), James Carroll writes,“When will the unempathetic Americansimagine what it feels like to havea robot monster bolt from the sky—thedrones of August—and, in one strike,turn a wedding feast into a funeral?”On the ethics of using robots in warfare,SCOPE’s Barrett said that, if wedon’t do it, our enemies will.It’s inevitable that robots will takeon more and more roles. The true ethicalquestion comes into play in howwe address the consequences, and sofar we have fallen short. We cannotcontinue turning weddings into funeralsand turning middle-class neighborhoodsinto violent banlieues of disaffected,unemployed losers meagerlysupported by taxing the winners.EDNREFERENCES1Clark, Gregory, “Tax and Spend,or Face The Consequences,” TheWashington Post, Aug 9, 2009,www.washingtonpost.com/wp-dyn/content/article/2009/08/07/AR2009080702043.html.2Carroll, James, “In search of empathy,”The Boston Globe, Aug 17,2009, www.boston.com/bostonglobe/editorial_opinion/oped/articles/2009/08/17/in_search_of_empathy.Contact me at rnelson@reedbusiness.com.6 EDN | SEPTEMBER 17, 2009
Plug in to our Power ToolsMake better, faster power inductor choiceswith Coilcraft’s powerful web tools.Start with an IC, a converter topology, ora list of electrical and mechanical specs. Inseconds you’ll get a list of every Coilcraft partthat could work for you, complete with priceinformation. A few more clicks gives you adetailed analysis of core and winding losses.Check out our complete power designer’stoolbox at www.coilcraft.com/powertools®RoHSCOMPLIANTwww.coilcraft.com 800/322-2645
Where power is hard to find,Tadiran has the solution.PROVEN20YEAROPERATINGLIFEDon’t let the cost and inconvenience of hard-wired ACpower send you up a tree. Instead, be as remoteas you want to be with the ONLY lithium cells tooffer PROVEN 20-year service life withoutrecharging or replacing the battery. Whenit comes to wireless powersolutions, Tadiran is takinginnovation to extremes.Tadiran Batteries2 Seaview Blvd.Port Washington,NY 110501-800-537-1368516-621-4980www.tadiranbat.com
PRESIDENT, BUSINESS MEDIA, REED BUSINESS INFORMATIONJeff DeBalko, jeff.debalko@reedbusiness.com1-646-746-6573PUBLISHER,EDN WORLDWIDERussell E Pratt, 1-781-734-8417;rpratt@reedbusiness.comASSOCIATE PUBLISHER,EDN WORLDWIDEJudy Hayes, 1-925-736-7617;judy.hayes@reedbusiness.comEDITOR-IN-CHIEF,EDN WORLDWIDERick Nelson, 1-781-734-8418;rnelson@reedbusiness.comEXECUTIVE EDITORRon Wilson, 1-510-744-1263;ronald.wilson@reedbusiness.comMANAGING EDITORAmy Norcross1-781-734-8436;fax: 1-720-356-9161;amy.norcross@reedbusiness.comContact for contributed technical articlesSENIOR ART DIRECTORMike O’Leary1-781-734-8307;fax: 1-303-265-3021;moleary@reedbusiness.comANALOGPaul Rako, Technical Editor1-408-745-1994;paul.rako@edn.comMASS STORAGE, MULTIMEDIA,PCs, AND PERIPHERALSBrian Dipert, Senior Technical Editor1-916-760-0159;fax: 1-303-265-3187;bdipert@edn.comMICROPROCESSORS, DSPs,AND TOOLSRobert Cravotta, Technical Editor1-661-296-5096;fax: 1-303-265-3116;rcravotta@edn.comNEWSSuzanne Deffree, Managing Editor1-631-266-3433;sdeffree@reedbusiness.comPOWER SOURCES,ONLINE INITIATIVESMargery Conner, Technical Editor1-805-461-8242;fax: 1-805-461-9640;mconner@reedbusiness.comDESIGN IDEAS EDITORMartin Rowe,Senior Technical Editor,Test & Measurement Worldedndesignideas@reedbusiness.comSENIOR ASSOCIATE EDITO<strong>RF</strong>rances T Granville1-781-734-8439;fax: 1-303-265-3131;f.granville@reedbusiness.comEDITORIAL/WEB PRODUCTIONDiane Malone, Manager1-781-734-8445; fax: 1-303-265-3024Steve Mahoney,Production/Editorial Coordinator1-781-734-8442; fax: 1-303-265-3198Melissa Annand,Newsletter/Editorial Coordinator1-781-734-8443; fax: 1-303-265-3279Adam Odoardi, Prepress Manager1-781-734-8325; fax: 1-303-265-3042CONSULTING EDITORJim Williams, Staff Scientist,Linear TechnologyCONTRIBUTING TECHNICAL EDITORSDan Strassberg,strassbergedn@att.netNicholas Cravotta,editor@nicholascravotta.comCOLUMNISTSHoward Johnson, PhD, Signal ConsultingBonnie Baker, Texas InstrumentsPallab Chatterjee, SiliconMapPRODUCTIONDorothy Buchholz,Group Production Director1-781-734-8329Joshua S Levin-Epstein,Production Manager1-781-734-8333; fax: 1-781-734-8096EDN EUROPEGraham Prophet, Editor, Reed PublishingThe Quadrant, Sutton, Surrey SM2 5AS+44 118 935 1650;fax: +44 118 935 1670;gprophet@reedbusiness.comEDN ASIALuke Rattigan, Chief Executive Officerluke.rattigan@rbi-asia.comKirtimaya Varma, Editor-in-Chiefkirti.varma@rbi-asia.comEDN CHINAWilliam Zhang,Publisher and Editorial Directorwmzhang@idg-rbi.com.cnJeff Lu, Executive Editorjefflu@rbichina.com.cnEDN JAPANKatsuya Watanabe, Publisherk.watanabe@reedbusiness.jpKen Amemoto, Editor-in-Chiefamemoto@reedbusiness.jpMAXIMUMRELIABILITYIn contact, stability and low-noise performanceMill-Max Mfg. Corp. spring-loaded connectors providesuperior reliability under the most rigorous environmentalconditions, offering: EDN. 225 Wyman St, Waltham, MA 02451. www.edn.com. Phone 1-781-734-8000.Address changes or subscription inquiries: phone 1-800-446-6551; fax: 1-303-470-4280; subsmail@reedbusiness.com. For a free subscription, go to www.getfreemag.com/edn. Reed Business Information, 8878 S Barrons Blvd, Highlands Ranch, CO 80129-2345. Include your mailing label.Stay in contact withMill-Max spring-loaded connectors.offerings and request a datasheetwww.mill-max.com/EDN596
Buck regulators with integrated MOS S FETs F simplify designs up to 25A/phaseEFFICIENCY (%)100EFFICIENCY VS. OUTPUT CURRENT90807060V OUT = 2.5VV OUT = 1.8VV OUT = 1.2V50SKIPV IN = 3.3VPWM400.1 1.010.0OUTPUT CURRENT (A)High efficiencyINPUT2.9V TO 5.5VPROGRAMMABLEINENV DDFREQSSMAX15038MAX15039Skip ModeLXOUTPGNDPWRGDOUTPUT1.8V, 4A/6A5V TO 12V INPUTMaxim’s PowerMind TM family of integrated buck regulators simplifies yourtoughest design challenges. These regulators provide high efficiency incompact, simple, cost-effective solutions that have the industry’s lowest EMI.HIGH VOLTAGEMAX15041*MAX17026*MAX86543A8A8A3.3V TO 5V INPUTMEDIUM VOLTAGE0.75AMAX1970 0.75AMAX1973MAX19511A2A2.5V TO 3.3V INPUTLOW VOLTAGEnewMAX15040*MAX8643MAX88334A3A3A3A1- to 8-phase buck regulator with 25A/phasein a 6mm x 6mm TQFNMAX15035MAX865515A25AMAX15021MAX150224A2A4A2ALDO CRTL1LDO CRTL2newMAX8855MAX86465A5A6AV IN4.5V TO 20V UP TO 200AV IN25ACOMP/OVPRS+CLKORS-V IN25ACOMP/OVPV INCLKIMAX8686MAX8686CSCS25AMAX86861 TO 8 PHASE25A/PHASEPowerMind is a trademark of Maxim Integrated Products, Inc.*Future product—contact factory for availability.newnewMAX8505MAX15038MAX1945MAX15039www.maxim-ic.com/buck-regulators3A4A6A6AMAX856610ACOMP/OVPCLKIMAX8686CSDIRECTwww.maxim-ic.com/shopwww.em.avnet.com/maximTMFor free samples or technical support, visit our website.Innovation Delivered is a trademark of Maxim Integrated Products, Inc. Maxim is a registered trademark of Maxim Integrated Products, Inc. © 2009 Maxim Integrated Products, Inc. All rights reserved.
pulseEDITED BY FRAN GRANVILLEINNOVATIONS & INNOVATORSOsram develops direct-emittinggreen-laser diodeThis lab prototype of Osram’s direct-emitting green-InGaNlaserdiode achieved an optical output of 50 mW in pulsedmodeoperation at room temperature.Samsung Electronics andIntrinsity have worked togetherto produce an ARM (www.arm.com) Cortex-A8 processor,Hummingbird, that can operateat 1 GHz in a 45-nm, lowpowerprocess and maintainthe same cycle-accurate andBoolean-equivalent operationas the original Cortex-A8RTL (register-transfer-level)specification. The companiesattained this clock rateOsram Opto Semiconductors announcedlab results for a direct-emitting green-InGaN (indium-gallium-nitride)-laserdiode. In pulsed-mode operation at room temperature,the laboratory prototype achieved anoptical output of 50 mW, with a threshold-currentdensity of approximately 9 kA/cm², emittinglight in true green—defined by the spectralrange of 515 to 535 nm—with a wavelength of515 nm. The current technology for green semiconductorlasers is to double the frequency of amaterial capable of lasing at 1060 nm to producea green laser at 531 nm.The highest output power for afrequency-doubling green laseris currently about 1.5W.Osram developed the greenlaserdiode in conjunctionwith the German Ministry forEducation and Research (www.bmbf.de/en) MOLAS researchproject, which involves technologiesfor ultracompact andmobile-laser-projection systems;green lasers also fi nduse in a range of medical andresearch applications.—by Margery ConnerOsram Opto Semiconductors,www.osram-os.com.1-GHz Cortex-A8 maintains cycle-accurate operationby applying custom-designedcircuit/memory structuresand Intrinsity’s enhanced RTLFastCore and Fast14 highspeeddomino logic to timingcriticalpaths in the Cortex-A8RTL core. The latest generationof Intrinsity’s NDL (oneof-N-domino-logic)implementationnow supports seamlessmixing between domino andstatic logic in a standard-cellsynthesizeddesign. This featureenables the core to useonly those NDL gates that are25 to 50% faster than staticlogicgates in those criticalpathcircuits in which theyaffect the overall clock rate.This mixing enables the systemto consume less staticand dynamic power than earlierimplementations of othercores that used the NDL gatesthroughout the entire core.The core includes 32-kbyteFEEDBACK LOOP“Einstein, Mozart,and every othergenius had contemporaries(andquite smart onesat that) whomthey were able tobounce ideas offof and refine theirbest work. … Verylittle would havebeen discoveredif that so-called‘genius’ werelocked in a cellby himself.”—Design engineer ChrisGammell, in EDN’s FeedbackLoop, at www.edn.com/article/CA6666234. Add your comments.instruction and data caches,512 kbytes of L2 memory,and an ARM Neon multimediaextension. The core supportsmultiple drain-to-drain-voltageand frequency operation thatincludes a minimum supplyvoltage of 1V so that the corecan target mobile devices.—by Robert CravottaIntrinsity, www.intrinsity.com.Samsung Electronics,www.samsung.com.SEPTEMBER 17, 2009 | EDN 13
pulseVersatile audio analyzer easily quantifiesperformance of audio components, productsDILBERT By Scott AdamsAgilent Technologieshas introduced theU8903A audio analyzer,a single unit that incorporatesthe broad range of capabilitiesyou need to quantifythe characteristics that affectsound quality in audio devices.The instrument is also themanufacturer’s next-generationreplacement for its widelyused legacy audio analyzer,the 8903B. The new, scalable,two-channel unit includesseveral measurement functionsand test signals, powerfulanalysis functions, andindustry-standard balancedand single-ended connectors.At dc and from 10 Hz to 100kHz, the analyzer helps you tomeasure performance in suchapplications as consumer andwireless audio and analogcomponentand -IC test. In addition,the manufacturer plansto soon offer upgrades that willequip the instrument with asmany as eight input channels.The U8903A offers increasedcapabilities, wider frequencycoverage, and greaterdynamic range than the8903B. The new model alsoincludes a graphical user interfacewith a 5.7-in. color displayand one-button selectionof major operating modes. Forautomated operation, an applicationnote describes equivalentsoftware commands andprovides sample test programsthat substitute the new unit’sSCPI (standard commands forprogrammable instruments) instructionsfor the older instrument’scommands.The U8903A is availablenow at a US list price of$12,000. For a limited time,You canpurchasea two-channelanalyzer forthe price ofa typical singlechannelunit.the manufacturer is offeringthe new instrument at a 20%discount when you trade inan 8903B. With this arrangement,you can purchase a twochannelanalyzer for the priceof a typical single-channelunit. Find further informationabout the discount program atwww.agilent.com/find/audioanalyzerpromo.—by Dan StrassbergAgilent Technologies,www.agilent.com/find/audioanalyzer.The compact U8903A audio analyzer integrates a range of single-ended and differential source andmeasurement capabilities that enable you to perform extensive testing of components and systemsat dc and at frequencies from 10 Hz to 100 kHz.MENTOR UNVEILSSTRATEGY AT DACMentor Graphics (www.mentor.com) announcedits acquisition of EmbeddedAlley Solutions as akey component of its Androidand embedded-Linuxstrategy last month at theDesign Automation Conferencein San Francisco.Mentor also announced theintegration of its Nucleusgraphical-user-interfacetool with the ARM (www.arm.com) Mali graphics-processingunit; it announcedthe availability ofa Linux/Nucleus operatingsystemcombination forthe Marvell (www.marvell.com) Sheeva MV78200 dual-coreembedded processor;and it plans to extendEmbedded Alley’s Androidmobile-applications platformto support FreescaleSemiconductor’s (www.freescale.com) QorIQ andPowerQUICC (quad-integrated-communicationscontroller)III processors.Mentor plans to combineits Nucleus RTOS (real-timeoperating system)and associated tools andservices with EmbeddedAlley’s Android and Linuxdevelopment systems tooffer device manufacturersone source for theoperating systems theyneed for embedded-systemdesigns. The goal isto support Mentor’s customersin supplying completesystems—not justsilicon—to their own customers.For more news onDAC, go to www.edn.com/090917pulsea.—by Rick NelsonDesign AutomationConference, www.dac.com.09.17.0914 EDN | SEPTEMBER 17, 2009
Quad 12-bit DAC hasinternal reference and EEPROMMicrochip Technologyhas announceda quad 12-bit DACthat remembers its settingsin an internal EEPROM. TheMCP4728 has a 2%, 2.048Vvoltage reference, but you canalso use an external reference.The reference has a 1.2-V/Hz noise floor with a 400-Hzflicker-noise corner. You communicatewith the part over anI 2 C (inter-integrated-circuit)bus in standard 100-kbps, fast400-kbps, and high-speed3.4-Mbps modes.You can individually shutdown each DAC; total shutdowncurrent is 0.04 A. Youcan set the outputs to go to alow-, medium-, or high-impedancestate during shutdown.Operating current is 800 A.The devices features rail-torailoutputs over the 2.7 to 5.5Vpower-supply range. TypicalDNL (differential nonlinearity)V DDV SSSDASCLRDY/BSYI 2 CINTE<strong>RF</strong>ACELOGIC2.048VINTERNALREFERENCEVOLTAGEThe MPC4728EV evaluationboard interfaces with the $49PICkit serial analyzer, whichconverts your computer’s USBbus to I 2 C, SMBus, SPI, orUSART protocols. You canalso use your own I 2 C interfaceboard.is 0.2 LSB with a maximum of0.75 LSB, which ensures thatthe device remains monotonicacross all input codes. The DACEEPROM AINPUTREGISTER AEEPROM DINPUTREGISTER DLDACOUTPUTREGISTER AOUTPUTREGISTER DV REF SELECTORV REF AV REF DV REFGAINCONTROLSTRINGDAC AGAINCONTROLSTRINGDAC Dcontains a power-on circuit forpredictable operation.The device has applicationsin consumer products,such as personal media players,digital cameras, and GPS(global-positioning-system)devices. Medical applicationsinclude portable glucose meters,blood-pressure monitors,and heart-rate monitors. It alsofinds use in industrial products,such as handheld instruments,motor-control applications, andtemperature and light control.The unit’s wide temperaturerange also makes it suitable forautomotive applications.The MCP4728EV evaluationboard is available nowand costs $15. The $1.36(10,000) MPC4728 operatesfrom 40 to 125C andcomes in a 10-pin MSOP.—by Paul RakoMicrochip Technology,www.microchip.com.OPAMP APOWER-DOWNCONTROLOPAMP DPOWER-DOWNCONTROLV OUTAV OUTB*V OUTC*V OUTDFEEDBACK LOOP“If people wantto play in theopen-hardwaresandbox, theyhave to sharetheir toys.”—Roboticist Zach HoekenSmith, in EDN’s FeedbackLoop, at www.edn.com/article/CA6676166. Add yourcomments.WORKSHOPADDRESSES LEDCHALLENGESThe US Department ofEnergy recently invested$6.4 million in LED designto encourage innovationin solid-state lighting (seewww.edn.com/article/CA6685811). But solid-statelighting is as much aboutthe surrounding electronicpower-controlcircuitry,thermal management, andoptics as it is about thebasic LED devices. EDN ishosting a free “Designingwith LEDs” workshop onOct 6 at the Westin Lombardhotel near Chicago.The workshop will focuson the electronic-circuit,thermal, and optical-designchallenges of designingwith solid-state lighting, aswell as other LED-basedlighting applications. Youcan register at the Web sitebelow.—by Margery Connerwww.edn.com/leds.09.17.09V DD*CHANNELS V OUT B AND V OUT C NOT SHOWN.You control the Microchip MPC4728 quad DAC with the I 2 C bus. The device’s nonvolatile memoryallows it to power up without microprocessor supervision.In October, EDN will host aworkshop on designing withLEDs.SEPTEMBER 17, 2009 | EDN 15
pulseVOICESNEWARK/PREMIER FARNELL’SGARY NEVISON:scoping out theEnvironmental Design ofElectrical Equipment ActAs a proposed amendment to the 1976 Toxic SubstancesControl Act, the Environmental Design of ElectricalEquipment Act (Bill HR2420) has stirred much controversywithin the electronics supply chain about whether a USversion of the EU (European Union) ROHS (restriction-of-hazardous-substances)directive is coming. Gary Nevison, legislationand environmental-affairs manager at global Premier Farnell(www.farnell.com) and its US business Newark, recently discussedthe bill and compared it with ROHS. The following textincludes excerpts from that conversation.Is the Environmental Designof Electrical EquipmentAct (HR2420), along with theUS Toxic Substances ControlAct, similar to a US versionof ROHS for electronicsdesigners?First, this new legislationis clearly not ROHS.AIt would be misleading to talkabout US ROHS compliance.The proposed HR2420 is differentfrom EU ROHS, and,although it is claimed to belegislation designed to controlhazardous substances, its mainaim appears to be preventingstates from imposing substancerestrictions on productswithin the scope of this legislation.It would introduce somelimited restrictions, but the exemptionslist is so comprehensivethat these [restrictions]would be few.So, this act is not the USROHS that many membersof the electronics supplychain have been waiting for?Initially, this [act] mightA appear to be so; however,closer scrutiny of the scopeand structure of HR2420 revealsthat it is … an effort tolimit uncoordinated piecemeallegislation on this issue by individualstates. At present, thereis no federal US equivalentto EU ROHS, although somestates have introduced limitedROHS-like laws. Many manufacturerswould welcome asingle uniform US ROHS law,as this [law] would remove theneed to meet multiple, changingrequirements across USstates, although others, whichpredominantly sell internally,may be concerned about thisextra restriction.The main implications ofHR2420 are for new chemicalsor major new uses of existingones. For example, productscontaining substances that arenot included in the TSCA [ToxicSubstances Control Act] registerrequire certification beforethey can be imported. TSCAcurrently imposes very few restrictionson substances, but itdoes affect lead-based paints,asbestos, and polychlorinatedbiphenyls.Please elaborate on howHR2420 differs from ROHS.Although the bill proposesto restrict the sameAsix substances as EU ROHSat the same concentration valuesin homogeneous materials,there are no apparent equivalentsto 17 EU ROHS exemptions,and there is no use of theEU ROHS product categories.The product scope appearsquite different from EU ROHS,as it seems to exclude householdor consumer products andinclude products not coveredby EU ROHS, such as electricity-distributionequipment. Anothersignificant difference isthat EU ROHS excludes productsdesigned for use withvoltages above 1000V ac or1500V dc, whereas HR2420has a limit of 300V. The restrictionswould apply to productsmanufactured after July 1,2010.So, in summary, the scopeof the bill is quite limited andis clearly different from EUROHS and mainly consistsof a detailed list of exclusionsand exemptions. As the scopeincludes many items currentlyexcluded from EU ROHS,HR2420 does not appear tobe a federal ROHS bill—morean attempt to avoid disjointedROHS requirements emergingfor products currently outsideor on the fringe of the scopeof EU ROHS.Many engineers think that,once the original ROHShubbub concluded, all ofthe environmental complianceissues were behind us.Why is it important for engineersand OEMs to continueto work with distributors onenvironmental complianceissues?That [ending of controversy]was never theAcase. Under the original ROHSscope, it was always going tobe reviewed, there were alwaysgoing to be more substances,product categories, etcetera.And now we have the REACH[registration/evaluation/authorization-of-chemicals]directive.There will be more andmore and more products andsubstance [limitations] in theshort, medium, and long term,[which] will probably be the biggestissue for design engineersaround the world. They will haveto seek suitable alternatives tothese ever-increasing [numberof] substances. This [situation]is kind of a nightmare. It is goingto roll and roll and roll formany years.Sure, engineers could lookat all of this on the Web, butthere is so much informationand so much misinformationthat is now dated that a distributoris a good route to take.There are good distributors doingthis [work] and there are notso good, but one of the biggestissues for me is whether thingsever get deleted from the Web.For the designer, the simplereason would be getting thelatest information, providing thedistributor is doing the job well.—interview conducted andedited by Suzanne Deffree16 EDN | SEPTEMBER 17, 2009
RAQ’sS PECIAL ADVERTISING SECTIONRarely Asked QuestionsStrange stories from the call logs of Analog DevicesWhat’s the (Converter) Frequency Kenneth?Q. How Do I Design a ConverterFront-end without Compromisingthe Performance?A. Designers that employ a converterfor high-frequency sampling have to facemany challenges. Designing a front-endisn’t simple, but the following commentscan guide the designer to a solution.Designers can choose amongst threetypes of front-ends: baseband, narrowband,or wideband; the application determineswhich should be applied.Baseband applications require bandwidthfrom dc or the low MHz to the Nyquistfrequency of the converter. In terms ofrelative bandwidth, this implies about 100MHz or less. These designs can employeither an amplifier or a transformer (balun).Narrowband applications (narrow beingrelative to the ADC’s full Nyquist bandwidth)usually operate at high intermediatefrequencies (IF). They typically useonly 5 to 20 MHz of bandwidth in the 2ndor 3rd Nyquist zone, with a center frequency>190 MHz. The design only needsa portion of the Nyquist bandwidth, butthe unused bandwidth is often needed toimplement an anti aliasing filter. A transformeror balun is typically used for theseapplications, but an amplifier can be usedif its performance is adequate at thesefrequencies.Wideband designs need it all, with theuser taking as much as the converterwill supply. These designs have thewidest bandwidth, making the frontenddesign the most challenging ofthe three types. These applicationsrequire bandwidth from dc or low MHzto several GHz. Currently, these designstypically employ a wideband balun, butamplifiers are catching up in bandwidthand performance.After choosing the converter, choosethe front-end amplifier (active) or transformer(passive). The tradeoffs betweenthe two are many and depend on theapplication, but can be distilled to a fewpoints. Amplifiers add noise, require apower supply, and burn power, but theyare not gain bandwidth dependent like atransformer. Also, they have better gainflatness within the pass band region.Transformers are passive, so they don’tadd noise or burn power, but their asymmetricalbehavior can cause spuriousissues. Transformers are not ideal devices;if not used properly their parasitics canundermine any design, particularly athigher frequencies (>100 MHz).Hopefully, this advice will keep thedesign on track. For additional information,please refer to the references orsend me an email.To Learn More AboutFront-end Designhttp://designnews.hotims.com/23118-101Contributing WriterRob Reeder is a seniorconverter applicationsengineer working inAnalog Devices highspeedconverter groupin Greensboro, NC since1998. Rob receivedhis MSEE and BSEEfrom Northern IllinoisUniversity in DeKalb,IL in 1998 and 1996respectively. In hisspare time he enjoysmixing music, art, andplaying basketball withhis two boys.Have a questioninvolving a perplexingor unusual analogproblem? Submityour question to:raq@reedbusiness.comFor Analog Devices’Technical Support,Call 800-AnalogDSPONSORED BYSEPTEMBER 17, 2009 | EDN 17
BAKER’S BEST,,BY BONNIE BAKERA designer’s guideto op-amp gain errorAs you sit down to select the proper operational amplifierfor your circuit, the first order of business is to determinethe signal bandwidth that your system will send throughthat amplifier. Once you settle on this parameter, youcan start to look for the right amplifier. The high-speedop-ampgurus warn that you should avoid using analogdevices that are too fast for your application. So you try to pick an amplifierwith a closed-loop bandwidth just a little higher than the maximumfrequency of your signal.This strategy may soundlike a good product-selec tionrecipe, but it will probablybring disaster to your applicationboard. In the lab, youmay find that, when you putan input-sine-wave signal atthe application’s maximumfrequency into your system,the output signal from youramplifier does not go acrossthe expected full-scale analogrange. The gain on thesignal is much less than youwould expect. If the slewratemagnitude of your amplifieris more than adequateand you are not driving theamplifier output into the power-supplyrails, then what has gone wrong?Stop double-checking your resistorvalues! When designing an amplifierinto a gain cell, you must know yoursignal’s maximum bandwidth, the amplifier’sclosed-loop noise gain, theamplifier’s gain-bandwidth product,and how much gain error your designcan tolerate. The closed-loop noisegain is the amplifier’s gain, as if a smallvoltage source were in series with theFigure 1 The open- and closed-loop gain of this voltage-feedbackamplifier has a gain-bandwidth product of 16 MHz and acircuit noise gain of 10V/V.op amp’s noninverting input.You can work this problem throughby example. For instance, start witha signal bandwidth of 1 MHz. Theamplifier’s circuit noise gain in Figure1 is 10V/V. Figure 1 also showsthe open-loop frequency responseof an amplifier that has just enoughbandwidth for this circuit—or so youthink. The amplifier has a 16-MHzgain-bandwidth product. The op amplooks as though it can support a gainof 10V/V, or 20 dB, out to 1 MHz, butlook a little closer. The gain of theopen-loop gain curve at the signal’sbandwidth isGBWPAOL SBW = ,SBWwhere A OLis the open-loop gain ofthe amplifier, SBW is the signal bandwidth,and GBWP is the gain-bandwidthproduct.In this case, the amplifier’s openloopgain, A OLSBW, is 16V/V at 1 MHz.But here’s the kicker: The closedloopgain error in this circuit is NG/(A OLSBWNG), where NG is thenoise gain. The closed-loop gain errorat 1 MHz in this example is 0.385, or again error of 38.5%.For this circuit, if you are willing totolerate a gain error of 0.05 from youramplifier and you understand that theGBWP of an amplifier can changea maximum of 30% from product toproduct and over temperature,you need an amplifierthat has a GBWP greaterthan 246 MHz. The guidingformula isGBWP OPA = 130 . ×NG× SBW ×( 1ERROR) ,ERRORwhere GBWP OPAis theop amp’s gain-bandwidthproduct.Use this formula duringyour first pass when youchoose an amplifier for yourcircuit. After you determinethe amplifier’s bandwidth,you can start to delve intothe other important amplifier characteristicsfor your application.EDNREFERENCES1 Bendaoud, Soufiane “Gain error affectsop amp choices,” Planet Analog,July 14, 2006, www.planetanalog.com/features/analog/showArticle.jhtml?articleID190400337.2 Mancini, Ron, “Op-amp-gain erroranalysis,” EDN, Dec 7, 2000, www.edn.com/article/CA56190.18 EDN | SEPTEMBER 17, 2009
THREE AIRCRAFT, A SINGLE MODEL,AND 80% COMMON CODE.THAT’S MODEL-BASED DESIGN.To develop the unprecedentedthree-version F-35, engineersat Lockheed Martin created acommon system model tosimulate the avionics, propulsion,and other systems andto automatically generatefinal flight code.The result: reusable designs,rapid implementation, andglobal teamwork. To learn more,visit mathworks.com/mbdTMAccelerating the pace of engineering and science©2008 The MathWorks, Inc.
TAPEOUT,,BY PALLAB CHATTERJEE, CONTRIBUTING TECHNICAL EDITORReference-tool flows andprocess-design kits, part oneMost of the advanced process technologies from waferfoundries include RTFs (reference-tool flows) to aid intool selection. In addition, PDKs (process-design kits)describe the electrical, yield, and performance aspectsof the process. These two pieces of support documentationhave been the basis of chip design since the startof the semiconductor industry. Although these files provided adequateinformation for engineers to complete designs on process geometries assmall as 0.25 micron, they alone do not represent all of the issues thatlithographic and advanced processing variation introduces.An RTF is usually a sequence ofrecommended tools that identify all ofthe steps—with the associated EDAtools—you need to verify that a physical-designview of a circuit meets allof the quality criteria to ensure manufacturability.The RTF also ensuresthat the chosen EDA tool will operatewith the information the foundryprovides and produce a “correct” result.Most of these tool flows involvetiming closure and detailed placementand routing of the circuits—capabilitiesthat all of the major EDA vendorsprovide. Supplementary tools includedevice simulators; parasitic RC(resistance/capacitance) extractors;physical verification, including DRC(design-rule checking), LVS (layoutversus schematic), and ERC (electrical-rulechecking); package modeling;high-capacity simulation; noisemodel ing; power analysis; and customlayout. The addition of these supplementarytools allows customers whowant analysis outside their all-in-oneflow to add independent checks or advancedanalysis.Most fabs try tomaintain more than2000 control filesfor a commercialfoundryoffering, soupdates take time,and the customermust verify them.A key misunderstanding about thetools in RTFs is that the foundry advocatesthem. Rather, the RTF is a list oftools with known interoperability witheach other. Furthermore, the foundrybelieves that these tools can producea result from a certain level of publiclyreleased information on the processwithin an acceptable amount of errorthat the wafer foundry determines.Participation in an RTF does not guaranteecorrect answers from any EDAtool for an arbitrary circuit application,so don’t blindly trust the tools on thelist. As a corollary, omission from thelist merely indicates that the foundryhas not established vendor-to-vendorinteroperability from a menu levelor that, to achieve higher accuracy orperformance, the tool requires supplementalinformation that is not availableto the general design communitybut may be available on a case-bycasebasis. Typically, mixed-signal, <strong>RF</strong>,memory, imaging, telecom, very-highspeed,multiprocessor-core, DSP, andlow-power designs do not use RTFs.The basics of RTFs are the technologyfiles that describe the maskingand design layers for the processand their functions. This technologyalso contains a naming conventionand information for operating a customphysical-design tool, or layout editor.The technology file also has therudimentary minimum-design-rule informationfor the process, so physicalsystemdesigners can use the rules asguidelines. There is also a place-androutecontrol file, which dictates device-to-deviceand wire-to-wire spacingon the design and is written in thedialect of the EDA vendor’s tools.The last major component is thephysical-verification files. These filesinclude the DRC deck, which comprisesminimum rules, recommendedrules, DFM (design-for-manufacturing)“increased-yield” rules, memory rules,I/O rules, power-supply rules, ESD(electrostatic-discharge) rules, analogrules, and special-device rules. A similarformat is available for ERC-, LVS-,and parasitic-extraction-rule decks. Asa result, most fabs try to maintain morethan 2000 control files for a commercial-foundryoffering, so updates taketime, and the customer must verifythem.EDNContact me at pallabc@siliconmap.net.+ Go to www.edn.com/090917pcand click on Feedback Loop to posta comment on this column.+ www.edn.com/tapeout20 EDN | SEPTEMBER 17, 2009
“I need to see the most signal detail.”Get15% OFFuntil 9/30.Starting at$4246Look no further.InfiniiVision scopes are engineered for the best signal visibility.It’s a bold claim — let us prove it to you.Agilent engineered the custom MegaZoom III technology in our Agilent Infi niiVision scopesso you get the best possible signal visibility on every measurement you make.1. Fastest uncompromised update rate shows you critical signal detail and givesyou statistically valid measurement results in seconds.2. Industry’s only hardware acceleratedmixed signal oscilloscope gives youanalog, logic, and protocol views.3. Customize your scope with softwareapplications including mask testing,RS-232/UART, CAN/LIN and more.100 MHz to 1 GHz MSO and DSO modelsTake the 5-minute Scope Challenge online andregister to win a free Agilent MSO7104 scope.www.agilent.com/find/bestscopesPrices in USD and subject to change without notice.© Agilent Technologies, Inc. 2009u.s. 1-800-829-4444 canada 1-877-894-4414
PRYING EYESBRIAN DIPERT • SENIOR TECHNICAL EDITORT-Mobile’s G1: Google’s Android OS emergesThe myTouch 3G (also known as the HTC Magic), T-Mobile’s second Google Android-OS-based and HTC-designedhandset, recently became available for purchase. More svelte and with a more responsive touchscreen than its predecessor,the G1 (also known as the HTC Dream), the myTouch 3G conversely relies exclusively on an on-screen virtualkeyboard. After peeling away the G1’s oft-preferred physical keyboard, my Prying Eyes partners at phoneWreck discoveredsome interesting facts about the two primary PCBs (printed-circuit boards) inside the premier Google Android offering.T-Mobile’s UMTS (Universal MobileTelecommunications System) cellular-datanetwork leverages the WCDMA (wideband-code-division/multiple-access),1700-and 2100-MHz bands in the United States,whereas GSM (global-system-for-mobile)-communication competitor AT&T, whichcurrently touts a more extensive 3G coveragefootprint, employs 850- and 1900-MHz spectrum slices. You’ll thereforesee some unique pieces of silicon in theG1 versus, say, the BlackBerry Bold (see“Trolling for gold in the BlackBerry Bold,”EDN, May 28, 2009, pg 20, www.edn.com/article/CA6659415). Avago’s ACPM-7381 and ACPM-7391 UMTS poweramplifiers, along with TriQuint’s ALM-1412quadband power-GSM amplifier, togethermate to Qualcomm’s RTR6285 <strong>RF</strong> transceiverand PM7540 power-managementIC. Avago also supplies the ALM-1412GPS (global-positioning-system) amplifier.The multidie, single-package memory subsystem sandwiches256 Mbytes of Samsung’s NAND flash memory and 128Mbytes of that company’s DDR SDRAM. T-Mobile bundlesa 1-Gbyte microSD (secure-digital) card with the handset;the memory-module interface is higher-capacity microSDHC(secure-digital-high-capacity)-compatible.AKM’s AK-79 86A sixaxiselectroniccompass supplementstheG1’s accelerometerandGPS capabilitiesto providethe handsetwith a rich setof location,motion, anddirectionstatistics.Qualcomm’s 528-MHzMSM7201A baseband processor,which embeds GPS and audioDAC/ADC functions, acts as thebrains of the G1. Curiously, HTCaugmented the G1’s logic witha Xilinx XC2C128 CoolRunner-II CPLD, which HTC has alsoemployed in past PDA and smartphonedesigns. Although theCPLD seemingly runs counter tothe integration- and cost-optimizedfocus of a high-volume consumerelectronicsdevice, it also enablesHTC to easily augment and updatehardware capabilities both on themanufacturing line and in the field.22 EDN | SEPTEMBER 17, 2009
PRY FURTHER AT EDN.COM+ Go to www.edn.com/pryingeyes for past PryingEyes write-ups.For those ofyou who preferold-fashionedbut reliablewires, SMSC’s(StandardMicrosystemsCorp’s)USB3316USB (UniversalSerial Bus)controller hasyou covered.Not shown are theG1’s 3.2M-pixel stillimage-onlycamerabased on Aptina’s(formerly Micron’s)MT9T013D sensorand Analog Devices’AD5398 autofocuscapablelens-coil driver;the Sharp-supplied3.2-in., 320480-pixelresolutionand 65,536-color LCD; and the1150-mAhr battery.The device has802.11b/gWi-Fi- andBluetooth-connectivityoptions.The TexasInstrumentsWL1251Btransceiver andcompanionWL1251FEpower amplifierhandle Wi-Fi,and the company’sB<strong>RF</strong>6300chip handlesBluetooth.•Over 2500 Std.DC-DC Converters•Surface Mount•From 2V to10,000 VDC OutputDC-DC Converters2V to 10,000 VDC Output•1-300 Watt Modules•Isolated/Regulated/Programmable ModelsAvailable•Military Upgrades Available•Custom Models,Consult FactorySee full Catalog immediatelywww.picoelectronics.comForEngineeringAssistanceCall Factoryor send directfor FREE PICO CatalogCall toll free 800-431-1064in NY call 914-738-1400Fax 914-738-8225PICO Electronics,Inc.143 Sparks Ave. Pelham, N.Y. 10803-18889INDUSTRIAL • COTS • MILITARYSEPTEMBER 17, 2009 | EDN 23
NEW T-CLAD ® PASTICKS IT TO HEAT.Peel and place thermal solution withstands the heat of solder reflow.Easier assembly, cooler LEDs.Thermally conductive insulated metal substrateboards specifically configured for LED applications.Bergquist’s T-Clad with pre-applied Bond-Ply ® 450 allows you to adhere yourmounted LEDs to a variety of heatsinks and surfaces while thermally optimizingyour application.This version of peel and place T-Clad canwithstand the high temperatures of solder reflow duringLED assembly and then be positioned in the lightingapplication using its strong thermally conductive adhesive.Call or visit to qualify for your FREE sample.Call 1.800.347.4572 or visit www.bergquistcompany.com/preappliedFREE T-Clad ® PASample Board18930 West 78th Street • Chanhassen, MN 55317 • A ISO9001:2000 registered facility(800) 347-4572 • Phone (952) 835-2322 • Fax (952) 835-0430 • www.bergquistcompany.comThermal Materials • Thermal Substrates • Fans and Blowers
BY UNDERSTANDING THE STAGES OF AN ANALOG-SIGNALPATH, DIGITAL-SYSTEM DEVELOPERS CAN MORE ACCURATELYCAPTURE SENSOR DATA FOR A VARIETY OF APPLICATIONS.BREAKING DOWNTHE SENSOR-SIGNAL PATHBY AARON GL PODBELSKI • CYPRESS SEMICONDUCTORSensors are increasingly finding use in embeddedsystems. Although industrial products havelong used them for manufacturing-control systems,consumer devices are now starting to employthem, as well. Manufacturers are integratingsensors into consumer products to createbetter user experiences—ranging from addingaccelerometers in mobile phones to adding water-vaporsensors in microwave ovens. Systemdesigners, who previously worked only in the digital domain, arenow finding themselves having to interface with analog sensors.You must digitize a sensor’s analog signal so that the systemcan use it, and the signal path goes through amplification,filtering, and digitization stages (Figure1). Each stage usually involves a componentwith passives around it to performproperly for an application. Onceyou digitize the signal, you can pass it toa control system on the microcontrolleror massage the data and pass it to a hostprocessor through a communicationprotocol. The protocol can use the sensordata as necessary.Every sensor has a different outputsignal and range. The output signal canbe voltage-, current-, resistive-, capacitive-,or frequency-based, but few standardsexist, and only specific industrialsystems use them. Even similar sensorsfrom the same manufacturer can havedifferent outputs, and these differencescan create problems for system designers.A designer must select a sensor thatmeets the requirements for the system. Ifthe requirements change during the design,however, a sensor change may alsobe in order. In addition, a new sensorwith a slightly different output wouldnecessitate altering the amplificationand filtering stages.Most sensors output a low-level current-or voltage-based signal, so a simpleresistive network adapts any currentbasedsignal into a voltage. This articleSEPTEMBER 17, 2009 | EDN 25
simplifies some concepts and the component-selectionprocess.AMPLITUDEThe output of a sensor can be as smallas several millivolts or as large as severalvolts. For proper digitization, the signalmust be large enough for the ADCto effectively read it. In most cases, thesensor signal requires amplification. Forexample, a typical type-K thermocoupleoutputs 41 V/C, which you mustgreatly amplify if your design requires1C granularity. Thus, you must takeADC resolution into account to ensurethat you sufficiently amplify the signalto obtain the desired granularity.You base the selection of an amplifiermainly on the type you need—be it aninstrumentation amplifier, a differentialamplifier, an operational amplifier, or aPGA (programmable-gain amplifier).You also must determine the amountof gain your amplifier requires. A resistivenetwork, with feedback, around theamplifier sets the amplifier’s gain. Themaximum gain for standard amplifiers isideally limitless. A digital signal to thedevice usually sets the gain for a PGA.This signal alters an internal resistivenetwork. The maximum possible gainfor a PGA is 0.5 to 1000 times less thanthat of a traditional amplifier, but thisrange is more than acceptable in mostcases.With amplifiers, you must take intoaccount another key specification: offsetvoltage. Offset voltage is the amount ofalteration in volts of a signal that passesthrough the amplifier. For example, ifyou put a 500-mV signal into an amplifierwith unity gain, or a gain of one, andan offset voltage of 10 mV, the resultingoutput would be 510 mV. If the outputrange of the sensor is 0 to 900 mV andthe system does not need a very granularreading of the sensor, this offset maybe negligible. If the range of the sensorSENSOR FILTERPROCESSING DIGITALAMPADCHOSTCORE INTE<strong>RF</strong>ACEACTUATORCONTROLAT A GLANCE Even similar sensors from thesame manufacturer can have differentoutputs, and these differencescan create problems for systemdesigners. Noise arises from a number ofsources, including board layout,radios, thermal components, andeven the sensor itself. To use the sensor’s filtered signal,you must quantify the analogsignal into the digital domain usingan ADC. You have the choice of using anexternal ADC or a microcontrollerwith an integrated ADC. ExternalADCs tend to have higher accuracyand higher perform ance in bothspeed and resolution.Figure 1 A sensor’s analog-signal path goes through several stages: amplification, filtering,and digitization.is 450 to 550 mV, this offset is probablyunacceptable. The smaller the offsetvoltage, the more costly the amplifieris. All amplifiers have an offset, but youneed to know whether the system cantolerate it. You can reduce or eliminatethe offset voltage using correlated doublesampling.FILTERINGAll systems impart some noise on thesensor’s signal. Noise arises from a numberof sources, including board layout,radios, thermal components, and eventhe sensor itself. Signal noise causes theADC to make inaccurate and unstablereadings, and the noise level increasesthrough the amplification stage, whichexaggerates the error in the signal. Youcan qualify signal noise as low frequency,high frequency, or a known frequency.You most often need to address highfrequency-noiseissues.You can filter noise using passive analogfilters, filter ICs, and digital filters(Figure 2). The most common method,passive filtering, involves creating a passivenetwork of resistors, capacitors, andinductors. You must design passive filters,however, and you cannot easily alterthem. Filter design can be as cumbersomeas the order of the filter you need;a first-order Chebyshev filter takes muchless effort to design than an eighth-orderBessel filter. So you should determinethe order of the filter you need beforeselecting the method of filtering youwill employ.Some ICs allow you to digitallyprogram the type of filter you need.These ICs use internal analog circuitryto create the filter and mayhave offset voltages associated withthem. They also allow you to movethe filter process after quantificationwith the ADC. Digital-filterdesign can be complex, but manyFigure 2 The sensor-signal path includes amplifiers, filters, and an ADC. You design the filter to remove noise and limit the bandwidthof the signal.26 EDN | SEPTEMBER 17, 2009
(a) (b) (c)(d) (e) (f)Figure 3 Combining INL error (a), DNL error (b), gain error (c), offset error (d), and total error (e) provides an understanding of theADC in use compared with an ideal ADC (f).tools allow for the easy design of highorderfilters. Digital filtering can be anideal means of removing noise; however,it often requires many CPU cycles,increasing power consumption. Thesystem normally incurs high-frequencynoise, necessitating the use of a lowpassfilter. This filter attenuates any part ofthe signal that is higher frequency thanthe set cutoff frequency. Some sensorsignals require the use of several typesof filters in tandem with each other.Most sensor data sheets specify a basicinterface circuit but do not mentionthe necessary filtering. System designersmust create the system before fullyunderstanding how much filtering isnecessary.SENSOR FILTERPROCESSING DIGITALAMPADCHOSTCORE INTE<strong>RF</strong>ACEACTUATORCONTROLSTANDARDMICROCONTROLLERPSOCFigure 4 Developers can implement the amplification and filtering stages, integrating theentire analog-signal chain onto one device.DIGITAL CONVERSIONTo use the sensor’s filtered signal, youmust quantify the analog signal into thedigital domain using an ADC. Selectionof an ADC mainly concerns the system’srequirements for sampling speed and resolution.The necessary sampling speedrelates to the sensor’s bandwidth or howoften the system needs updating. Resolutiondepends on the granularity you needfor the ADC to react to the sensor’s information.The system’s usage model definesthis speed and resolution requirement.For example, a generic gyroscopemeasures 360 of rotation at 0.67 mV/,resulting in an output range of 241 mV.To remain upright, a hobbyist’s helicoptermight need information from a gyroscopeat a granularity of 1 but witha throughput of 10k samples/sec. Thisrequirement would necessitate a 10-bitADC, which would provide 0.35/bit.Note that the signal still has noise on it,however, and 1 bit is acceptable. Conversely,a digital camera with image stabilizationmight require a granularity of0.02 but with a throughput of 5k samples/secto adjust the image sensor as acamera shakes. This requirement wouldnecessitate the use of a 16-bit ADC,which would provide 0.005/bit.Manufacturers measure the accuracyof ADCs in terms of INL (integralnonlinearity), DNL (differentialnonlinearity), offset error, gain error,and SNR (signal-to-noise ratio).When you combine these terms, theyoffer an understanding of the ADC’stotal error (Figure 3). For most applications,it is not necessary to lookinto these ADC specifications, butengineers should have a thoroughunderstanding of these values for theADC in use. You have the choice ofusing an external ADC or a microcontrollerwith an integrated ADC.SEPTEMBER 17, 2009 | EDN 27
External ADCs tend to have higher accuracyand higher performance in bothspeed and resolution. Most sensor-applicationrequirements align well with integrated-microcontrollerADCs, however.Another option is to use configurableADCs, which comprise programmablelogicblocks within a microcontroller. Integrateddigital and analog programmableblocks allow for the dynamic definitionof configurable peripherals for eachsensor application. These blocks includecounters, PWMs (pulse-width modulators),UARTs, SPIs (serial-peripheral interfaces),amplifiers, filters, ADCs, andDACs. Developers can also implementthe amplification and filtering stages, integratingthe entire analog-signal chainonto one device (Figure 4). Using configurableADCs can result in cleaner designsthan those using passive components.In addition, you can dynamicallyreconfigure these blocks, so you have theoption of reusing these system resourcesfor other functions.Sensors continue to penetrate a rangeof markets, bringing you more controlSENSORS CONTINUETO PENETRATE ARANGE OF MARKETS,BRINGING YOUMORE CONTROL ANDGREATER FLEXIBILITY.and greater flexibility. Sensors increasereliability through management of theenvironment, including, for example,temperature monitoring; improved performancethrough feedback mechanisms;and enabling new types of user interfaces.For many of these designs, the integratedADCs on microcontrollers providesufficient granularity and accuracy.Developers who are unfamiliar with analogdesign may encounter pitfalls alongthe analog-signal chain between thesensor and the microprocessor.Implementing the multiple stages ofthe analog-signal path can seem convoluted,especially to engineers accustomedto designing primarily in the digitaldomain. However, by breaking downthe analog-signal path into the variousamplification, filtering, and ADC stages,digital-system developers can moreeasily and accurately capture sensor datafor a variety of industrial and consumerapplications. In addition, readily availableICs, configurable ADCs, and filterdesigntools can greatly simplify sensordesign.EDNAUTHOR’S BIOGRAPHYAaron GL Podbelski is a project manager atCypress Semiconductor. He manages severalof the company’s PSoC (programmable-system-on-chip)products and focuseson marketing the PSoC’s analog capabilitiesfor customers using sensors. Podbelski hasa bachelor’s degree in computer engineeringfrom Marquette University (Milwaukee,WI). His personal interests include playingthe guitar and drums, snowboarding, surfing,playing with tech gadgets, and smiling.A version of this article appeared ina May 2009 supplement published byEDN’s sister publication Design News.MPEM AGNA-POWERELECTRONICS2 KW TO 900 KW+ AC/DC POWER SUPPLIESSpecification XR Series TS Series III MS Series II MT Series IIIModels 54 150 84 140Power 2 to 6 kW 5 to 45 kW 30 to 75 kW 100 to 900 kW+Voltage 0-1000 Vdc 0-4000 Vdc 0-4000 Vdc 0-4000 VdcCurrent 0-375 Adc 0-2700 Adc 0-4500 Adc 0-24000 AdcApplicationsElectric PowertrainsWater TreatmentMilitaryBattery ChargingAviation SystemsSource EmulationLaboratory TestingPLC IntegrationMany moreTS Series IIIAnnouncing the XR Series: The Highest Power Density 2U Programmable DC Power SupplyMT Series IVUp to 88% efficiencyLow output rippleCurrent-fed processingConstant voltage and constant current programmingVoltages up to 1000 Vdc, Currents up to 375 AdcRS-232 standard, LXI Ethernet, GPIB, USB, RS485 availablewww.magna-power.com (908) 237-220028 EDN | SEPTEMBER 17, 2009
“I need a faster, more accurate audio analyzer.”We knew you’d say that.The new Agilent U8903A audio analyzer is four times more accurate andover seven times faster than its popular HP 8903B ancestor. In fact, it isamong the fastest and most accurate currently in its class. It’s also a morefully capable audio analyzer, letting you make crosstalk measurements andphase measurements, FFT analysis, and graph sweep. And it will soon beupgradable to eight channels. You can’t find that anywhere else. Go onlineto learn more about this remarkable new audio analyzer.Save $ 2,400when you trade in your old HP 8903Bfor a new Agilent U8903A audio analyzer*www.agilent.com/find/U8903Apromo©Copyright Agilent Technologies 2009Trade in offer good for a limited time.u.s. 1-800-829-4444 canada: 1-877-894-4414
<strong>RF</strong><strong>SWITCHING</strong> <strong>OPTIONS</strong>:THE RIGHT FIT MIGHT COME WITH A LOSS
MANUFACTURERS ARE OFFERING SOI AND MEMS ALTER-NATIVES TO PIN-DIODE, GaAs, AND ELECTROMECHANI-CAL SWITCHES FOR A VARIETY OF <strong>RF</strong> APPLICATIONS,BUT YOU NEED TO UNDERSTAND <strong>RF</strong>-SWITCH SPECSBEFORE YOU COMMIT TO A NEW TECHNOLOGY.BY RICK NELSON • EDITOR-IN-CHIEFWhether you are building “big-iron” <strong>RF</strong>testequipment or tiny, multiband mobiledevices, you need to route <strong>RF</strong> andmicrowave signals among instrumentsand devices under test or between antennasand amplifiers. To accomplishthat task, you can turn to various solid-stateimplementations, includingSOI (silicon-on-insulator) devices andMEMS (microelectromechanical-system) switches, both of whichvendors were touting at the June IEEE MTT-S (Microwave Theoryand Techniques) International Microwave Symposium. Highly integratedbulk-CMOS devices represent another alternative. However,you need to know when those optionsmake sense as alternatives to the moretraditional electromechanical, PIN-diode,and GaAs (gallium-arsenide) FETversions. Solid-state and MEMS switchestake up much less real estate and havelonger lifetimes than do electromagneticswitches. Further, SOI and MEMS devicescan be easier to integrate than GaAsdevices with other components. Beforeembarking on switch selection, however,you need to understand whether thespecs of the device you select will meetyour application requirements.GETTING TO KNOW SPECSYou need to consider switch bandwidth.The switch you choose must, at itsmaximum operating frequency, conducta signal. Unfortunately, when choosinga switch, you can’t rely on the “3-dBdown”—thatis, noise that is 3 dB lowerthan peak noise—rule of thumb that youmight use to determine the upper operatinglimit of an amplifier, for example.You need to determine the acceptableperformance with respect to bandwidthin the context of the other specifications,including characteristic impedance andVSWR (voltage-standing-wave ratio),crosstalk, and isolation (Reference 1).You should also consider whether youneed to transmit low frequencies or dc.Electromechanical and FET switchesgenerally pass low frequencies; PIN-diodeswitches and capacitive MEMS switchesgenerally do not (Reference 2).Other <strong>RF</strong>-switch specifications includecharacteristic impedance. You typicallyuse a lumped-element model to representtransmission lines (Figure 1). In thesemodels, successive infinitesimal segmentsof the line, which ideal transmission linesinterconnect, have series resistance R, seriesinductance L, shunt conductance G,and shunt capacitance C. Characteristicimpedance, then, isZ 0 = R + j ω L .G+jωCFor zero resistance and zero conductance,characteristic impedance isZ 0 =LC .Any switch you select should presentthe same characteristic impedanceyour system exhibits to prevent signalreflections. The reflection coefficientquantifies reflections. This coefficient isequivalent to the s 11S parameter (Reference3), which can contribute to apoor VSWR. With regard to VSWR,you want to know whether you are selectingan absorptive switch or a reflectiveswitch. Absorptive versions apply aresistive shunt to ground in the off stateto provide a good impedance match andminimize VSWR, regardless of switchposition; reflective versions serve applicationsin which VSWR doesn’t matteror in which you control impedance elsewhere(Reference 4).INSERTION LOSS, ISOLATIONOther key specs include isolation,crosstalk, and insertion loss:PIL = 10log OUT10 ,PINwhere IL is insertion loss and P OUTandP INare output power and input power,respectively.You can also represent insertion lossin terms of S parameters:⎛ sIL 10 21 ⎞= log10⎜ ⎟.⎝ 1s 11 ⎠The S-parameter notation to theright of the equal sign shows that theswitch’s insertion loss is inherent to theswitch and does not depend on inputmismatches. The 1s 11denominator indicatesthat you subtract any reflectedSEPTEMBER 17, 2009 | EDN 31
signal that the s 11term represents fromthe normalized input-signal level beforeyou calculate insertion loss.Insertion loss is critical, says DavidHall, an <strong>RF</strong>-product manager for NationalInstruments, which uses <strong>RF</strong>switches in its lineup of <strong>RF</strong>-switch-matrixtest-and-measurement products. Aswitch must transmit your signal withoutundue attenuation at the frequencyor frequencies of operation. If youchoose a switch with too much insertionloss, he says, you may have to amplify aswitched signal to make it usable, addingto system complexity and potentiallyintroducing linearity errors.AT A GLANCE You need to determine theacceptable performance with respectto bandwidth in the context of theother specifications, including insertionloss, crosstalk, and isolation. Absorptive versions applya resistive shunt to groundin the off state to provide goodimpedance match and minimizeVSWR (voltage-standing-wave ratio)regardless of switch position. Bulk-CMOS devices offer costadvantages and fit within a smallfootprint. One MEMS (microelectromechanical-system)switch is rated for100 million operations and can workfor 1 billion cycles. The handset market offers highvolumeopportunities for nine-throwswitches going into pentabandphones.SPECIFYING LINEARITYA switch can exhibit linearity errorseven in the absence of amplification.Specs that indicate linearity includethe 1-dB compression point and the IP3(third-order intercept point). Measuringthe 1-dB compression point involvesa power sweep on the input to a deviceunder test; when the output drops 1 dBfrom the level you would expect basedon the small-signal response, you’vereached the 1-dB compression point. IP3measurements involve applying closelyspaced tones to a nonlinear device undertest, which generates third-order intermodulationproducts. The IP3 pointis typically hypothetical because it mightlie outside the safe operating region ofthe device under test. You obtain the IP3using extrapolation. It occurs when thepower of the third-order intermodulationproducts would equal that of the desiredsignal (Reference 5).Other specs include crosstalk and isolation.Crosstalk represents the magnitudeof a signal that couples from anactive switch to an adjacent inactiveswitch, and isolation represents the magnitudeof a signal that a switch transmitsfrom its input to its output when in theopen position. A switch can be subjectto various deleterious effects, includingimpedance mismatch, insertion loss,crosstalk, and isolation (Figure 2).Electromechanical switches as wellas PIN diodes and GaAs FET switchestraditionally have excelled at meetingthese specs, but each offers drawbacks.Electromechanical switches are largeand can occupy significant amounts ofPCB (printed-circuit-board) real estate.PIN diodes include a high-resistivity intrinsicregion between their P- and N-type semiconductor regions (Reference6). The intrinsic region becomes conductivewhen you forward-bias the device,essentially closing the switch to allow<strong>RF</strong> signals to pass. PIN diodes arerugged, compact devices that can handleR L R L R LG C G C G CFigure 1 Lumped-element models typically represent transmission lines, with successiveinfinitesimal segments, interconnected by an ideal transmission line, having seriesresistance R, series inductance L, shunt conductance G, and shunt capacitance C,from which you can calculate characteristic impedance. Your switch should offer thesame characteristic impedance.high voltages and currents; drawbacksinclude the need for external bias circuitryand the fact that the required dcbias current—and the higher the bias,the lower the insertion loss—makesthem problematic for use in power-sensitivebattery-operated system.GaAs DEVICES ARRIVEGaAs devices remain popular. NECmarkets the devices for wireless applications(Reference 7). At the June InternationalMicrowave Symposium, <strong>RF</strong> MicroDevices introduced its 6-GHz <strong>RF</strong>3021,<strong>RF</strong>3023, <strong>RF</strong>3024, and <strong>RF</strong>3025 switches,which it fabricated in its GaAs PHEMT(pseudomorphic-high-electron-mobility-transistor)technology. The symmetricSPDT (single-pole/double-throw)<strong>RF</strong>3021 and <strong>RF</strong>3025 switches featurehigh isolation; the <strong>RF</strong>3023 and <strong>RF</strong>3024symmetric SPDT switches feature low insertionloss and moderate isolation. Thisyear, Skyworks Solutions Inc introducedits Sky13317-373LF, a PHEMT GaAsSP3T (single-pole/three-throw) antennaswitch that operates from 0.1 to 6 GHz.Last month, Hittite Microwave Corpintroduced its GaAs PHEMT, MMIC(monolithic-microwave-integrated-circuit),SP4T (single-pole/four-throw),nonreflective HMC-C071 switch module,which targets microwave radio,VSAT (very-small-aperture-terminal),military and aerospace, fiber-optic, andbroadband-test applications requiringoperation from dc to 20 GHz.GaAs FET switches have their drawbacks,however, in that they require externalcomponents in the form of blockingcapacitors, and they can be difficultto integrate with other components(Reference 8). Alternatives to GaAsswitches, as well as to electromechanicalswitches and PIN diodes, include bulk-CMOS, MEMS, and SOI switches.BULK-CMOS SWITCHESAnalog Devices touts its ADG9XXfamilysubmicron, 1-GHz bulk-CMOSswitches (Figure 3), which the companyoffers for use in the 900-MHz ISM (industrial/scientific/medical)band (Reference9). Ray Goggin, staff design engineer,and John Quill, engineering manager,both of Analog Devices’ switchesandmultiplexer-product line, cite severaladvantages of the devices. For exam-32 EDN | SEPTEMBER 17, 2009
INCIDENT SIGNALOUTPUT SIGNALATTENUATED DUE TOINSERTION LOSSSWITCHREFLECTED SIGNALDUE TO IMPEDANCEMISMATCH(a)INCIDENT SIGNALSWITCHOUTPUT SIGNALCOUPLED SIGNALDUE TO CROSSTALK(b)INCIDENT SIGNALSWITCHCOUPLED SIGNALDUE TO POORISOLATION(c)Figure 2 A switch can be subject to various deleterious effects. An impedance mismatchcan result in input reflections, and the switch’s insertion loss attenuates the outputsignal (a). In addition, an active switch can couple a signal to an adjacent inactiveswitch (b), and an open switch can couple a signal from its input to its output (c).ple, they require no dc-blocking capacitorsand come in small packages, suchas the ADG919’s 33-mm QFN. Theyalso operate on supply voltages of 1.65to 2.75V, and they consume less than 1A of current. In addition, the devicesaccept TTL (transistor-transistor-logic)inputs, simplifying their integration withother system components. At 1 GHz, insertionloss is 0.8 dB and isolation is 43dB. As for power-handling capability,the 1-dB compression point is 17 dBm.The devices successfully pass 1-kV HBM(human-body-model) ESD (electrostatic-discharge)tests on <strong>RF</strong> pins and 2-kVHBM ESD tests on non-<strong>RF</strong> pins. Theswitches are available in absorptive andreflective versions.
PICO Transformersand Inductors• SU<strong>RF</strong>ACE MOUNT • PLUG-IN • AXIAL LEADS • TOROIDAL • INSULATED LEADS •PICO Electronics, Inc.143 Sparks Ave. Pelham, NY 10803 •Tel:(914) 738-1400 • Fax: 914-738-8225 • E-mail: info@picoelectronics.comCOTS/INDUSTRIAL/MILITARYwww.picoelectronics.comCustomized Models Available...Optimized for your Application!Goggin and Quill note that bulk-CMOS switches have lower bandwidthand maximum switched-<strong>RF</strong>power than do other switches, but,for applications whose requirementsfall within bulk-CMOSlimitations, the bulk-CMOS devicesoffer cost advantages. Theyalso note that each device exhibitsa footprint as small as 9 mm 2 in aCSP (chip-scale package) and canintegrate driver circuitry, therebyeliminating the need for separateICs.INPUT 1GATEDRIVER<strong>RF</strong>CCTRL5050<strong>RF</strong> 1<strong>RF</strong> 2Figure 3 The Analog Devices SPDT ADG918absorptive switch has 50-terminated shuntresistors to ground, minimizing reflections back tothe <strong>RF</strong> source. It is available in a 33-mm chipscalepackage.V CCMEMS AND SOS DEVICESIf your application requires performancelevels that lie beyondwhat bulk-CMOS devices can provide,you might consider a MEMSswitch. Omron’s 2SMES-01, forexample, operates to 10 GHz, offering30-dB isolation and 1-dB insertionloss (Figure 4). It offers maximum powerconsumption of 10 W, which, accordingto Donna Sandfox, product managerat Omron Electronic Components LLC,is about one ten-thousandth that of anequivalent electromagnetic relay.The device targets high-throughputATE (automated-test-equipment) applications.An electrostatic-drive systempowers the device, which performs 100million operations switching a resistiveload at 0.5 mA and 0.5V dc. Sandfoxsays the company has tested the deviceover 1 billion cycles. Each switch consistsof two normally open SPST (single-pole/single-throw)silicon switchesin a 5.23.01.8-mm housing offeringSPDT or DPST (double-pole/singlethrow)normally open operation.Peregrine Semiconductor addressesthe <strong>RF</strong>-switch market with its Ultra-CMOS SOS (silicon-on-sapphire) technology,which integrates ultrathin-silicon-CMOScircuitry on an insulatingdielectric sapphire substrate. Peregrinealso targets ATE and other applications,including digital TV, cable and satelliteset-top boxes, game consoles, andcellular communications. For <strong>RF</strong> trans-R<strong>RF</strong> 1Over 5000FROMLOW PROFILE.19"ht.INPUT 2DRIVER ICR<strong>RF</strong> COM<strong>RF</strong> 2Figure 4 An electrostatic-drive system powers Omron’s 2SMES-01, which operates to 10GHz, offering 30-dB isolation and 1-dB insertion loss. It performs 100 million operationsswitching a resistive load.34 EDN | SEPTEMBER 17, 2009
SWITCHLOW-NOISEAMPceivers, the company complements itsswitches with UltraCMOS quad MOS-FETs, PLLs, and prescalers (Figure 5),enabling Peregrine parts to make upa substantial portion of the <strong>RF</strong>-signalchain. Rodd Novak, vice president ofsales, marketing, and business development,says that the handset marketoffers high-volume opportunities, withthe company’s nine-throw switches goinginto pentaband phones.For ATE applications, Peregrine offersthe PE42552 absorptive SPDT device,which operates to 7.5 GHz with a 1-dBcompression point of 34.5 dBm. Insertionloss is 0.65 dB at 3 GHz. Also forATE applications, the company complementsthe PE42552 switch with thePE43703 7-bit digital step attenuator.Mark Schrepferman, director of salesand marketing for communications andindustrial markets at Peregrine, says thatPeregrine has learned a lot about how toserve the test-equipment market fromlooking at its own test needs. ChristianSteele, product-development sectionmanager at Peregrine, recounts onemajor problem. Big-box ATE systems,he says, often include one or two sourcesand one or two receivers and rely onelectromechanical switches to route signalsamong the available instrumentsand the multiport devices under test. “Alot of those mechanical switches have areliability of very few throws—often lessthan 2 million,” he explains. “At thevolumes in which we are shipping ourhandset switches, we would be replacingvery expensive mechanical switches inPOWERAMPQUADMOSFETQUADMOSFETPLLPRESCALE<strong>RF</strong>igure 5 For <strong>RF</strong>-transceiver applications, Peregrine Semiconductor complements itsswitches with UltraCMOS quad MOSFETs, PLLs, and prescalers.PLLless than a month.” The cover story ofthe October issue of EDN’s sister publicationTest & Measurement World willdescribe how Peregrine engineers performengineering characterization andproduction test of SOS devices.Schrepferman says he has learned a lotfrom Steele and colleagues involved intest at Peregrine and has put the informationto good use. He notes that, evenin the down economy, the company hasseen “a tremendous number of designwins” in the test-equipment market.As <strong>RF</strong>-switch technology advances,there is unlikely to be a clear-cut winnerserving all applications, and Peregrineis evaluating multiple technologies. AtPeregrine, there is one restriction. RonReedy, co-founder and chief technologyofficer, says that he put one edict inplace when he founded the company:“Don’t get in front of the CMOS steamroller.”If you can implement somethingin bulk silicon CMOS, which he callsmankind’s greatest volume accomplishmentby any standard, then that’s howyou should implement it.Reedy says that Peregrine is investigatingMEMS, noting that, “from a performancepoint of view, it’s pretty hardto beat metal contact.” It’s counterproductive,however, to put a high-performanceswitch on a lossy substrate, andwork is under way to see whether MEMSbelong on sapphire. “We would not saythere is a single technical solution thatcovers all applications,” he adds.EDN
REFERENCES1Rowe, Martin, “Get to know <strong>RF</strong>switch specifications,” Test & MeasurementWorld, October 2007, www.tmworld.com/article/CA6482911.2 “Microwave switches,” Microwaves-101 Microwave Encyclopedia, www.microwaves101.com/encyclopedia/switches.cfm.3Nelson, Rick, “What are S-parameters,anyway?” Test & MeasurementWorld, February 2001, www.tmworld.com/article/CA187307.4Corrigan, Theresa, “Ask the ApplicationEngineer—34: WidebandCMOS Switches,” Analog Dialogue,October 2004, Analog Devices, www.analog.com/library/analogDialogue/archives/38-10/wideband_switch.html.5Kundert, Ken, “Accurate and RapidMeasurement of IP 2and IP 3,” The Designer’sGuide Community, May 2002,www.designers-guide.org/Analysis/intercept-point.pdf.6The PIN Diode Circuit Designers’Handbook, Microsemi Corp, 1998,www.ieee.li/pdf/pin_diode_handbook.pdf.7Iwata, N, and M Fujita, “GaAs SwitchICs for Cellular Phone Antenna ImpedanceMatching,” NEC TechnicalJournal, March 2009, www.nec.co.jp/techrep/en/journal/g09/n01/090108.html.8“<strong>RF</strong> Switch Performance Advantagesof UltraCMOS Technology over GaAsTechnology,” Application Note AN18,Peregrine Semiconductor, 2007, www.psemi.com/pdf/app_notes/an18.pdf.9Corrigan, Theresa, “ADG9xx WidebandCMOS Switches: FrequentlyAsked Questions,” Application NoteAN-952, Analog Devices, 2008, www.analog.com/static/imported-files/application_notes/AN_952.pdf.+ Go to www.edn.com/090917csand click on Feedback Loop to posta comment on this article.+ For more technical articles, go towww.edn.com/features.Electrical engineers agree: with a Protomat S-Seriesprototyping machine at your side, you’ll arrive at thebest solutions, fast. These highly accurate benchtopPCB milling machines eliminate bread-boarding andallow you to create real, repeatable test circuits—including plated vias—in minutes, not days.• Declare your independence from board houses• Affordable, entry-level price tag• The best milling speed, resolution, and accuracyin the industry• Single-sided, double-sided, and multilayeredmachining without hazardous chemicals• Optional vacuum table and autosearch camerafor layer alignmentProtoMat ® S-SeriesPCB Milling MachinesFor complete details visit:www.lpkfusa.comor call:1-800-345-LPKFFOR MORE INFORMATIONAnalog Deviceswww.analog.comHittite MicrowaveCorpwww.hittite.comNational Instrumentswww.ni.comNECwww.nec.comOmron ElectronicComponentswww.components.omron.comYou can reachEditor-in-ChiefRick Nelsonat 1-781-734-8418and rnelson@reedbusiness.com.PeregrineSemiconductorwww.psemi.com<strong>RF</strong> Micro Deviceswww.rfmd.comSkyworks SolutionsIncwww.skyworksinc.com36 EDN | SEPTEMBER 17, 2009
© 2009 Silicon Laboratories Inc. All rights reserved.HIGH JITTERRMS off the chartsLOW JITTER0.3 ps (12kHz - 20MHz) RMS to be exactSilicon Labs has a complete portfolio of industry-leading XOs, VCXOs,any-rate, any-output clock generators, any-rate jitter attenuating clocks andclock buffers that set a new standard for exibility, performance and lead time.Any-rate frequency synthesisFlexible clocks and oscillatorssupport the widest frequencyrange in the industry, makingthem ideal for multi-rateapplications. Available inindustry-standard, drop-incompatible packages.Ultra-low jitterBased on our patentedDSPLL ® and MultiSynthtechnologies, these low jitterproducts improve systemperformance, reduce BOM,minimize board space andsimplify system design.Shortest lead timesSilicon Labs changes theoscillator manufacturingmodel, enabling short,predictable two weeklead times for anyfrequencyXO andVCXOs.Find out which technology is right for your application based onyour jitter performance needs. Download the technical paper:When to Use a Clock vs. an Oscillator at www.silabs.com/cvo
BY SERGUEI OKHONIN, PhD • INNOVATIVE SILICONIn search of a better DRAM:evolving to floating bodiesWIDELY INVESTIGATED FLOATING-BODY MEMORIES APPEAR TOBE COMPELLING REPLACEMENTS FOR CONVENTIONAL DRAMs.A NEW FLOATING-BODY MEMORY USES THE INTRINSIC BIPOLARTRANSISTOR TO STORE SIGNIFICANTLY GREATER CHARGE.The memory industry hascrammed more and more memorybits onto ever smaller dieand is selling those slivers of siliconfor a few cents each. Currently,1-Gbit and even 4-GbitDRAMs (dynamic random-access memories)are available. Process engineers havebeen able to achieve these goals, especiallywith respect to the capacitor element, whichhas become more difficult to scale, a problemthat gets worse as device geometries shrink.These advances employ the basic one-transistorDRAM, which Robert H Dennard,PhD, a fellow at the IBM Thomas J WatsonResearch Center (www.watson.ibm.com),created in 1966. In 1970, Intel (www.intel.com) released thefirst DRAM chip, a 1-kbit PMOS device. Since that time, thebasic DRAM building block has comprised a single transistorand an increasingly complex capacitor.Scaling introduces yet another major problem for DRAMmanufacturers: leakage current. In both the bit cell and its supportingcircuitry, leakage becomes more significant as CMOS(complementary metal-oxide-semiconductor) processing nodesprogress from 90 nm through 78, 50, and 45 nm. Manufacturersare now discussing building memory chips at the 32-nm node.At this point, leakage in traditional designs will become a difficultproblem and prohibitively expensive to counteract, requiringnew architectures, changes to standard operating specifications,and significant process evolutions.The problems of scaling and leakage, as well as device size,rest fundamentally with the basic transistor-plus-capacitorSG+ + + + +BOXDN + N +PSMOS TRANSISTORINTRINSICBIPOLAR TRANSISTORbuilding block. Although the transistor elementis theoretically infinitely scalable, atleast for the foreseeable future, the capacitoris not. Manufacturers can fabricate capacitorsas either high stacks or deep trenches. However,if the overall bit cell shrinks due to increaseddensity or a smaller process node, thecapacitor must become higher or deeper tomaintain the minimum charge necessary forreliable operation.The memory industry is fast approachingthe scaling limits for the capacitor element,and it is therefore time for a new approach.There appears to be a groundswellof opinion in the commercial world and theanalyst community that floating-body-basedmemories may provide the answer. The floating-body effectis the dependence of the body potential of a transistor, usingthe SOI (silicon-on-insulator) technology, on the history ofits biasing and the carrier-recombination processes. The transistor’sbody forms a capacitor against the insulated substrate.The charge accumulates on this capacitor and may cause adverseeffects—for example, opening parasitic transistors in thestructure and causing off-state leakages, resulting in highercurrent consumption and, in DRAM, the loss of informationfrom the memory cells.Multiple significant companies, notably Samsung (www.samsung.com), Intel, and STMicroelectronics (www.st.com),have published work on the subject, and at least three pa-FLOATING BODYFigure 1 The use of an intrinsicbipolar transistor results in a superiorfloating-body-memory design.G+ + + + +N + N +PBOXDFigure 2 The write cycle fora one data pattern generatescurrent flow throughthe transistor body.Figure 3 A read-cyclemechanism senses bipolarcurrent flow.Figure 4 Read and write currents are nearly identical.SEPTEMBER 17, 2009 | EDN 39
pers presented at the International Electron Devices Meeting,which took place in December 2008 in San Francisco,addressed floating bodies. Equally significant, at least one majorstand-alone-memory maker, Hynix (www.hynix.com), andone processor company, AMD (www.amd.com), have signed(a)licenses with a floating-body-memory company to developproducts.The floating-body effect naturally occurs in transistors fabricatedon SOI substrates, leading to the accumulation of chargein the transistor body. Recently introduced device types suchas FinFETs (fin-shaped field-effect transistors) and surroundgate,or pillar, transistors also demonstrate a floating-bodyeffect, even when you implement them on traditional bulk-CMOS substrates. For these reasons, engineers no longer regardthe floating-body effect as a parasitic nuisance. Many engineershave tried to manipulate and enhance the body chargeso that they can use it to reliably store a logical “state” andfunction as a memory element. The objective is a memory bitcell that comprises only one transistor, fundamentally the simplestand most scalable of all semiconductor devices.Although a number of companies have investigated floating-bodymemories, the conventional approach is unlikelyto lead to a manufacturable product. A second generationof floating-body memory must be able to store a significantlylarger charge in a smaller transistor. This increased amount ofcharge greatly improves the amount of time the memory canretain its state as well as the signal margin between a one stateand a zero state. Other improvements include faster reads andwrites and reduced write power consumption.PRINCIPLES OF OPERATIONEarly attempts at realizing usable floating-body memoriesemployed a MOS transistor to pass current and create chargein the body using impact ionization. Although you can demonstratememory performance using such a technique, theamount of charge you create with this method is insufficientto create a robust and manufacturable memory device. A superiorapproach is to use the bipolar transistor intrinsic in(b)(c)Figure 5 A retention-measurement comparison of first-generation(MOS-transistor element only) and second-generation (addedbipolar element) floating-body memories reveals the secondgeneration’s superiority (a). The bipolar-inclusive floating-bodyapproach also delivers a more substantial programming window(b), along with a 25-fold improvement in retention time (c).Figure 6 A FinFET design is amenable to floating-body techniques.40 EDN | SEPTEMBER 17, 2009
the SOI-MOS structure to create charge (Figure 1). This approachallows the creation of a much larger charge and theability to store more charge because of the increased capacitanceof the memory cell.If you consider an N-channel device, the N source, the P-type body, and the N drain form the emitter, base, and collector,respectively, of an NPN (negative-positive-negative) bipolartransistor. The body of the MOS transistor is the baseof the bipolar transistor and acts as a storage node. Writing aone into a second-generation bipolar floating-body-memorycell triggers the intrinsic bipolar transistor, causing current toflow throughout the transistor body. This approach differs significantlyfrom the behavior of MOS, in which current flowsonly at the interface. Charge collects at the interface due to theslight bias at the gate. The impact-ionization effect that createsan excess of majority carriers in the floating body is more efficientin this bipolar bit-cell structure, quickly charging thebody and therefore resulting in rapid writes (Figure 2).You can read a floating-body memory using a similar mechanism,which senses the bipolar current through the transistor(Figure 3). Write current is close in value to the read current,and the latch-up characteristic of the intrinsic bipolar qualitycauses the behavior of the memory cell to appear nearly digital(Figure 4).Figure 7 The read and write currents for a FinFET with a lengthof 55 nm and width of 11 nm reveal its robust capabilities.CELL MARGIN AND SCALABILITYBipolar floating-body memories have a significantly higheroperating margin than traditional floating-body devices, allowingthem to create faster memory bit cells. The operatingmargin is better for two reasons: The storage charge is higher,and the difference between the one and the zero states is muchlarger because the bipolar element is a better amplifier than aMOS device (Figure 5). A high cell margin eases sensing usinga simple sensing scheme. The approach also simplifies thesense amplifier’s design. When you implement floating-bodymemories with conventional planar transistors, the memories’GAIN Experience. SAVE Time.Your Power Questions Answered in the LabEMI: How to Get the Lowest NoiseThermal & Mechanical ConsiderationsView Videos Nowatvicorpower.com/PT3Input Overvoltage ProtectionImproving Output FilteringPO ERTECHTORIALS Vicor's PowerTechtorial Series concentrateson important, real-world technical issues inpower system design. Power questions are answeredby senior applications engineers through concise,expert instruction in the lab. Gain access to view anever-growing number of Vicor PowerTechtorialvideos at vicorpower.com/PT3 .Vicor Applications Engineering: 800-927-9474
large voltage margin helps to mitigatethe negative effects of processfluctuations, which become increasinglysignificant at smaller processgeometries. Perhaps just as important,the new bipolar floating-bodymemorydesigns are compatible withadvanced, nonplanar devices, such asFinFET, multigate FET, and gate-allaroundFET. Older floating-body designswork on only thin-film planardevices.FINFETs AND ARRAYSFinFETs, surround-gate transistors,and other similar 3-D structuresshould provide the basis for futurestand-alone memories as the industryscales to ever smaller process geometries.Working at the 54-nm node,some DRAM companies are usingFinFET designs, and most suppliersshould follow in the next five years.Manufacturers can fabricate Fin-FETs and pillar transistors on bothSOI and conventional bulk substrates.In a FinFET or trigate-basedZ-RAM (zero-capacitor RAM), theTECHNOLOGYMEMORYTRANSISTORCELL SIZEMACRO SIZEDENSITYSUBARRYREDUNDANCYECCSENSINGROW DRIVERSINTE<strong>RF</strong>ACEPE<strong>RF</strong>ORMANCEFigure 8 You can implement a 4-Mbit macro in a 45-nm SOI process.45-NM PARTIALLY DEPLETEDSOIWIDTH: 112 NM, LENGTH:112 NM, OXIDE THICKNESS:3 NM0.4800.192 MICRONS(0.0922 MICRONS 2 )1323718 MICRONS(FOR 4.44 MBITS)0.21 mm 2 /MBITLOCAL SENSING WITH 260ROWS/BIT LINE AND 1120BITS/ROWFOUR ROWS AND 16COLUMNS PER SUBARRAY9 ECC CHECK BITS PER WORDVOLTAGE, EIGHT-BIT-LINEPITCH, CORE TRANSISTORSFOUR-WORD-LINE PITCH,WORD-LINE DRIVERS WITHBT DUAL-GATE OXIDE DEVICES140-BIT DATA WORD, 15-BITADDRESS, 7-BIT CONTROL4-NSEC RANDOM ACCESS,2-NSEC* READ LATENCY, 2-GHzCYCLE TIME*WORD LINE HIGH TO SENSE AMPLIFIER’S OUT TIME42 EDN | SEPTEMBER 17, 2009
charge accumulates under the transistor gate, andthe current flows in the middle of the fin structure.The fin stores charge throughout the structure,permitting excellent control of the bipolarcurrent (Figure 6). Bipolar floating-body-memoryimplementations using FinFET structures tendto exhibit an approximately 30-fold increase inmargin, leading to a proportional increase in signaland device speed. Figure 7 shows the cell currentduring write and read. Even an undoped cellwith an 11-nm fin shows clean digital behavior and a goodmargin. Floating-body memory continues to scale with technologywhether you fabricate on planar or 3-D structures.For the successful implementation of floating-body memories,it is crucial to create memory-bit-cell arrays. Original floating-bodyimplementations, due to the limited possible marginbetween the one and the zero states, are relatively slow andtherefore unsuitable for combining into useful arrays. However,new bipolar floating-body-memory cells, which produce superiorsignal margins due to the large current gain available fromthe bipolar device, enable more robust arrays (Figure 8).The macro for these memories supports a 2-nsec read latencyand a 4-nsec write latency. It also implements a high-speedpage mode to read a 140-bit word during every 2-GHz-processorclock cycle, using a shared-I/O bus and data latches at theinterface. This macro supports consecutive memory operationsat 4 nsec and a burst of four words during page mode. The 4-Mbit memory macro comprises 16 subarrays of 2561024-bit,+ Go to www.edn.com/ms4332 and clickon Feedback Loop topost a comment onthis article.+ For more technicalarticles, go to www.edn.com/features.or 256-kbit, cells (Figure 8). The exact implementation,2601120 bits, includes redundancyand ECC (error-correcting code). Subarrays sharetwo banks of sense amplifiers at each end of everysubarray and each bank with adjacent subarrays.A combination of column and row operations allowsthe access of four 140-bit words by simultaneouslyactivating two subarrays on the right andleft sides of a shared block of row drivers.A subset of combined read- and write-restore operationsconstitutes a refresh operation. A control block outsidethe 4-Mbit macro initiates refresh. A low-power refresh modeminimizes the voltage swings on inactive bit lines between theread- and write-restore operations of the refresh cycle. Thememory macro has a refresh cycle of 1 msec at 105C, alongwith three external power supplies of 1.1, 2.6, and 0.5V. A dedicatedon-chip voltage-generation block attaches to the memorymacro and supplies the required operational voltages.EDNAUTHOR’S BIOGRAPHYSerguei Okhonin, PhD, is co-founder and chief scientistof Innovative Silicon, where he has worked forsix years. He has a master’s degree in physics fromNovosibirsk State University (Novosibirsk, Russia)and a doctorate from the Swiss Federal Institute ofTechnology (Zurich, Switzerland). His personal interests includeskiing, rock climbing, and mountaineering. You can reach him atsokhonin@z-ram.com.DC- DC EMI NOISE ...CANCELLATIONPicor QUIETPOWER Active Filters suppress EMI AND save PCB spaceQUIETPOWER Typical DC-DC Converter Input EMIONLY 1 in 2High Performance Up to 60 dB CM attenuation Up to 80 dB DM attenuation Up to 14 A and 80 V Supports EN55022, Class B limitsVery Small Size
Maximum Efficiency.Minimum Power Loss.Energy-Efficient Wireless Basestation SolutionsDesigning for wireless basestations poses several design challenges,including how to optimize efficiency and deliver robust systemperformance. These challenges need to be addressed on thereceive side of a basestation when designing with amplifiersfor signal conditioning, timing products for frequency translationand clock generation, ADCs for high dynamic range dataconversion, and power supplies for powering signal pathand digital processing systems.MainDiversityPLLPLLDist.Shown: Wireless Basestation Application DiagramLow Noise, Low PowerIn high-performancebasestation designs, linearityand low-noise operation arekey to maximizing receiversensitivity. National’sLMH6517 DVGA, combinedwith the ADC16V130 16-bitADC, LMK04000 clock jittercleaner, and LMX2541frequency synthesizeroptimize next-generationmulti-carrier GSM, LTE,UMTS, and WiMax basestations.1:NADCADCPLLDACASIC/FPGASERDESPowerHigh-Power DensityHigh-power density andefficiency are criticalto reduce total powerconsumption, minimize heatgenerated by power losses,and improve system reliabilityand safety. National’s diverseportfolio of LM5000 powermanagement solutionsmaximize power density andend-to-end power chainefficiency.SignalConditionersDesign FlexibilityNational’s DS64BR401quad 6.4 Gbps transceiversperform signal conditioningon both input (EQ) and output(De-Emphasis) stages tomaximize basestation designflexibility and recover fromtransmission losses inducedby backplane or cableinterconnects.© 2009, National Semiconductor Corporation. National Semiconductor, , and PowerWise are registered trademarks. All rights reserved.national.com/comms
EDITED BY MARTIN ROWEdesignideasAND FRAN GRANVILLEREADERS SOLVE DESIGN PROBLEMSMissing pulse detects positionor produces a delayMichael C Page, Chelmsford, MAConsider an application that needs a series of pulses to indicateposition in which the lack of apulse “indexes” the count. To achievethat goal, the application uses a rotating,36-tooth sprocket with one missingtooth. Rotational speed rangesfrom 500 to 7000 rpm. The mechanismuses an inductive pickup to sensethe sprocket’s teeth. With one of thesprocket’s 36 teeth missing, the detectorsenses 35 pulses, and then a pulsedisappears.Unfortunately, the mechanism frequentlybreaks down or simply breaksapart. Because the application usesthis wheel just to trick the computerby simulating an operating engine,the application’s designers replacedthe rotating gear with a simulator circuit(Figure 1). Given the rotationalspeed and number of teeth, the maximumpulse frequency is 7000/6036,or 4200 Hz. The circuit works wellfrom single stepping to more than 1MHz before starting to break down.The maximum frequency dependson the logic family and constructionmethods you use.Figures 2 and 3 show the outputsDIs Inside47 Emulate SPI signalswith a digital-I/O card48 Resistive DAC and op ampform hybrid divider49 Connect two buttonswith just two wiresTo see all of EDN’s DesignIdeas, visit www.edn.com/designideas.running at 100 Hz and 1 MHz, respectively.At power-up, capacitor C 1remainsthe same, which forces RST onSIGNAL INR 61kR 71kIC 4BHC08SENSER 31kIC 1AHC14IC 1BHC14IC 4CHC08R 43.3kDETECTCLOCKD 4BZD23-4V7D 1D1N4148R 21kR 56.8k1CLKRSTIC 24024Q0Q1Q2Q3Q4Q5Q6D 2D1N41482D 3D1N414832DSET QIC 3A4013QNRSTPOWERRESETR 11kC 11 nF5VV 1IC 4AHC08Figure 1 IC 2combines with three diodes to produce a stream of 36 pulses before resetting.SEPTEMBER 17, 2009 | EDN 45
designideasVOLTAGE(V)5432100 50 100 150 200 250 300 350 400TIME (mSEC)50 mSEC/DIVFigure 2 Operating at 100 Hz, the circuit signals include the clock-sine-wave signal (red), the sense-square-wave signal(green), and the detect signal (blue), which indicates the missing pulse.54VOLTAGE(V)32100 5 10 15 20 25 30 35 40TIME (SEC)5 SEC/DIVFigure 3 The pulse train at a clock frequency of 1 MHz still shows the missing 36th pulse along with the power-resetsignal (blue).IC 3Alow. That action puts the D flipflopinto a known state. As C 1chargesthrough R 1, the voltage at the powerreset falls, letting clock pulses set IC 3A’soutputs. You must keep the small valuefor the C 1/R 1combination if youuse a high input frequency with a lowcount rate. As Figure 3 shows, the desiredcount must exceed the durationof the power reset. The values in Figure1 provide a time of approximately0.661 k (the value of R 1)1 nF(the value of C 1), or 0.66 sec, leavinga minimum count of approximatelythree at 1 MHz.For the clock signal, the circuituses a sine-wave signal with an amplitudeof 5 to 10V from the system.The clock signal goes through R 3toD 4and IC 1Ato produce a 5V squarewavesignal. The signal goes to counterIC 2and to one input of AND gateIC 4B. With the other input of IC 4Bcoming from IC 3A’s QN output, whichis high from power reset at start-up,the input-pulse train passes throughIC 4B, which simulates sprocket teethat the sensor. Resistors R 6and R 7halve the clock-signal amplitude justto make the graphics clear at “signalin.” Diodes D 1, D 2, and D 3pull up to5V through R 2and form an AND gateto select the desired count. CounterIC 2’s outputs are binary, so, for a36-tooth sprocket with one missingtooth, outputs Q0, Q1, and Q5 correspondto 123235.You can produce any count as highas 128 by adding the appropriate diodeson the Q outputs on IC 2. In otherwords, you need to generate onemissing pulse of 36 to simulate the36-tooth sprocket. Thus, you select acount of 35; the circuit automaticallyadds a count of one due to the oneclockdelay of the counter. Becauseyou reset IC 2at power-up, all outputsare low, keeping the D input of IC 3Alow, with a count of zero.THE CIRCUIT AUTO-MATICALLY ADDS ACOUNT OF ONE DUETO THE ONE-CLOCKDELAY OF THECOUNTER.As clock pulses continue into IC 2and when outputs Q0, Q1, and Q5 areall high, with a count of 35, IC 3A’s Dpin pulls high through R 2. On the nextclock pulse, the Q output of IC 3Agoeshigh and the QN output goes low, stoppingthe pulses from passing throughIC 4B. This action indicates the missingtooth and produces the sense condition(the missing pulses in figures 2 and 3).Meanwhile, the Q output of IC 3A’s outputgoes high, yielding a single detectpulse at IC 4Cthrough R 4and R 5. Onthe next clock pulse, with IC 3A’s Qoutput high, IC 2resets logic zero and isready for the next count cycle. R 4andR 5halve the clock signal just to makethe graphics clear at “detect.”The 4024 is an eight-stage binaryripplecounter. You can replace it witha 4040 counter to achieve a count of2048, and you can cascade counters toget longer counts or delays. The 4040’spinout differs from that of the 4024,but their operation is identical. Somesystems have an extra tooth instead ofa missing tooth, and some have multiplemissing teeth at odd locationsaround the sprocket, all waiting for replacementby this simple circuit.EDN46 EDN | SEPTEMBER 17, 2009
Emulate SPI signalswith a digital-I/O cardAndy Street, Autoliv Electronics, Lowell, MAA design-verification tester for millimeter-wave SOC (systemon-chip)devices needed to combineswitching, electrical measurements,temperature measurement, a paralleldigitalinterface, and a serial-digital interfaceinto one instrument. To minimizerack space, the circuit uses an AgilentTechnologies (www.agilent.com)34980A multifunction mainframe becauseits plug-in cards could support aforce/sense dc matrix and multiplexedtemperature measurements. The additionof an Agilent 34950A 64-bitdigital-I/O card formed the basis of asystem that could provide both an SPI(serial-peripheral-interface) bus and asimple parallel bus. The 34950A groupsits I/O lines into two banks of four 8-bit channels. It provides 64 kbytes ofmemory per bank for pattern generationor signal capture. It also has threeI/O lines per bank for handshaking.888834950ABANK 1D00CH01D07D08CH02D15D16CH03D23D24CH04D31H0H1H2D CLKD SSD DATAH 0START/STOPH 1STROBEDATAT PDb O7T/2YOU CAN STOREA MAXIMUM OF 32TRACES IN THE PAT-TERN RAM PER BANK.Tb O6b O5SPI PROTOCOLb O4INVALID VALID VALID VALID VALID VALID VALID VALID VALIDb O3b O2b O134950A SYNCHRONOUS BUFFERED OUTPUTFigure 1 The 34950A synchronous buffered output uses the falling edge, making it unsuited torising-edge SPI implementations.However, the card’s handshake linesprovide insufficient control for implementingSPI transactions. To get adequatecontrol, you can emulate the SPIbus using three of the data-I/O lines.Motorola (www.motorola.com)micro controllers first used the SPImaster-slave protocol. Today, it’s becomethe control interface in a varietyof ICs, including PLLs (phaselockedloops) and <strong>RF</strong> ASICs (references1 and 2). The SPI bus uses theclock, SS (slave-select), MOSI (master-out/slave-in),and MISO (masterin/slave-out)lines. The clock line is asignal from the master to the slave. AllSPI signals are synchronous with thisclock. The SS line selects the slave forcommunication. The SPI specificationdefines four modes of operation, whicheffectively specify the clock edges fortoggling and sampling and the clockidlelevel. The specification makes norequirements on voltage levels or datarates, and many SPI implementationscan clock in excess of 10 MHz. Figure1 shows a block and timing diagramof the 34950A’s Bank 1, configuredfor synchronous, buffered output.H0 through H2 denote the handshakelines. The figure also shows an 8-bitSPI transaction for reference.You cannot use the 34950A’s handshakelines to emulate all modes of theSPI bus because the bus latches dataon the falling edge of the clock, makingthe bus unsuitable for slaves thatuse the rising edge. Inverting the clockpolarity is not a solution because youmay lose the last data bit. Furthermore,if you write a number of transactionsto a slave, you must store each transactionas a separatetrace memory in the34950A. Althoughb O0CLOCKSLAVESELECTMASTEROUT/SLAVE INeach bank supports64k8 bits, you canstore a maximum of32 traces in the patternRAM per bank,thereby limiting thenumber of SPI transactions.In addition,the card lacks a sequencer,so you cannotdownload a numberof bit patterns andthen play them in sequence.You must loadeach pattern into theI/O card’s memoryand then play eachpattern under SCPI(standard commandsfor program mable instruments)from ahost computer, slowingtransactions.Instead of using thehandshake lines, thissolution uses three da-SEPTEMBER 17, 2009 | EDN 47
designideasta-I/O lines to emulate the SPI clock,SS, and MOSI. The software driver forthe I/O cards then has the responsibilityof translating the data to be sentinto an SPI-compatible bit stream.Listing 1, which is available at www.edn.com/090917dia, contains the algorithmin pseudocode, which translatesa hexadecimal string, DH, of charactersto an SPI signal. LD, LSS, and LCLKare integers to define which data outputsrepresent the MOSI, CLK, and SS,respectively.Assuming a 24-bit register writewith two bits of overhead for the SSprefix and postfix, the 64-kbyte memorycan support more than 1000 SPItransactions. The approach has twoadditional advantages: The three linesthat form the SPI bus are under softwarecontrol, which provides cablingflexibility, and the implementationcan support multiple slaves throughthe use of additional SS lines. Figure2 shows an MSO (mixed-signaloscilloscope)screen that shows theSPI transaction. The SPI clock rate is5 MHz, which the 34950A’s internal10-MHz clock limits. The differentpayload sizes correspond to writingdata to 16- and 24-bit registers withinthe slave.EDNFigure 2 An MSO screen shows the SPI transactions using the digital-I/O lines.REFERENCES1 Leens, Frederic, “An Introductionto I²C and SPI Protocols,” IEEE Instrumentation& Measurement,Volume 12, No. 1, February 2009,pg 8, www.imm.ieee-ims.org/docs/ColumnsFebruary2009.pdf.2 “SPI Block Guide V04.01,” FreescaleSemiconductor, July 2004, www.freescale.com/files/microcontrollers/doc/ref_manual/S12SPIV4.pdf.Resistive DAC and op ampform hybrid dividerMarián Štofka, Slovak University of Technology, Bratislava, SlovakiaA resistive DAC in a resistivefeedbackloop of an op amp letsyou create an analog-digital-analogdivider. The resistance, R WA, betweenthe W and A terminals of the AnalogDevices (www.analog.com) AD5293(Figure 1) decreases linearly with increasingthe digital-control data, D:1024DRWA( D)=RAB1024 × ,and the value of the R WB, the resistancebetween the W and B terminals of theDAC, rises proportionally to D asDRWB( D) = × RAB.1024R ABis a constant value of resistance betweenthe ends of the digital potentiometer.The circuit uses resistance R WAas a feedback resistor, and resistanceR WBconnects between the invertinginput of the op amp and ground. Thevoltage gain of the noninverting amplifierbecomesRAV= 1+ WA = 1024RWBD .The output voltage isV = V × 1024OUT IN .DBoth the input voltage and the digitalinputdata can be time variables, andthe clock frequency for fetching digitalinputdata can be as high as 50 MHz.The potentiometer’s data sheet providesthe ground-referred parasitic capacitancesat the A, B, and W terminalsof the potentiometer. Thoroughmeasurement of the capacitances atthese terminals provides enough datato determine capacitances between theterminals. An evaluation of the measureddata shows that the direct capacitancebetween the A and W terminalsat the midscale position of the wiper isjust 2.4 pF:CAW ( X = ½) 24 . pF .If you assume that the five segmentsof the potentiometer are ordered topologicallyinto a chain, then the directintercapacitance between the A and Bends of potentiometer isCAB( X = ½) ½CAW( X = ½) 12. pF.The capacitance per segment of thefive segments of the potentiometer is48 EDN | SEPTEMBER 17, 2009
CSEGM5 CAB( X = ½)6pF,where X 1 ⁄2 denotes the midscale ofthe resistive DAC.The five-step distributed RC line ofthe potentiometer has a time constantofRτ SEGM = AB × CSEGM= RAB×5CAB = 24 NSEC,where R ABis 20 k. The ground-referredwiper capacitance, C W, of 40 pFis much higher than the intercapacitancesand creates a time constant:τW RWB× CW.DIGITAL INPUTSPIEXT_CAPYNC1 F 100 nFANALOG INPUTX10111213147AD52935 OR 3.3VV LOGICRESET8 6 1 3C ASDIN V DDASCLKSYNCSDOW 4RDYC WGND V SS B9 2 5The feedback network of the amplifieris frequency-compensated for SEGM W. Thus, you can calculate the valueof R WBas 600, meaning that the voltagegain of the amplifier, A V, is 32.3.For gains higher than 32.3, the effect ofC Wbecomes negligible, and you neednot bother about amplifier stability. Tosuppress the derivative behavior of theamplifier for gain values of two to 32.3,you can add a 40-pF compensating capacitorin parallel to feed back part ofthe potentiometer. The amplifier thushas an integrating character for allgains down to a value of two.You fetch the divisor, Y, which is adigital-data word, D, through a standardSPI (serial-peripheral interface).After power-on, you must initially neutralizethe write-in protection of theresistive DAC. You have to first programthe control bit C 1to the value ofone, whereas it is zero by default. Youachieve this task by clocking in theword containing C 3, C 2, C 1, and C 0,C B47 pFNCAD867715VNC1287 563 415V100 nF100 nFFigure 1 The resistive DAC-potentiometer forming the feedback for an op ampcontrols the op amp’s gain as inversely proportional to the digital-input-dataword. The circuit thus becomes a two-quadrant divider.OUTwhich equals 0110, and you put thedesired C 2and C 1values at data positionsD 2and D 1. After performingthese steps, you change the wiper positionin which the control bit is C 3, C 2,C 1, and C 0, which equals 0001, and thedata bits, D 9to D 0, represent the gainas 1024/D.EDNConnect two buttonswith just two wiresFikret Yilmaz, Mobil Elektronik, Istanbul, TurkeySometimes, you need to read the status of pushbuttons that are as much as 5m awayfrom your electronic circuit. That task is easy ifyou have just one button. You simply design aconstant-current source, connect the current linefrom your button, and measure the current in theline. If you press the button, current flows throughit. Otherwise, current does not flow.Problems occur, however, when you need toread two or more buttons. Several approachesto this problem are available. For example, youcould use an RS-485 interface with two wiresfor communication and two for power. Alternatively,you could use a single-wire connectionR 11 T 1 1k1312 TO 12VAT 30 mA7 9R 21kS 11 2S 21 2D 11N4004D 21N40041212IC 14N35IC 24N35Figure 1 You can connect two buttons using diodes.654654V CCR 310kR 410kD OUTS1D OUTS2SEPTEMBER 17, 2009 | EDN 49
designideasR 11 T 1 1k1312 TO 12VAT 30 mA7 9R 21kS 11 2S 21 2S 31 2 D 31N4004S 41 21212IC 14N35IC 24N35654654V CCR 310kR 410kD IC 131N40044N35 61D 21N4004R 51k254D 41N4004R 61k12IC 44N35654D OUTS1D OUTS2Figure 2 By adding a third wire, you can connect four pushbutton switches.V CCR 710kR 810kD OUTS3D OUTS4with one wire for communication andtwo for power. Another option is touse separate wires for each button.In that case, you would use one morewire than there are buttons. Finally,you could use a POE (power-over-Ethernet) approach, employing fourwires for communication and power.All of these approaches require a buttonreader or a controller, which youmust program, adding complexity andcost.The circuit in Figure 1 shows youhow to connect two buttons using diodes.Because the diodes steer the current,the circuit needs just two wires.On a positive cycle from the transformersecondary and with switch S 2closed, current flows through IC 1, R 2,and D 2. Thus, output D OUTS2pulls low.Conversely, if S 1closes on a negativecycle, then current flows through D 1to R 1and IC 2, which pulls D OUTS1low.The circuit in Figure 2 extends theconcept to four pushbutton switchesby adding a third wire.EDNBest-in-Class Device ProtectionAngular Magnetic EncoderAS5163 - 14-bit, Single Wire InterfaceAutomotive Magnetic Rotary Encoder27V overvoltage protection-18V reverse polarityIntelligent short circuit protection-40°C to +150°CGet your freesamples today!5VclampinghighT 2yOutput VoltageT 1yclampinglow0V0°Programmable OutputSet point highSet point lowT 1 Angle Range T 2 360°West Coast (408) 345-1790 · East Coast (919) 676-5292www.austriamicrosystems.com/AS516350 EDN | SEPTEMBER 17, 2009
RISE ABOVENew Xilinx Targeted Design Platforms give designers a boost to takeinnovation to a much higher level. These comprehensive platforms combine Virtex ® -6or Spartan ® -6 devices, the ISE Design Suite 11, development hardware, IP, referencedesigns, documentation, and service and support—so you’re way ahead of the competitionbefore you even start. Plus, everything works together seamlessly, so design teams canfocus on product differentiation, add more value, and make the whole project moresuccessful. Learn more now at www.xilinx.com/6.© Copyright 2009 Xilinx, Inc. XILINX, the Xilinx logo, Virtex, and Spartan, are trademarks of Xilinx in the United States and other countries. All other trademarks are the property of their respective owners.
For RoHS CompliantConsult FactorySurface MountAll PICO surface mount unitsutilize materials and methodsto withstand extremetemperatures of reflowprocedures withoutdegradation of electricalor mechanical characteristics.PICO Electronics, Inc.Call Toll Free… 800-431-1064 • FAX: 914-738-8225E-Mail: info@picoelectronics.com143 Sparks Ave., Pelham, NY 10803-1837PICOTransformersSU<strong>RF</strong>ACE MOUNTSU<strong>RF</strong>ACE MOUNT • PLUG-IN • AXIAL LEADS • TOROIDAL • INSULATED LEADSCLASS V (155°C)AVAILABLELow Profilefrom.24″ht.AUDIO TRANSFORMERSImpedance Levels 10 ohms to250k ohms, Power Level to 3Watts, Frequency Response±3db 20Hz to 250kHz. All unitsmanufactured and tested toMIL-P<strong>RF</strong>-27. QPL units available.Special order class V (155°C).POWER & EMI INDUCTORSUltra-Miniature inductors areideal for Noise, Spike andPower Filtering Applicationsin Power Supplies, DC-DCConverters and SwitchingRegulators. All unitsmanufactured and testedto MIL-P<strong>RF</strong>-27. QPL unitsavailable. Special orderclass V (155°C).PULSE TRANSFORMERS10 Nanoseconds to 100Microseconds. ET Rating to150 Volt-Microsecond. Allunits manufactured and tested toMIL-P<strong>RF</strong>-21038. QPL unitsavailable. Special order class V(155°C).400HZ TRANSFORMERS115 volt or 26 volt primary,Split secondary voltages, 5 voltto 310 volts, 150VA, Manufacturedand test to MIL-P<strong>RF</strong>-27. Specialorder class V (155°C).Delivery–stock to one weekfor sample quantitieswww.picoelectronics.comCLASS V (155°C) AVAILABLEActualSizeSurface Mountelectrical equivalentsof QPL-MIL-P<strong>RF</strong>-21038/27DC-DC CONVERTER TRANSFORMERSPOWER INDUCTORSCOMMON MODE EMI INDUCTORSThese units have gull wing constructionwhich is compatible with tube fedautomatic placement equipment orpick and place manufacturing techniques.Transformers can be used forself-saturating or linear switching applications.The Inductors are ideal fornoise, spike and power filtering applicationsin Power Supplies, DC-DCConverters and Switching Regulators.See PICO’s full Catalog immediatelyon the internetwww.picoelectronics.comPICO’S MULTIPLEXDATA BUS PULSETRANSFORMERSMEET MIL-P<strong>RF</strong>-21038/27These pulse transformers manufactured toMIL-STD-1553. COMMAND/ RESPONSEMULTI-PLEX DATA BUS requirements.They also are manufactured to MIL-P<strong>RF</strong>-21038/27 specifications and are designedto meet MAC AIR SPECIFICATIONSA3818, A5690, A5232, and A4905.All of these trasformers exhibit superiorelectrical performance. Common moderejection ratio is greater than 45dB at 1MHz. Input impedance is greater than3000 ohms over the band from 75 kHz to1MHz at 1V rms. This series possessesexceptional waveform integrity: Rise timeand fall time is less than 100 nanoseconds.Overshoot and ringing is less than±1V peak. Droop is less than 20%.J LEADGULL WINGCall toll Free 1-800-431-1064or send direct for FREE 178 pg.PICO catalog featuringTransformers, Inductors,DC-DC Converters,AC-DC Power SuppliesISO 9001:2000 CertifiedQPL UnitsAvailableMIL-P<strong>RF</strong>-27/76MIL-P<strong>RF</strong>-27/96MIL-P<strong>RF</strong>-27/97MIL-P<strong>RF</strong>-27/103MIL-P<strong>RF</strong>-27/277MIL-P<strong>RF</strong>-27/290MIL-P<strong>RF</strong>-27/357MIL-P<strong>RF</strong>-27/358MIL-P<strong>RF</strong>-27/359MIL-P<strong>RF</strong>-27/172
productroundupPOWER SOURCESOffline BCM arraysinclude a verticallymounted heat sinkClaiming 95% efficiency, the 650W,vertically mounted VI Brick BCMarrays provide isolation and conversionfrom 380V to 12 or 48V. Aiming atfront-end applications, the devices have352 and 384V nominal voltages and 11,12, 44, and 48V-dc output voltages. Themodules yield 290W/in. 3 power densityand fast transient response. The BCMbuses offline power to the motherboardand converts it to 12 or 48V, minimizingdistribution losses and reducing conversionsteps and overall cost. Devicesin the VI Brick BCM-array family cost$109 each.Vicor Corp, www.vicorpower.comTriple-output dc/dc module includesswitch-mode and linear regulatorsThe LTM4615 triple-output dc/dc Module regulatorsystem contains two 4A switch-mode regulators anda 1.5A low-dropout linear regulator. The switching regulatorshave an adjustable 0.8 to 5V output voltage and regulatethree outputs at 1.5, 4, and 4A or two outputs at 1.5 and8A while operating from one, two, or three input supplies.The regulator system features short-circuit and overtemperatureprotection and provides a 2% total dc-output errorfor switch-mode regulators and 1% error for low-dropoutlinear regulators. Claiming 90% efficiency, the device operatesover a 40 to 125C temperature range. Available ina 15152.8-mm LGA package, the LTM4615 triple-outputregulator costs $17.20 (1000).Linear Technology Corp, www.linear.comSwitching regulator providesalternative to linear regulatorsThe 96%-efficient, 0.5A V78XX-500-SMT dc-switching-regulatorseries provides a high-performance alternativeto linear regulators. Features include a 4.5 to 28V-dcinput range; 3.3, 5, 12, and 15V-dc regulated output voltage;500-mA output current; 40 to 75C temperature range at100% load; and 60% load derating at 85C. The converter providesshort-circuit protection, thermal shutdown, 10-mV p-pModFLEX gives you thePOWER & FREEDOMto bring your wirelessproduct to marketYOUR WAY.OPTION 1 :Purchase certifiedoff the shelf modules.OPTION 2 :License certified designfiles at anytime.*LS Research offers royalty free license options.The leader in wireless product developmentLearn more at: www.lsr.com/productsSEPTEMBER 17, 2009 | EDN 53
productroundupPOWER SOURCEStypical ripple and noise, and a 2 millionhourMTBF. Measuring 15.248.57mm, devices in the surface-mountV78XX-500-SMT dc-switching-regulatorseries cost $8.24 each.V-Infinity, www.v-infinity.com10A synchronousbuck converter hasintegrated FETsCombining a 10A synchronousbuck converter with integratedFETs, the 1-MHz TPS51315 dc/dcswitcher uses a D-Cap Mode controlscheme, allowing fast transient responseand reducing the required number of externaloutput capacitors by 32%. Theconverter meets Energy Star/90 Plusguidelines using an auto-skip mode andan Eco-mode light-load-control scheme,achieving an improved efficiency acrossthe entire load range. Additional featuresinclude a 3 to 14V input voltage,and a 0.75 to 5.5V output-voltage rangeenables flexible design in 3.3, 5, and12V power systems. Available in a 57-mm QFN package, the TPS51315 dc/dcswitcher costs $3.80 (1000).Texas Instruments, www.ti.com240W one-eighth-brickmodule providesdigital interfaceAllowing monitoring using a PM-Bus interface, the BMT454 eighthbrickfamily delivers 240W and a 36 to75V input-voltage range. The devices are95% efficient at 53V input and 12V outputat 20A full loads. They target information-and communication-technologyapplications requiring 48V power. Theseapplications include radio base stations,servers, and routers. The devices also suituse as intermediate-bus converters withregulators or as dc/dc modules poweringhard drives and fans. The modules featurea 1500V-dc input-to-output isolation fig-AAA, AA, C and D-CellBattery HoldersFor Primary & Stand-By powerMPD offers a wide diversity of popular sizes, styles, terminations,options and accessories. The fact is, our website features theindustry’s largest array of battery holders, including extensivemedical and military types.Key Features • Engineered designs assure optimum batterytightness • Low contact resistance • Accepts NiCd, CarbonZinc, & Alkaline batteries • Molded of flammability rated highimpact resistant plastic • Broad temperature service range• Molded-in contacts for PC board wave soldering • 6'' UL-CSArated wire • Male/Female 9V snap fastener terminalsFor details: write, call, fax or visit our websitewww.batteryholders.com54 EDN | SEPTEMBER 17, 2009
ure, allow output-voltage adjustmentsbetween 8.1 and 13.2V with a 2%output tolerance, and have 18W/cm 2power density. Available in 9V/20A,12V/20A, and 12V/2A models withouta digital interface, the BMR454costs $37 (10,000).Ericsson Power Modules,www.ericsson.com15W converters come inopen-frame and shielded-metal-caseoptionsThe fully isolated, 15W PXAopen-frame and PXB shieldedmetal-casedc/dc converters provide88% efficiency and have a 40 to85C operating temperature. ThePXA series has single-output modelswith 24V nominal inputs and 48Vdc in 2-to-1 and wide 4-to-1 versions.The PXB series comes in single- anddual-output models with 12V-dc nominalinputs in a 2-to-1 version and 24and 48V-dc inputs in 2-to-1 and wide4-to-1 versions. Single-output voltagesinclude 3.3, 5, 12, and 15V dc, and thePXB series offers dual-output modelsproviding 5, 12, and 15V-dc outputs.Standard models include remoteon/off and output adjustment; singleoutputmodels and devices with overvoltageand overcurrent/short-circuitprotection are also available. Measuring11 in. each, devices in the PXAand PXB series cost $28 (500).TDK-Lambda Corp,www.us.tdk-lambda.comproductmartThis advertising is for newand current products.INTEGRATED CIRCUITSOp amp uses EMIsuppressionfiltersConsuming 560 nW, the LPV521op amp suits wireless remotesensors, power-line monitoring, andmicropower oxygen and gas sensors.Features include a 1.6 to 5.5V voltagerange with a 0.4-A maximum supplycurrent, a maximum 1-mV input offsetvoltage, and a 3.5-V/C input offsetvoltagedrift. The op amp comes withEMI-suppression filters, reducing <strong>RF</strong>Ifrom external sources. Available in afive-pin SC-70 package, the LPV521op amp costs 65 cents (1000).National Semiconductor,www.national.comVoltage-reference ICscome in multipleaccuracy gradesThe CAT8900 voltage referenceaims at handheld medicaldevices, high-resolution ADCsand DACs, and precision-regulatorsystems. The device offers 1.024,1.2, 1.25, 1.8, 1.048, 1.5, 2.6, 3, and3.3V voltage-reference options; customvoltages are available on request.The device provides 0.02% initial accuracyat 0.5 mV, a 20-ppm/C tem-perature coefficient, and an 800-nAmaximum supply current. The voltagereference sources or sinks as much as10 mA of load current with 50 mV ofdropout; the devices require no output-bypasscapacitor for most applications.The devices have an eight- to12-week leadtime and come in threeleadSOT-23 packages. The CAT8900voltage reference comes in 5-, 2.5-,ADVERTISER INDEXCompanyPageAchronix Semiconductor C-3Agilent Technologies C-221, 29Analog Devices Inc 17austriamicrosystems AG 50Bergquist Co 24Coilcraft 7Digi-Key Corp 1Express PCB 54International Rectifier Corp 5Keil Software 33Linear Technology Corp C-4LPKF Laser & Electronics 36LS Research 53Magna-Power Electronics Inc 28MathWorks Inc 19Maxim Integrated Products 10, 11Memory Protection Devices 541-, and 0.5-mV initial accuracygrades and costs 60 cents, 92 cents,$1.55, and $2.44 (1000), respectively.On Semiconductor,www.onsemi.comCompanyPageMentor Graphics 12Micro Crystal 35Mill Max Manufacturing Corp 9Mouser Electronics 4National Instruments 2National Semiconductor 44Pico Electronics 2334, 52Renesas Technology Corp 37Samsung Electro-Mechanics 42Silicon Labs 38Tadiran Electronic Industries 8Trilogy Design 55Vicor 41, 43Xilinx Inc 51EDN provides this index as an additional service.The publisher assumes no liability for errors oromissions.SEPTEMBER 17, 2009 | EDN 55
TALES FROM THE CUBEJACOB BRODSKY • WASHINGTON SUBURBAN SANITARY COMMISSIONDANIEL VASCONCELLOSLightning strikes sewage setupIn early 1988, we had finished the start-up of a brand-newSCADA (supervisory-control-and-data-acquisition) systemat the water-and-sewer utility where I work. It was summer,and we were busy repairing the damage from frequentthunderstorms. A co-worker, George, was supposed to providepostmortem repair and analysis of the RTU (remotetelemetry-unit)processor boards. Most of the 1-foot-square,multilayer processor boards had obvious scorch marks fromlightning damage. We’d typically see aburned switching-power-supply module,PCB (printed-circuit-board) burnsat the phone-line interface, or burns onan analog-input card. We’d then takea closer look at each of the sites wherethese boards originated and try to improvethe situation through more carefulgrounding, better surge suppressors,or another means.One morning, I noticed that Georgehad set aside a few RTU processor cards.When I asked about it, he told me thatthey were malfunctioning but that he’dbeen having a hard time with figuringout what was wrong, so he’d set themaside to work on when he had time. “Ijust repaired this one over here,” hesaid. “It was a single blown 74HC00gate; the other three in the same chipwere just fine.” I asked where anothersimilar card had come from. “You’re goingto love this,” he said, “All of theseproblem cards, except for one, are fromthe Cabin John interceptor.” I walkedaway, befuddled.I made a detour in my travels laterthat week to take a look at the CabinJohn interceptor, a waste-water meteringvault. Resembling a small brick outhouse,it sat at the bottom of a valley inthe woods. The RTU processor’s onlypurpose at that site was to gather flowdata from a large venturi tube, whichmeasures fluid pressures and velocities,in the vault below the building. Thisflow meter was important because itwas a change-of-custody billing meter.The saying “sewage flows downhill”isn’t just an expression; it’s also a fact oflife. The sewage in this area was headedfor a plant in another jurisdiction.The only two points of data comingfrom the site were flow and a loss-of-acpoweralarm. The RTU processor wasin a robust fiberglass box. At the backof the box, all the PCBs mounted flaton a grounded steel plate with ¼-in.standoffs and isolated screws. The setupalso included an I/O card, the processorcard, a battery-charge controllerboard, some terminal strips, an ac circuitbreaker, and the 928/952-MHz telemetryradio. A 100-foot tower stoodnext to the building to get the antennaabove the tree line.The installation gathered the flowdata from a 4- to 20-mA-current-looppressure transmitter next to the cabinet.The pressure-transducer electronics,batteries, and PCBs were isolated fromground. Sure that George must havebeen wrong about something, I wentback to his shop. He had set aside a fewmore cards, all with different failures.The next morning, as I stumbledinto the shower, I had a flash of insight:Maybe lightning was striking the towerand causing the ground potential ofthe backplate to rise. The board wasquite close to the backplate. Perhaps astatic charge was forming between thegrounded metal backplate and the isolatedRTU processor. A small arc couldbe destroying the weakest chips.I got to work early and snagged some1-in.-long nylon screws and standoffsfrom the maintenance shop. Replacingthe ¼-in. screws with the longer onesprotected the isolated board from randomchip damage, fixing the problem,and we haven’t seen failures of this sortsince.EDNJacob Brodsky is a control-system engineerwith the Washington SuburbanSanitary Commission (Laurel, MD).+ www.edn.com/tales56 EDN | SEPTEMBER 17, 2009
Breaking Through Performance BarriersSpeedster is theWorld’s Fastest FPGANew Speedster 1.5 GHz FPGAsBreaking through the speed barriers of traditional FPGAs,fully reprogrammable Speedster devices from AchronixSemiconductor are capable of operating at 1.5 GHzpeak performance.Speedster’s unique 40 lanes of 10.3 Gbps embedded SerDesand four independent 1066 Mbps DDR2/DDR3 controllersenable extremely high I/O throughput to match the device’soutstanding internal performance. At 3X faster, Speedsteropens up new worlds of application design previouslyunavailable to engineers using traditional FPGAs.Familiar Silicon. Familiar Tools. Fast Time-to-Market.Speedster uses familiar LUT-based fabric and standardsynthesis and simulation tools, so designers can usestandard RTL.High-Performance ApplicationsSee what’s possible:www.achronix.com/speedster
125Msps ADC at 125mW1.8VLTC226114-BitCMOSCMOS DDRLVDS DDROne-Third the Power of Comparable High Speed ADCsLinear’s newest high-speed ADC family achieves 1 /3rd the power consumption of existing solutions without compromising ACperformance. Operating from a low 1.8V supply, the 14-bit, 125Msps LTC ® 2261 dissipates 125mW while maintaining 73.4dBSNR and 85dBc SFDR. Digital outputs can be configured as DDR CMOS, DDR LVDS or standard CMOS for minimizing FPGApin count. Setting a new low power industry standard enables a wide variety of applications to “cut the cord” and move intothe portable world.FeaturesLowest Power ADC FamilyInfo & Free Samples• 73.4dB SNR, 85dB SFDR• Single 1.8V Supply• CMOS, DDR CMOS or DDR LVDSDigital Outputs• Optional Data Output RandomizerPart NumberLTC2261-14LTC2260-14LTC2259-14Speed125Msps105Msps80MspsPower125mW106mW89mWSNR73.4dB73.4dB73.4dBwww.linear.com/22611-800-4-LINEA<strong>RF</strong>ree High SpeedADC Brochure• Shutdown (0.5mW) andNap (9mW) Modes• 40-Pin (6mm x 6mm) QFN PackageLTC2261-12LTC2260-12LTC2259-12125Msps105Msps80Msps125mW106mW89mW70.8dB70.8dB70.8dBwww.linear.com/adcsolutions, LTC, LT and LTM are registered trademarks of LinearTechnology Corporation. All other trademarks are the property oftheir respective owners.
![[270].pdf 37407KB Sep 02 2010 09:55:57 AM - ElectronicsAndBooks](https://img.yumpu.com/50350834/1/185x260/270pdf-37407kb-sep-02-2010-095557-am-electronicsandbooks.jpg?quality=85)
![draaien, A Viruly 1935 OCR c20130324 [320]. - ElectronicsAndBooks](https://img.yumpu.com/49957773/1/190x252/draaien-a-viruly-1935-ocr-c20130324-320-electronicsandbooks.jpg?quality=85)
![20051110 c20051031 [105].pdf 35001KB Feb 18 2009 08:46:32 PM](https://img.yumpu.com/48687202/1/190x253/20051110-c20051031-105pdf-35001kb-feb-18-2009-084632-pm.jpg?quality=85)
Change language