26.12.2014 Views

Cognitive Radio Chip-Hardware Challenges - Hardware Conference

Cognitive Radio Chip-Hardware Challenges - Hardware Conference

Cognitive Radio Chip-Hardware Challenges - Hardware Conference

SHOW MORE
SHOW LESS

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

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

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Chip</strong>-<strong>Hardware</strong><br />

<strong>Challenges</strong><br />

Eric Klumperink<br />

Mark Oude Alink, Amir Ghaffari, Dlovan Mahrov, Saqib Subhan,<br />

Zhiyu Ru, Niels Moseley, Michiel Soer, Rameswor Shresta,<br />

Andre Kokkeler, Bram Nauta<br />

IC Design group, CTIT, University of Twente, Enschede<br />

http://icd.ewi.utwente.nl (all papers available on-line)<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

1


Overview<br />

• Introduction <strong>Cognitive</strong> <strong>Radio</strong> (CR)<br />

• <strong>Radio</strong> <strong>Hardware</strong> <strong>Challenges</strong> & Ideas<br />

• <strong>Cognitive</strong> <strong>Radio</strong><br />

• Flexible Receiver<br />

• Flexible Transmitter<br />

• Spectrum sensing<br />

• Conclusions<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

2


Static Frequency Allocation<br />

Almost all spectrum allocated - how find spectrum for mobile internet<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

3


Spectrum Allocated, but not in use!<br />

[Swisscom]<br />

<strong>Cognitive</strong> <strong>Radio</strong>: find & use “white spaces”<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

4


Applications<br />

more capacity more users, reliability, speed, ....<br />

Examples:<br />

• E.g. Broadband wireless access (IEEE 802.22)<br />

– 50-860 MHz (old TV-bands), Range: 17-30km<br />

• Reliable communication for Emergency Services<br />

other future<br />

applications ..<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

5


Overview<br />

• Introduction <strong>Cognitive</strong> <strong>Radio</strong> (CR)<br />

• <strong>Radio</strong> <strong>Hardware</strong> <strong>Challenges</strong> & Ideas<br />

• <strong>Cognitive</strong> <strong>Radio</strong> – Software Defined <strong>Radio</strong><br />

• Flexible Receiver – Clock Controlled Filter<br />

• Flexible Transmitter – Multi-path Architecture<br />

• Spectrum sensing – Mixed Signal X-correlator<br />

• Conclusions<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

6


<strong>Cognitive</strong> <strong>Radio</strong> <strong>Hardware</strong> Challenge<br />

Needed: Spectrum Sensor + Transmitter + Receiver<br />

Desired: flexibility<br />

– Frequency<br />

– Bandwidth<br />

– Modulation<br />

– ...<br />

“Software <strong>Radio</strong>”<br />

ADC<br />

DSP<br />

Analog-<br />

Digital<br />

Converter<br />

Digital<br />

Signal<br />

Processing<br />

How fast can we go<br />

DAC<br />

Digital-Analog<br />

Converter<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

7


ADC Bandwidth versus SNDR<br />

Frequency BW [Hz] [Hz]<br />

1,E+11<br />

1,E+10<br />

1,E+09<br />

1,E+08<br />

1,E+07<br />

1,E+06<br />

1,E+05<br />

<strong>Radio</strong> frequencies<br />

ISSCC 1997-2008<br />

VLSI 1997-2008<br />

ISSCC 2009<br />

VLSI 2009<br />

Jitter=1psrms<br />

Jitter=100fsrms<br />

1,E+04<br />

1,E+03<br />

10 20 30 40 50 60 70 80 90 100 110 120<br />

SNDR [dB]<br />

[Murmann, 2009]<br />

Is 40-50dB Signal to Noise-and-Distortion Range enough<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

8


How much Path Loss: “inverse square law”<br />

Antenna aperture<br />

~ (lambda/4) 2<br />

• Total power through sphere is constant<br />

• Received power in fixed area A reduces 1/R 2<br />

• Also less due to ground-reflection/multi-path: 1/R n with n~3<br />

10x more distances 1000x more power loss<br />

(10x power “10dB”, so 30dB per factor 10x distance)<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

9


Reduce path loss Parabolic antenna!<br />

+ Sensitivity<br />

- Big dish needed<br />

- Alignment needed, unpractical for mobile<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

10


Long distance large loss High SNDR<br />

SNDR<br />

• Assumptions for CR device: no antenna gain (non-directional)<br />

• Desired range for communication: at least e.g.1m..100m<br />

Result: Path loss ~ 10 … 70 dB + 10..30dB = 20..100dB<br />

(frequency-dependent 100MHz ..1GHz)<br />

6dB per ADC bit => >16 bits ADC needed<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

11


ADC Bandwidth versus SNDR<br />

BW [Hz]<br />

1,E+11<br />

1,E+10<br />

1,E+09<br />

1,E+08<br />

1,E+07<br />

1,E+06<br />

1,E+05<br />

1,E+04<br />

1,E+03<br />

10 20 30 40 50 60 70 80 90 100 110 120<br />

SNDR [dB]<br />

ISSCC 1997-2008<br />

VLSI 1997-2008<br />

ISSCC 2009<br />

VLSI 2009<br />

Jitter=1psrms<br />

Jitter=100fsrms<br />

<strong>Cognitive</strong> <strong>Radio</strong><br />

(6GHz and 100dB)<br />

[Murmann, 2009]<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

12


Downconversion / filtering required<br />

Resolution (bit)<br />

20<br />

16<br />

12<br />

8<br />

4<br />

1mW<br />

1 W<br />

1 kW<br />

1 W<br />

downconversion<br />

filter<br />

1kHz 1MHz 1GHz<br />

Signal Bandwidth (Hz)<br />

band<br />

filter<br />

Amplify<br />

f-shift<br />

(mixer)<br />

channel<br />

filter<br />

ADC +<br />

software<br />

“Software Defined <strong>Radio</strong>” if still very flexible through software<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

13


CR and SDR definitions by ITU<br />

• ITU = International Telecommunication Union<br />

• <strong>Cognitive</strong> <strong>Radio</strong> (CR):<br />

– Obtains knowledge from environment/user<br />

– Adjusts <strong>Radio</strong> parameters<br />

– Learns from results<br />

• Often a CR uses a Software Defined <strong>Radio</strong><br />

• Software Defined <strong>Radio</strong> (SDR):<br />

– Adjusts <strong>Radio</strong> parameter and protocols by software<br />

– At least frequency, modulation type, power<br />

– More flexible than predefined set by standard<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

14


Overview<br />

• Introduction <strong>Cognitive</strong> <strong>Radio</strong> (CR)<br />

• <strong>Radio</strong> <strong>Hardware</strong> <strong>Challenges</strong> & Ideas<br />

• <strong>Cognitive</strong> <strong>Radio</strong> – Software Defined <strong>Radio</strong><br />

• Flexible Receiver – Clock Controlled Filter<br />

• Flexible Transmitter – Multi-path Architecture<br />

• Spectrum sensing – Mixed Signal X-correlator<br />

• Conclusions<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

15


Mixer multiplication frequency shift<br />

sin<br />

f<br />

t<br />

sinf<br />

t sinf<br />

f t sinf<br />

f <br />

LO<br />

in<br />

LO<br />

in<br />

LO<br />

in<br />

t<br />

<br />

Downconversion<br />

(receiver)<br />

Upconversion<br />

(transmitter)<br />

f IN<br />

Difference and Sum<br />

frequencies<br />

f LO<br />

Lowpass<br />

Filter<br />

f LO -f IN<br />

Down-conversion<br />

f IN<br />

f LO<br />

Up-conversion<br />

f LO +f IN<br />

Band<br />

Filter<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

16


How implement RF band-filtering<br />

Q<br />

1<br />

R<br />

L/ C)<br />

AM-radio Receiver (1929):<br />

• Frequency selection by coil<br />

inductance L & capacitor C<br />

resonator, swing<br />

• Tuned mechanically by<br />

variable plate-capacitor C<br />

1<br />

center frequency : fc<br />

<br />

L C<br />

Now: fixed passive bandfilter<br />

(external SAW/BAW bandfilter)<br />

+ mixer with tunable frequency<br />

CMOS chip: limited Q=f c /bandwidth


Do we really need Filters Why<br />

Filter<br />

harmonics and<br />

much more ....<br />

2 3 4 5<br />

non-linearity!<br />

BB<br />

RF<br />

Filter<br />

Base Band (BB)<br />

ω<br />

(digital LO-generator)<br />

Time-variant BB<br />

to RF transfer!<br />

3 5<br />

But: good RF filters are costly and inflexible<br />

<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

3<br />

Can we implement suppression more flexibly<br />

5<br />

18


Flexible Integrated Filter Idea<br />

f S<br />

RC >> T on<br />

V in<br />

C 1 C 1<br />

average<br />

t<br />

V out<br />

C 1........<br />

C 1 ...<br />

C 4<br />

f S<br />

1.5f S<br />

0<br />

C 1 C 1<br />

[Franks, ISSCC 1960]<br />

“N-path filters”<br />

• High ohmic @ switching frequency<br />

• Short circuit @ other frequencies<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

19


Filter Properties<br />

Band-pass around f s<br />

Tunable (clock f s )<br />

High Q for high RCf s<br />

Good linearity & noise<br />

Unwanted harmonics<br />

V out / V in<br />

f S<br />

Frequency<br />

V out<br />

More paths Less Distortion<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

20


Filter Properties: flexibly tunable<br />

Selectivity (Q)<br />

Compression (P 1dB<br />

)<br />

Linearity (IIP3)<br />

Noise Figure<br />

3 to 29<br />

+2dBm<br />

+19dBm<br />


Overview<br />

• Introduction <strong>Cognitive</strong> <strong>Radio</strong> (CR)<br />

• <strong>Radio</strong> <strong>Hardware</strong> <strong>Challenges</strong> & Ideas<br />

• <strong>Cognitive</strong> <strong>Radio</strong> – Software Defined <strong>Radio</strong><br />

• Flexible Receiver – Clock Controlled Filter<br />

• Flexible Transmitter – Multi-path Architecture<br />

• Spectrum sensing – Mixed Signal X-correlator<br />

• Conclusions<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

22


<strong>Cognitive</strong> <strong>Radio</strong> Transmitter<br />

Current Dream: One multi-band Flexible transmitter Upconverter<br />

LO<br />

PA<br />

Filter<br />

BB<br />

Signal<br />

LO<br />

PA<br />

Filter<br />

LO<br />

PA<br />

Filter<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

23


Multipath Polyphase Technique<br />

Divide the nonlinear circuit into ‘N’ equal smaller pieces<br />

Nonlinear<br />

circuit<br />

Nonlinear<br />

circuit<br />

Nonlinear<br />

circuit<br />

Nonlinear<br />

circuit<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

24


Multipath Polyphase Technique (2)<br />

Add equal but opposite phase shift before and after<br />

Use equidistant phases: path i uses i*2π/N<br />

path i=1<br />

0×2π/N<br />

Nonlinear<br />

circuit<br />

-0×2π/N<br />

y 1<br />

x(t)<br />

path i=2<br />

1×2π/N<br />

Nonlinear<br />

circuit<br />

-1×2π/N<br />

y(t)<br />

y 2<br />

y N<br />

path i=N<br />

Nonlinear<br />

circuit<br />

( N-1)×2π/N -(N-1)×2π/N<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

25


Example: Output for 3-Path case<br />

Phase rotation increases with order of nonlinearity/harmonic:<br />

y 3<br />

y<br />

120<br />

120<br />

1 y 2 y 3 y 1<br />

y 1<br />

y 1 y 2 y 3<br />

0<br />

y 0<br />

0<br />

0<br />

3<br />

y 2<br />

Fundamental<br />

(y x)<br />

Magnitude:<br />

A<br />

240<br />

y 2<br />

2 nd Harmonic<br />

(y x 2 )<br />

240<br />

3 rd Harmonic<br />

(y x 3 )<br />

First non-cancelled<br />

Harmonic<br />

4 th Harmonic<br />

(y x 4 )<br />

2 3 4 5 6<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

26


N-PATH: First Non-cancelled Component<br />

A<br />

2 Path<br />

A<br />

3 Path<br />

A<br />

4 Path<br />

A<br />

‘N’ Path<br />

2 3 4 5 6<br />

2 3 4 5 6<br />

2 3 4 5 6<br />

2 3 4 5 (N+1)<br />

(N+1) th<br />

Harmonic<br />

Multi-tone input: cancels p· 1 ± q· 2 , except if: p+q=j×N+1<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

27


Concept of Flexible Transmitter<br />

BB<br />

path i=1<br />

Digital baseband<br />

phase shifters:<br />

0×2π/N<br />

path i=2<br />

1×2π/N<br />

Nonlinear<br />

circuit<br />

Nonlinear<br />

circuit<br />

Analog up-mixers<br />

with different phases<br />

LO<br />

LO<br />

0<br />

-2π/N<br />

y 1<br />

y 2<br />

y N<br />

-(N-1)×2π/N<br />

path i=N<br />

(N-1)×2π/N<br />

Nonlinear<br />

circuit<br />

LO<br />

[Shrestha et al, JSSC 2006]<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

28


Cancelling Harmonics/Sidebands<br />

1-path Desired signal, LO + BB LLO ±B BB<br />

Power<br />

18-path<br />

Power<br />

Desired signal<br />

3 LO +3 BB<br />

(3=0×18+3)<br />

Non-cancelled terms<br />

L=j×N+B<br />

17 LO - BB<br />

j=…-2,-1,0,1,2,…<br />

15 LO -3 BB<br />

(15=1×18-3)<br />

1 3 5 7 9 11 13 15 17<br />

[Shresta et al, JSSC 2008]<br />

/ LO<br />

29<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink


Overview<br />

• Introduction <strong>Cognitive</strong> <strong>Radio</strong> (CR)<br />

• <strong>Radio</strong> <strong>Hardware</strong> <strong>Challenges</strong> & Ideas<br />

• <strong>Cognitive</strong> <strong>Radio</strong> – Software Defined <strong>Radio</strong><br />

• Flexible Receiver – Clock Controlled Filter<br />

• Flexible Transmitter – Multi-path Architecture<br />

• Spectrum sensing – Mixed Signal X-correlator<br />

• Conclusions<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

30


Spectrum Sensing: shadowing problem<br />

C does not detect A due to “shadowing” of building<br />

need for sensing below noise floor<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

31


Linearity vs Noise<br />

Ideal output = actual input ( ) spectrum<br />

Difference 80dB<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

32


Linearity vs Noise<br />

Effect of non-linearity (too strong signals):<br />

(NF=20dB, IIP3=+10dBm, RBW=100kHz, att.=0dB)<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

33


Linearity vs Noise<br />

Apply attenuation: distortion OK but too much noise<br />

(NF=20dB, IIP3=+10dBm, RBW=100kHz, att.=48dB)<br />

Each dB of attenuation:<br />

- Noise floor up by 1dB<br />

- Distortion peaks down by 2dB<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

34


f test<br />

[Oude Alink et al, DySPAN 2010]<br />

Idea to improve Spectrum Analysis<br />

Idea to “get rid of the noise”:<br />

– a) Duplicate receiver<br />

– b) Cross-correlate outputs<br />

Result: correlated signal (and noise)<br />

Atten. ADC FFT<br />

out<br />

f test<br />

X-corr<br />

out<br />

Atten. ADC FFT<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

35


Cross-correlation measurement<br />

Normalized Measurement Time=2 0<br />

23<br />

45<br />

67<br />

89<br />

10 11 12 13 14<br />

test tone<br />

RBW=10kHz<br />

-100dBm<br />

DC-offset<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

36


XCSA: Measurements III<br />

RBW=10kHz<br />

signal not<br />

visible with<br />

autocorrelation<br />

more accurate<br />

amplitude estimation<br />

NF=17.1dB<br />

NF=3.1dB<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

37


Conclusions<br />

CR,SDR hardware less fixed analog filters<br />

stronger interference, nonlinearity, ...<br />

Ideas to address challenges:<br />

Clock controlled switched RC-filter<br />

Cancel via Multipath Polyphase transmitter<br />

Exploit mixed-signal cross-correlation<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

38


Questions<br />

<strong>Cognitive</strong> <strong>Radio</strong> <strong>Challenges</strong>, Eric Klumperink<br />

39

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!