CMOS Switched-Capacitor Circuits: Recent Advances in Bio-Medical and - - PowerPoint PPT Presentation

ASU August 17, 2011

CMOS Switched-Capacitor Circuits: Recent Advances in Bio-Medical and RF Applications

David J. Allstot

• Univ. of Washington
• Dept. of Electrical Engineering

Seattle, WA 98195-2500

ASU August 17, 2011

• 2010: 4.6 B

subscribers

• 2012: 1 B WiFi
• US mobile phones:
• Use yearly

energy of 638,000 US Homes

• Emit 6K tons

CO2

• Demand increases

with newer data phones

• PA is dominant

energy hog

Motivation

PA Metropolitan Seattle Area

2

ASU August 17, 2011

CMOS PA Trends: Pout

3

J.S. Walling, S.S. Taylor and D.J. Allstot, “A class-G supply modulator and class-E PA in 130 nm CMOS,” IEEE JSSC, pp. 2339-2347, Sept. 2009. S.-M. Yoo, J.S. Walling, E.C. Woo and D.J. Allstot, “A switched-capacitor power amplifier for EER/Polar transmitters,” IEEE ISSCC Dig. Tech. Papers, pp. 428-429, 2011.

ASU August 17, 2011

CMOS PA Trends: PAE

4

ASU August 17, 2011

Outline

• Challenges in CMOS RF PA Design
• Switched-Capacitor PA Solution
• Analog-domain Compressed Sensing for

Bio-signal Acquisition

5

ASU August 17, 2011

 



2

= 2

V V

• ut

DD SAT L

P R

VDD

• VDD Scaling
• Impedance Transformation

Challenge: Max Power Out

RL = 50 

Vout

1 : n Ropt = RL/n2

• 45 nm CMOS

1W, VDD = 1.0 V VSAT = 0.2 V Ropt 0.3 

Parasitic R Limit)

Linear PAs



6

ASU August 17, 2011

• Linear Power Amplifiers

– AM Signals (i.e., non- Constant Envelope) – Class-A: – Class-B: – Class-AB – Class-C: Peak = 100%

@ Pout = 0 (Attractive for Body Area Networks)

Challenge: Efficiency

 =

L DC

P P 

       

2

V = 0.5

• ut

DD

V

 

       

2

= 4

• ut

DD

V V

7

ASU August 17, 2011

ON OFF

• Switching Power

Amplifiers – PM and FM Signals (i.e., Constant Envelope)

• Class D, E, F, etc.
• Zero-V Switching

– Rise in vD delayed until switch OFF – vD = 0 @ switch ON

• dvD/dt = 0 @ switch

OFF

• Ideal = 100%

Impedance Transformer & Wave-Shaping Network

= = 0

DC D D

P v i

Challenge: Efficiency

Class-E PA

8

ASU August 17, 2011

0.2 0.4 0.6 0.8 1 20 40 60 80 100 Normalized Envelope (V) Ocurrences (%)

0.2 0.4 0.6 0.8 1 5 10 15 20 25 Normalized Envelope (V) Ocurrences (%) 0.2 0.4 0.6 0.8 1 5 10 15 20 25 Normalized Envelope (V) Ocurrences (%)

FM QAM 64-QAM

Spectral vs. Energy Efficiency

9

ASU August 17, 2011

Linearization Techniques

• Feedforward
• Feedback
• LINC – Linear Amp with Nonlinear Components
• EER – Envelope Elimination and Restoration

– Can use highly-efficient switching PA; e.g., Class-E – Pout  VDD for Switching PA – Split signal into envelope (A) & phase () paths – Improved overall efficiency – Distortion from delay mismatches in A &  paths

10

ASU August 17, 2011

A



Kahn EER Technique (1952)

• Polar conversion in DSP using CORDIC Algorithm
• DAC and supply modulator needed

Original Kahn Modern Kahn

11

ASU August 17, 2011

LDO = Low-dropout Reg. iout

LDO Modulator & Efficiency

LDO

• Overall efficiency is product
• f supply modulator and PA

efficiencies

• Increased over Linear PA
• LDO Characteristics

– Vout ≈ ENVin

– – –

• ut
• ut out

P v i 

DD out

DC

P v i  / /

• ut

DC

• ut

DD

P P v V

 



12

ASU August 17, 2011

Dual-Supply Modulator

Class-G: Spectral vs. Energy Efficiency

• Small envelope:

Use Vdd/x

• Large envelope:

Use Vdd

• Extend to

more than two power supplies? Class-H?

13

ASU August 17, 2011

0.2 0.4 0.6 0.8 1 20 40 60 80 100 Vout (V) Drain Efficiency (%)

Class-G Class-B OFDM PDF

0.2 0.4 0.6 0.8 10 2 4 6 8 10 Probability (%)

Avg. Class-B

Class-G: Spectral vs. Energy Efficiency

Avg. Class-G

• Overall efficiency

is product of class-G modulator and class-E PA efficiencies

• Ideally 5X higher

average  than linear PA for this probability density function

Class-E

14

ASU August 17, 2011

Class-E PA and Driver

Interstage tuning inductors reduce driver power Driver taper of 2 – custom stages

15

ASU August 17, 2011

130nm Class-G PA

16

ASU August 17, 2011

Class-G Static Measurements

0.2 0.4 0.6 0.8 1 200 400 600 800 1000 Input Envelope2 (V2) Output Power (mW) 0.2 0.4 0.6 0.8 1 16 32 48 64 80 PAE (%)

0.2 0.4 0.6 0.8 1 20 40 60 80 Normalized Envelope (V) Efficiency (%) 0.2 0.4 0.6 0.8 1 2 4 6 8 Probability (%)

Class G PAE 64QAM OFDM PDF Theory Avg PAE

• Meas. Avg PAE

64 QAM OFDM Symbol Period = 4 s  Theoretical avg. PAE = 24%  Measured avg. PAE = 22% Freq = 2 GHz

17

ASU August 17, 2011

Class-G Dynamic Measurement

• 80 -60 -40 -20

20 40 60 80

• 80
• 60
• 40
• 20

Frequency Offset (MHz)

• Norm. Output Power (dB)

rms EVM = 2.5% Freq = 2 GHz

18

ASU August 17, 2011

• DAC, supply modulator functions combined

– No supply modulator: Higher efficiency and smaller area

• Multiple unit current-cell-based PAs as DAC

19

PA based on digital modulation Unit current cells

[Kavousian, et al., ISSCC 2007 ] [Presti, et al., JSSC 2009]

Digitally-Modulated PA

ASU August 17, 2011

• Accuracy / Efficiency Tradeoff
• Accurate current cell requires high rout

– Cascode more headroom: Lower efficiency

• Extra resolution required for predistortion
• Efficiency:

20

OUT OUT OUT DC OUT Ideal

P V P P P    

Nonlinear VOUT Input Code VOUT

Linear Saturated

Current-Cell-Based PA

ASU August 17, 2011

Switched-Capacitor Basics

• Energy is lost w/ precharge and reset
• No energy lost in charge redistribution w/o precharge

(b) Charge Redistribution w/o precharge (a) Precharge and Reset

21

ASU August 17, 2011

SCPA in Polar Transmitter

22

ASU August 17, 2011

Basic SCPA Concept

23

• SC technique can be used for voltage generation
• Easy to split into capacitor bank (small area & loss)

– Resonant frequency maintained (Constant C) Constant envelope Good efficiency

ASU August 17, 2011

Switched-Capacitor PA

• Capacitor can be arrayed

– Single capacitor can be split into many – Each capacitor is switched to VDD or GND – Constant resonant frequency – RF Switched-Capacitor DAC

24

Constant Capacitance

ASU August 17, 2011

Thevenin Equivalent Circuit

• Digitally-controlled output voltage
• Constant top-plate capacitance vs. the

number of switched capacitors

25

CU=C1=C2=Cn=CN= N

C

Constant Capacitance = C

ASU August 17, 2011

Output Power

Pout delivered to ROUT

26

• VOUT  n/N
• POUT  (n/N)2
• 4/ for 1st harmonic

component

R V N n

DD 2 2 2

2       

POUT =

DD

V N n        4 2 1

VOUT =

ASU August 17, 2011

Power Dissipated in SC

27

• Charging & discharging

with switch → CV2f dynamic power

• Assume fast tr,tf with

constant current through L

• Effective switched

capacitance varies with envelope code

ASU August 17, 2011

Ideal Efficiency

• Higher efficiency with

higher QLoaded

• Higher QLoaded:
• Smaller Capacitance
• Less CV2f dynamic power
• Efficiency tradeoff

due to L & switch

28

fCR R fL Q Loaded   2 1 2  

SC OUT OUT

P P P   

Loaded

Q n N n n n ) ( 4 4

2 2

   

ASU August 17, 2011

Ideal  vs. Practical 

Normalized POUT (dBm)

Practical Efficiency

29

CLOCK DR SWC OUT SC OUT

P P P P P P          

OUT SC OUT Ideal

P P P   

• Practical implementation:

− Lossy inductor:  →  − SW parasitic R: →  − SW parasitic C: − Switch driver: − Clock distribution:

f V C N n P

DD SW SWC 2

) / ( 

f V C P

DD CLOCK CLOCK 2

 f V C N n P

DD DR DR 2

) / ( 

Benefit from scaling

Ideal  (%) Practical  (%)

ASU August 17, 2011

CMOS Switch as Voltage Source

30

0.25 0.5 0.75 1 n/N CB Voltage (CB) AM-PM AM-AM

time

1/fs VDD

• Faster switch improves both AM-AM and AM-PM

distortion performance (e.g., better with CMOS scaling)

ASU August 17, 2011

6-bit Switched-Capacitor Array

• Split into 4-bit unary and 2-bit binary arrays
• Additional bits possible

– More unary/binary bits or C-2C ladder

• Unit-cell switch and switch-driver

31

ASU August 17, 2011

Switch Implementation

32

• Cascode  More
• utput power with

same Rout

• Total supply

voltage of 2VDD

• All thin-gate

devices

• Separate driver

voltage ranges for NMOS & PMOS

ASU August 17, 2011

Switched-Capacitor PA Schematic

33

C= 8.2pF Bandpass Matching Network

ASU August 17, 2011 34

• 90 nm RF LP CMOS process (MIM cap and UTM)

Output Matching Network Capacitor Array Switch, Drivers, Logic & Bypass Capacitor

1430 m 730 m

Chip Microphotograph

ASU August 17, 2011

PA Measurement: Pout & PAE

35

• 6-bit implementation
• Fewer Pdriver at backoff
• Peak = 45%

ASU August 17, 2011

AM-AM & AM-PM / Pout vs. Freq.

36

• Different impedance seen

from source depending

• n input code
• Scaling friendly
• Peak Pout ≥ 24dBm
• Peak  ≥ 45%

ASU August 17, 2011

Constellation / Spectral Mask

37

• 64 QAM/OFDM
• EVM = 2.9%
• Pout = 17.7 dBm

ASU August 17, 2011

Reference

Degani, et. al. ISSCC 2008 Presti, et. al. JSSC 2009 Xu, et. al.

ESSCIRC 2010

Walling, et. al. JSSC 2009 This work

Architecture Class-AB

DPA Current Cell

Outphasing Class-G Switched- Capacitor Process 90nm 0.13um 32nm 0.13um 90nm Power Supply 3.3V 1.2V/2.1V 2V 3.3V 1.5V/3V Peak Power 25 dBm 25 dBm 25.1 dBm 29.3 dBm 25 dBm Peak Efficiency 50% 47% 40.6% 69% 45%

• Avg. Power

(OFDM) 15.5 dBm 15.3 dBm 18.6 dBm 19.6 dBm 17.7 dBm

• Avg. Efficiency

(OFDM) 19% 22% 18.1% 22.6% 27% Output Matching NW N/A Ext. Matching On-Chip Balun On-Chip Matching On-Chip Matching

38

Performance Comparison

• What’s next? Class-G SCPA in package – high PAE.

ASU August 17, 2011

Outline

• Motivation for Compressed Sampling (CS)
• Compressed Sampling and three key ideas
• CS reconstruction
• Experimental Procedures and Results
• Conclusions

39

ASU August 17, 2011

• Body Area Network

 Many wireless sensors linked to personal Smartphone, etc.  Personal mobile units linked to Dr. via internet/cellular network  Dr. feedback for real-time control of detail vs. energy efficiency

• Reduce data rates to increase sensor lifetime and energy efficiency

40

Motivation for Compressed Sampling

ASU August 17, 2011

Compressed Sampling Sensor System

• Ultra-low power CS analog front-end (AFE)
• RF power amplifier is energy hog; ADC is energy piglet
• CS reduces data rates with commensurate energy

savings for PA, ADC, etc; i.e., only [Y] is digitized and transmitted

41

LNA ADC Power Amplifier Antenna CS AFE

Electrode

Compressed Sampling Bio-Signal Acquisition System

Sensor

x(t) [Y]

Compressed Data Rate Feedback

ASU August 17, 2011 42

Intuition for CS – Conventional

ASU August 17, 2011 43

Intuition for CS – Group Sampling

• R. Dorfman, “The detection of defective members of large populations,” The

Annals of Mathematical Statistics, vol. 14, no. 4, pp. 436-440, Dec. 1943.

• M. Sobel and P.A. Groll, “Group testing to eliminate efficiently all defectives

in a binomial sample,” Bell System Technical Journal, vol. 38, no. 5, pp. 1179- 1252, Sept. 1959.

ASU August 17, 2011

Intuition – II: Sub-Nyquist Sampling

• Intuitive explanation of three key ideas
• Nyquist sample a sinusoid; i.e., 2 samples/period
• Only 2 amplitude values (i.e., looks like sawtooth

waveform)

• How to get enough amplitude values to infer sinusoid?

W

W

44

ASU August 17, 2011

W

W

r1 r2

• Key Idea #1: Randomize Sampling
• Multiply original analog samples by random weights to obtain

many more analog amplitudes

45

Intuition – II: Sub-Nyquist Sampling

ASU August 17, 2011

W r1 r2 r3 r4 r5 r6 r7 r8

W

46

• Key Idea #2: Reconstruction (e.g., 8! possible solutions)
• Key Idea #3: Optimization assuming known class of signal; e.g.,

sinusoid). 8! Solutions—CS finds best with high probability.

• What about compression?

Intuition – II: Sub-Nyquist Sampling

ASU August 17, 2011

Formal Compressed Sampling

[X]NX1 [Y]MX1

• [X]: Analog input sample vector (e.g., N = 16)
• []: Measurement matrix of (e.g., 6-bit Gaussian or

Uniform) random coefficients (M rows and N columns)

• [Y]: Compressed analog output vector (e.g., M = 8)
• Compression Factor C = N/M (e.g., C = 2)

[]MXN

[Y] = [Φ][X]

47

ASU August 17, 2011

Compressed Sampling - I

[X]NX1 = [X1, …, XN] [Y]MX1 = [Y1, …, YM]

• [X]16X1; []8X16; [Y]8X1; C = 2
• []8X16 is Measurement Matrix;

e.g., Gaussian or Uniform random coefficients each quantized to n = 6 bits

• Multiply and sum for each Yi is a Random Linear Projection
• [Y] is a compressed analog signal with global information
• Typically K < M < N (i.e., signal is sparse such as ECG)

[]MXN = [11, …, N ] [ [ [ ] ] ] M1, …, N …

48

1 1 1 N i i i

Y X



 



K = 3

[Y] = [Φ][X]

ASU August 17, 2011

Compressed Sampling - II

• [X]1024X1: Analog samples from ECG signal
• [Y]256X1: Compressed analog output signal
• []256X1024: Measurement Matrix
• C = 4X in this example; (C = 2X – 16X possible

for ECG) [X] [Y]

49

ASU August 17, 2011

CS Reconstruction

• Reconstruction/optimization of compressed signal (e.g., Smartphone)
• [Φ] is non-square and non-invertible; under-determined system with

many solutions

• Optimize exploiting knowledge of signal; e.g., ECG bio-signals are

time-domain sparse

50

LNA DAC Antenna Baseband DSP CS Optimization/ Reconstruction

Compressed Sensing Bio- Signal Reconstruction System y(t)

Original Nyquist Data Rate

ASU August 17, 2011

Accuracy Requirments for ECG

• AAMI—American Institute for Advancement of

Medical Instrumentation (Standards Vary)

• Ambulatory Quality ECG—8-10 bits (48-60 dB)
• Diagnostic Quality ECG—10-12 bits (60-72 dB)

51

ASU August 17, 2011

• Accuracy depends on:

 Compression Factor, C = N/M  PDF of random coefficients and # bits  Algorithm—Convex Optimization with L1 Norm

CS Reconstruction - II

[X] [Y]

52

ASU August 17, 2011

Sparsity vs. Compressibility

• Theoretical Limit: M > K log(N/K) with K nonzero

input samples (Heuristic: M > 2K)

53

50 60 70 80 90 100 Sparsity (%) 2 6 10 14 18 22 Compression Factor, C = N/M

ASU August 17, 2011

Quantization of Random Coefficients - I

• Gaussian []: Choose n = 6 bits for C = 2X – 16X

54

ASU August 17, 2011

Switched-Capacitor CS CODER

• For ECG signal:

BW = 2 KHz fS = 4 KHz C = 100 fF PDYN ≈ 0.4 nW

• C-2C in MDAC/ADC

[Y] = [Φ][X]

55

LNA ADC Power Amplifier Antenna CS AFE

Electrode

Compressed Sensing Bio- Signal Acquisition System

Sensor

Ultra-low Power Analog Circuits

SC Multiplying Digital-Analog Converter

ASU August 17, 2011

CS-ADC Chip-photo

IBM8RF 0.13 µm CMOS 3 mm x 3 mm M = 64 N=128 to 1024 Testing Underway: Expect ~ 1 uW total power with C = 16

ASU August 17, 2011

Thank you very much!

57