1

Low-power Multimedia Wireless Communication Systems

Naresh R. Shanbhag

Coordinated Science LaboratoryDepartment of Electrical and Computer Engineering

University of Illinois at Urbana-ChampaignURL://uivlsi.csl.uiuc.edu/~vips/

2

Trends and Problems in Integrated Circuits

(Source: Semiconductor Industry Association 1997 Roadmap)

• Problems (due to Systems-on-a-chip in deep submicron)– Time-to-market/Design productivity vs. Design complexity => Solutions:

Design reuse via Intellectual Property; Design exploration; H/S codesign, Reconfiguration

– Reliability: noise, signal integrity, process variations => Solutions: noise analysis, smart place&route, noise-tolerance

– Design efficiency: low-power and high-speed techniques & bounds

Current day Integrated Circuit Integrated Circuit in 2012*

0.2 micron 0.035 micron 3.7M/sq.cm 180M/sq.cm750 MHz 10 GHz250 MHz 1.54 GHz1.8-2.5V 0.5-0.6V70 Watts 175 Watts1.2 Watts 3.2 Watts

FEATURE SIZE

DENSITY

ON-CHIP CLOCK

OFF-CHIP CLOCK

SUPPLY VOLTAGE

POWER (HIGH PERF.)

POWER (MOBILE)

0.5-0.75 micron 0.1-0.15 micronWIRE PITCH

3

(Source: www.ti.com/sc/docs/wireless/97/issues.htm)

Integrated Multimedia Communication Systems

Source(video/speech/

data)

ChannelRFSource

Proc./Coding

ChannelCoding

Modulation

• Additional Problems:– mixed-signal issues: coupling, low-voltage analog, etc.

• Additional Solutions: DSP and communication theory; joint source-channel coding, signal-adapted DSP, multiresolution DSP; well-defined system/algorithm design

4

Standards(VDSL, ATM

Wireless, Video)Specifications

Multimedia Communication System

Design

• Projects:– Wireline system design: ATM-LAN, VDSL, cable modem (Goel, Hegde,

Tschanz)– Wireless system design: (Wang)– Hermitian decoder ASIC (Profs. Blahut and Kotter, Ashbrook, Feng)– Low-power transforms and synthesis: (Profs. Hajj and Najm, Ramprasad,

Hegde)

System Design IC Design

IC Fab&Test

5

Noise-tolerant VLSI

Algorithmic Noise-Tolerance

Circuit

Architecture

Logic

No

ise

Pro

pa

gatio

n

Noise Analysis/Measurements

Noise Models

Architectural Noise-Tolerance

Logic Noise-Tolerance

Noise-Tolerant Circuits

Algorithm

• Projects:– Algorithmic noise-tolerance (Hegde, Wang)– Noise-tolerant circuits (Wang, Ganesh)– Noise-tolerant distributed arithmetic filters (Anders)

6

Deep Submicron (DSM) Noise

• Ground bounce

• IR drop

• Crosstalk

• Charge sharing

• Charge leakage

• Process variations

• Alpha particles

• Electro-magnetic radiation

(b)

VDD

VSS

D Q NIC-GB

M1

Error

Q

D

Noise Problems: aggressive architectural (deep pipelining) and circuit (dynamic, low-voltage) styles.

Noise Sources:

7

Mirror Technique For Dynamic Circuit

Vin1 Vin2

Vout

Vout

NMOSLOGIC

NMOSLOGIC

Vin1

VinN

NMOSLOGIC

Vin1

VinN

Vout

(a) Conventional domino (b) Mirror technique (c) NAND gate design

8

ANTE: A Noise-Tolerance Metric

Cnic1

Cnic2

Error Free

Noise Immunity Curve

• Average Noise Threshold Energy

• Energy Normalized ANTE

noisenoiseTVEANTE 2

ANTE

NANTE

: Energy dissipated per cycle

9

Simulation Results: Full Adder

Mirror technique

Static

Conventional dynamic

Design Specifications:

(1) Power supply: 3.3V(2) Load capacitor: 20fF(3) Clock cycle: 1GHz

Area( 2m)

Energy(pJ)ANTE(nJ)

EnergyNormalized ANTE

Static 574.32.202 3.115 1414Conventional dynamic288.80.889 1.405 1580Mirror tech. 487.21.693 5.203 3073

Technology: 0.35 micron CMOS

10

Noise-Tolerant ASIC

Technology: 0.35m CMOS Pin #: 48

Transistor #: ~ 20K Area: ~ 5mm2

Technique: dynamic, mirror noise-tolerant dynamic

Measured Noise immunity improvement: 34.1% ~ 69.5%, average: 55.2%

11

Soft DSP

DSP BLOCKLATCH LATCH

FROM A/D TO A/D

CLOCK

SUPPLY VOLTAGE

DELAY

Vdd-crit

CLOCK-PERIOD

TRADITIONAL DSP DESIGN• critical-path-delay of the DSP block < sample period.

• reduction in supply voltage to the DSP block is limited by Vdd-crit.

Soft DSP

• reduce Vdd beyond Vdd-crit.

• detect/correct errors in output via Algorithmic Noise-Tolerance. MEET THE SNR/BER CRITERION AT REDUCED ENERGY DISSIPATION.

12

Error Probability in Arithmetic Units

PATH-DELAY DISTRIBUTION OF 8-BIT RIPPLE CARRY ADDER

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

x100

% in

pu

t co

mb

inat

ion

s

1 2 3 4 5 6 7 8

delay in multiples of FA delay

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

probability of errorfa

ctor

by

whi

ch e

nerg

y di

ssip

atio

n re

duce

s

• 48% reduction in energy dissipation possible with prob. of err. = 0.1• distribution with long tail leads to smaller error penalty• errors at the AU level lead to SNR/BER degradation• Algorithmic Noise-Tolerance to enhance performance

13

Algorithmic Noise-Tolerance (ANT)

SOFT DSP Error Control

x(n) (n)erryy(n)(n)

^ y

e(n)

0(n) y when0e(n)

0(n) y when0 e(n)

:

err

err

IDEAL

ANT via LINEAR PREDICTION:

hN

• exploit the correlation in filter output to perform error-control• optimum predictor for error detection - minimize MSE e(n)

e (n) when y (n) = 0err

0 e (n)+ y (n) errp

p

SOFT DSP FILTER

x(n) D D D(n)y

^

(n)eph1 hN-1

(n) y e(n) (n)e errp

14

Frequency Selective Filtering via Soft DSP

35

40

45

50

55

60

65

70

0 5 10 15 20 25

SNR at the filter output

% r

ed

uc

tio

n in

en

erg

y d

iss

ipa

tio

n

difference-based ANT-based

• difference-based scheme :

• 52% power savings with 1dB SNR loss (effective for high correlation).

• prediction-based scheme:

• 44% power savings with 0.7dB SNR loss (overhead: 2-tap predictor)

• effective for low correlation (higher BW).

SNR desired = 20dB

15

Information-Theoretic VLSI Framework

• Information Transfer Rate: R (bits/sec)• Information Transfer Capacity: C (bits/sec)• For reliability : C > R; For energy-efficiency: C R

Algorithm:input stats.,I/O map

Implementationarch., circuit, tech.,DSM noise

R

C

Achievable boundson reliability and efficiency

• Projects:– Lower-bounds on energy-efficiency of noisy digital circuits (Hegde)– Lower-bounds on signal transition activity and coding schemes

(Ramprasad, Prof. Hajj)– Bounds on: throughput and energy-efficiency; adaptive systems (Goel)– Design techniques for ultra efficient VLSI

Soft DSP

16

Lower Bound on Energy Dissipation

• Minimize: Eb = (Pdyn+ Pstat)/R bits/sec

subject to: C R

• operating point: fc = kmVdd/ CL

N. R. Shanbhag, University of Illinois at Urbana-Champaign

C L

Information Transfer Rate : R

N

Transition Activity : t

Operating Speed : f c

Q(Vdd/2

17

Dynamic Power

N. R. Shanbhag, University of Illinois at Urbana-Champaign

Energy dissipation at minimum supply voltageis greater than minimum achievable energy dissipation

18

Controller(ROM,microprocessor,

ASIC)

Reconfigurable

data-path(ASIC, FPGA,

multi-processor)

Signal ProcessingAlgorithm

(SPA)

Input Output

Signal MonitoringAlgorithm

(SMA)

AuxiliarySignals

ControlSignals

Reconfigurable DSP

• Dynamic Algorithm Transforms:

• Projects:– Low-power adaptive filtering, VDSL equalizer ASIC (Goel, Tschanz)– Domain-specific reconfigurable DSP processors (Tschanz)– Reconfigurable DSP for video processing (Minocha)– FPGA board design (Park)– Video over wireless (Profs. Jones and Ramchandran)

min Energy/Throughputsubject to: DSP constraint

19

min Energys.t. J < Jo

DSPmodels

Energymodels

Input state-space S

Configurationspace C

Energy-optimumconfiguration

Dynamic Algorithm Transforms (DAT)

A framework for designing low-power reconfigurable DSP systems

20

Input State-space S

• Set of all possible input states

pt(s2,s2)s1=G s2=B

• s(n) = received signal power• Two-state model• State s1 : Good channel• State s2 : Bad channel• p(si) : steady-state probabilities• pt(si,sj) : transition probabilities

pt(s1,s2)

pt(s1,s1)

pt(s2,s1)p(s1) p(s2)

0.01

0.1

1

10

100

1000

0 0.1 0.2 0.3time

Re

ce

ive

d s

ign

al p

ow

er

Good channel G

Bad channel B

Signal power received at the mobile unitmobile speed = 60 miles/hr

RF signal frequency = 2 GHz

21

Configuration-space C

• Set of all possible configuration vectors the reconfigurable datapath can support

Dx(n)

0

w1

Dx(n-1)

0

w2

Dx(n-N+1)

0

wN

Nth TAP

y(n)

c(n)=[…] : N-bit configuration vector

C: Set of all N-bit tuples (2N vectors)

22

ONU

TWISTED PAIR DISTRIBUTION CABLE

FIBER

System Data-rate Distance

ADSL 1.544 Mb/s 18 kft

ADSL 8.448 Mb/s 9 kft

VDSL 12.96 Mb/s 4.5 kft

VDSL 51.84 Mb/s 1 kft

Very High-speed Digital Subscriber Loop

• Cable length– 100ft to 1kft (worst-

case)

• Far-end crosstalk– 4-11 interferers

• Desired BER=10-7

– SNR=21.5dB

23

51.84 Mb/s VDSL Transmitter

In-phase shaping filter

Q-phase shaping filter

16-CAPEncoder

Scrambler DAC LPF51.84Mb/s

– square-root raised cosine– excess bandwidth=36%– center frequency=12.96 MHz

Real

Imag

16-CAP signal constellation

51.84 MHz

Analog Front End

24

51.84 Mb/s

I-PHASE EQUALIZER

SMA BLOCKA/D

SLICER

Q-PHASE EQUALIZER

SLICER

DECODER

DSCRAMB

W,e(n)

W,e(n)

FBF

+

+TimingRecovery

PGAControl

PGA

DAT-based 51.84 Mb/s VDSL Receiver

– I/Q-phase equalizers: 48 taps each– FBF: 10 complex strength-reduced taps– Powers of two LMS + Blind Equalization

25

-20

020

4060

8010

0

0 0.2 0.4 0.6 0.8 1

Cable length (kft)

En

erg

y S

av

ing

s (

%)

VARIATIONS IN CABLE LENGTH VARIATIONS IN FEXT INTERFERERS-2

00

2040

6080

100

2 4 6 8 10 12

Number of FEXT interferers

En

erg

y S

av

ing

s (

%)

Energy Savings: 51.84 Mb/s VDSL

AVERAGE ENERGY SAVINGS=53%

26

Wireless Environment

MOBILE

MULTIPATH CHANNEL

MULTI-USER

INTERFERENCE

ANTENNA

BASE

STATION

ANTENNA

• Adaptability of system to time-varying propagation and traffic environments

• Adaptation to different spectrum allocations

• Ability to accommodate mixed-cell (pico, micro and macro) architecture

• Ability to handle different services: audio, video, speech,data, multimedia

Flexibility Features of IMT-2000 systems

27

ChannelEncoder

t

PowerAmplifier

Pt

ChannelDecoder

t

RAKE Receiver

crake

OUTER TRANSCEIVERINNER

TRANSCEIVER

SNRDistortion and BER

Reconfigurable Wireless Communication System

SourceEncoder

Rs

SourceDecoder

Rs

Image

Video

Wireless

Wireline

• Energy-optimum configuration via Dynamic Algorithm Transforms and Joint Source-Channel Coding

otot

oi

RcR

DscDts

cEnergy

)(

),(..

)(min arg)(sc iopt

• With Doug Jones and Kannan Ramchandran

28

Source-Channel Variabilities

(Rate-Distortion Curves)

SOURCE VARIABILITIES

(BER Curves)

CHANNEL VARIABILITIES

110

010

000

0 1 2 3

Source Rate (in bits per pixel)

Ave

rag

e D

isto

rtio

n p

er

pix

el

carphone

akiyo

coastguard

00.

51

4.5 5.5 6.5 7.5

Channel SNR (Eb/No)

Pro

ba

bili

ty o

f e

rro

r

t=16

t=24

t=32

29

Simulation Results: QCIF Images and IMT-2000

Test channels

• Energy Savings: maximum 93% (average 59%)• Fraction of Energy due to the digital blocks:

– ranges from 40-10% (for distance: 10-100m)

Channel A (low delay spread) Channel B (medium delay spread)

0

0.0004

0.0008

0.0012

0.0016

0.002

0 20 40 60 80 100

distance (in m)

En

erg

y (i

n J

/pix

el)

akiyo carphone

claire coastguard

container hall_objects

mother_and_daughter silent

0

0.0004

0.0008

0.0012

0.0016

0.002

0 20 40 60 80 100

distance (in m)

En

erg

y (i

n J

/pix

el)

akiyo carphone

claire coastguard

container hall_objects

mother_and_daughter silent

30

Summary

• Evolving next generation (3G) wireless standards: flexibility and

energy-efficiency. • Evolving integrated circuit technology: deep submicron noise, complex

system-on-a-chip (SOC).• DSP via Soft Computations (Soft DSP): energy-efficient,

noise-tolerant circuit design and algorithmic noise-tolerance• Dynamic low-power techniques are required

– inter-application dynamism => domain-specific processors

– intra-application dynamism => run-time reconfiguration

• Dynamic algorithm transforms: input space, configuration space,

DSP models, energy models, joint-optimization of energy

& performance