German Aerospace Center (DLR)Institute of Comm. and NavigationOberpfaffenhofen, Germany

On the On the AdaptivityAdaptivity in Downin Down-- and Uplink and Uplink MCMC--CDMA Systems CDMA Systems

Ivan Ivan CosovicCosovic

IEEE Communication Theory Workshop 2005Park City, USA, June 15, 2005

Stefan KaiserStefan Kaiser

DoCoMo Euro-LabsWireless Solution LaboratoryMunich, Germany

2

IntroductionIntroduction

• Adaptivity in MC-CDMA

• Concept of Adaptivity in Uplink MC-CDMA

• Concept of Adaptivity in Downlink MC-CDMA

• Conclusions

3

AdaptivityAdaptivity in MCin MC--CDMACDMA

MC-CDMAreceiver

MC-CDMAtransmitter

SNRMAI

fadingchannel

4

AdaptivityAdaptivity in MCin MC--CDMACDMA

MC-CDMAreceiver

MC-CDMAtransmitter

fadingchannel

CSI

5

AdaptivityAdaptivity in MCin MC--CDMACDMA

MC-CDMAreceiver

MC-CDMAtransmitter

fadingchannel

CSI

6

AdaptivityAdaptivity in MCin MC--CDMACDMA

MC-CDMAreceiver

MC-CDMAtransmitter

fadingchannel

CSI

Adapt transmission signal to CSI:

1) Eliminate MAI Zero-forcing (ZF)2) Maximize SNR Maximum ratio transmission (MRT)3) Trade-off between ZF and MRT MMSE

7

+

MC-CDMAreceiver

MC-CDMAtransmitter

fadingchannel

CSI CSI

AdaptivityAdaptivity in MCin MC--CDMACDMA

8

MC-CDMAreceiver

MC-CDMAtransmitter

fadingchannel

AdaptivityAdaptivity in MCin MC--CDMACDMA

CSI at Tx and Rx – more degrees of freedom

Mobile station - low complexityBase station - higher complexity

CSI CSI

9

AdaptivityAdaptivity in Uplink MCin Uplink MC--CDMACDMACSI Available Only at CSI Available Only at TxTx

MMSE (trade-off between ZF and MRT)

MRT (maximizes SNR)

ZF (eliminates MAI)

CoefficientsPre-equalization Technique

)(

1

)()()(pre

)( lwdK

k

kkkk +=∑=

cGHr

WHG kl

kl )*()(

pre, =

WH

HGkl

klk

,l 2)(

)*()(

pre =

WH

LK

HGn

kl

klk

,l 1 22)(

)*()(

pre

σ+−=

Input/output relation:RxTx fading

channel

CSI

10

CSI

AdaptivityAdaptivity in Uplink MCin Uplink MC--CDMACDMACSI Available at CSI Available at TxTx and Rxand Rx

RxTx fadingchannel

CSI• Concept of combinedcombined--equalizationequalization

• Aim at maximizing SNR at Tx and suppressing MAI at Rx

New improved MCimproved MC--CDMA singleCDMA single--user boundsuser boundsClosely approached even in a fully-loaded system

• ComplexityLow complexity at Tx – one-tap pre-equalizationHigher complexity at Rx – non-linear parallel interference cancellation (PIC)

• Optimization algorithmClosed form solution not available in a multi-user caseA simple and effective approach; A single optimization parameter - parameter ‘p’ of the generalized pregeneralized pre--equalizationequalization used to trade SNR and MAI

WHHG kl

pkl

kl )*()()(

pre, =

Controls power distributionover subcarriers

11

0 2 4 6 8 10 12

SNR in dB

10-4

10-3

10-2

10-1

100

BE

R

MF bound, L=16(state of the art)

AWGN channel

• Uncoded system• QPSK• Uncorrelated Rayleigh fading• Spreading length L=16• Single-user• Ideal CSI

0

0

NELNE

t

t

+=γ

l

l l

l

+

+−

−

= ∑−

= 211

21 1

0MF

γγ LL LP

MF – matched filter

AdaptivityAdaptivity in Uplink MCin Uplink MC--CDMACDMACSI Available at CSI Available at TxTx and Rxand Rx

12

CombinedCombined--EqualizationEqualizationFrequency Domain SingleFrequency Domain Single--User Transmission SystemUser Transmission System

)( pre,post, lll HGfG ⋅=

Tx Rx

x

1c

x

pre,1G

x x

post,1G

x

*1c1H 1n

x

2c

x

pre,2G

x x

post,2G

x

*2c2H 2n

x

Lc

x

pre,LG

x x

post,LG

x

*LcLn

+

+

+

... ... ... ... ... ... +d d̂

channel

LH

channel

13

CombinedCombined--EqualizationEqualizationSelection Diversity (SD) CombinedSelection Diversity (SD) Combined--EqualizationEqualization

Tx Rx

x

1c

x x x x

*1c1H 1n

x x x x x

x

Lc

x x x x

*LcLn

+

+

+

... ... ... ... ... ...

+d d̂

channel

LH

0

0 0

0

... ... ... ... ... ...

*lclc 1

l

l

HHL*

lH ln

Best Best propagationpropagation conditionsconditions

channel

14

0 2 4 6 8 10 12

SNR in dB

10-4

10-3

10-2

10-1

100

BE

R

MF bound, L=16(state of the art)

SD bound, L=16(new MC-CDMA bounds)

4.5 dB

AWGN channel

• Uncoded system• QPSK• Uncorrelated Rayleigh fading• Spreading length L=16• Single-user• Ideal CSI

++−

+−

−= ∑

−

= 0

01

0SD 1

11)1(1

2 NENELLPt

tL

lll

l

l

0

0

NELNE

t

t

+=γ

l

l l

l

+

+−

−

= ∑−

= 211

21 1

0MF

γγ LL LP

SD – selection diversity

MF – matched filter

AdaptivityAdaptivity in Uplink MCin Uplink MC--CDMACDMACSI Available at CSI Available at TxTx and Rxand Rx

15

Requirements:– Appropriate for application in multimulti--user uplinkuser uplink MC-CDMA– Capable to closely approach selection diversity bound– Preferably flexible

Solution: Generalized preGeneralized pre--equalization (Gequalization (G--prepre--eqeq))

CombinedCombined--EqualizationEqualization

Drawback of selection diversity: Not suitable for the multiNot suitable for the multi--user caseuser case

16

CombinedCombined--Equalization in MultiEqualization in Multi--User MCUser MC--CDMACDMA

0 1 2 3 4 5 6 7 8 9

p

10-6

10-5

10-4

10-3

10-2

10-1

100

BE

R

SNR = 1.5 dB, G-pre-eq + PIC (1 iter.)SNR = 0.0 dB, G-pre-eq + PIC (3 iter.)SNR = 0.0 dB, G-pre-eq + PIC (5 iter.)

• Convol. coded system (R=1/2)• QPSK • Uncorrelated Rayleigh fading• Spreading length L=16• Number of active users K=16• Ideal CSI

As p grows single-user bounds are better, but more MAI is inducedOptimization problem reduced to determination of parameter p

17

-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6

SNR in dB

10-5

10-4

10-3

10-2

10-1

100

BE

R

pre-eq max SINR, pre-eq onlyPIC (3 iter.), post-eq onlyMF bound, L=16AWGN channelG-pre-eq (p=3) + PIC (3 iter.)G-pre-eq-MRC bound, L=16, p=3SD comb-eq bound, L=16

Pre-eq only(CSI at Tx)

Post-eq only(CSI at Rx – no adaptivity)

Comb-eq(CSI at Tx and Rx)

4 dB

Concept of Concept of AdaptivityAdaptivity in Uplink MCin Uplink MC--CDMACDMANumerical ResultsNumerical Results

• Convol. coded R=1/2• QPSK• Uncorelated Rayleigh fading• Spreading length L=16• Number of active users K=16• Ideal CSI

18

Concept of Concept of AdaptivityAdaptivity in Downlink MCin Downlink MC--CDMACDMACSI Available Only at CSI Available Only at TxTx

MMSE (trade-off between ZF and MRT)

MRT (maximizes SNR)

ZF (eliminates MAI)

Pre-equalization matrixPre-equalization Technique

wUdFz += H

( ) ( )( ))(H)()1(H)1( KK cHcHF K=

WFU =

Input/output relation:

RxTx fadingchannel

CSI

( ) W1H −= FFFU

( ) Wn12H −

+= IFFFU σ

Pre-equalization(& spreading)

Fading & despreading

19

AdaptivityAdaptivity in Downlink MCin Downlink MC--CDMACDMACSI Available Only at CSI Available Only at TxTx

R

bRRRdb )(1 −−= −

wUbFz += H

H QU =

RxTx fadingchannel

CSINonNon--linear interference prelinear interference pre--cancellation:cancellation:

RQF =H QR-decomposition; lower triangular; unitary matrixQ

wdRwdRRRQQz +=+= −1H

( )IQQ =H

R diagonal matrix - contains diagonal elements of R

MAI is completely eliminated

20

AdaptivityAdaptivity in Downlink MCin Downlink MC--CDMACDMACSI Available at CSI Available at TxTx and Rxand Rx

CSI

RxTx fadingchannel

CSI• ComplexityHigher complexity at Tx – non-linear interference pre-cancellationLow complexity at Rx – one-tap post-equalization, i.e., single-user detection (SUD)

• Optimization algorithmClosed form solution is given by a complicated iterative procedureSimple and effective approach PrePre--equalizationequalization is performed assumingassuming that a certain SUD is employeda certain SUD is employed at the receiver

)( )()( kk f Hg ′′=

( ) ( )( ))(H)()1(H)1( KK gHgHF K=fading, SUD & despreading

)( FU f ′=pre-equalization

SUD

21

AdaptivityAdaptivity in Downlink MCin Downlink MC--CDMACDMANumerical ResultsNumerical Results

0 2 4 6 8 10 12 14 16

SNR in dB

10-5

10-4

10-3

10-2

10-1

100

Bit

Err

or R

ate

S&GS&MMMSE detectionnon-linear precoding +PDnon-linear precoding + MRCnon-linear precoding + EGCAWGN

• Uncoded system• QPSK• Uncorelated Rayleigh fading• Spreading length L=16• Number of active users K=16• Ideal CSI

22

ConclusionsConclusions

• Combined equalization can properly be designed for uplink and downlink, allowing low complexity terminal stations and higher low complexity terminal stations and higher complexity base stationscomplexity base stations.

• Optimization based on generalized pregeneralized pre--equalizationequalization.

• Simulation results show that in multi-user scenarios combined combined equalization significantly outperforms systems with only preequalization significantly outperforms systems with only pre-- or postor post--equalizationequalization.

• Prerequisite is to have channel knowledge available at the transmitter channel knowledge available at the transmitter and the receiverand the receiver.

23

5th International Multi5th International Multi--Carrier Spread Carrier Spread Spectrum Workshop (MCSpectrum Workshop (MC--SS 2005)SS 2005)

September 14-16, 2005 Oberpfaffenhofen, Germany

www.mcss2005.org