You are on page 1of 51

SPACESPACE-TIME CODING AND SIGNAL PROCESSING FOR HIGH DATA RATE WIRELESS COMMUNCATIONS

Ayman F. Naguib, Vahid Tarokh, Tarokh, Nambi Seshadri, Seshadri, and A. R. Calderbank
AT&T Labs - Research 180 Park Avenue Florham Park, NJ 07932 naguib@research.att.com

AT&T Labs-Research

Tutorial Overview
AT&T Labs

! ! ! !

PART 1: : Preliminaries. PART 2: : SpaceSpace-Time Coding. PART 3: : Applications of SpaceSpace-Time Coding. Summary.

PART 1: Preliminaries
AT&T Labs

! !

Wireless channels. Outage capacity of fading channels. Diversity Options: time, frequency, and space.

Receive and transmit diversity techniques.

Wireless Channels
AT&T Labs
Remote Dominant Reflector

Local Scatterers to Base Co-Channel Mobile

Base Station

Local Scatterers to Mobile

Local Scatterers to Base

Local Scattering Multipath Propagation Mobile Motion Cellular Spectrum Reuse


Remote Dominant Reflector

Fading Intersymbol Interference Time Varying Channel Co-channel Interference

Signal Level in Wireless Channels


AT&T Labs
Short Term Fading Signal Level (dB) Mean Path Loss

Long Term Fading

Distance (dB)

! ! !

Slow fading (shadowing) is caused by large obstructions between transmitter and receiver. Fast fading is due to scattering of the signal by object near transmitter. Path loss is proportional to 1/r, is between 2.5 and 5.

Classification of Wireless Channels


AT&T Labs

The multipath channel:

h(t, , z) = i (t ) e j i ( t , i ,z ) (t i )
i

i (t, i , z) = 2 fc i (t ) fd ,it + ( z / ) sin i


!

i z

The three spreads: spreads


Channel Spread
Delay Spread Doppler Spread Angle Spread

Channel Selectivity
Frequency Selective Time Selective Space Selective

Measure of Selectivity
Coherence Bandwidth Coherence Time Coherence Distance

Space-Time Fading
AT&T Labs

10log10 h(t,z)
Di sta nc

e(

in w

av ele

ng

th

s em Tim

Angle Spread d = 0, Doppler Spread fd = 200 Hz

Space-Time Fading .
AT&T Labs

10log10 h(t,z)
Di sta

nc

e(

in w

av ele

ng

th

s em Tim

Angle Spread d = 5, Doppler Spread fd = 200 Hz

PART 1: Preliminaries
AT&T Labs

! !

Wireless channels. Outage capacity of fading channels. Diversity Options: time, frequency, and space.

Receive and transmit diversity techniques.

Fundamental Limits: Outage Capacity


AT&T Labs

Tx 1

Rx 1

Tx 2

Rx 2

LM H =M MM " N

11 21

12 ! 1 N 22 ! 2 N " # " M 2 ! MN

M1

OP PP PQ

Tx N

Rx M

! ! !

Multiple input multiple output (MIMO MIMO) channel: N transmitters, M receiver ij is the complex channel gain from i-th transmit antenna to j-th receive antenna. H is the NM channel matrix.

Fundamental Limits: Outage Capacity


AT&T Labs

Channel capacity for SISO channel:


C = lo g 2 (1 + )

bits/sec/use, is the SNR

Channel capacity for MIMO channel:


C = log2 det I + HH* bits/sec/use n H is the n m channel matrix

LM N

OP Q

Outage capacity C x :
Pr C > Cx

q=

H :C ( ) = C x

C ( H ) fH ( H ) d H = x

Fundamental Limits: Outage Capacity


AT&T Labs
Achievable Data Rate (kbits/sec) per 30 kHz Channel

Achievable Data Rates with Multiple Antennas


225 200 175 150 125 100 75 50
IS-136+

We can be here
1 Tx 1 Rx Antennas 1 Tx 2 Rx Antennas 2 Tx 1 Rx Antennas 2 Tx 2 Rx Antennas

25

We are here now


0
0 5 10 15

IS-136

20

SNR per Rx Antenna (dB)

It is dangerous to put limit on wireless


Guglielmo Marconi (1932)

PART 1: Preliminaries
AT&T Labs

! !

Wireless channels. Outage capacity of fading channels. Diversity Options: time, frequency, and space.

Receive and transmit diversity techniques.

Diversity in Wireless Radio


AT&T Labs

! !

Receive the signal via independently fading channels. Improves the performance in a fading environment.

Space diversity: Multiple antennas are used to receive the signal. Antenna spacing must be such that the fading at each antenna is independent (coherence coherence distance). distance Frequency diversity: Signal is transmitted in several frequency bands (coherence coherence BW). BW Not effective on flat (non-frequency selective) channels. Techniques that exploit frequency diversity include: RAKE Receivers, OFDM, equalization. Time diversity: Signal is transmitted in different time slots (coherence coherence time). time Channel coding plus interleaving is used to provide time diversity. Not effective over slow fading channels. Polarization diversity: use two antennas with different polarization for reception and/or transmission.

Receive Diversity
AT&T Labs

Make use of a number of receive antennas that are well separated (> > coherence distance) distance to generate independent receptions of the transmitted signal.

Selection diversity: choose received signal with largest received power, S/N, etc. Switched diversity: choose alternate antenna if signal falls below a certain threshold. Linear combining: linearly combine a weighted replica of all received signals.

There is a dramatic improvement even with two branch selection diversity. The outage with MRC improves linearly with the number of diversity branches M, but complexity becomes prohibitive. prohibitive

Rx: Selection and Switched Diversity


AT&T Labs

Front End 1

Data Detector 1 Data

Selection Diversity
Front End 2

Selector

Data Detector 2

Front End 1

Switched Diversity
Front End 2

Switch Data Detector Data

Receive Diversity: Linear Combining


AT&T Labs

1
r1 Front End
1

r1 = A1e j1 d + n1 r2 = A2e j2 d + n2
Data Detector Data

r2 Front End
2

Equal gain: Maximal ratio: MMSE: training decoding

1 = e j1 , 2 = e j2 1 = A1e j1 , 2 = A2e j2 2 1 , 2 = arg min 1r 1 + 2r 2 d

d = S r + * 2r 2

1 , 2 * 1 1

Receive Diversity Performance


AT&T Labs

Transmit Diversity
AT&T Labs

Provide diversity benefit to a mobile using base station antenna array for frequency division duplexing (FDD FDD) schemes. Cost is shared among different users. Order of diversity can be increased when used with other conventional forms of diversity. Two kinds of transmit diversity techniques:

Transmit diversity with feedback from receiver Transmit diversity without feedback from receiver:

No training. Feedforward information.

Transmit Diversity with Feedback


AT&T Labs

w 2 (t ) S2 ( t )
Mod

Tx 1 Rx

r1 ( t )
Tx 2

r (t )

r (t ) w1 ( t ), w 2 ( t )
variations

r1 ( t ) r2 ( t )

Demod

r2 ( t ) S1 ( t ) w1 ( t )

w1 ( t ) + w 2 ( t )
2

=1

! !

w1(t) and w2(t) are varied such that |r(t)|2 is maximized. maximized w1(t) and w2(t) are adapted with feedback information from the receiver.

10

Tx Diversity with Frequency Weighting


AT&T Labs

Tx 1 Rx Tx 2

r2 ( t ) r1 ( t )
(A) Constructive Interference

Channel Encoder

Channel Decoder

r1 ( t )

r2 ( t )

(B) Distructive Interference

e j ( t )

( kT ) = 2 fm kT

! !

Use frequency weighting to mitigate the harm of scenario B. Simulate fast fading can use conventional channel coding and interleaving techniques.

Tx Diversity with Antenna Hopping


AT&T Labs
Tx 1

Tx 2

Rx

Repetition Code

s1 s2 ! s N
T Tx N

ML Detection

! !

At time i, 1 i N , transmit s from antenna i. Achieves a diversity order of N using ML detection or MRC at the receiver. Bandwidth efficiency is 1/N.

11

Tx Diversity with Channel Coding


AT&T Labs
Tx 1

Tx 2

Rx

Channel Code

c1c 2 ! c N
T Tx N

ML Decoding

! ! !

The channel code has a minimum Hamming distance dmin N. Transmit code symbol i from antenna i. After receiving the N symbols, the decoder performs ML decoding to decode the received codeword.

Transmit Diversity via Delay Diversity


AT&T Labs

Tx 1

s(t )
Tx 2 Information Channel Encoder Delay

s(t T )
Tx N

( N 1) T

Delay

s ( t ( N 1) T )

! ! !

Provide diversity benefit by introducing intentional multipath. Receiver uses an equalizer or MLSE for detection. Provides a diversity order of N. No loss of BW efficiency.

12

Transmit Diversity Options


AT&T Labs

p1 ( t )
a1 ( t ) e j 1 ( t )
Information bits

p2 ( t )
a 2 ( t ) e j6 2 ( t )
Channel Encoder

! ! ! !

Antenna hopping Frequency weighting Delay diversity SpaceSpace-Time Coding

pN (t )

a N ( t ) e j N ( t )
Waveform Shaping

Part 2: Space-Time Coding


AT&T Labs

SpaceSpace-Time coded modulation (STCM): Probability of error analysis and design criterion. SpaceSpace-time trellis codes examples. Decoding of spacespace-time trellis codes. Comparison with delaydelay-diversity. SpaceSpace-Time block codes (STBC): Basic idea and decoding of STBC. Incoherent detection of STBC. STBC for more than 2 transmit antennas. Interference suppression with STBC: Zero forcing IC and ML decoding of STBC. MMSE IC and ML decoding of STBC. Iterative IC and ML decoding of STBC.

13

Space-Time Coded Modulation


AT&T Labs

Space-Time Encoder Information Source Receiver

For each input symbol, the space-time encoder chooses the constellation points to simultaneously transmit from each antenna so that coding and diversity gains are maximized. maximized Space-Time trellis codes: coding and diversity gain. Space-Time block codes: diversity gain + some coding gain (depending on the rate of the code).

Space-Time Coding: The Model


AT&T Labs

! ! !

! !

N transmit and M receive antennas. The overall channel is made up of NM slowly varying subchannels. Each sub-channel is Rayleigh fading (same analysis applies to Rician fading too). At any time interval, N signals are transmitted simultaneously, simultaneously one from each transmit antenna. The sub-channels undergo independent fading. The fade coefficients are assumed to be fixed during a slot and independent from slot to another.

14

STC: The Model .


AT&T Labs

Transmitted Code Vector:

c l = [ c1 ( l ), c2 ( l ), ! , c N ( l ) ]T
!

Channel Matrix:

LM H =M MM " N

11 21

M1

12 ! 1 N 22 ! 2 N " # " M 2 ! MN

OP PP PQ

Received Signal Vector:

r (l ) = H c l + n (l )

STC: Probability of Error Analysis


AT&T Labs

! !

Transmitted code vector sequence: C = {c1,c2, , cL}. Probability of error: assuming perfect knowledge of CSI,

~ ~ Pr C C H exp d 2 (C, C )
M L 2 j =1 l =1

t
2

~ ~ (l ) +!!+, c (l ) - c ~ (l ) d (C, C ) = , j1 c1 (l ) - c jN N N 1 ~ = Gs h j A (C, C ) h *j ,


j =1 M

where

Gs = Es / 4 N o and h j = [ , j1 , j 2 ! , jN ]T ~ ~ ~ A (C, C ) = B(C, C )B* (C, C )

15

STC: Probability of Error Analysis .


AT&T Labs

The matrix B is the error matrix between the transmitted ~ code vector sequence C and the decoded code vector C sequence .

~ (1) LM c (1) c ~ (1) c (1) c M B= MM " ~ (1) Nc (1) c


1 1 2 2 N N

~ (2 ) c1 (2) c 1 ~ c2 (2) c2 (2 ) " ~ (2) cN (2) c N

~ ( L) ! c1 ( L ) c 1 ~ ! c2 ( L ) c2 ( L ) # " ~ ( L) ! cN ( L ) c N

OP PP PQ

STC: Probability of Error Analysis .


AT&T Labs

The matrix A=BB* is Hermitain A can be written as UU* , where = diag{l1,l2,,lN} and U is orthonormal, orthonormal U* U=I, and its columns are the eigenvectors of A. Let j = U*h*j , then we will have
M M N ~ d 2 (C, C ) = s - *j L - j = s i | ij |2 j =1 j =1 i =1 2 The random variables ij = | ij | are i.i.d. , each has a 2 distribution with 2 degrees of freedom, that is

ij ~ fv ( ) = e v for > 0 and 0 otherwise

16

STC: Probability of Error Analysis .


AT&T Labs

Probability of error: average over all | ij |2

~ Pr C C
!

t FGH 1 + 1 IJK
N i =1 i s r M

Let r be the rank of the matrix A and l1,l2,,lr be the nonzero eigenvalues of A. Then

~ Pr C C
!

1I t FGH IJK FGH JK


i i =1 s

rM

Thus a diversity gain of rM and a coding gain of (l1l2lr)1/r are achieved.

STC: Design Criteria


AT&T Labs

Rank Criterion: In order to achieve the maximum diversity NM, the matrix B(C1,C2) has to be full rank for any two code vector sequences C1 and C2. If B(C1,C2) has a minimum rank r over the set of two tuples of distinct code vector sequences, then a diversity rM is achieved. Determinant Criterion: The minimum of the r-th roots of the sum of determinants of all rr principal cofactors of A(C1,C2) taken over all pairs of distinct code vector sequences C1 and C2 corresponds to coding gain, r being the rank of A(C1,C2) . The target of code design is making this sum as large as possible. If a code is designed to give a diversity gain of NM, for a better coding gain, the minimum of the determinant of A(C1,C2) taken over all pairs of distinct code vector sequences C1 and C2 must be maximized.

17

STC: Design Criteria .


AT&T Labs

! !

Space-time trellis codes were designed using this criteria. The trade-off between diversity, rate, and trellis complexity has also been studied. It is proved that our designs are optimal in terms of the trade-off between complexity, constellation size, diversity, and rate. It is not known whether our codes provide the optimum coding gains at a given complexity, diversity, and rate.

STC: 4-PSK Example


AT&T Labs

00 01 02 03

10 11 12 13

2 Input: Tx 1: Tx 2: 0 1 2 3 2 2 0 0 1 2 3 2 0 1 2 3 2 2

20 21 22 23

30 31 32 33

4-PSK 4-State Space-Time Code with 2 Tx Antennas

18

STC: 8-PSK Example 1


AT&T Labs

2 3 Input: Tx 1: Tx 2: 0 1 5 7 6 4 0 0 5 1 3 6 0 1 5 7 6 4 4 5 6 0 1 2 3 4 5 6 7 00,01,02,03,04,05,06,07 50,51,52,53,54,55,56,57 20,21,22,23,24,25,26,27 70,71,72,73,74,75,76,77 40,41,42,43,44,45,46,47 10,11,12,13,14,15,16,17 60,61,62,63,64,65,66,67 30,31,32,33,34,35,36,37 8-PSK 8-State Space-Time Code with 2 Tx Antennas 7 1 0

STC: 8-PSK Example 2


AT&T Labs
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 00 51 22 73 44 15 66 37 37 00 51 22 73 44 15 66 22 73 44 15 66 37 00 51 51 22 73 44 15 66 37 00 01 52 23 74 45 16 67 30 30 01 52 23 74 45 16 67 23 74 45 16 67 30 01 52 52 23 74 45 16 67 30 01 02 53 24 75 46 17 60 31 31 02 53 24 75 46 17 60 24 75 46 17 60 31 02 53 53 24 75 46 17 60 31 02 03 54 25 76 47 10 61 32 32 03 54 25 76 47 10 61 25 76 47 10 61 32 03 54 54 25 76 47 10 61 32 03 04 55 26 77 40 11 62 33 33 04 55 26 77 40 11 62 26 77 40 11 62 33 04 55 55 26 77 40 11 62 33 04 05 56 27 70 41 12 63 34 34 05 56 27 70 41 12 63 27 70 41 12 63 34 05 56 56 27 70 41 12 63 34 05 06 57 20 71 42 13 64 35 35 06 57 20 71 42 13 64 20 71 42 13 64 35 06 57 57 20 71 42 13 64 35 06 07 50 21 72 43 14 65 36 36 07 50 21 72 43 14 65 21 72 43 14 65 36 07 50 50 21 72 43 14 65 36 07

Input: Tx 1: Tx 2:

0 1 5 7 6 4 0 0 5 4 6 3 0 1 6 3 4 3

2 3 4 5 6 7 1 0

8-PSK 32-State Space-Time Code with 2 Tx Antennas

19

STC: 16-QAM Example


AT&T Labs
0 1
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

Input: Tx 1: Tx 2:

0 1 A C 5 7 0 0 B A C F 0 1 A C 5 7

2 3 4 5 6

7
4 5 6 7

8 9

A B

C D E F

16-QAM 16-State Space-Time Code with 2 Tx Antennas

STC: Maximum Likelihood Decoder


AT&T Labs

Given the received vector sequence R = {r1,r2, , rL}, the ~ receiver chooses the code vector sequence C as the one that was transmitted such that the a posteriori probability
~ Pr C R , H ( l ) , l = 1, ! , L

t
2

! !

is maximized. maximized The space-time maximum likelihood decoder (STMLD STMLD):


L ~ C = arg m~in r ( l ) C l =1

Es H( l ) ~ cl

The STMLD is implemented as a vector Viterbi algorithm (v vVA) VA where the trellis path with the smallest accumulated metric is chosen.

20

STC: Performance with Perfect CSI


AT&T Labs

Performance of 4-PSK Space-Time Codes with 2 Transmit and 1 Receive Antennas


100

Frame Length = 130 Symbols

Frame Error Rate

10-1

10-2

4-State Code 8-State Code 16-State Code 32-State Code 64-State Code

10

12

14

16

18

20

SNR (dB) per Receive Antenna

STC: Performance with Perfect CSI .


AT&T Labs

Performance of 8-PSK Space-Time Codes with 2 Transmit and 1 Receive Antennas


100

Frame Length = 130 Symbols

Frame Error Rate

10-1

8-State Code 16-State Code 32-State Code 64-State Code

10-2 10 12 14 16 18 20

SNR (dB) per Receive Antenna

21

STC: Performance with Perfect CSI .


AT&T Labs

Performance of 16-QAM Space-Time Codes with 2 Transmit and 2 Receive Antennas


100

Frame Length = 130 Symbols


16-State Code 64-State Code

Frame Error Rate

10-1

10-2

10-3 10 12 14 16 18 20

SNR (dB) per Receive Antenna

Delay Diversity
AT&T Labs

Signal 2

Constellation Mapper

Information Source

Rate Repetition Code

One Symbol Delay

Constellation Mapper

Signal 1

Delay Diversity

22

Delay-Diversity as a ST Code
AT&T Labs

2 3 Input: Tx 1: Tx 2: 0 1 5 7 6 4 0 0 1 5 7 6 0 1 5 7 6 4 4 5 6 0 1 2 3 4 5 6 7 00,01,02,03,04,05,06,07 10,11,12,13,14,15,16,17 20,21,22,23,24,25,26,27 30,31,32,33,34,35,36,37 40,41,42,43,44,45,46,47 50,51,52,53,54,55,56,57 60,61,62,63,64,65,66,67 70,71,72,73,74,75,76,77 8-PSK 8-State Delay Diversity Code with 2 Tx Antennas 7 1 0

ST Code as a Delay-Diversity
AT&T Labs
Signal 2 Constellation Mapper Best Block Code of Length 2

Information Source

One Symbol Delay

Constellation Mapper

Signal 1

Space-Time Coding as a Delay Diversity

For the 8-PSK 8-state ST code:

C = 0 0 ,1 5 ,2 2 , 3 7 , 4 4 , 5 1, 6 6 , 7 3

23

Part 2: Space-Time Coding


AT&T Labs

SpaceSpace-Time coded modulation (STCM): Probability of error analysis and design criterion. SpaceSpace-time trellis codes examples. Decoding of spacespace-time trellis codes. Comparison with delaydelay-diversity. SpaceSpace-Time block codes (STBC): Basic idea and decoding of STBC. Incoherent detection of STBC. STBC for more than 2 transmit antennas. Interference suppression with STBC: Zero forcing IC and ML decoding of STBC. MMSE IC and ML decoding of STBC. Iterative IC and ML decoding of STBC.

SpaceSpace-Time Block Codes


AT&T Labs

Signal 2 Constellation Mapper ST Block Code Information Source

c1 c 2

c1 c2

* -c2 c* 1

Constellation Mapper

Signal 1

Idea:

c1 c2
!

Lc M Nc

c2 c1

OP Q

Assumption: channel is quasiquasi-static. static

24

Decoding of STBC
AT&T Labs

Received Signal:

r1 = h1c1 + h2 c2 + n1
* * + h2 c1 + n2 r2 = h1c2 1 * 2

r=
!

LM r OP = LM h Nr Q N h
1 * 2

h2 c1 n1 + * = Hc + n h1* c2 n2

OPLM OP LM OP QN Q N Q
2

H is orthogonal:

H*H = h1 + h2 I = h I 2 ~ ~ r = H *r = h c + n
2 2

Decoding of STBC .
AT&T Labs

~ is still white c and c are detected The noise term n 1 2 independently lower complexity. complexity With M receive antennas:

~ r = H* i ri =
i =1

FG h IJ c + n ~ H K
M 2 i i =1

! !

With M receive antennas, a diversity order of 2M is achieved. Only simple linear processing at the receiver is required. Complete CSI is required at the receiver. In practice channel estimation is used to obtain CSI.

25

Decoding of STBC .
AT&T Labs

H1

soft decision for c1

r1

Linear Combiner

~ r1
c1

ML Decision

$1 c

r1 * ~ r = H1 H 2 r2

r2

LM OP NQ
~ r2

ML Decision c2

$2 c

H2

soft decision for c2

STBC: Incoherent Detection


AT&T Labs

Received Signal:

r1 c1 c2 h1 n1 ~ = * + = Ch + n r= * r2 c2 c1 h2 n2
!

LM OP LM NQ N

OPLM OP LM OP QN Q N Q

Given c1 and c2, h1 and h2 can be estimated as

~ h1 + 1 h1 1 ~ * h= ~ = 2 C r = 2 h2 + 2 c1 + c2 h2
* c1 n1 + c2 n2 1 = 2 2 c1 + c2

LM OP N Q

LM N

OP Q

and

* c2 n +cn 2 = 21 1 22 c1 + c2

26

STBC: Incoherent Detection .


AT&T Labs

If the channel is quasiquasi-static, static then the received signal for the next code symbol is:

r3 = h1c3 + h2 c4 + n3
* * r4 = h1c4 + h2c3 + n4

! !

~ Use the channel estimate h for detecting c3 and c4. Need to transmit two known symbols at the beginning of each time slot. If the channel variation is very slow, then at each time we average the channel estimate less noisy estimates. This scheme will perform within 3 dB of the case when ideal CSI is available at the receiver.

SpaceSpace-Time Block Codes for N > 2


AT&T Labs

Is there a rate 1 STBC with simple linear processing for more than two antennas (N N >2) ? Answer: theory of generalized orthogonal designs.

Complex constellation: NO. Real constellation: YES.

! ! !

Rate 1 codes with simple linear processing for arbitrary number of transmit antennas and real constellation. Rate 1/2 codes with simple linear processing for arbitrary number of transmit antennas and complex constellation. Rate 3/4 codes with simple linear processing for N=3 3 and N=4 4 transmit antennas and complex constellation.

27

SpaceSpace-Time Block Codes for N > 2 .


AT&T Labs

Example 1: Complex constellation, rate=1/2, 4 transmit antennas (N=4)

LMc OP MMc P c MNc PPQ


1 2 3 4

LM c MMc c MNc

2 3

c2 c3 c4 c1 c4 c3 c4 c1 c2 c3 c2 c1

* * * * c1 c2 c3 c4 * * * * c2 c1 c4 c3 * * * * c3 c1 c2 c4 * * * * c4 c3 c1 c2

OP PP PQ

Will provide 4 branch diversity performance + 3 dB coding gain.

Space-Time Block Codes for N > 2 .


AT&T Labs

Example 2: Complex constellation, rate=3/4, 4 transmit antennas (N=4)

LMc OP MMc PP Nc Q
1 2 3

LM c MM MM c MM c2 MM c N2
1 2 3 3

* * c3 c3 2 2 * * c c * 3 c1 3 2 2 * * * * + + + + c3 c c c2 c2 c2 c2 c1 c1 1 1 2 2 2 * * * * c3 c2 + c2 c1 + c1 c1 + c1 + c2 c2 2 2 2 * c2

OP PP PP PP PP Q

Will provide 4 branch diversity performance + 1.25 dB coding gain.

28

SpaceSpace-Time Block Codes for N > 2 .


AT&T Labs

Example 3: Real constellation, rate=1, 4 transmit antennas (N=4)

LMc OP MMc P cP MNc PQ


1 2 3 4

LM c MMc c MNc

2 3

c2 c3 c4 c1 c4 c3 c4 c1 c2 c3 c2

OP PP cP Q
1

Will provide 4 branch diversity performance and no coding gain.

Part 2: Space-Time Coding


AT&T Labs

SpaceSpace-Time coded modulation (STCM): Probability of error analysis and design criterion. SpaceSpace-time trellis codes examples. Decoding of spacespace-time trellis codes. Comparison with delaydelay-diversity. SpaceSpace-Time block codes (STBC): Basic idea and decoding of STBC. Incoherent detection of STBC. STBC for more than 2 transmit antennas. Interference suppression with STBC: Zero forcing IC and ML decoding of STBC. MMSE IC and ML decoding of STBC. Iterative IC and ML decoding of STBC.

29

Interference Cancellation with STBC


AT&T Labs
Burst 2

Information Source

Space-Time Block Encoder Burst 1 Interference Cancellation and ML Decision c1 c2 Terminal 2 Information Terminal 1 Information

Terminal 1
Burst 2

Information Source

Space-Time Block Encoder Burst 1

Terminal 2

K users, N transmit antennas per user. Classical IC techniques need N(K-1) +1 receive antennas to suppress interference from K-1 co-channel users. Exploit code structure to suppress interference only K receive antennas are required. Assumption: full synchronization between terminals. Can be implemented using adaptive LMS or RLS. Can be used for increasing system capacity or increasing data rate.

Zero Forcing IC with STBC


AT&T Labs

Received Signal:

r1 = H1 c + G1 s + n1 r2 = H 2 c + G 2 s + n2
!

LMr OP = LMH Nr Q NH
1 2

1 2

G1 c n1 + G2 s n2

OP LM OP LM OP QNQ N Q

r = H~ c+
Zero forcing solution:

LM H NH

1 2

G1 G2

$ O OP LM r OP = LM r$ OP = LMcOP + LM n $ P Q Nr Q Nr$ Q N s Q Nn Q
-1 1 2 1 2 1 2

30

Zero Forcing IC with STBC .


AT&T Labs

LMH NH
2

1 2

G1 G2

~ ~ 1 1 H = H1 G1G 2 H 2 , G = G 2 H 2 H1 G1

~ OP = LMH Q N0
1

0 ~ 1 G

OP L I Q MNH H
2 1 2

1 G1G 2

1 1

OP Q
1 2

LM I N H H
&

1 1

1 G1G 2 I

~ ~O r O LH 0 O Lc O L n OP Lr O = L ~ =M ~ P M s P + M~ P M P M P ~ r r Q N Q N Q N 0 G Q N Q Nn Q
1 2

~ ~ H and G are orthogonal and have the same structure and n2 are still white can detect c and s as before. n1 and ~ noise ~ .

Zero Forcing Interference Cancellation .


AT&T Labs

ZF Interference Cancellation:
( c, c ) = ZFIC r1 , r2 , H1 , H 2 , G1 , G 2 {
1 ~ r = r1 G1G 2 r2 ~ 1 H = H1 G1G 2 H2 ~ $ = arg min ~ c r Hc c C

~ $ r Hc c = ~ }

31

ZFIC Performance
AT&T Labs

FER Performance of 8-PSK with STBC and Zero Forcing Interference Cancellation
100

10-1

Frame Error Rate

10-2

10-3

10-4

10-5

STBC( 2 Tx, 2 Rx) + ZFIC, SIR = 0 dB STBC( 2 Tx, 1 Rx) STBC( 2 Tx, 2 Rx) , No Interference

10-6 10

15

20

25

30

SNR per Rx Antenna (dB)

MMSE Interference Cancellation


AT&T Labs

Error Criteria:

D (w, - 1 , - 2 ) = w*r - 1 c1 - - 2 c2
!

MMSE IC: find a linear combiner w, 1, and 2 such that E{(w, 1, 2)}is minimized. 1 or 2 should be set equal to 1 or else we will get a zero solution.
* - 1 = 1 D1 (w1 , - 2 ) = w1 r c1 - - 2c2 2

&

~* ~ = w 1 r c1

2 * ~* ~ - 2 = 1 D 2 (w2 , - 1 ) = w2 r c2 - - 1c1 = w 2 r c2

32

MMSE Interference Cancellation .


AT&T Labs

MMSE IC solution:

~ = M h2 w 1 h* 2 1

LM N

OP LMh OP , w ~ = L M h O Lh O MNh 1 PQ MN 0 PQ Q N0Q


-1 -1 1 1 2 2 * 1 1 SNR

M = HH* +

h1 = first column of H h2 = second column of H

MMSE Interference Cancellation .


AT&T Labs

MMSE solution:

w1 = M - h 2 h w2 = M - h h
!

( (

* -1 2

-1 h* 2 M h1 h1 , - = -1 1 - h* 2 M h1 * 2 * h1 M -1h 2 h2 , - = * 1 - h1 M -1h 2 * 1

* -1 1 1

Using the structure of M and the fact that h1 and h2 are orthogonal, orthogonal we can show that either 1=0 and 2 =1 or 1=1 and 2 =0. Two different combiners w1 and w2 for c1 and c2, respectively.

33

MMSE Interference Cancellation .


AT&T Labs

( c , c ) = M M SE_ IC r1 , r2 , H 1 , H 2 , G 1 , G 2 , { ~ r = r1T r1T


T

H1 G1 ~ H= H2 G2 M = HH* +

LM N

OP Q
T T , h 2 = h 12 hT 22 T

1 I

T h 1 = h 11 hT 21

* 1 1 = M 1h 1 , * 2 = M h2 2 *~ *~ $1 , c2 = arg min 1 $2 c1 = arg min 1 rc rc $1 C c $2 C c *~ $1 c = 1 rc 2 *~ $2 + 1 rc 2

MMSE IC Performance
AT&T Labs

FER Performance of 8-PSK with STBC and MMSE Interference Cancellation


100

10-1

Frame Error Rate

10-2

10-3

10-4 STBC( 2 Tx, 1 Rx) STBC( 2 Tx, 1 Rx) + MMSE IC, SIR = 0 dB STBC( 2 Tx, 1 Rx) + MMSE IC, SIR = 10 dB STBC( 2 Tx, 1 Rx) + MMSE IC, SIR = 20 dB STBC( 2 Tx, 2 Rx), No Interference

10-5

10-6 10

15

20

25

30

SNR per Rx Antenna (dB)

34

Iterative Interference Cancellation


AT&T Labs

$, s $ ) = IT _ IC_ STDEC r1 , r2 , H1 , H 2 , G 1 , G 2 (c {

g g
2

$ 0 , c ,o ) = IC_ STDEC r1 , r2 , H 1 , H 2 , G1 , G 2 (c $o , x 1 = r1 H 1 c $ o = arg min s


s S

$o x 2 = r2 H 2 c
2

ex

G1 s + x 2 G 2 s
2

s ,o = x 1 G 1 s + x 2 G 2 s
2

$ 1 , s ,1 ) = IC_ STDEC r1 , r2 , G 1 , G 2 , H1 , H 2 (s $1 , y1 = r1 G 1 s $ 1 = arg min c


c C 1

g
2

$1 $ 2 = r2 G 2 s y
2 1

e y H c
2

+ y2 H 2 c
2

c ,1 = y1 H 1 c + y 2 H 2 c

$, s $ ) = (c $o, s $o ) If ( c ,o + s ,o ) < ( c ,1 + s ,1 ) ( c $, s $ ) = (c $1, s $1 ) Else ( c }

Iterative MMSE IC Performance


AT&T Labs

FER Performance of 8-PSK with STBC and MMSE Iterative Interference Cancellation
100

10-1

Frame Error Rate

10-2

10-3

10-4 STBC( 2 Tx, 1 Rx) STBC( 2 Tx, 2 Rx) + Iterative MMSE IC, SIR =0 dB STBC( 2 Tx, 2 Rx) , No Interference

10-5

10-6 10

15

20

25

30

SNR per Rx Antenna (dB)

35

Iterative MMSE IC Performance .


AT&T Labs

FER Performance of 8-PSK with STBC and MMSE Iterative Interference Cancellation
100 STBC( 2 Tx, 2 Rx) + MMSE IC, SNR = 20 dB STBC( 2 Tx, 2 Rx) + ITMMSE IC, SNR = 20 dB

Frame Error Rate

10-1

10-2

10-3 -30

-20

-10

10

20

30

SIR (dB)

Part 3: Space-Time Coding Applications


AT&T Labs

! !

Narrowband TDMA Cellular. Broadband Wireless OFDM. Applications of STBC with Interference Suppression. Suppression

36

Base-Station Transmitter
AT&T Labs
Burst 2

Concatenated Space-Time Encoder


ST Encoder Information Source Block Encoder (Reed Solomon) Interleaving Interleaving

Burst Building

Pulse Shaper

Burst Building

Pulse Shaper

Burst 1

! ! !

High rate outer block code will clean up burst errors of the ST code and will provide an error detection mechanism. Same interleaving for both bursts. Synchronization and pilot sequences are inserted in both bursts for timing and frequency synchronization as well as channel estimation at the receiver.

Slot Structure
AT&T Labs

Lb
Training Sequence Data Pilot Data Pilot Pilot Data

Burst 2
Training Sequence

S2 Lt
Training Sequence

P2

P2 Ls
Pilot

P2

S2

Lp
Pilot

S1

Data

P1

Data

Pilot

P1

P1

Data

Training Sequence

S1

Lb
!

Burst 1

Training sequences S1 and S2 as well as pilot sequences P1 and P2 are orthogonal sequences. In addition these sequences are constant energy sequences, i.e. from M-PSK constellation (this will give best MSE performance).

37

Mobile Receiver
AT&T Labs
Matched Filter Timing Synchronization

Channel Estimation and Interpolation

Deinterleaving

Space-Time Vector Viterbi Decoder

Reed Solomon Decoder

Matched Filter

Timing Synchronization

Concatenated Space-Time Decoder


Channel Estimation and Interpolation
Deinterleaving

! ! !

Design assumes flat fading channel. Frequency offset is compensated for as part of the channel estimation. Channel estimates at the pilot and training symbol positions are interpolated to obtain CSI for the whole slot.

Channel Estimation
AT&T Labs

Complex channel gain: ~ ( t ) = ( t ) e j 2 fo t , f is the frequency offset ij ij o ~ ~ R ( ) = E ij ( t ) ( t + ) = J ( 2 f ) e j 2 fo

ij

Pilots and synch sequences should be positioned in the slot such that the channel sampling frequency fs satisfies 2(fd +fo) fs = LsTs. Channel Estimation:
~ (l ) + z (l ) Y j ( l ) = A P1 P2 ! P N a j j $ ij ( l ) = Pi* Y j ( l ) A Pi
2

j = 1, 2 , ! , M

i = 1, 2 , ! , N

38

Channel Interpolation
AT&T Labs

Channel estimates are interpolated to obtain complete CSI for the whole time slot. Wiener interpolation filter (WIF):
! !

Optimum interpolation filter needs the knowledge of fd , fo, and SNR. SNR A different filter for each symbol in the burst. Poor performance at low Doppler with worst case design. design Similar to WIF, except that we use same filter for all symbols in the burst. Poor CSI near the ends of the slot when only pilot symbols in that slot are used. Lower complexity than WIF, performance is, however, poorer. poorer

Optimum lowlow-pass timetime-invariant interpolation filter:


! ! !

Channel Interpolation ...


AT&T Labs
Matched Filter

Channel Estimation and Interpolation

Channel Estimates

Interpolation Filters

Select Interpolation Filter

Estimate Doppler Range

Matched Filter

Channel Estimation and Interpolation

Channel Estimates

Quasi-adaptive interpolation
! !

Doppler range is divided into different regions and pre-designed WIF is used for each range. Receiver selects the filter use based on the quality of the interpolated CSI for the previous slot

39

Modem Simulation
AT&T Labs

! ! ! ! ! ! !

STCM with 16-QAM and 8-PSK. Timing and carrier frequency offset synchronization. fc = 1.9 GHz, IS-136 basic channelization and slot structure. /4-DQPSK modulation for synchronization and pilot symbols. Oversampling factor of 8. SR raised cosine pulse shaping with 0.35% excess BW. FER Performance:
! ! ! !

FER vs. SNR and Doppler Spread. Space-Time code vs. Delay-Diversity code. Effect of interpolation filter and transmit antenna correlation. Effect of delay spread.

Slot Structure
AT&T Labs

Burst 2

6.67 ms=162 Symbols


P2 2 P1 2 D2 P2 D2 21 2 21 D1 P1 D1 21 2 21 P2 D2 2 21 P1 D1 2 21 P2 D2 P2 D2 P2 2 21 2 21 2 P1 D1 P1 D1 P1 2 21 2 21 2 D2 O2 S2 12 5 14 D1 O1 S1 12 5 14

S2 14 S1 14

O2 D2 9 5 O1 D1 9 5

Burst 1

6.67 ms=162 Symbols


S1,S2 : Synchronization Sequence (S1 and S2 are orthogonal, S1 as in IS-136) P1,P2 : Pilot Symbols (P1 and P2 are orthogonal) D1,D2 : Data O1,O2 : Overhead Symbols

40

Error Distribution
AT&T Labs

Error Histogram for the 16-QAM 16-State ST Code with 2 Tx 2 Rx Antennas and Optimized WIF @ fd = 180 Hz
1.0

Fraction of Total Number of Byte Errors per Frame

0.8

SNR = 12.5 dB SNR = 15.0 dB SNR = 17.5 dB SNR = 20.0 dB

0.6

0.4

0.2

0.0 1-2 3-4 5-6 7-8 9-10 11-12 >12

Number of Byte Errors per Frame

Frame Error Rate Performance


AT&T Labs

16-QAM 16-State ST Code with 2 Tx and 2 Rx Antennas @ fd = 180 Hz with Optimized WIF
100

PF, Frame Error Rate (FER)

10-1

10-2

10-3

ST code only ST code + RS1 ST code + RS3 ST code + RS5

10-4 14 16 18 20 22 24

SNR (per Rx antenna) dB

41

Frame Error Rate Performance .


AT&T Labs

8-PSK 32-State ST Code with 2 Tx and 2 Rx Antennas @ fd = 180 Hz and Optimized WIF
100

PF, Frame Error Rate (FER)

10-1

10-2

10-3

ST Code Only ST Code + RS1 ST Code + RS3 ST Code + RS5

10-4 8 9 10 11 12 13 14 15

SNR (per Rx antenna) dB

Frame Error Rate Performance .


AT&T Labs

16-QAM 16-State Space-Time Code with 2 Tx and 2 Rx Antennas


17
ST Code + RS5, LPIF ST Code + RS5, Optimized WIF

SNR for 0.1 Frame Error Rate (dB)

16

15

14

13

12 20 40 60 80 100 120 140 160 180

Maximum Doppler Frequency fd (Hz)

42

Frame Error Rate Performance .


AT&T Labs

8-PSK 32-State Space-Time Code with 2 Tx and 1 Rx Antennas


20
ST code + RS5, LPIF ST code + RS5, Optimized WIF

SNR for 0.1 Frame Error Rate (dB)

19

18

17

16 20 40 60 80 100 120 140 160 180

Maximum Doppler Frequency fd (Hz)

Frame Error Rate Performance .


AT&T Labs

8-PSK 32-State Space-Time Code with 2 Tx and 2 Rx Antennas


13 ST code + RS5 and LPIF ST code + RS5 and Optimized WIF

Es/No for 0.1 Frame Error Rate

12

11

10

9 20 40 60 80 100 120 140 160 180

Doppler Spread fd (Hz)

43

Frame Error Rate Performance .


AT&T Labs

8-PSK 32-State ST Code vs. Delay Diversity with 2 Tx and 1 Rx Antennas


30
DD Code + RS5, LPIF ST Code + RS5, LPIF DD Code + RS5, Optimized WIF ST Code + RS5, Optimized WIF

SNR for 0.1 Frame Error Rate (dB)

28

26

24

22

20

18

16 20 40 60 80 100 120 140 160 180

Maximum Doppler Frequency fd (Hz)

Frame Error Rate Performance .


AT&T Labs

8-PSK 32-State ST Code vs. Delay Diversity with 2 Tx and 2 Rx Antennas


20 19
DD Code + RS5, LPIF ST Code + RS5, LPIF DD Code + RS5, Optimized WIF ST Code + RS5, Optimized WIF

SNR for 0.1 Frame Error Rate (dB)

18 17 16 15 14 13 12 11 10 9 20

40

60

80

100

120

140

160

180

Maximum Doppler Frequency fd (Hz)

44

Frame Error Rate Performance .


AT&T Labs

8-PSK 32-State Space-Time Code with Optimized WIF


20

SNR for 0.1 Frame Error Rate

18

16
2 Tx 1 Rx, = 0 2 Tx 1 Rx, = 0.75 2 Tx 2 Rx, = 0 2 Tx 2 Rx, = 0.75

14

12

10

8 20 40 60 80 100 120 140 160 180

Maximum Doppler Frequency fd (Hz)

Delay Spread Performance


AT&T Labs

8-PSK 32-State ST Code with 2 Tx and 2 Rx Antennas and Optimized WIF @ fd = 180 Hz and 5 s delay spread (TU model)
100

PF , Frame Error Probability

10-1

10-2

10-3

ST code, 0 s delay spread ST code, 5 s delay spread ST code+RS5, 0 s delay spread ST code+RS5, 5 s delay spread

10-4 8 9 10 11 12 13 14 15

SNR (per Rx Antenna) dB

45

Delay Spread Performance .


AT&T Labs

8-PSK 32-State Space-Time Code with Optimized WIF @ fd = 180 Hz and 5 s delay spread (TU model)
18

SNR for 0.1 Frame Error Rate

16
2 Tx 1 Rx, 0 s delay spread 2 Tx 1 Rx, 5 s delay spread 2 Tx 2 Rx, 0 s delay spread 2 Tx 2 Rx, 5 s delay spread

14

12

10

8 20 40 60 80 100 120 140 160 180

Maximum Doppler Frequency fd (Hz)

Performance Summary
AT&T Labs
DIVERSITY MODULATION & CODING 8PSK/CST 8PSK/CST 8PSK/CST 8PSK/CST DOPPLER SPREAD 10 Hz 10 Hz 180 Hz 180 Hz SNR FOR 55.8 KBPS (ST) 18.3 dB 11.2 dB 20.1 dB 11.8 dB SNR FOR 50.4 KBPS (ST+RS3) 17.3 dB 10.3 dB 17.6 dB 10.6 dB SNR FOR 46.8 KBPS (ST+RS5) 16.9 dB 10.1 dB 17.1 dB 10.2 dB

2Tx 1Rx 2Tx 2Rx 2Tx 1Rx 2Tx 2Rx

DIVERSITY

MODULATION DOPPLER & CODING SPREAD 16-QAM/CST 16-QAM/CST 16-QAM/CST 16-QAM/CST 10 Hz 10 Hz 180 Hz 180 Hz

2Tx 1Rx 2T x 2Rx 2Tx 1Rx 2T x 2Rx

SNR FOR 74.4 KBPS (ST) 22.4 dB 15.7 dB 30.6 dB 18.1 dB

SNR FOR 67.2 KBPS (ST+RS3) 21.1 dB 13.8 dB 25 .0 dB 15.3 dB

SNR FOR 62.4 KBPS (ST+RS5) 20.5 dB 13.2 dB 23.0 dB 14.3 dB

46

Part 3: Space-Time Coding Applications


AT&T Labs

! !

Narrowband TDMA Cellular. Broadband Wireless OFDM. Applications of STBC with Interference Suppression. Suppression

Broadband Wireless OFDM


AT&T Labs
Burst 2

Space-Time Coded OFDM Tx


ST Encoder Information Source Block Encoder (Reed Solomon) IFFT IFFT

RF

RF
Burst 1

RF

IFFT

Space-Time Vector Viterbi Decoder

Reed Solomon Decoder

RF

IFFT

Space-Time Coded OFDM Rx

47

Performance of STST-Coded OFDM


AT&T Labs

1 MHz bandwidth. 256 tones. ! QPSK 16-state ST code with 2 transmit antennas. ! 2-Ray fading channel model with equal strength. ! Ideal CSI. CSI 1.5 Mbps data rate (1.5 bits/sec/Hz) For 10% FER, FER an Eb/No between 2.72.7-4 dB is required (depending on delay spread). With 16-QAM, 4 Tx and 2 Rx antennas, 6 bits/sec /Hz is achievable. Can be used for downlink.
! !

Frame Error Probability for Concatenated Space-Time Coded OFDM with 4-PSK 16-State ST Code and 2 Tx and 2 Rx Antennas
100 0 s Delay Spread 5 s Delay Spread 40 s Delay Spread

Frame Error Probability

10-1

10-2 5 6 7 8 9 10

SNR per Rx Antenna (dB)

Part 3: Space-Time Coding Applications


AT&T Labs

! !

Narrowband TDMA Cellular. Broadband Wireless OFDM. Applications of STBC with Interference Suppression.

48

Increase Capacity
AT&T Labs

Channel Encoder

Space-Time Block Code

Modulator

Concatenated Encoder

Modulator

Space-Time Processor: ST Decoder, Interference Suppression, and MLSE

Soft Decision Generator

Channel Decoder

Concatenated Decoder

Outer FEC code provides immunity against channel errors. errors Inner STBC provides interference suppression and fading mitigation through transmit diversity.

Performance
AT&T Labs

FER Performance of 8-PSK with FEC + STBC and MMSE Interference Cancellation at 0 dB SIR
100

10-1

Frame Error Rate

10-2

10-3

STBC Only STBC + RS(57,63) over GF(64) STBC + RS(53,63) over GF(64)

10-4 10

15

20

25

30

SNR per Rx Antenna (dB)

49

Increase Data Rate


AT&T Labs

Channel Encoder R1 Serial to Parallel

Space-Time Block Code

Modulator

Concatenated Encoder

Modulator

Channel Encoder R2

Space-Time Block Code

Modulator

Concatenated Encoder

Modulator

Throughput:

7 = 0.5 R1 1 - FER1 + R2 1 - FER2

Performance
AT&T Labs

Throughput Performance of 8-PSK with STBC and Unequal Error Protection with RS(53,63)
1.0 0.9 0.8 0.7

Throughput

0.6 0.5 0.4 0.3 R1= 1 and R2 =1 0.2 0.1 0.0 10 R1 = 53/63 and R2 = 1

15

20

25

SNR per Rx Antenna (dB)

50

Summary
AT&T Labs

! !

Space-Time coding: good idea with great potential. Current work in space-time coding:

Systematic design of space-time trellis codes. Multi-channel equalization for space-time coding. Applications of space-time coding to WCDMA.

51