Space - Time Coding and Signal Processing For High Data Rate Wireless Communcations

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
AT&T Labs
! !
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.

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

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)
AT&T Labs
! !
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

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

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 Error Rate
10-1
8-State Code 16-State Code 32-State Code 64-State Code
10-2 10 12 14 16 18 20
21
STC: Performance with Perfect CSI .

AT&T Labs
Performance of 16-QAM Space-Time Codes with 2 Transmit and 2 Receive Antennas

100

16-State Code 64-State Code
Frame Error Rate
10-1
10-2
10-3 10 12 14 16 18 20
Delay Diversity
AT&T Labs
Signal 2
Constellation Mapper
Information Source
Rate Repetition Code
One Symbol Delay
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
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

AT&T Labs
SpaceSpace-Time Block Codes

AT&T Labs
Signal 2 Constellation Mapper ST Block Code Information Source
c1 c 2
c1 c2
* -c2 c* 1
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.

AT&T Labs
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
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

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

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
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
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

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 Estimates
Interpolation Filters
Select Interpolation Filter
Estimate Doppler Range
Matched Filter
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

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

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
19
18
17
16 20 40 60 80 100 120 140 160 180

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

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
28
26
24
22
20
18
16 20 40 60 80 100 120 140 160 180

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
18 17 16 15 14 13 12 11 10 9 20
40
60
80
100
120
140
160
180
44

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
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
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

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

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
49
Increase Data Rate

AT&T Labs
Channel Encoder R1 Serial to Parallel
Modulator
Modulator
Channel Encoder R2
Modulator
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
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

Space - Time Coding and Signal Processing For High Data Rate Wireless Communcations

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Space - Time Coding and Signal Processing For High Data Rate Wireless Communcations

Uploaded by

Copyright:

Available Formats

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

Receive and transmit diversity techniques.

Local Scatterers to Base Co-Channel Mobile

Local Scatterers to Mobile

Local Scatterers to Base

Local Scattering Multipath Propagation Mobile Motion Cellular Spectrum Reuse

Fading Intersymbol Interference Time Varying Channel Co-channel Interference

Signal Level in Wireless Channels

Long Term Fading

Classification of Wireless Channels

The multipath channel:

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

The three spreads: spreads

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

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

Receive and transmit diversity techniques.

Fundamental Limits: Outage Capacity

Fundamental Limits: Outage Capacity

Channel capacity for SISO channel:

bits/sec/use, is the SNR

Channel capacity for MIMO channel:

Fundamental Limits: Outage Capacity

Achievable Data Rates with Multiple Antennas

We are here now

SNR per Rx Antenna (dB)

It is dangerous to put limit on wireless

Receive and transmit diversity techniques.

Diversity in Wireless Radio

Rx: Selection and Switched Diversity

Data Detector 1 Data

Switch Data Detector Data

Receive Diversity: Linear Combining

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

Receive Diversity Performance

No training. Feedforward information.

Transmit Diversity with Feedback

Tx Diversity with Frequency Weighting

(B) Distructive Interference

Tx Diversity with Antenna Hopping

Tx Diversity with Channel Coding

Transmit Diversity via Delay Diversity

Transmit Diversity Options

Antenna hopping Frequency weighting Delay diversity SpaceSpace-Time Coding

Part 2: Space-Time Coding

Space-Time Coded Modulation

Space-Time Encoder Information Source Receiver

Space-Time Coding: The Model

STC: The Model .

Transmitted Code Vector:

Received Signal Vector:

STC: Probability of Error Analysis

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

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

STC: Probability of Error Analysis .

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

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

STC: Probability of Error Analysis .

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

STC: Probability of Error Analysis .

Probability of error: average over all | ij |2

1I t FGH IJK FGH JK

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

* 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