Th

1
Iterative receivers for multi-antenna systems

• Thèse présentée devant lINSA de Rennes en vue de
  lobtention du doctorat dÉlectronique

Pierre-Jean BOUVET Le 13 décembre 2005
2
Foreword
Foreword

• RD Unit
• Broadband Wireless Acces / Innovative Radio
  Interface (RESA/BWA/IRI)
• Supervisor
• Maryline HELARD, RD engineer HDR at France
  Telecom RD division
• Context
• Internal project SYCOMORE (research on digital
  communications)
• European project IST 4-MORE (4G demonstrator
  based on MIMO and MC-CDMA techniques)

3
Outline
Outline

1. Introduction
2. Multi-antenna techniques
3. Generic iterative receiver
4. Optimal space-time coding
5. Application to MC-CDMA
6. Conclusion

4
Part I Introduction
5
Context

• Digital wireless communications
• High spectral efficiency
• Robustness
• Radio-mobile application
• Multi-path propagation
• Mobility
• Multi-user access

Time and frequency selective channel
6
Multi-antenna (MIMO) transmissions

• Principle
• Multi-antenna at transmitter and receiver
• MIMO capacity Telatar 95

covariance of rank of singular values of
SISO capacity
7
Multi-antenna (MIMO) transmissions

• Motivations
• Spectral efficiency gain
• Performance gain
• Spatial diversity gains
• Antenna array gains
• Limits
• Interference terms
• Co Antenna Interference (CAI)
• Spatial correlation
• Antennas must be sufficiently spaced
• Rich scattering environment required
• Optimal MIMO capacity exploitation
• Complex algorithm not well suited for practical
  implementation
• Lack of generic schemes

Capacity gain linear in min(Nt, Nr)
8
Objectives

• Multi-antenna transmission
• Spectral efficiency gain
• Arbitrary antenna configuration
• Near-optimal reception
• MIMO capacity exploitation
• Iterative (turbo) principle
• Low complexity algorithm
• Multi-user access

9
Part II MIMO techniques
10
Transmitter
Information bits
Modulation symbols
Coded bits
Convolutional code
BICM scheme Caire et al. 98
11
MIMO channel

• Multi carrier approach (OFDM)

Equivalent flat fading MIMO channels
Reduced complexity MIMO equalization (no ISI
treatment)
12
MIMO channel

• Equivalent flat fading MIMO channel

• By assuming ideal symbol interleaving
• T-block Rayleigh fading model
• Represents the optimal performance of a MIMO-OFDM
  system over a radio-mobile channel

13
Classification of MIMO techniques
Channel State Information (CSI)

• CSI required at Tx and Rx
• Eigen beam forming
• Water-filling
• Pre-equalization
• CSI required only at Rx
• Treillis based
• Block based
• No CSI required
• Differential STC
• USTM

14
Classification of MIMO techniques
Channel State Information (CSI)

• CSI required at Tx and Rx
• Eigen beam forming
• Water-filling
• Pre-equalization
• CSI required only at Rx
• Treillis based
• Block based
• No CSI required
• Differential STC
• USTM

15
Classification of MIMO techniques

• Spatial Data Multiplexing (SDM)
• Foschini et al. 96, Wolniansky et al. 98

• CSI required at Tx and Rx
• Eigen beam forming
• Water-filling
• Pre-equalization
• CSI required only at Rx
• Treillis based
• Block based
• No CSI required
• Differential STC
• USTM

• Space Time Block Coding (STBC)
• Alamouti 98, Tarokh et al. 99

• Linear Precoded STBC
• Da Silva et al. 98

• Algebraical STBC
• Damen et al. 03, El Gamal et al. 03

16
LD Code
STC latency
Input block length
STC rate
17
Equivalent representation
Joint space-time coding and channel representation
18
Special LD Code



Examples
19
Solution

• Transmission matrices
• Reception matrices
• Equivalent channel matrix

20
Example Alamouti Code over channel

• Transmission matrices
• Equivalent model

21
Co-antenna interference
Desired signal
Noise
CAI terms
Multi-antenna transmission provides CAI terms
CAI terms can be treated like ISI terms (which
were due to the frequency selectivity in SISO
transmission)
22
Part III Generic iterative receiver
23
Reception state of the art

• Optimal solution joint detection
• ML detection based on a super trellis
• Sub-optimal solution
• Disjoint decoding MIMO detection ? channel
  decoding
• MAP MIMO detection
• SIC, OSIC, PIC detection
• MRC, MMSE, ZF equalization
• Iterative decoding MIMO detection ? channel
  decoding Berrou et al. 93
• MAP MIMO detection
• Tonello 00, Boutros et al. 00, Vikalo et al. 02
• Filtered based MIMO equalization
• Sellathurai et al. 00, Gueguen 03, Witzke et al.
  03

Optimal performance
Very high complexity
Relative low complexity
Optimal performance for orthogonal STC (Alamouti)
Sub-optimal performance for non-orthogonal STC
Near optimal performance
High complexity
Near optimal performance
reduced complexity
24
Reception state of the art

• Optimal solution joint detection
• ML detection based on a super trellis
• Sub-optimal solution
• Disjoint decoding MIMO detection ? channel
  decoding
• MAP MIMO detection
• SIC, OSIC, PIC detection
• MRC, MMSE, ZF equalization
• Iterative decoding MIMO detection ? channel
  decoding Berrou et al. 93
• MAP MIMO detection
• Tonello 00, Boutros et al. 00, Vikalo et al. 02
• Filtered based MIMO equalization
• Sellathurai et al. 00, Gueguen 03, Witzke et al.
  03

Near optimal performance
reduced complexity
25
Principle
Channel decoding stage
MIMO equalization stage

• Application of the turbo-equalization concept to
  MIMO

26
MIMO equalizer (1)

• MMSE based soft interference cancellation
  (MMSE-IC)
• Glavieux et al. 97, Wang et al. 99, Reynolds et
  al. 01, Tüchler et al. 02, Laot et al. 05
• MMSE optimization of both filters

27
MIMO equalizer (2)

• Optimal solution MMSE-IC
• Time invariant approximation MMSE-IC(1)

TNr x TNr matrix inversion
28
MIMO equalizer (3)

• Matched filter approximation MMSE-IC(2)
• Zero-Forcing solution ZF-IC

Iteration 1
Iteration p
Iteration 1
Iteration p
29
Complexity analysis (MIMO equalizer)
Proposed iterative receivers provide complexity
gain
30
Asymptotical analysis

• Asymptotical performances Genie aided receiver
• Asymptotical equivalent channel

31
Asymptotical diversity

• Pair-wise error probability
• Chi-square approximation and Chernoff bound

32
Asymptotical diversity

• Proposed definition of the space-time diversity
• Total diversity exploited by both channel and
  space-time coding
• Modified Singleton Bound Gresset et al. 04

Full channel diversity can only be achieved by
using jointly channel coding and space-time coding
33
Performance results simulation conditions

• Theoretical independent T-Block Rayleigh flat
  fading MIMO channel
• Non recursive non systematic convolutional code
  (133,171)o, K7
• SOVA algorithm for channel decoding
• No spatial correlation
• Normalized BER
• Asymptotical curve Matched filter Bound (MFB)
• Optimal curve AWGN decoupled

Receive array gain not taken into account
Genie aided receiver
Min(Nt,Nr) parallel AWGN channels
34
Performance results Jafarkhani code
Iterative decoding
Disjoint decoding
MFB is reached whichever iterative algorithm is
used
5 iterations are sufficient
0.8 dB gain at 10-4 versus disjoint MAP receiver
(state of the art)
35
Performance results SDM
Disjoint decoding
Iterative decoding
MFB is reached only with the MMSE-IC(1) receiver
7 dB gain at 10-4 versus disjoint MMSE receiver
36
Performance results SDM overloaded
Disjoint decoding
Iterative decoding
MFB is reached only with the MMSE-IC(1) receiver
The Iterative receiver still converges although
the rank of is degenerated
37
Synthesis

• Derivation of a MMSE iterative receiver for
  generic MIMO transmission
• Reduced complexity versus MAP based iterative
  algorithm
• Asymptotical analysis
• Proposition of an estimation of the space-time
  coding diversity
• Simulation results
• MMSE-IC(1) tends towards the MFB curve whichever
  space-time coding scheme is used
• MMSE-IC(1) still works in case of rank
  degenerated channel matrix
• MMSE-IC(2) and ZF-IC converge when CAI terms are
  quite low and/or for small order modulation

38
Part IV Optimal space-time coding
39
Optimality conditions

1. Maximizing data rate
2. Maximizing space-time coding diversity
3. Minimizing and
4. Minimizing the non orthogonal terms of

40
Optimality conditions

1. Maximizing data rate
2. Maximizing space-time coding diversity
3. Minimizing and
4. Minimizing the non orthogonal terms of

41
Maximizing data rate

• Ergodic Capacity
• High SNR approximation (Foschini et al. 96)

42
Maximizing the diversity

• Assuming ML detection
• Pairwise error probability analysis
• Diversity gain maximization
• TAST El Gamal et al. 03, FDFR Ma et al. 03
• Assuming MMSE-IC reception
• Asymptotical analysis
• Space-time coding diversity maximization
• Sufficient condition Along a space-time coded
  block, each data symbol must be transmitted
  uniquely by each antenna

43
Summary

• Conditions
• STC construction rule
• During Nt symbol durations, min(Nt,Nr) data
  symbols have to be uniquely transmitted by the Nt
  antennas

1
2
3
44
Diagonal Threaded Space Time (DTST) coding
45
Example over a channel
Optimal with iterative decoding
46
Performance results

• 4 transmit antennas and 2 receive antennas
• Channel model T-block Rayleigh flat fading
• No spatial correlation
• Reception
• If S is orthogonal MRC
• If S is non orthogonal MMSE-IC with 5 iterations
• Optimal performance AWGN decoupled
• Corresponds to virtual
  parallel AWGN channels

47
System Parameters
Alamouti AS
Jafarkhani
Double Alamouti (DA)
DTST
48
Ergodic capacity
Near optimal exploitation for DA and DTST schemes
49
BER Performance
2 bps/Hz
Best performance achieved with DTST (and DA)
50
Capacity at BER10-4
When increasing the spectral efficiency, only the
iterative system is able to exploit the MIMO
capacity
51
Synthesis

• Construction criteria of optimal LD code
• DTST code
• Check the optimality criteria
• Subset of special linear dispersion code family
• Generic construction scheme
• Simulation results
• DTST codes lead to near optimal exploitation of
  MIMO capacity and spatial diversity

52
Part V Application to MC-CDMA
53
MC-CDMA

• Introduced in 93 Yee et al. 93, Fazel et al.
  93
• Aim
• to spread multi-user information in the frequency
  domain
• Principle
• Combination of CDMA and OFDM techniques
• Benefits
• Robustness against multi-path channels
• Multi-user flexibility
• Low multi-access interference (MAI) in downlink
  scenario

54
MIMO MC-CDMA Transmitter
55
Equivalent model

• Equivalent channel matrix
• Receive signal
• Receiver algorithm
• Since S is a special LD code, proposed MMSE-IC
  receiver can be used

Desired signal
noise
MAI CAI terms
56
Multi-user iterative receiver

• MMSE-IC (1) solution
• Full load approximation

Nu x Nu matrix inversion
TNr x TNr matrix inversions
Complexity of the equalization stage equivalent
to the OFDM case
multi-user complexity each user must be channel
decoded
57
Performance results 4x2 Bran E channel

• Bran E model

• Transmission parameters specified by the IST 4
  MORE project for DL transmission

Bit Interleaving depth 512/user for QPSK 1024/user for 16-QAM
FFT size 1024
Nc 695
CP size 256 samples
W 41.7 MHz
Fo 5 GHz
Velocity 16.6 m/s
Number of taps 12
Fs 50 MHz
No spatial correlation
58
Performance results 4x2 Bran E channel

• DA code

• Alamouti AS

CAI MAI terms
MAI terms
Multi user MMSE-IC receiver
Single user MMSE receiver
59
Performance results 4x2 Bran E channel
Perfect channel estimation
Alamouti AS SU MMSE receiver
Small degradation compared to Rayleigh i.i.d.
channel
DA code outperforms Alamouti code
60
Performance results 4x2 Bran E channel
Imperfect channel estimation Basic pilots aided
algorithm with 1D interpolation (16 of pilots)
same impact whichever receiver is used
DA code still outperforms Alamouti AS code
2.1 dB
1.9 dB
61
Synthesis

• MIMO MC-CDMA systems with iterative decoding
• Exploitation of MC-CDMA advantages and MIMO
  capacity
• Multi-user algorithm complexity (each user must
  be individually decoded)
• Equalization stage based on linear filters
• Near-optimal performance no matter what the load
• Application to realistic channels
• Small degradation compared to theoretical channel
• Impact of channel estimation is satisfactory

62
Part VII Conclusion
63
General conclusion

• MIMO capacity can be efficiently exploited by
  iterative processing
• MMSE-IC based solutions lead to low complexity
  algorithm (especially comparing to MAP based
  solution)
• High order modulations are suitable
• High number of antennas can be considered
• MMSE-IC receiver can be derived for MC-CDMA
  transmission
• The behavior of MMSE-IC receiver over realistic
  channel including channel estimation is
  satisfactory

64
Major contributions

• Proposition and analysis of a MIMO iterative
  receiver
• Generic structure
• Reduced complexity algorithms
• Theoretical analysis (complexity and asymptotical
  behavior)
• Proposition of new optimal LD codes
• DTST
• Application of iterative reception
• MC-CDMA
• Linear precoding
• Performance results
• Theoretical channels
• Realistic channels (channel estimation and
  spatial correlation)

65
Future prospects

• Iterative channel estimation
• Joint channel estimation and decoding
• Turbo-codes instead of convolutional codes as
  channel coding
• Multi-loop iterative scheme
• Real channels
• Realistic spatial correlation model
• Application to OFDMA
• Implementation issues

66
Publications and patents

• International Conference
• P-J. Bouvet and M. Hélard, Near optimal
  performance for high data rate MIMO MC-CDMA
  scheme, MC-SS 05
• B. Le Saux, M. Hélard and P-J. Bouvet,
  Comparison of coherent and non-coherent space
  time schemes for frequency selective fast-varying
  channels , IEEE ISWCS 05
• P-J. Bouvet, M. Hélard and V. Le Nir, Low
  complexity iterative receiver for linear precoded
  OFDM, IEEE WiMob 05
• P-J. Bouvet and M. Hélard, Efficient iterative
  receiver for spatial multiplexed OFDM system over
  time and frequency selective channels, WWC 05
• P-J. Bouvet, M. Hélard and V. Le Nir, Low
  complexity iterative receiver for non-orthogonal
  space-time block code with channel coding, IEEE
  VTC Fall 04
• P-J. Bouvet, V. Le Nir, M. Hélard and R. Le
  Gouable, Spatial multiplexed coded MC-CDMA with
  iterative receiver IEEE PIMRC 04

67
Publications and patents

• International Conference (contd)
• P-J. Bouvet, M. Hélard and V. Le Nir, Low
  complexity iterative receiver for linear precoded
  MIMO systems, IEEE ISSSTA 04
• M. Hélard, P-J. Bouvet, C. Langlais, Y. M. Morgan
  and I. Siaud, On the performance of a Turbo
  Equalizer including Blind Equalizer over Time and
  Frequency Selective Channel. Comparison with an
  OFDM system, Symposium Turbo 03
• C. Langlais, P-J. Bouvet, M. Hélard and C. Laot,
  Which Interleaver for turbo-equalization system
  on frequency and time selective channels for high
  order modulations ? , IEEE SPAWC 03
• National conference
• B. Le Saux, M. Hélard and P.-J Bouvet,
  Comparaison de technique MIMO cohérents et
  non-cohérentes sur canal rapide sélectif en
  fréquence, MajeSTIC 05

68
Publications and patents

• Patents
• P-J Bouvet and M. Hélard, Procédé démission
  dun signal ayant subi un précodage linéaire,
  procédé de réception, signal, dispositifs et
  programmes dordinateur correspondant , Nov. 05
• J-P. Javaudin and P-J. Bouvet, Procédé de codage
  d'un signal multiporteuse de type OFDM/OQAM
  utilisant des symboles à valeurs complexes,
  signal, dispositifs et programmes d'ordinateur
  correspondants, May 05
• J-P. Javaudin and P-J. Bouvet, Procédé de
  décodage itératif d'un signal OFDM/OQAM utilisant
  des symboles à valeurs complexes, dispositif et
  programme d'ordinateur correspondants, May 05
• P-J. Bouvet and M. Hélard, Procédé de réception
  itératif d'un signal multiporteuse à annulation
  d'interférence, récepteur et programme
  d'ordinateur correspondants, March 05

69
Publications and patents

• Patents (contd)
• P-J. Bouvet, M. Hélard and V. Le Nir, Procédé
  de réception itératif pour système de type MIMO,
  récepteur et programme d'ordinateur
  correspondants , Nov. 04
• P-J. Bouvet, V. Le Nir and M. Hélard, Procédé
  de réception d'un signal ayant subi un précodage
  linéaire et un codage de canal, dispositif de
  réception et produit programme d'ordinateur
  correspondants , Jun. 04
• M. Hélard, P-J. Bouvet, V. Le Nir and R. Le
  Gouable, Procédé de décodage d'un signal codé à
  l'aide d'une matrice espace-temps, récepteur et
  procédé de codage et décodage correspondant ,
  Sept. 03

70
Questions
Questions?