Combined Multiuser Detection and Channel Decoding with Receiver Diversity

1
Combined Multiuser Detection and Channel
Decodingwith Receiver Diversity

• IEEE GLOBECOM
• Communications Theory Mini-Conference
• Sydney, Australia November 10, 1998
• Matthew C. Valenti and Brian D. Woerner
• Mobile and Portable Radio Research Group
• Virginia Tech
• Blacksburg, Virginia

2
Outline of Talk

• Multiuser detection for TDMA systems.
• Macrodiversity combining for TDMA.
• Turbo-MUD for convolutionally coded asynchronous
  multiple-access systems.
• Proposed System.
• The Log-MAP algorithm.
• For decoding convolutional codes.
• For performing MUD.
• Simulation results for fading channels.

3
Multiuser Detection for the TDMA Uplink

• For CDMA systems
• Resolvable interference comes from within the
  same cell.
• Each cochannel user has a distinct spreading
  code.
• Large number of (weak) cochannel interferers.
• For TDMA systems
• Cochannel interference comes from other cells.
• Cochannel users do not have distinct spreading
  codes.
• Small number of (strong) cochannel interferers.
• MUD can still improve performance for TDMA.
• Signals cannot be separated based on spreading
  codes.
• Delay, phase, and signal power can be used.

4
Macrodiversity Combining for the TDMA Uplink

• In TDMA systems, the cochannel interference comes
  from adjacent cells.
• Interferers to one BS are desired signals to
  another BS.
• Performance could be improved if the base
  stations were allowed to share information.
• If the outputs of the multiuser detectors are
  log-likelihood ratios, then adding the outputs
  improves performance.

BS 1
MS 1
BS 3
MS 3
MS 2
BS 2
5
Macrodiversity Combiner

• Each of M base stations has a multiuser detector.
• Each MUD produces a log-likelihood ratio of the
  code bits.
• The LLRs are added together prior to the final
  decision.

Multiuser Estimator 1
Multiuser Estimator M
6
Turbo Multiuser Detection

• Most TDMA systems use forward error correction
  (FEC) coding.
• The process of multiuser detection and FEC can be
  combined using iterative processing.
• Turbo-MUD
• This is analogous to the decoding of serially
  concatenated turbo codes, where
• The outer code is the convolutional code.
• The inner code is an MAI channel.
• The MAI channel can be thought of as a time
  varying convolutional code with complex-valued
  coefficients.

7
Turbo MUD System Diagram
multiuser interleaver
Convolutional Encoder 1
interleaver 1
MAI Channel
MUX
n(t) AWGN
Convolutional Encoder K
interleaver K
Turbo MUD
multiuser interleaver
APP
Bank of K SISO Decoders
SISO MUD
multiuser deinterleaver
Estimated Data
8
Macrodiversity Combining for Coded TDMA Systems

• Each base station has a multiuser estimator.
• Sum the LLR outputs of each MUD.
• Pass through a bank of Log-MAP channel decoder.
• Feed back LLR outputs of the decoders.

9
The Log-MAP Algorithm

• The Viterbi Algorithm can be used to implement
• The MUD (Verdu, 1984).
• The convolutional decoder.
• However, the outputs are hard.
• The iterative processor requires soft outputs.
• In the form of a log-likelihood ratio (LLR).
• The symbol-by-symbol MAP algorithm can be used.
• Bahl, Cocke, Jelinek, Raviv, 1974. (BCJR
  Algorithm)
• The Log-MAP algorithm is performed in the Log
  domain,
• Robertson, Hoeher, Villebrun, 1997.
• More stable, less complex than BCJR Algorithm.
• We use Log-MAP for both MUD and FEC.

10
MAI Channel Model

• Received signal at base station m
• Where
• a is the signature waveform of all users.
• Assumed to be a rectangular pulse.
• ?k,m is a random delay of user k at receiver m.
• Pk,mi is power at receiver m of user ks ith
  bit.
• Matched filter output for user k at base station
  m

11
Log-MAP MUD AlgorithmSetup

• Place y and b into vectors
• Place the fading amplitudes into a vector
• Compute cross-correlation matrix for each BS
• Assuming rectangular pulse shaping.

12
Log-MAP MUD AlgorithmExecution
S3
S2
S1
S0
i 0
i 6
i 3
i 2
i 1
i 4
i 5
Jacobian Logarithm
Branch Metric
13
Simulation Parameters

• The uplink of a TDMA system was simulated.
• 120 degree sectorized antennas.
• 3 cochannel interferers in the first tier.
• K3 users.
• M3 base stations.
• Fully-interleaved Rayleigh flat-fading.
• Perfect channel estimation assumed.
• Each user is convolutionally encoded.
• Constraint Length W 3.
• Rate r 1/2.
• Block size L4,096 bits
• 64 by 64 bit block interleaver

14
Performance for Constant C/I 7dB
15
Performance for Constant Eb/No 6dB
16
Conclusion and Future Work

• MUD can improve the performance of TDMA system.
• Performance can be further improved by
• Combining the outputs of the base stations.
• Performing iterative error correction and
  multiuser detection.
• This requires that the output of both the MUDs
  and FEC-decoders be in the form of log-likelihood
  ratios.
• Log-MAP algorithm used for both MUD and FEC.
• The study assumes perfect channel estimates.
• The effect of channel estimation should be
  considered.
• Decision directed estimation should be possible.
• Output of each base station can assist estimation
  at the others.

17
Uncoded Performance for Constant C/I

• C/I 7 dB
• Performance improves with MUD at one base
  station.
• An additional performance improvement obtained by
  combining the outputs of the three base stations.

18
Uncoded Performance for Constant Eb/No

• Performance as a function of C/I.
• Eb/No 20 dB.
• For conventional receiver, performance is worse
  as C/I gets smaller.
• Performance of single-base station MUD is
  invariant to C/I.
• Near-far resistant.
• For macrodiversity combining, performance
  improves as C/I gets smaller.