1. West Virginia University
7/8/99
An Optimal Soft-Output Multiuser Detection Algorithm and its Applications
Matthew C. Valenti
Assistant Professor
Comp. Sci. & Elect. Eng.
West Virginia University
Morgantown, WV
U.S.A.
Introduce yourself.
Funded by the Bradley Fellowship and the Office of Naval Research.
Iterative Detection and Decoding for Wireless Communications

2. Outline of Talk Turbo multiuser detection. System model.
Related work.
System model.
The optimal SISO MUD algorithm
Applications of SISO MUD
Antenna arrays
Distributed multiuser detection.
Introduction

3. Turbo Multiuser Detection
Time-varying FIR filter
“multiuser interleaver”
FEC
Encoder #1
interleaver #1
Channel
MAI
Channel
Model
Parallel to
Serial
n(t)
AWGN
FEC
Encoder #K
Introduction
interleaver #K
Turbo
MUD
Extrinsic Info
multiuser
interleaver
Bank of
K SISO
Decoders
SISO
MUD
multiuser
deinterleaver
Estimated
Data

4. Some Developments in Turbo Multiuser Detection
Gialllorenzi and Wilson
1996: Trans. Comm.
Hypertrellis approach. Not iterative. No interleaving.
Vojcic, Shama, Pickholtz
1997: ISIT
Optimal Soft Output MAP. Asynchronous. Not iterative.
No noise whitening.
Reed, Schlegel, Alexander, Asenstorfer
1997: Turbo Code Symposium, PIMRC.
Several Journal Papers (Trans. Comm., JSAC, ETT)
Early work considered synchronous, later asynchronous.
M. Moher
1998: Trans. Comm. (synchronous), Comm. Letters. (asynchronous)
Based on cross entropy minimization.
Introduction

5. complex Rayleigh fading
System Model
encoder
interleaver
modulator
bank of
matched
filters
transmitter 1
receiver 1
asynchronous channel
AWGN or
complex Rayleigh fading
bank of
matched
filters
encoder
interleaver
modulator
receiver M
transmitter K

6. Whitened Matched Filter Output
Matrix notation for output of matched filter at mth receiver
Cholesky decomposition
Whitened matched filter output
colored noise
crosscorrelations
transmitted symbols (round-robin)
channel gains (diagonal)
Optimal SISO MUD
white noise, variance = No/(2Es)
lower triangular, only K diagonals

7. Metric for Optimal SISO MUD
Trellis representation:
Noiseless Reconstruction of the signal:
Branch metric:
Now, just use MAP algorithm.
Optimal SISO MUD
constant
ignore for LLR
Squared Euclidian distance
between received symbol and
noiseless reconstruction of signal
Term incorporating
the extrinsic information Z

8. Turbo MUD for Direct Sequence CDMA
CDMA: Code Division Multiple Access
The users are assigned distinct waveforms.
Spreading/signature sequences
All users transmit at same time/frequency.
Use a wide bandwidth signal
Processing gain Ns
Ratio of bandwidth after spreading to bandwidth before
MUD for CDMA
The resolvable MAI originates from the same cell.
Intracell interference.
MUD uses observations from only one base station.
M=1 case.
Applications

9. Performance of Turbo-MUD for CDMA in AWGN
K = 5 users
Spreading gain Ns = 7
Convolutional code: Kc = 3, r=1/2
Eb/No = 5 dB
1  K  9

10. Performance of Turbo-MUD for CDMA in Rayleigh Flat-fading
K = 5 users
Fully-interleaved fading
Eb/No = 9 dB
1  K  9

11. Turbo MUD for TDMA TDMA: Time Division Multiple Access
Users are assigned unique time slots
All users transmit at same frequency
All users have the same waveform, g(t)
TDMA can be considered a special case of CDMA, with gk(t) = g(t) for all cochannel k.
MUD for TDMA
Usually there is only one user per time-slot per cell.
The interference comes from nearby cells.
Intercell interference.
Observations from only one base station might not be sufficient.
Performance is improved by combining outputs from multiple base stations.
Applications

12. Performance of Turbo-MUD for TDMA in AWGN
K = 3 users
Convolutional code: Kc = 3, r=1/2
Observations at 1 base station
Eb/No = 5 dB
1  K  9

13. Performance of Turbo-MUD for TDMA in Rayleigh Flat-Fading
K = 3 users
Fully-interleaved fading
Eb/No = 9 dB
1  K  9

14. Antenna Arrays Consider an antenna array with M elements.
In this case, M>1
Each element has its own multiuser detector.
Can use the SISO MUD algorithm.
Antenna elements should be far enough apart that the signals are uncorrelated.
Applications
array
element #1
Multiuser
Detector #1
array
element #M
Multiuser
Detector #M

15. Distributed Multiuser Detection
Why must the elements of an antenna array be located at the same base station?
We could synthesize an antenna array by using the antennas of spatially separated base stations.
A benefit is now signals will be uncorrelated.
Applications
base
station #1
Multiuser
Detector #1
base
station #M
Multiuser
Detector #M

16. Cellular Network TopologyF2 F1 F3 F4 F5 F6 F7 F3 F2 F4 F1 F7 F5 F2 F1 F3 F4 F5 F6 F7 F6
Alternative layout
120 degree sectorized antennas
Located in 3 corners of cell
Frequency reuse factor 3
Conventional layout
Isotropic antennas in cell center
Frequency reuse factor 7

17. Performance of Distributed MUD
Eb/No = 20 dB
1  K  9
For conventional receiver:
Performance degrades quickly with increasing K.
Only small benefit to using observations from multiple BS.
With multiuser detection:
Performance degrades very slowly with increasing K.
Order of magnitude decrease in BER by using multiple observations.
Now multiple cochannel users per cell are allowed.

18. Cooperative Decoding for the TDMA Uplink
Now consider the coded case.
The outputs of the MUD’s are summed and passed through a bank of decoders.
The SISO decoder outputs are fed back to the multiuser detectors to be used as a priori information.
Applications
Extrinsic
Info
Multiuser
Detector #1
Bank of
K SISO
Channel
Decoders
Estimated
Data
Multiuser
Detector #M

19. Performance of Cooperative Decoding
K = 3 transmitters
Randomly placed in cell.
M = 3 receivers (BS’s)
Corners of cell
path loss ne = 3
Fully-interleaved Rayleigh flat-fading
Convolutional code
Kc = 3, r = 1/2

20. Performance of Cooperative Decoding
Eb/No = 5 dB
1  K  9
Randomly placed in cell.
M = 3 receivers
For conventional receiver:
Performance degrades quickly with increasing K.
Only small benefit to using observations from multiple BS.
With multiuser detection:
Performance degrades gracefully with increasing K.
No benefit after third iteration.
Could allow an increase in TDMA system capacity.

21. Conclusion An optimal SISO MUD algorithm has been derived.
Complexity is exponential in the number of users.
For many applications, the SISO MUD is too complex.
Traditional turbo-MUD for CDMA systems.
However, there are many applications where the SISO MUD is suitable.
Turbo-MUD for TDMA, hybrid CDMA/TDMA, WCDMA
SISO MUD can be used to achieve distributed detection.
Future work.
Comparison against suboptimal approaches.
Other applications of SISO MUD algorithms.
Conclusion