1. Institute for Infocomm Research (I2R)
Results Updates
Institute for Infocomm Research (I2R)
Sumei SUN, WU Yan, LI Yuan,
Patrick FUNG, Ying-Chang LIANG,
Zhongding LEI
( to:
16 November 2004

2. Outline CDD versus SM and GSTBC Unequal Error Protection (UEP) in LDPC
November 2004
Outline
CDD versus SM and GSTBC
Unequal Error Protection (UEP) in LDPC
Long Preamble Design
Beamforming and FEC
Institute for Infocomm Research (I2R)

3. Outline CDD versus SM and GSTBC Unequal Error Protection (UEP) in LDPC
November 2004
Outline
CDD versus SM and GSTBC
Unequal Error Protection (UEP) in LDPC
Long Preamble Design
Beamforming and FEC
Institute for Infocomm Research (I2R)

4. November 2004
PER Comparison of 2-rx MRC, CDD and STBC, Conv Code (chan B perfect CE and FOE ) -5 5 10 15 20 25 30 -4 -3 -2 -1
SNR per receive antenna (dB)
PER
0.65dB
0.71dB
1.4dB 1
2 SIMO 2
2 CDD delay=32
2 STBC
108 Mbps
48 Mbps
24 Mbps
Diversity order of 2-rx MRC
Diversity order of STBC
Diversity order of CDD
STBC gives the lower bound of CDD!
Institute for Infocomm Research (I2R)

5. November 2004
PER Comparison of 2x2 SM, 4x2 CDD and 4x2 GSTBC, Conv Code (chan B perfect CE and FOE ) 5 10 15 20 25 30 35 40 45 -3 -2 -1
coded PER
SNR per receive antenna
1.6dB
2.5dB
5.1dB 2
24Mbps SM
48Mbps SM
108Mbps SM
24Mbps GSTBC
48Mbps GSTBC
108Mbps GSTBC
24Mbps CDD
48Mbps CDD
108Mbps CDD
CDD outperforms SM, but gap becomes smaller for higher order modulation;
GSTBC outperforms CDD; and gain becomes higher for higher order modulation.
Institute for Infocomm Research (I2R)

6. November 2004
PER Comparison of 2x2 SM, 4x2 CDD and 4x2 GSTBC, Conv Code (chan E perfect CE and FOE ) 5 10 15 20 25 30 35 40 45 -3 -2 -1
40MHz BW,
1000B packet, Channel Model E
coded PER
SNR per receive antenna
2.4dB
3.2dB
6dB 2
24Mbps SM
48Mbps SM
108Mbps SM
24Mbps GSTBC
48Mbps GSTBC
108Mbps GSTBC
24Mbps CDD
48Mbps CDD
108Mbps CDD
CDD has same performance as SM at target PER;
GSTBC outperforms CDD; the higher the modulation, the more the performance gains.
Institute for Infocomm Research (I2R)

7. Effect of CE and FOE on 4x2 CDD, Conv Code (chan B )
November 2004
Effect of CE and FOE on 4x2 CDD, Conv Code (chan B ) 5 10 15 20 25 30 35 -3 -2 -1
Coded PER of 4
2 CDD with CE, FOE and PC
SNR per receive antenna (dB)
PER
2.1dB
1.4dB
0.95dB
With CE, FOE and PC
Perfect CE and Syn 2
108 Mbps
48 Mbps
24 Mbps
Recall that performance degradation for GSTBC is 2.54, 1.82, and 1.67 dB, respectively for the three setup.
Institute for Infocomm Research (I2R)

8. November 2004
PER Comparison of 2x2 SM, 4x2 CDD and 4x2 GSTBC, Conv Code (chan B with practical CE and FOE ) 5 10 15 20 25 30 35 -3 -2 -1
SNR per receive antenna (dB)
PER
0.98dB
2.1dB
4.8dB 2
24Mbps VBLAST
48Mbps VBLAST
108Mbps VBLAST
24Mbps GSTBC
48Mbps GSTBC
108Mbps GSTBC
24Mbps CDD
48Mbps CDD
108Mbps CDD
40MHz BW,
1000B packet, Channel Model B Convolutional coded with CE, FOE and PC
GSTBC still shows much better performance than CDD and SM.
Institute for Infocomm Research (I2R)

9. November 2004
PER Comparison of 2x2 SM, 4x2 CDD and 4x2 GSTBC, LDPC (chan B with perfect CE and FOE ) 5 10 15 20 25 30 35 -4 -3 -2 -1
40 MHz BW 4
2 two group CDD Channel B
DC11 LDPC coded perfect CE FOE delay=32
SNR per receive antenna (dB)
PER
0.99dB
1.1dB
2.6dB 2
24Mbps VBLAST
48Mbps VBLAST
108Mbps VBLAST
24Mbps GSTBC
48Mbps GSTBC
108Mbps GSTBC
24Mbps CDD
48Mbps CDD
108Mbps CDD
A more powerful code can better exploit the advantage of CDD.
Institute for Infocomm Research (I2R)

10. November 2004
PER Comparison of 2x2 SM, 4x2 CDD and 4x2 GSTBC, LDPC (chan B with practical CE and FOE ) 5 10 15 20 25 30 35 -4 -3 -2 -1
40 MHz BW 4
2 two group CDD Channel B DC11 LDPC coded with CE FOE delay=32
SNR per receive antenna (dB)
per
0.24dB
0.94dB
1.7dB 2
24Mbps VBLAST
48Mbps VBLAST
108Mbps VBLAST
24Mbps GSTBC
48Mbps GSTBC
108Mbps GSTBC
24Mbps CDD
48Mbps CDD
108Mbps CDD
CDD is less sensitive to channel estimation and synchronization errors than GSTBC.
Institute for Infocomm Research (I2R)

11. November 2004
PER Comparison of 2x2 SM, 4x2 CDD and 4x2 GSTBC, LDPC (chan E with practical CE and FOE ) 5 10 15 20 25 30 -4 -3 -2 -1
40 MHz BW 4
2 two group CDD Channel E DC11 LDPC coded with CE FOE delay=32
SNR per receive antenna (dB)
per
0.97dB
1.6dB
2.6dB 2
24Mbps VBLAST
48Mbps VBLAST
108Mbps VBLAST
24Mbps GSTBC
48Mbps GSTBC
108Mbps GSTBC
24Mbps CDD
48Mbps CDD
108Mbps CDD
The diversity gain from CDD can not be very well exploited.
Institute for Infocomm Research (I2R)

12. Diversity Order of 2x2 SM, 4x2 CDD, and 4x2 GSTBC, Linear Detection
November 2004
Diversity Order of 2x2 SM, 4x2 CDD, and 4x2 GSTBC, Linear Detection
Assume linear detection : i.e. no diversity gain from detection.
SM: dsysmin(dpath, dfree);
CDD : dsys  min(2×dpath, dfree);
GSTBC: dsys  2×min(dpath, dfree);
dSM  dCDD  dGSTBC
Institute for Infocomm Research (I2R)

13. Further Work CDD + SS; CDD + SS + frequency domain transform.
November 2004
Further Work
CDD + SS;
CDD + SS + frequency domain transform.
Institute for Infocomm Research (I2R)

14. Outline CDD versus SM versus GSTBC
November 2004
Outline
CDD versus SM versus GSTBC
Unequal Error Protection (UEP) in LDPC
Long Preamble Design
Beamforming and FEC
Institute for Infocomm Research (I2R)

15. The Scheme Conventional scheme—direct mapping
November 2004
The Scheme
Conventional scheme—direct mapping
Pre-designed UEP permutation embedded in LDPC encoder
Systematic bits in LDPC require more protection than parity check bits
Different from bit-interleaver: no latency, no complexity
Write coded sequence (systematic + parity) into a rectangular block of put systematic bits into MSB positions.
Institute for Infocomm Research (I2R)

16. Simulation Setup for LDPC Coded QAM with 2x2 SM
November 2004
Simulation Setup for LDPC Coded QAM with 2x2 SM
PD-LDPC (2304, 1152) of R=1/2;
UEP vs Direct mapping
H1 designed using progressive edge growth (PEG);
Three kinds of H1 column weights: 4, 6, 8.
64QAM with Gray mapping;
Single encoder for two spatial streams in 2x2 SM
Practical Receiver
FOE, PC, CE and SNR estimation
Channel model B and E
Institute for Infocomm Research (I2R)

17. 2x2 SM in Channel B UEP vs Direct mapping Weight, w Gain,dB 4 0.28 6
November 2004
2x2 SM in Channel B
UEP vs Direct mapping
Weight, w
Gain,dB 4
0.28 6
0.46 8
1.2
w=6 vs w=4
UEP: gain=0.18dB
Direct: gain=0dB
PD-LDPC of w=8 does not introduce performance gain at PER=10-2, although it shows higher diversity order.
Institute for Infocomm Research (I2R)

18. 2x2 SM in Channel E UEP vs Direct mapping Weight, w Gain,dB 4 0.29 6
November 2004
2x2 SM in Channel E
UEP vs Direct mapping
Weight, w
Gain,dB 4
0.29 6
0.67 8
1.6
Gain from w=6 to w=4
UEP: gain=0.15dB
Direct: gain=-0.25dB
PD-LDPC of w=8 does not perform better than w=4 and 6 at PER=10-2.
Institute for Infocomm Research (I2R)

19. UEP permutation is superior to direct mapping
November 2004
Summary
UEP permutation is superior to direct mapping
Performance gain at target PER
gain increases with column weight of PD-LDPC
At best choice of w=6, gains are 0.46dB and 0.76dB for Channel Model B and E
Performance gain achieved at no extra cost
not extra redundancy;
no extra latency (embedded in LDPC encoder);
no extra encoder/decoder complexity (pre-designed).
Institute for Infocomm Research (I2R)

20. Outline CDD versus SM versus GSTBC
November 2004
Outline
CDD versus SM versus GSTBC
Unequal Error Protection (UEP) in LDPC
Long Preamble Design
Beamforming and FEC
Institute for Infocomm Research (I2R)

21. Integration of Long Preamble (1/3)
November 2004
Integration of Long Preamble (1/3)
Ready integration of I2R long preamble design into HT-LTF structure to achieve
overhead savings;
improved residual frequency estimate
in all modes
Identical channel estimation performance +
more accurate residual freq offset estimate GI
s a v i n g
0.8s
13.6s
One single GI per transmit antenna
Institute for Infocomm Research (I2R)

22. Integration of Long Preamble (2/3)
November 2004
Integration of Long Preamble (2/3)
More reduction with more transmit antennas GI
s a v i n g
2.4s
Identical channel estimation performance +
more accurate residual freq offset estimate
26.4s
Institute for Infocomm Research (I2R)

23. Integration of Long Preamble (3/3)
November 2004
Integration of Long Preamble (3/3)
13.6s
HT-DATA GI
13.6s
Institute for Infocomm Research (I2R)

24. Outline CDD versus SM versus GSTBC
November 2004
Outline
CDD versus SM versus GSTBC
Unequal Error Protection (UEP) in LDPC
Long Preamble Design
Beamforming and FEC
Institute for Infocomm Research (I2R)

25. A Simpler Beamforming Scheme
November 2004
A Simpler Beamforming Scheme
Common Beamforming I2R Scheme
Instantaneous
Dedicated PPDU
Frequency domain (per tone)
Fast (dependant on the channel fading frequency)
CSI
CSI estimation
Beamforming
Update frequency
Channel covariance matrix
(2nd order statistics)
No PPDU
(beamforming is transparent to the RX which needn’t know if the TX is in beamforming mode)
Time domain (per user w/ much less complexity)
Slow (subject to position or DOA change frequency)
Institute for Infocomm Research (I2R)

26. November 2004
Further Studies
For adaptive modulation and coding in SVD beamforming, we propose to compare the following:
use separate encoders for different MIMO streams;
use a single encoder, but transmit different codeword for different spatial streams;
use a single encoder and single codeword for all the spatial streams, but optimize the puncturing pattern and bit assignment to different spatial streams.
Institute for Infocomm Research (I2R)