Sunteți pe pagina 1din 15

July 2004 doc.: IEEE 802.

11-04/682r0

LDPC versus Convolutional Codes


in MIMO-OFDM over 11n channels
Huaning Niu and Chiu Ngo
Samsung Electronics

Submission Slide 1 Huaning Niu, Chiu Ngo, Samsung Electronics


July 2004 doc.: IEEE 802.11-04/682r0

Outline
• Background and motivation
• System model
• Simulation evaluation
– Simulation methodology
– Simulation results
• Conclusions

Submission Slide 2 Huaning Niu, Chiu Ngo, Samsung Electronics


July 2004 doc.: IEEE 802.11-04/682r0

Background and Motivation


• LDPC codes for 802.11n [1-4]
– First proposed in [1]
– Over SISO link: LDPC codes have 2-4 dB gain over CC6
(Convolutional code memory 6) in AWGN. The gain decreases by
1-2 dB in fading model D (50 ns RMS) due to the limited time
diversity [2,3]
– Over uncorrelated MIMO link: LDPC codes have similar
performance gain over CC6 in both AWGN and spatial
uncorrelated fading channel [4]
• Motivation: What is the LDPC gain over CC6 over
correlated 11n channels where frequency diversity and
spatial diversity are present

Submission Slide 3 Huaning Niu, Chiu Ngo, Samsung Electronics


July 2004 doc.: IEEE 802.11-04/682r0

System Model
Add Remove
IFFT FFT
CP CP
Add
IFFT Remove
CP FFT MIMO Demod.
Coding H CP
MIMO Det Decoding
and
Mapper
Mod.

Add Remove
IFFT FFT
CP CP

• One code for all data streams


• Interleaver between coding and modulation is needed for
CC6 only
• Soft APP processing after MIMO detector
Submission Slide 4 Huaning Niu, Chiu Ngo, Samsung Electronics
July 2004 doc.: IEEE 802.11-04/682r0

Simulation Methodology
• PHY modes based on 11a spec
– QPSK ½ (mode 3), 16QAM ½ (mode 5) with packet size 600 Byte
– 16QAM ¾ (mode 6), 64 QAM ¾ (mode 8) with packet size 1500
Byte
• Channel models
– AWGN channel
– 802.11n channel models B and D with antenna spacing  at both
the transmitter and receiver [5]
• MIMO architecture
– 2x2 MIMO system
– ZF detection

Submission Slide 5 Huaning Niu, Chiu Ngo, Samsung Electronics


July 2004 doc.: IEEE 802.11-04/682r0

FEC parameters
• Convolutional codes
– Defined by 802.11a
– Soft Viterbi decoding
• LDPC codes
– Randomly generated regular LDPC codes with column
weight 3
– About 5000 information bits per coding block
– Rate ½ and ¾ are simulated
– Decoding algorithm using min-sum with a simple
correction term
– Maximum 100 iteration passes

Submission Slide 6 Huaning Niu, Chiu Ngo, Samsung Electronics


July 2004 doc.: IEEE 802.11-04/682r0

Performance over AWGN channels

BER performance comparison of PER performance comparison of


regular LDPC and CC6 over AWGN regular LDPC and CC6 over AWGN
channels channels

Submission Slide 7 Huaning Niu, Chiu Ngo, Samsung Electronics


July 2004 doc.: IEEE 802.11-04/682r0

Simulation Results over 11n: Mode 3

BER performance comparison of PER performance comparison of


regular LDPC and CC6 with QPSK, ½ regular LDPC and CC6 with QPSK, ½
coding with packet size 600 byte over coding with packet size 600 byte over
11n channels. 11n channels.
Submission Slide 8 Huaning Niu, Chiu Ngo, Samsung Electronics
July 2004 doc.: IEEE 802.11-04/682r0

Simulation Results over 11n: Mode 5

BER performance comparison of PER performance comparison of


regular LDPC and CC6 with 16QAM, regular LDPC and CC6 with 16QAM,
½ coding with packet size 600 byte over ½ coding with packet size 600 byte over
11n channels. 11n channels.
Submission Slide 9 Huaning Niu, Chiu Ngo, Samsung Electronics
July 2004 doc.: IEEE 802.11-04/682r0

Simulation Results over 11n: Mode 6

BER performance comparison of PER performance comparison of


regular LDPC and CC6 with 16QAM, regular LDPC and CC6 with 16QAM,
¾ coding with packet size 1500 byte ¾ coding with packet size 1500 byte
over 11n channels. over 11n channels.
Submission Slide 10 Huaning Niu, Chiu Ngo, Samsung Electronics
July 2004 doc.: IEEE 802.11-04/682r0

Simulation Results over 11n: Mode 8

BER performance comparison of PER performance comparison of


regular LDPC and CC6 with 64QAM, regular LDPC and CC6 with 64QAM,
¾ coding with packet size 1500 byte ¾ coding with packet size 1500 byte
over 11n channels. over 11n channels.
Submission Slide 11 Huaning Niu, Chiu Ngo, Samsung Electronics
July 2004 doc.: IEEE 802.11-04/682r0

Observations
• LDPC coding gain Rate Channel Gain
over CC6 is
significant in PER 1/2 AWGN 2-3 dB
performance 11n B 5-6 dB
• Coding gain is larger
11n D 7-8 dB
with higher code rate
3/4 AWGN 4-5 dB
• Coding gain is larger
with longer delay 11n B 6-7 dB
spread 11n D 8-9 dB
Submission Slide 12 Huaning Niu, Chiu Ngo, Samsung Electronics
July 2004 doc.: IEEE 802.11-04/682r0

Conclusions
• Performance comparison of regular LDPC codes
and convolutional codes in 11n channel models B
and D is presented.
• The indoor fading channel provides limited timing
diversity, and cause degraded coding gain for
LDPC codes in SISO link [2,3]
• LDPC codes can effectively utilize the spatial
diversity in MIMO link and the frequency
diversity
Submission Slide 13 Huaning Niu, Chiu Ngo, Samsung Electronics
July 2004 doc.: IEEE 802.11-04/682r0

References
[1]. Eric Jacobsen, "LDPC FEC for 802.11n application", IEEE 802.11-
03/0865r1, Intel Labs
[2]. Aleksandar Purkovic, Nina Burns, Sergey Sukobok, Levent Demirekler,
"LDPC vs. Convolutional Codes for 802.11n Applications: Performance
Comparison", IEEE 802.11-04/007lr1, Nortel Networks
[3]. Aleksandar Purkovic, Sergey Sukobok, Nina Burns, "LDPC vs.
Convolutional Codes: Performance and Complexity Comparison", IEEE
802.11-04/337, Nortel Networks, Mar. 2004
[4]. Ravi Mahadevappa, Stephan ten Brink, "Different Channel Coding Options
for MIMO-OFDM 802.11n”, IEEE 802.11-04-0014-00-000n, Realtek
Semiconductors
[5]. Laurent Schumacher and Bas Dijkstra, “Description of a MATLAB®
implementation of the Indoor MIMO WLAN channel model proposed by
the IEEE 802.11 TGn Channel Model Special Committee,” Implementation
note version 3.2 – May 2004

Submission Slide 14 Huaning Niu, Chiu Ngo, Samsung Electronics


July 2004 doc.: IEEE 802.11-04/682r0

Backup: LDPC decoding algorithm


dv 1

• Variable node: VC   C V i


i 0
 dc 1 
• Check node: CV  2 tanh  tanh( V j C / 2) 
1

 j 1 
• Min-Sum approximation with a simple correction
C  2 tanh 1  tanh( A / 2) tanh( B / 2) 
 1  exp( A  B ) 
 sgn( AB) min{| A |,| B |}  ln  
 1  exp(  A  B ) 
 0.5, A  B  1, A  B  1

 sgn( AB) min{| A |,| B |}  0.5, A  B  1, A  B  1
 0, else

Submission Slide 15 Huaning Niu, Chiu Ngo, Samsung Electronics

S-ar putea să vă placă și