1. Submission Title: [FEC & Modulation Options and considerations]
Sept 2005
Project: IEEE P Working Group for Wireless Personal Area Networks (WPANs)
Submission Title: [FEC & Modulation Options and considerations]
Date Submitted: [5 Oct 2005]
Source: [Francois Chin, Sam Kwok, Sai-Ho Wong]
Company: [Institute for Infocomm Research, Singapore]
Address: [21 Heng Mui Keng Terrace, Singapore ]
Voice: [ ]
Abstract: []
Purpose: [Assist the group in the selection of a modulation scheme]
Notice: This document has been prepared to assist the IEEE P It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein.
Release: The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P
Francois Chin (I2R)

2. Outline Baseline What we propose : FEC 7 Modulation to go with FEC 7:
Sept 2005
Outline
Baseline
What we propose : FEC 7
Modulation to go with FEC 7:
8-PPM + BPSK
Complexity scalability consideration
Francois Chin (I2R)

3. Baseline - Modulation(from 582r0)
Sept 2005
Baseline - Modulation(from 582r0)
Francois Chin (I2R)

4. Systematic Convolutional Encoder
Sept 2005
FEC Options (from 602r0)
SOC code
K= 3,4 or 5
R = 1/4
Coherent Receiver: True Rate = ¼
Non Coherent Receiver: Equivalent to Rate = ½
(Rate ¼ with erasures)
FEC 1
Convolutional Encoder
K= 3,4 or 5
R= 1/4
Coherent Receiver: True Rate = ¼
Non Coherent Receiver: Equivalent to Rate = ½
(Rate ¼ with erasures)
FEC 2
Systematic
Convolutional
Encoder
K= 3,4 or 5
R = 1/2
Convolutional
Encoder
K=3, R= 1/2
FEC 3
Coherent Receiver: Concatenated code, Rate = ¼
Non Coherent Receiver: Convolutional code, Rate = ½
Systematic
Convolutional
Encoder
K= 3,4 or 5
R = 1/2
BCH or RS
GF(28): RS(40,32)
GF(26): RS(53,43)
Coherent Receiver: Concatenated code Rate = 0.4
Non Coherent Receiver: RS code, Rate = 0.8
FEC 4
Systematic Convolutional Encoder
K= 3,4 or 5
R= 1/2
FEC 5
Coherent Receiver: Convolutional code Rate = ½
Non Coherent Receiver: Uncoded
Convolutional Encoder
K= 3,4 or 5
R= 1/3
FEC 6
Coherent Receiver: Convolutional code Rate = 1/3
Non Coherent Receiver: Convolutional code Rate = 1/2
Coherent Receiver: True Rate = ½
Non Coherent Receiver: Equivalent to Rate = 2/3
(Rate ½ with puncturing)
Systematic Convolutional Encoder
K= 3,4 or 5
R= 1/2
FEC 7
Francois Chin (I2R)

5. Summary of rates (modified from 602r0)
Sept 2005
Summary of rates (modified from 602r0)
Same ave PRF as the preamble section
Francois Chin (I2R)

6. doc.: IEEE 802.15-<doc#>
<month year>
doc.: IEEE <doc#>
Sept 2005
FEC 6: Burst Pulses
S = = S
1 chip ~ 2 ns
burst duration = TB
= 15 chips ~ 30 ns
symbol duration ~ 1.0us = 480 chips = 32 TB
Proposed peak PRF of 494 MHz, with S code of length 15
S code duration = 15 / peak PRF ~ 30 ns
Another candidate, S = (any 15-chip M-seq)
Chip duration is 1/peak PRF
Francois Chin (I2R)
<author>, <company>

7. doc.: IEEE 802.15-<doc#>
<month year>
doc.: IEEE <doc#>
Sept 2005
Proposal 6 : 4-PPM + BPSK
Guard time for channel delay spread (120ns) C NC
000
00 -
001
010
01 -
011
110
11 -
111
100
10 -
101 S -S 15 16 31 3 8 19 20 4 7 12 11 23 24 27 28
Francois Chin (I2R)
<author>, <company>

8. doc.: IEEE 802.15-<doc#>
<month year>
doc.: IEEE <doc#>
Sept 2005
FEC 7: Codec & Modem
s1 s2
8-PPM demod
Common
Coder + Modulation Mapping
Viterbi
½ rate
r1 r2
Systematic Bit
s1 s2
8-PPM + BPSK modulatorwith input [s1 s2 r1 r2]
Non Coherent Receiver
Info. Bit
Convolutional
1/2
a1 a2
s1 s2
Redundant Bit
8-PPM + BPSK demod
Viterbi
½ rate
r1 r2
r1 r2
Coherent Receiver
Francois Chin (I2R)
<author>, <company>

9. Modulation : 8-PPM + BPSK
Guard time (120 ns)
Sept 2005
Scrambling Range (120 ns)
Coh
S1s2r1r2
N- Coh s1s2r1 -
0000
000-
0001
0010
001-
0011
0110
011-
0111
0100
010-
0101
1100
110-
1101
1110
111-
1111
1010
101-
1011
1000
100-
1001
Modulation : 8-PPM + BPSK 1 3 2 5 4 7 6 9 8 11 10 13 12 15 14 17 16 19 18 21 20 23 22 25 24 27 26 29 28 31 30 S -S
Gray coded
Francois Chin (I2R)

10. doc.: IEEE 802.15-<doc#>
<month year>
doc.: IEEE <doc#>
Sept 2005
FEC 7: Burst Pulses
1 chip ~ 2 ns
burst duration = TB
= 31 chips ~ 63 ns
symbol duration ~ 2.0us = 992 chips = 32 TB
Proposed peak PRF of 494 MHz, with S code of length 31
S code duration = 31 / peak PRF ~ 63 ns
S can be binary version of the BPTS (meant for preambles)
Via Ternary to Binary: +/- → 1; 0 → -1
Chip duration is 1/peak PRF
Francois Chin (I2R)
<author>, <company>

11. So what we propose in FEC 7 is:
Sept 2005
So what we propose in FEC 7 is:
Coherent
Non-Coherent Tx
(Common for Coherent and Non Coherent)
Symbol Time
~2µs
PHY-SAP
~1 Mbps
FEC 7
Convolutional 1/2
Modulation
8-PPM + Polarity
(4 bits per symbol)
Channel
Raw Rate
2 Mbps Rx
Demodulation
8PPM + Polarity
(4 bits/symbol)
8PPM only
(3 bits/symbol)
Decoder
Viterbi 1/2
~1Mbps
 Perfect compatibility between Tx and coherent and
non-coherent Rx
Francois Chin (I2R)

12. Features of FEC 7 Options 7: Longer burst S
Sept 2005
Features of FEC 7
Options 7:
Longer burst S
better inter pulse interference suppression for coherent receiver
Can have specific burst sequence for different piconet, thus further SOP interference suppression with set of nearly orthogonal sequences for the burst
Large burst energy gives better SNR in non-coherent receiver
8-PPM > 4-PPM demod > 2-PPM, the former has better uncoded BER performance
Base rate of codec is ½
Lower codec complexity to meet low cost / power requirement
Francois Chin (I2R)

13. Piconet-Specific Burst Sequences
Sept 2005
Piconet-Specific Burst Sequences
PBTS for Preambles S1 S2 S3 S4 S5 S6
Burst Sequences S1 S2 S3 S4 S5 S6
Burst Sequence can be derived from Preamble sequences
Via Ternary to Binary: +/- → 1; 0 → -1
Francois Chin (I2R)

14. RX Complexity Scalability
Sept 2005
RX Complexity Scalability
All options allow signal detection in low-cost low-power implementation without viterbi decoder, especially in Non-coherent receiver
For such non-coherent RX, FEC 1~4 with 2PPM+BPSK requires 2PPM demod on every alternate ~0.5us symbol to detect systematic bit, 3dB loss of energy as other burst (or symbol) carrying redundant bit is not used for detection
FEC 5 with 2PPM+BPSK requires 2PPM demod on every symbol to detect systematic bit
Francois Chin (I2R)

15. RX Complexity Scalability
<month year>
doc.: IEEE <doc#>
Sept 2005
RX Complexity Scalability
FEC 6 with 4PPM+BPSK requires 2PPM demod on every ~1us symbol to detect 1 systematic bit
~1us S -S 31 32 63 7 16 39 40 8 15 24 23 47 48 56 57
detect information bit ‘0’
detect information bit ‘1’
Francois Chin (I2R)
<author>, <company>

16. RX Complexity Scalability
<month year>
doc.: IEEE <doc#>
Sept 2005
RX Complexity Scalability
FEC 7 with 8PPM+BPSK requires 4PPM demod on every ~2us symbol to detect 2 systematic bit
~2us 1 3 2 5 4 7 6 9 8 11 10 13 12 15 14 17 16 19 18 21 20 23 22 25 24 27 26 29 28 31 30 S -S
detect ‘00’
detect ‘01’
detect ‘11’
detect ‘10’
Francois Chin (I2R)
<author>, <company>

17. RX Complexity Scalability
Sept 2005
RX Complexity Scalability
given low-cost low-power receiver implementation (uncoded bit detection only, either without or turning off, viterbi decoder) is an attractive feature,
2PPM+BPSK modulation option is not energy efficient, as only alternate burst (or symbol) of energy is used for detection
Using same energy integration / capture interval for uncoded bit detection in non-coherent rxr, 8PPM + BPSK option using 4PPM demod (on every ~2us symbol to detect 2 systematic bit) will capture less noise than 4PPM + BPSK option using 2PPM demod (on every ~1us symbol to detect 1 systematic bit)
Francois Chin (I2R)

18. Statistics of Collected Burst Energy
Sept 2005
Statistics of Collected Burst Energy
Let’s look at the signal+noise and noise-only statistics for 2-, 4- and 8-PPM, with symbol period of 0.5us, 1us and 2us respectively and 8, 16 & 32 pulses / symbols, respectively.
Francois Chin (I2R)

19. Statistics of Collected Burst Energy
Sept 2005
Statistics of Collected Burst Energy
2-PPM case
symbol period = 0.5us
8 pulses / symbol
Pulse SNR = 0dB
Noise-only slot
Mean = 8
Std-Dev = √8
Signal+Noise slot
Mean = 16
Std-Dev = √16*√3
Francois Chin (I2R)

20. Statistics of Collected Burst Energy
Sept 2005
Statistics of Collected Burst Energy
4-PPM case
symbol period = 1us
16 pulses / symbol
Pulse SNR = 0dB
Noise-only slot
Mean = 16
Std-Dev = √16
Signal+Noise slot
Mean = 32
Std-Dev = √32*√3
Francois Chin (I2R)

21. Statistics of Collected Burst Energy
Sept 2005
Statistics of Collected Burst Energy
8-PPM case
symbol period = 2us
32 pulses / symbol
Pulse SNR = 0dB
Noise-only slot
Mean = 32
Std-Dev = √32
Signal+Noise slot
Mean = 64
Std-Dev = √64*√3
Francois Chin (I2R)

22. Statistics of Collected Burst Energy
Sept 2005
Statistics of Collected Burst Energy
Generally, for N pulses / symbol with Pulse SNR = 
Noise-only slot
Mean = N, Std-Dev = √N
Signal+Noise slot
Mean = ( +1)N, Std-Dev = √((2 +1)N)
For more pulses/symbols, Std-Dev / Mean decreases, means the statistics are more confined to the mean value
Francois Chin (I2R)

23. Symbol Error Performance
Sept 2005
Symbol Error Performance
Depends on the extense of overlaps of the Noise-only and signal+noise p.d.f. curves
More overlaps means there is higher probability that the noise-only slot to exceed the signal+noise slot, therefore, higher SER
Less overlaps for higher order PPM means lower SER
Francois Chin (I2R)

24. PPM Demod BER (Non-Coherent in AWGN)
Sept 2005
PPM Demod BER (Non-Coherent in AWGN)
2- & 4- PPM receivers
Will further be passed into rate ½ Viterbi decoder
8-PPM receivers
De-punctured and further passed into rate ½ Viterbi decoder
Francois Chin (I2R)

25. Sept 2005
AWGN Uncoded BER at PPM Demod output (Non-Coherent w/o Viterbi Decoder)
Francois Chin (I2R)