1. January 19
April 2009
Project: IEEE P Working Group for Wireless Personal Area Networks (WPANs)
Submission Title: Multi-Rate PHY Proposal for the TG4g PHY Amendment
Date Submitted: March 2009
Source: Michael Schmidt, Dietmar Eggert, Frank Poegel, Torsten Bacher, Sascha Beyer, Atmel
Contact: Michael Schmidt, Atmel
Voice: ,
Re: TG4g Call for proposals
Abstract: PHY enhancements towards TG4g supporting multiple data rates
Purpose: PHY proposal for the TG4g PHY amendment
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
Slide 1
Michael Schmidt
Page 1

2. Overview DSSS versus Narrow-Band Coding and Modulation Preamble Design
January 19
April 2009
Overview
DSSS versus Narrow-Band
Coding and Modulation
Preamble Design
Europe
Slide 2
Michael Schmidt
Page 2

3. DSSS versus Narrow Band
April 2009
DSSS versus Narrow Band
Myth: DSSS obtains more range due to processing gain.
In principal, there is no clear advantage of either one. The range is mainly determined by the data rate itself.
DSSS with a spreading factor k introduces k-times more noise due to bandwidth extension (-174 dBm/Hz) which is again compensated by the processing gain.
DSSS may yield an additional coding gain but it is usually far away from the Shannon limit. Current powerful capacitiy achieving codes (LDPC, Turbo Codes) are usually ½ rate codes.
Slide 3
Michael Schmidt

4. DSSS versus Narrow Band
April 2009
DSSS versus Narrow Band
Myth: DSSS outperforms NB in case of Multipath
Current IEEE O-QPSK PHYs apply C(16,4) coding and C(32,4) coding, leading to a fairly moderate spreading factor k of 4 and 8, respectively.
A Rake receiver is not applicable for small k. (It is well known that DSSS with a Rake receiver approaches the Matched Filter Bound, assuming long and good PN DSSS sequences.)
For DSSS-(G)MSK, detection suffers from additional distortions due to inter-chip interference, since it operates at a k-fold chip rate according to the spreading factor k. This may not be completely compensated by the processing gain if k is small or if M-ary code sets are applied.
Slide 4
Michael Schmidt

5. DSSS versus Narrow Band
April 2009
DSSS versus Narrow Band
Cont. ~
Myth: DSSS outperforms NB in case of Multipath
For DSSS with a spreading factor k >= 4, chip based equalization is not really applicable, since the SNR at the chip level is decreased by k, leading to unreliable estimates.
chip based equalization is only useful for very mild spreading (1/2 rate coding) or no spreading at all.
Example: PHY according to the GSM standard
applies GMSK-BT=0.3 with ½ rate convolutional coding
trellis based SISO equalization is supported:
trainings sequences for channel estimation
trailing zeros
Slide 5
Michael Schmidt

6. DSSS versus Narrow Band
April 2009
DSSS versus Narrow Band
Cont. ~ Myth: DSSS outperforms NB in case of Multipath
When applying (G)MSK-DSSS:
seek for improved code sets compared to C(32,4) and C(16,4) of IEEE
consider traditional large spreading gains (64 … 128) based on PN sequences. This, however, may lead to low data rates, conflicting with a target PSDU length of 1500 octets.
Slide 6
Michael Schmidt

7. DSSS versus Narrow Band
April 2009
DSSS versus Narrow Band
Targeting low data rates
NB:
Low data rates are obtained by a low chip rate equal to the bit rate.
This requires appropriate narrow band filtering, which is challenging for both low-IF and ZIF receivers.
Requires low IF in order to avoid a high pole-Q
Reduced IF increases 1/f noise
More complicated DC removal
Requires advanced AD conversion and base band filtering when targeting a high IF/BW ratio
Slide 7
Michael Schmidt

8. DSSS versus Narrow Band
April 2009
DSSS versus Narrow Band
Cont. ~ Targeting low data rates
DSSS:
Low data rates can be obtained by appropriate spreading at a certain fixed chip rate.
Different data rates can be obtained by simply changing the spreading gain.
Bandwidth of the receive filter can be kept constant with a fixed IF
Simplifies design of the analog receiver front end
Slide 8
Michael Schmidt

9. DSSS versus Narrow Band
April 2009
DSSS versus Narrow Band
Cont.~ Targeting low data rates
Slide 9
Michael Schmidt

10. DSSS versus Narrow Band
April 2009
DSSS versus Narrow Band
Impairments
Narrow band interferer (e.g. distortions due to CW signal )
Clock offset
Interference Avoidance
NB is more flexible when FH or AFA is applied
Slide 10
Michael Schmidt

11. DSSS versus Narrow Band
April 2009
DSSS versus Narrow Band
The choice of DSSS or NB is influenced by Regularity Requirements
FCC CFR 47:
For a BW < 500 kHz, FH is mandatory
BW >= 500 kHz avoids the need for hopping but reduces the number of hopping channels for AFA
ETSI, ERC
Output power:
6.2 dBm/100 kHz for restricted wideband operation within MHz
DSSS appears to be useful here
Operation within MHz is too limited for DSSS
Pure GFSK modulation seems to be useful as suggested in [3]
Out of Band Power: -36 dBm/100 kHz
Slide 11
Michael Schmidt

12. April 2009
Coding and Modulation
Slide 12
Michael Schmidt

13. April 2009
Coding
Current IEEE DSSS codes C(32,4) and C(16,4) obtain certain constrains which are obsolete when applying a chip scrambler
The preamble should be independently designed (must not be necessarily assembled with code words)
Slide 13
Michael Schmidt

14. Coding Options for improved DSSS block codes C(64,4) C(32,4) C(16,4):
April 2009
Coding
Options for improved DSSS block codes C(64,4) C(32,4) C(16,4):
sub codes of near extended BCH codes
sub codes of Reed Muller Codes
Improve existing IEEE codes by considering MSK modulation rather than O-QPSK and remove constrains
C(8,4) coding seems to have little potential for an additional coding gain (code length is too short).
Either skip ½ rate coding or apply ½ convolutional coding
Consider DSSS codes with a large spreading gain, e.g. C(64,1). This, however, will have a severe impact on the preamble design. (Preamble detection must be able to cope with this kind of spreading gain.)
Slide 14
Michael Schmidt

15. GMSK Why GMSK? Constant envelope signal
April 2009
GMSK
Why GMSK?
Constant envelope signal
Permits direct modulation transmitter design
GMSK considerably reduces side lopes compared to MSK
Reduced interference
MSK hardly applicable for the EU band
GMSK (GFSK with modulation index h = 0.5) is the preferred choice but a modulation index h < 0.5 may be unavoidable for the EU band MHz
GMSK is detectable for IEEE legacy devices using O-QPSK with appropriate coding
Slide 15
Michael Schmidt

16. April 2009
GMSK
Known Issues:
Transmitter design is more complicated compared to MSK
Accurate I/Q based GMSK shaping can be implemented within the base band but a mixer is required.
Avoid mixed architecture by direct modulation. This, however, might be subject to patent issues.
Receiver may require Trellis based detectors
depends on the choice of BT
less relevant for DSSS-GMSK
less relevant if channel equalization is considered anyway (see, e.g. GSM PHY)
Slide 16
Michael Schmidt

17. Preamble Design The length of the preamble should be related to
April 2009
Preamble Design
The length of the preamble should be related to
the maximum frame length
the target of the receiver sensitivity
the maximum clock offset
PHY
Data rate
[kbps]
Preamble length Tp [ms]
Maximum payload time Td [ms]
R = Tp/Td
DSSS-BPSK
IEEE 20
1.6
50.8
(for 127 octets)
0.0315
DSSS-O-QPSK
IEEE
100
0.32
10.16
NB or DSSS (G)MSK
2.4 … 3.6
120
(for 1500 octets)
0.02 … 0.03
Slide 17
Michael Schmidt

18. April 2009
Preamble Design
Good AKF properties for timing synchronisation and clock offset estimation
Channel estimation (useful for equalization when applying ½ rate coding or no coding)
Dissolve IQ images during preamble detection if IF is symmetric w. r. t. channel spacing.
Slide 18
Michael Schmidt

19. Preamble Design Cont.~ Dissolving IQ Images Problem:
April 2009
Preamble Design
Cont.~ Dissolving IQ Images
Problem:
Avoid false preamble detection on IQ images, since the transceiver is usually occupied with SFD search during this time period.
Slide 19
Michael Schmidt

20. Preamble Design IQ imbalance: Since is approximately left-analytic
April 2009
Preamble Design
IQ imbalance:
Since
is approximately left-analytic
Slide 20
Michael Schmidt

21. Preamble Design Cont.~ Dissolving IQ Images
April 2009
Preamble Design
Cont.~ Dissolving IQ Images
A conjugate signal can be differentiated from its non-conjugate in case of frequency modulation
This, however, is not possible with a trivial {… …} preamble sequence, since the time reference is unknown during preamble detection.
Slide 21
Michael Schmidt

22. Preamble Design Possible candidates:
April 2009
Preamble Design
Possible candidates:
multiple (4…8) extended M-sequences of length:
32 for NB-GFSK
64 or 128 for DSSS-GFSK
Slide 22
Michael Schmidt

23. European Frequency Band (863 – 870 MHz)
April 2009
European Frequency Band (863 – 870 MHz)
Harmonized standard ETSI EN defines frequency band of 863 – 870 MHz with different power limits for digital modulation systems using DSSS
Duty cycle limitations do not apply to systems using Listen Before Talk techniques combined with Adaptive Frequency Agility (AFA)
Slide 23
Michael Schmidt

24. ERC Recommendation 70-03 Regulatory parameters for 863 – 870 MHz [1]
January 19
April 2009
ERC Recommendation 70-03
Regulatory parameters for 863 – 870 MHz [1]
Slide 24
Michael Schmidt
Page 24

25. ERC Recommendation 70-03 Regulatory parameters for 863 – 870 MHz [1]
January 19
April 2009
ERC Recommendation 70-03
Regulatory parameters for 863 – 870 MHz [1]
Slide 25
Michael Schmidt
Page 25

26. EU DSSS-Proposal April 2009 DSSS-GMSK BT = 0.3…0.5
Chip rate: 400 kchip/s
Spreading factors: 1, 2, 4, 8,16, 32, 64
Data rates: 400, 200, 100, 50, 25, 12.5, 6.25 kbps
Ch# 1 2 3 4 5 6 7 8 9 10
Frequency
[MHz]
863.5
864.1
864.7
865.6
866.2
866.8
867.4
868.3
868.9
869.5
6.2 dBm/100 kHz when restricting to channels {4, 5, 6, 7}
0.8 dBm/100kHz when restricting to channels {4, 5, 6, 7, 8, 9, 10}
-4.5 dBm/100kHz when applying all channels
LBT with AFA
Michael Schmidt

27. EU DSSS-Proposal April 2009 Spectral properties: GMSK
Chip rate: 400 kchip/s
BT = 0.3
Michael Schmidt

28. April 2009
EU NB-Proposal
Utilize band – MHz, similar as suggested in [3]
GFSK, modulation index < 0.5 possibly unavoidable
Chip rate: 100 kchip/s
Data rates: 100 kbps or 50 kbps (preferably by ½ rate convolutional coding)
Channel spacing >= 250 kHz
Output power 14 dBm and possibly 27 dBm for a channel with MHz, however:
Complex PA design
out of band limit of -36 dBm /100 kHz is quite challenging
LBT with AFA
Michael Schmidt

29. April 2009
Bibliography
[1] ERC RECOMMENDATION (Tromsø 1997 and subsequent amendments); RELATING TO THE USE OF SHORT RANGE DEVICES (SRD); Recommendation adopted by the Frequency Management, Regulatory Affairs and Spectrum Engineering Working Groups,Version of 18 February 2009.
[2] Draft ETSI EN V2.2.1 ( ) Electromagnetic compatibility and Radio spectrum Matters (ERM); Short Range Devices (SRD); Radio equipment to be used in the 25 MHz to MHz frequency range with power levels ranging up to 500 mW; Part 1: Technical characteristics and test methods.
[3] Khanh Tuan Le, “Preliminary Proposal for a Multi-Regional Sub-GHz PHY for g”, 802 Plenary Meeting 12th March, 2009 Vancouver BC.
Michael Schmidt