1. Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs)
Submission Title: Reduced Duty Cycle Multiband OFDM
Date Submitted: March 3, 2003
Source: Eric Ojard and Jeyhan Karaoguz Company: Broadcom Corporation
Address: 190 Mathilda Place, Sunnyvale, CA 94086
Voice:
Re: [ a Call for proposal]
Abstract: Reduced duty cycle Multiband OFDM can dramatically improve performance in SOP environments, effectively solving the near-far problem.
Purpose: [TG3a-Broadcom-CFP-Presentation .]
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
Eric Ojard, Broadcom Corporation

2. Reduced Duty-Cycle Multiband OFDM
Eric Ojard
Eric Ojard, Broadcom Corporation

3. Introduction Multiband OFDM has some well-known advantages for UWB:
Collects all multipath energy without high complexity
Decouples equalization & coding w/o transmitter channel knowledge
Facilitates suppression of narrowband interference (egress & ingress)
It’s less well-known that it has the potential to solve the near-far problem with Simultaneously Operating Piconets (SOPs)
This presentation shows how Reduced Duty-Cycle MB-OFDM can achieve excellent performance in SOP environments.
Eric Ojard, Broadcom Corporation

4. SOP Interference Characteristics
White-Noise-Like Interference:
interference looks like gaussian white noise.
Matched filter (rake) coherently sums desired signal, while interference sums incoherently.
Spreading gain is realized.
Techniques:
different codesets for different piconets
long PN sequence
short PN sequence
time-hopping
baud/chip-rate offsets
Structured Interference:
Lower entropy rate than AWGN.
example: frequency-hopping, (a.k.a. time-frequency interleaving), with coding across frequencies. Interference level varies with time & frequency.
Well-designed receivers can exceed the nominal spreading gain.
Techniques:
use reliability metrics in viterbi decoder (smart receiver)
nonlinear limiter in receive path (dumb receiver)
Orthogonal Interference:
Examples:
- FDM (transmit power penalty)
Truly orthogonal codes (requires synchronization)
No theoretical limit to interference distance.
Correlated Interference:
interference looks similar to the desired signal.
Matched filter (rake) coherently sums desired signal... and undesired signal too.
Spreading gain is not realized.
In practice, multipath tends to decorrelate interference, but this isn’t guaranteed.
Fast Frequency Hopping
Slow Frequency Hopping
lowest capacity
highest capacity
longer channels
faster hopping
less structure
shorter channels
slower hopping
more structure
perfectly decorrelated
unstructured
perfectly correlated
perfectly orthogonal
Eric Ojard, Broadcom Corporation

5. Fast Hopping w/ Multipath & Interference
blue = piconet #1
red = piconet #2
Free Space:
one “collision” per cycle (if synchronized)
good immunity to near-far problem
single RF front-end captures all energy
CM1 channels:
interference is smeared over several pulses (not uniformly)
less immunity to near-far problem
single RF front-end can’t capture all energy
CM3 channels:
interference starts to like gaussian white noise
This behaves a lot like a wideband system, requiring a wideband front-end
Eric Ojard, Broadcom Corporation

6. Slow Hopping OFDM
collisiions
blue = piconet #1
red = piconet #2
Interference is highly structured even in the presence of multipath.
Unfortunately, for a 3-band system w/ one interferer, the receiver will typically see 2 collisions per 3 symbols.
requires a very low-rate code to work at high ISR
won’t achieve target rates at high ISR
Eric Ojard, Broadcom Corporation

7. Reduced Duty-Cycle SH-OFDM
collisiions
blue = piconet #1
red = piconet #2
Idea: Increase # information bits per symbol by eliminating conjugate symmetry, then reduce the symbol rate.
A 3-band system sees at most 1 collision per 3 symbols.
Advantages:
With well-designed codes, interleavers & receivers, we’ll show that we can reach target rates even at high ISR.
Bonus: reduced power consumption at receiver (turn off radio & analog front end during silence).
Disadvantages (compared to conjugate-symmetric QPSK):
3 dB higher PAR
lost diversity on frequency-selective channels (not quantified yet... ~1 dB average loss?)
requires 2 DACs
Eric Ojard, Broadcom Corporation

8. 3-Band Simulations (1) Modulation Parameters Derived Parameters
3 bands (fc = GHz, GHz, GHz)
128 tones per band
100 information tones per band
subcarrier spacing = 528 MHz / 128
16-sample cyclic prefix (30.3 nsec)
½ duty cycle
QPSK
rate 1/3 64-state convolutional code G = [117, 155, 127]8
coded bits interleaved across tones and bands (details not shown). For the 3-band case, each CC output stream is mapped to a different band.
Length 3 TFI pattern: signal [1 2 3], interference [1 3 2]
Derived Parameters
symbol length = ( ) / (528 MHz) = nsec
hop length = nsec * 2 = nsec
TFI cycle length = 3 * 2 * nsec = usec
Bit Rate = 100 * 2 * (1/3) / (545.5 nsec) = Mbps
Eric Ojard, Broadcom Corporation

9. 3-Band Simulations (2) Simulation Details one interferer
byte packets
signals over CM3 (cycled through all 100 channels in order)
interference over CM1 (one chosen randomly for each packet)
channels normalized to get desired ISR (no shadowing)
For interference simulations, we add AWGN 4 dB below the received signal PSD (~ 6 dB below sensitivity limit)
time-offsets between signal & interference randomly chosen for each packet
LPF: Transmitter & Receiver both use a 3rd order LPF
Other practical constraints (e.g. front-end dynamic range) not simulated
Channel Estimation & Timing Recovery not simulated (perfect channel estimate & timing assumed)
BER is averaged over 90 best channels
Eric Ojard, Broadcom Corporation

10. 3-Band Simulation Results (1)
AWGN-like interference will always have a near-far problem
With a conventional receiver, structured interference looks worse than white noise (CC was designed for AWGN)
~16 dB
But with a simple smart receiver, performance is much better than white noise (limited only by out-of-band roll-off and front-end dynamic range)
Here, SNR is the ratio of the received signal PSD level to the AWGN PSD level.
Eric Ojard, Broadcom Corporation

11. 3-Band Simulation Results (2)
Here, we hold the ISR constant and vary the noise.
When a high-level interferer is present, the coding scheme behaves exactly like a punctured code.
performance is predictable and easy to analyze
It’s important to choose a code that performs well for all possible puncturing patterns that collisions can cause
Example, for G=[ ], G’=[ ] is a catastrophic code
In free space, high-level interference introduces a penalty of only ~2.1 dB.
~2.1 dB
Note that this graph is for free space (hence better performance than CM3)
Eric Ojard, Broadcom Corporation

12. 3-Band Comments
In the 2-SOP case, this simple scheme can theoretically eliminate the near-far problem.
The simulated system is limited by the small total bandwidth.
At 122 Mbps, only 2 SOPs can co-exist at very close range.
Increasing the number of close SOPs requires increasing the total bandwidth or decreasing the data rate.
but note that > 2 SOPs are supported at reduced ISR
Increasing the data rate in a 2-SOP environment would requires either...
higher SNR and a higher-rate code, or
increasing the total bandwidth
Eric Ojard, Broadcom Corporation

13. 7-Band Simulation (1) Modulation Parameters Derived Parameters
7 bands (fc = 3.432, 3.960, 4.488, 5.016, 5.544, 6.072, GHz)
128 tones per band
100 information tones per band
subcarrier spacing = 528 MHz / 128
16-sample cyclic prefix (30.3 nsec)
½ duty cycle
QPSK
rate 1/3 64-state convolutional code G = [117, 155, 127]8
coded bits interleaved across tones and bands (details not shown). For the 7-band case, we made no attempt to associate bands with CC output streams
Length-7 TFI patterns
Derived Parameters
symbol length = ( ) / (528 MHz) = nsec
hop length = nsec * 2 = nsec
TFI cycle length = 7 * 2 * nsec = usec
Bit Rate = 100 * 2 * (1/3) / (545.5 nsec) = Mbps
Parameters that differ from 3-band tests are shown in RED
Eric Ojard, Broadcom Corporation

14. 7-Band Simulation (2) Simulation Details
one, two, and three interferers
byte packets
signals over CM3 (cycled through all 100 channels in order)
interference over CM1 (chosen randomly for each packet)
channels normalized to get desired ISR (no shadowing)
For interference simulations, we add AWGN 0 dB below the received signal PSD (~ 6 dB below sensitivity limit)
time-offsets between signal & interference randomly chosen for each interferer for each packet
LPF: Transmitter & Receiver both use a 3rd order elliptical filter (see appendix for response)
Other practical constraints (e.g. front-end dynamic range) not simulated
Channel Estimation & Timing Recovery not simulated (perfect channel estimate & timing assumed)
BER is averaged over 90 best channels
Parameters that differ from 3-band tests are shown in RED
Eric Ojard, Broadcom Corporation

15. 7-Band Simulation
AWGN-like interference will always have a near-far problem
1 interferer (2 SOPs including desired signal): performance is much better than w/ AWGN-type interference
2 interferers (3 SOPs): still works at high ISR
3 interferers (4 SOPs): still works at high ISR
Here, SNR is the ratio of the received signal PSD level to the AWGN PSD level.
Eric Ojard, Broadcom Corporation

16. 7-Band Comments
7-band solution appears to support 4 SOPs with much better performance than any current proposal.
4 overlapping SOPs 122 Mbps, SNR = 4 dB, ISR ~= 16 dB
2 overlapping SOPs should allow much higher rates (250 Mbps?)
Eric Ojard, Broadcom Corporation

17. General Comments
Well-designed hopping patterns, codes, and interleavers are required to achieve the theoretical benefits of structured interference.
The performance will be limited by filter roll-off and front-end dynamic range.
Interfering piconets don’t have to be “coordinated”, but they do have to be well-behaved. This only works if all piconets use a low duty-cycle when they sense other close piconets.
Eric Ojard, Broadcom Corporation

18. Conclusions
Low Duty Cycle MB-OFDM shows promise for vastly improved performance with SOPs.
3-band mode supports 2 overlapping 122 Mbps
7-band mode supports 4 overlapping 122 Mbps
These results are a proof-of-concept. There may be potential to improve performance by optimizing the code, interleaver & modulation parameters.
Eric Ojard, Broadcom Corporation