1. 40 MHz Vs 20 MHz for video Date: 2009-07-13 Authors: July 2009
doc.: IEEE /yyyyr0
July 2009
40 MHz Vs 20 MHz for video
Date:
Authors:
Robert Stacey, Intel
Robert Stacey, Intel

2. July 2009
Abstract
Compare
Two links using shared 40 MHz channel
Two links using two independent 20 MHz channels
Show that shared 40 MHz is at least as efficient as 2 independent 20 MHz channels
List additional advantages and disadvantages of 40 MHz operation
Robert Stacey, Intel

3. PHY rates: 40 MHz Vs 20 MHz (normal GI)
July 2009
PHY rates: 40 MHz Vs 20 MHz (normal GI)
1 stream
2 streams
3 streams
20 MHz
40 MHz
6.5
13.5
13.0
27.0
19.5
40.5
26.0
54.0
39.0
81.0
58.5
121.5
52.0
108.0
78.0
162.0
117.0
243.0
104.0
216.0
156.0
324.0
175.5
364.5
65.0
135.0
130.0
270.0
195.0
405.0
40 MHz rate is slightly more than 2x the 20 MHz rate (2.08x) because of guard band tone filling
Robert Stacey, Intel

4. MAC efficiency (simulated)
July 2009
MAC efficiency (simulated)
MAC efficiency assuming 3ms TXOP limit, 10% PER, A-MPDU aggregation (from [1])
Robert Stacey, Intel

5. July 2009
MAC efficiency
Sequence overhead is the same irrespective of whether 20 or 40 MHz is used
For equal number of spatial streams, preamble lengths are the same
SIFS, backoff, BA duration etc. are the same
For equivalent efficiency, a 40 MHz sequence needs twice the number of MSDUs per aggregate
There is a 64 MSDU limit on block ack window
At very high PHY rates, this affects efficiency (see plot)
This can be overcome with a combination of A-MPDU and A-MSDU aggregation
Aggregate up to B MSDUs to form a 7.5KB A-MSDU and then form A-MPDU
Conclusion: MAC efficiency is same for both 20 MHz and 40 MHz operation
Robert Stacey, Intel

6. Channel access efficiency
July 2009
Channel access efficiency
Backoff
STA1
Data
DataDataData
Data
DataDataData
STA2
ACK BA
ACK BA
Collision
Backoff
Backoff
STA1
Data
DataDataData
Data
STA2
ACK BA
TXOP
STA3
Data
Data
DataDataData
STA4
Access delay
ACK BA
With 1 STA accessing channel:
Average access delay is CWmin/2 slots
With 2 STAs accessing channel:
Access is gained by one of the two STAs in less than CWmin/2 slots
But occasional collisions increase access delay
Robert Stacey, Intel

7. Channel access efficiency (simulated)
July 2009
Channel access efficiency (simulated)
Access delay = time between successful (non-colliding) channel accesses
Includes time lost to collision and exponential backoff
Assumes there is a collision detect exchange at beginning of TXOP (i.e. Data/ACK or RTS/CTS)
Collision duration (2 or more STAs transmitting) is assumed to be 10 slots (approx 90uS)
Conclusion is not significantly different with higher collision penalty
With default CWmin=15, channel access is more efficient with 2 STAs than for 1 STA
Robert Stacey, Intel

8. Statistical multiplexing gain
July 2009
Statistical multiplexing gain
Sum of the 2 flows
2 individual video flows
Peak single stream: 450 packets per second
Peak dual stream: <900 packets per second
 Unlikely (but not impossible) that the peaks coincide and sum
Robert Stacey, Intel

9. Statistical multiplexing gain
July 2009
Statistical multiplexing gain
For independent random variables X and Y, the variance of the sum is the sum of their variance:
Standard deviation is
i.e. less than the sum
We don’t eliminate the maximums (peaks), just make them occur much less frequently
i.e. Xmax + Ymax occurs much less frequently than Xmax or Ymax individually
Robert Stacey, Intel

10. 40 MHz operation efficiency
July 2009
40 MHz operation efficiency
We expect shared 40 MHz to be more efficient independent 20 MHz
Slightly higher PHY data rate (4% gain)
Slightly more efficient channel access (<1% gain)
The overall gain, however, is very slight: ~5%
In addition, there is some gain from statistical multiplexing
Standard deviation of the sum of two VBR streams is less than the sum of the standard deviation of the individual streams
Robert Stacey, Intel

11. Other advantages/disadvantages
July 2009
Other advantages/disadvantages
Advantage:
Robustness: the more throughput headroom available, the easier it is to build and maintain a buffer against interference
Single flow benefits from having entire 40 MHz available
Dual flow benefits from statistical multiplexing gain and, to a small degree, increased efficiency
Disadvantage:
Interference that only affects one video stream may have an impact on the other due to increased use channel use from retransmissions
Robert Stacey, Intel

12. Throughput & Buffering
July 2009
Throughput & Buffering
Throughput headroom determines how quickly
the receive buffer is filled or refilled
throughput
Throughput headroom
User latency tolerance
Sustained throughput
time
Buffer depth determines the length and degree of throughput degradation that can be tolerated
Start streaming
Start playout
Robert Stacey, Intel

13. July 2009
References
E. Perahia, R. Stacey, Next Generation Wireless LANs: Throughput, Robustness and Reliability in n, Cambridge University Press, September 2008
Robert Stacey, Intel