1. More Simulations on Secondary CCA
July 2007
doc.: IEEE /2090r0
May 2008
More Simulations on Secondary CCA
Date:
Authors:
Eldad Perahia (Intel)
Eldad Perahia (Intel)

2. July 2007
doc.: IEEE /2090r0
May 2008
Abstract
Simulation results to address Coex CIDs 6227, 6228, 6229, 6230 regarding CCA on the secondary channel.
Eldad Perahia (Intel)
Eldad Perahia (Intel)

3. Brief Overview of the CIDs
July 2007
doc.: IEEE /2090r0
May 2008
Brief Overview of the CIDs
CIDs 6228, 6229, 6230 raise concerns regarding fairness to 20 MHz OBSS on the secondary channel in the presence of 40 MHz.
Several proposed changes are given:
Specify CCA backoff on the secondary independent from the primary
Allow devices on the secondary to signal intolerance
CID 6227 states that the language “idle during the times the primary channel CCA is performed (defined in ) during an interval of a PIFS” could be interpreted as sensing for only 4 usec in a 13 usec period during SIFS
Proposed resolution is to always track the secondary (when not transmitting)
Eldad Perahia (Intel)
Eldad Perahia (Intel)

4. Benchmarking New Simulations
July 2007
doc.: IEEE /2090r0
May 2008
Benchmarking New Simulations
New simulations attempted to replicate 100% loading conditions in 06/608r1 and 08/0145r0
My Simulation parameters – comparable to 06/608r1
EDCA w/ AC_BE, no TXOP
PER=.01
11n data rate = 270Mbps; 11a data rate = 54 Mbps
fully loaded network
1500 byte MPDU w/ 5 MPDU’s in an aggregate for 11n STA
Benchmark results
Goodput for 40 MHz 11n (no OBSS on secondary)
06/608r1: 138 Mbps
My simulation: 143 Mbps
Goodput for 11a (no 11n on secondary)
06/608r1: 30 Mbps
My simulation: 29 Mbps
Goodput with one 11n STA and one 11a STA on secondary
06/608r1: 11a, 23 Mbps; 11n, 97 Mbps
My simulation: 27 Mbps; 11n, 94 Mbps
Eldad Perahia (Intel)
Eldad Perahia (Intel)

5. Tx Options when Secondary Channel is Occupied
July 2007
doc.: IEEE /2090r0
May 2008
Tx Options when Secondary Channel is Occupied
In 11n D4.0, clause
“If a STA was unable to transmit a 40MHz mask PPDU because the secondary channel was occupied during this PIFS interval, it has two choices:
a) Transmit a 20 MHz mask PPDU.
b) Restart the channel access attempt. In this case, the STA shall invoke the backoff procedure as specified in as though an internal collision had taken place.”
Simulations will address fairness for each option separately
Eldad Perahia (Intel)
Eldad Perahia (Intel)

6. Objective Measure for Fairness
July 2007
doc.: IEEE /2090r0
May 2008
Objective Measure for Fairness
Does the presence of a 40 MHz TGn STA cause any more reduction in available bandwidth to an existing 20 MHz legacy BSS than does the presence of an additional 20 MHz legacy STA?
Benchmark
Goodput for one 11a (no 11n on secondary): 29 Mbps
Total Goodput for two 11a STAs (no 11n on secondary): 29 Mbps (each STA averages 14.5 Mbps)
Eldad Perahia (Intel)
Eldad Perahia (Intel)

7. Simulation of Choice a)
July 2007
doc.: IEEE /2090r0
May 2008
Simulation of Choice a)
In this simulation mode, 11n transmits 20 MHz PPDU when secondary is sensed to be busy
Goodput with one 11n STA and one 11a STA on secondary
11a: 27 Mbps
11n: 94 Mbps
11a STA goodput reduces from 29 Mbps to 27 Mbps in presence of 11n on secondary, compared to 14.5 Mbps with another 11a STA
11n STA performance reduces from 143 Mbps to 94 Mbps
11n STA only transmits 40 MHz 10% of the time
Medium access is much more fair to 11a on OBSS than to 40 MHz 11n
Eldad Perahia (Intel)
Eldad Perahia (Intel)

8. Simulation of Choice b)
July 2007
doc.: IEEE /2090r0
May 2008
Simulation of Choice b)
In this simulation mode, 11n never transmits 20 MHz PPDU, always invokes backoff procedure in primary
Goodput with one 11n STA and one 11a STA on secondary
11a: 14.3 Mbps
11n: 77 Mbps
Goodput of each of 11n and 11a reduces by half
Medium access is equally fair to 11a on OBSS as to 40 MHz 11n
Eldad Perahia (Intel)
Eldad Perahia (Intel)

9. Simulation for CID 6227 May 2008 Choice a) Choice b)
July 2007
doc.: IEEE /2090r0
May 2008
Simulation for CID 6227
Previous simulations were run with the CCA sensing on the secondary during the last 25 usec of the primary backoff period
This simulation mode models sensing during smaller portions of time
Choice a)
CCA sensing on secondary for 4 usec at 9 usec prior to end of backoff period
11n: 95.8 Mbps
11a: 25.7 Mbps
CCA sensing on secondary for 4 usec at 18 usec prior to end of backoff period and 4usec at 9 usec
11n: 94.7 Mbps
11a: 26.3 Mbps
CCA sensing on secondary for 4 usec at 27 usec prior to end of backoff period and for 4 usec at 18 usec and 4usec at 9 usec
11n: 93.8 Mbps
11a: 26.8 Mbps
Choice b)
CCA sensing on secondary for 4 usec at 9 usec prior to end of backoff period
11n: 85.7 Mbps
11a: 12.9 Mbps
CCA sensing on secondary for 4 usec at 18 usec prior to end of backoff period and 4usec at 9 usec
11n: 81.1 Mbps
11a: 13.7 Mbps
CCA sensing on secondary for 4 usec at 27 usec prior to end of backoff period and for 4 usec at 18 usec and 4usec at 9 usec
11n: 76.9 Mbps
11a: 14.5 Mbps
Eldad Perahia (Intel)
Eldad Perahia (Intel)

10. Summary With most conservative sensing, 11a and 11n share equally
July 2007
doc.: IEEE /2090r0
May 2008
Summary
With most conservative sensing, 11a and 11n share equally
Choice a) is actually very unfair to the 40 MHz 11n STA
Even with aggressive sensing, 11a goodput decreases slightly and 11n goodput increase slightly
This slight difference is not significant given so many other variables that have not been modeled (i.e. variable length aggregates, duration of TXOP, variable MCSs …)
On balance, current CCA mechanism is fair and no changes are required to the draft
Eldad Perahia (Intel)
Eldad Perahia (Intel)