Last Post: June 2, 2004:
Thought you guys might like to see some of the offline conversations between Criss Hyde and me. :-)
These are very informative snipits.
Yep, you nailed 99% of it. 802.11a defines a minimum contention window size of 15 slots (0-15). 802.11g is modeled directly after 802.11a, and thus supports 0-15...however, for backwards compatibility, 0-31 has to be supported in case an 802.11g client is attaching to an 802.11b AP (I'm assuming). Check out 9.2.12 in the 802.11g standard. aCWmin(0) is for rate sets of 1, 2, 5.5, & 11 Mbps. aCWmin(1) is for ERP PHYs. As I understand, this gives 802.11g stations operating in ERP-OFDM mode (associated to an 802.11g AP that is supporting ERP-OFDM) a statistical advantage of about 2:1 on capturing use of the medium over non-ERP stations. If an 802.11g station associates to an 802.11b AP, then the 802.11g station will be using DSSS, and thus 0-31 slots. Please verify this for me.
The relevant pages in the standards are:
IEEE 802.11g: page 46, section 19.8.4 and Table 123G (continued on following page) notice two values for aCWmin, 31 and 15 marked aCWmin(0) and aCWmin(1) respectively. I find no clue regarding the
meaning of (0) and (1). If the lower aCWmin value, 15, applies to interoperating with 802.11b stations, then Intersil is correct.
IEEE 802.11b: page 28, section 18.3.3 and Table 101
IEEE 802.11a: page 39, section 17.4.4 and Table 93
Does this help? And please let me know if you find anything more definite.
In the attached document, these guys from Intersil say that 802.11g uses an initial number of slot times from 0-15 instead of 0-31 which gives the 802.11g stations an edge in capturing the medium more of the time for "upstream" traffic. I've been unable to find this in either the 802.11a or 802.11g standards. Can you help?
I agree, with the exception of:
SIFS = 10 us
PIFS = 30 us
DIFS = 50 us
I haven't seen anything that says 802.11a/g use anything different than 802.11b for IFS. If you can find that, holler back.
802.11g 9.2.12 provides the information I was missing. I think your description is correct.
Another way of describing this >>might be<< that if a G station associates with an AP that only supports B rates, then the G station, being backward compatible with B, takes on all the slot and IFS and
contention window (minimum 31) characteristics of a B station. It becomes in effect a B station.
If a G station associates with a G rate capable AP that has B stations associated as well, then the G station takes on all the slot and IFS
characteristics of a B station but a half size contention window (minimum 15). The G stations may have to wait ten times as long for
the channel to clear after a B frame transmission as compared to a G frame transmission, but G stations will have twice the odds of gaining the channel during contention with B stations.
If a G station associates with a G capable AP that has no B stations associated as well, then the G stations takes on the slot time of an A
station (9 microseconds), the SIFS of a B station (10 microseconds), PIFS and DIFS peculiar to a G station (19 and 28 respectively), and a
contention window of an A station (minimum 15).
IEEE 802.11g: page 46, section 19.8.4, second paragraph, says the slot time is 20 microseconds, like 802.11b uses, or 9 microseconds (the optional Short Slot Time), like 802.11a uses.
In the table on the same page you will find that all 802.11g stations use a SIFS of 10 microseconds. Since PIFS and DIFS are built on slots and SIFS you will not find them listed in the table. You have to calculate them yourself. The resulting chart should make a nice projector slide.
I completely forgot about that little detail...
Will add it.