Cliff Skolnick writes:
> Theory is great, but in practice I've had problem with OFDM as
> specified for 802.11g. Depending on how you tune your gear multipath
> can kill you, but timings can be adjusted to handle it well, even use
> it. I've not experienced this personally as I've only use 802.11a and
> 802.11g indoors, but I've heard stories.
OFDM is essentially unaffected by delay spread (ahem, "multipath")
until you hit the guard interval, and then it collapses under a TON of
ISI.
The GI in both 802.11a and 802.11g is 800ns. I don't know what it
might be in Alvarion's gear.
> The biggest problem which I have totally experienced is the current
> lack of adjacent channel rejection for OFDM of the current generation
> for adjacent yet non-overlapping channels (1, 6, 11). I was involved
> with a large roll out a will tons of 802.11b on 1 and 11, and the
> 802.11g on 6. The 802.11g was getting beat up performance wise so we
> looked into it. The numbers I was quoted were a 30db rejection for
> 802.11b (orinoco gear) and a < 6db rejection (I'm being nice here) for
> linksys and other G cards. Next time we will pick 1 or 11 for the
> 802.11g gear :/
Well, its not "total lack", but the ACR spec in 802.11g (and 802.11a)
is significantly less in the 802.11b specification. At 54Mbps, both
11g and 11a require "-1dB" of ACR (at a minimum). The 802.11b spec
calls for at least 35dB of ACR at 11Mbps.
Intersil claims 41dB of ACR for Prism 2.5.
Next time pick 1 or 11 for 802.11g, and run 802.11b on the other
'alternate' channel.
> What kind of adjacent channel rejection does your gear have? and have
> you adjusted your OFDM timings to handle multipath where the bit shift
> is larger than 802.11g? So far I've not been impressed with the
> stability of 802.11a or 802.11g gear, but I think the issues were not
> OFDM's fault, more implementation issues.
I'm not sure what "bit shift is larger than 802.11g" might mean.
Jim
--
"Speed, it seems to me, provides the one genuinely modern pleasure."
-- Aldous Huxley (1894 - 1963)
--
general wireless list, a bawug thing <http://www.bawug.org/>
[un]subscribe: http://lists.bawug.org/mailman/listinfo/wireless