[WLANware] Antenna diversity + wave guide antennas; books; performance data

[Daniel Nitzpon]

hi!
has been a while since that post, but i thought i should still drop you 
a couple of words..

dondavis at reglue.org schrieb:
> I had thought of creating a mesh backbone using wrt's with wave guide
> antennas.  The center (gateway) would have a large omni.

creating a mesh with a star topology is a very bad idea for several 
reasons: when an antenna receives too many different signals, you will 
basically only get packets interfering with themselves and will have 
very low throughput and bad reliability. even the rts/cts mechanism 
doesn't help too much. plus you loose the advantage of a mesh, that 
outages of single node can be compensated by the rest.
if you want to distribute internet acces from one source on a large 
scale, better build a backbone mesh, where you have one node per 
direction with directional antennae and located far enough away from 
each other in order not to interfere. connect these via lan cables and 
let another pair (one backbone, one usual mesh node, also connected via 
lan) take care of the handover in the usual mesh.
put the backbone on a channel that does not interfere with the regular 
mesh or use 5ghz or optolink for it.

> The thought was with the wrt true diversity antennas that one wave guide
> antenna connected to the left antenna would be pointed at the source and
> the right antenna would be pointed at the next node.  Or possibly one wave
> guide pointed at the source and the other with an omni.

with antenna diversity you can never be sure which antenna the outgoing 
signals are sent to. plus the diversity chip in linksys devices tends to 
burn out. better use 2 cheap devices instead or a device with 2 minipci 
slots.

> Is there an easily referenced table or a graphic that shows the
> correlation between gateway speed and node count?  (I understand a lot of
> factors affect this, but I'm thinking of in an 'ideal' bandwidth sharing
> scenario tempered with real world experience?)  For example at 3MB/sec,
> 10MB/sec, 25MB/sec down what sort of speed can I expect for 10, 20, 50,
> 100 nodes?

the basic calculations for an optimum scenario should be obvious, 
everything else really depends on how you do things exactly. it seems to 
me that if you have several hops on the same channel you will end up 
with approx. 50 kb/s after 4 or 5 hops. but if you have a cable 
connection or another channel at every 2nd hop you ca get much better 
results. on the contrary, if you have too many aps in direct contact, 
you can be worse off as well.
keep in mind that experience shows that upload can be a much more 
limiting factor than download.

greetz
daniel