Maximum of how many wireless cards

[Jim Thompson]

(I just love hearing from these people(vendors) who hide behind 
pseudo-anonymous yahoo and hotmail accounts.)

> Maximum of how many wireless cards I can use in a hostap box?

All on 2.4GHz?  One.

> I have 6 PCI slots, can the driver support loading
> upto 6 wireless cards?

yes, but you're still going to be in wireless hell.

> Is it a goood idea to use 4 PCI cards with channel 1,
> 5, 10 and 14 (licensed) when using outdoor directional
> antenna for each interface, or the overlaps are too
> much?

Its a terrible idea.

Normal adjacent channel (ch1 on ch6 or ch6 on ch11) rejection on a 
Prism2/2.5/3.0 card is about 41dB.
(IEEE minimum is 35dB).

Lets say you transmit at 15dBm, into 2.2dBi of antenna gain, and the rx 
sensivity of the card on at 11Mbps is -84dBm.

Free space path loss at 2.4GHz can be expressed as: Path loss = 40 dB + 
20log(distance in meters).  Since log(1) = 0,
its easy to see that the pathloss for 2.4GHZ at 1m is 40dB.

So, your signal leaves the card at 17.2dBm (modulo some amount of coax 
loss).  It is received, presumably less
than 1m away, but lets say you have a 1m separation on the 4 antennas, 
at 17.2 - 40, or -22.8dBm, **plus** the 2.2dBi gain of the antenna on 
the 2nd receiver, for -20.6dBm.

This is more than 60dBm higher than the sensivity limit @ 11Mbps, and 
is certainly higher than the CCA limit set on your card (likely set to 
the Intersil 3861B CR38 default of 0xc).   You can play with CR48, and 
CR9 bits 6:5 as well, but its not going to help.)

So at the least, you'll set CCA on an adjacent channel when any of your 
radios begins to transmit.  At worst,
you will generate a *much* higher noise floor on the other radios, and 
the result is that you will either have
the 4 radios smashing each other's packets, or (at the least) you will 
significantly decrease range.

Note that you may not get CCA set on a received packet on the 
off-channel radios, and in this case, a radio in your box on an 
adjacent channel will smash the received packet (as its is being 
received) if it chooses to tx during...

Moreover, you'll have this problem with 'alternate' channels 1 and 11) 
too (I can't reveal the alternate channel rejection, lets just say that 
its not low enough, and I know this to be true.)

Your goal, if you're going to run multi-channel, is to reduce the 
received signal on all other radios to well below
the noise floor.   Assuming the NF @ -100dBm, you'll likekly need to 
find 120dB of isolation (or more) between
the radios.   If you're not flinching, then you don't understand just 
how large a number 120dB is.   You might as
well read "not in our lifetime".

Things you can try:

1) Put hella-strong 'channel filters' in front of your radios.   This 
will provide additional in-channel rejection.
2) Separate the antennas more, though at 10m, the path loss is 60dB, so 
its still far too high.
3) attempt to decrease the antennas mutual gain toward each other.  
You'll find that this is increasingly difficult with higher gain, since 
most antennas with a lot of gain have significant side and back lobes.
4) Attempt to co-ordinate operation of the radios.   Keep *any* radio 
from transmitting while *any* other radio is
	transmitting or receiving.   (Parallel receive is fine.)

	Here you've
	a) build some more expensive hardware (along with custom Prism2/2.5/3 
based cards),
	b) reduced the throughput to likely 1.0X that of a single-radio system
	c) likely violated the patents of my (now-ex) employer

> Anyone have any ideas or first-hand experience on these topics!

Yes (real-world experience).

Jim