placement of 3 cards on a PC box

[Jim Thompson]

the receivers don't have that much selectivity.

Your cards output at least 15dBm (likely, though there are cards that  
will output 23dBm or more based on Prism2.5).

The ACR of a Prism2.5 chipset is 46dB   
(http://systems.cs.colorado.edu/grunwald/MobileComputing/Papers/Prisim- 
Wireless/pc-card-info.pdf) at a FER of 0.08.

So lets assume that you've got 2 Prism2.5 radios next to each other in  
a box, and that you've got 1m of separation between the antennas..

on ch1 you shove 15dBm of power out
on ch6 you receive this 40dB down (due to the path loss of the 1m  
separation between the (presumed 0dBi!) antennas), for -25dBm, and then  
you get another 46dB of adjacent channel rejection, for -71dBm.

note that its also presumed that range matters to you, and that you  
really did want to be able to receive and decode the client that is  
hitting you with a signal at -86dBm or so (typical prism2/2.5-based  
designs have a rx sensitivity of around -86 to -89dBm @ 11Mbps at a FER  
of 0.08.)

The result is that if a distant client transmits, and the radio on the  
adjacent channel transmits anytime inside the packet that is arriving,  
the packet will be destroyed (and eventually retransmitted).

Oh yeah, you're going to need around 14dB SNR just to decode an 11Mbps  
DSSS/CCK signal.

See how unlikely it is that two radios inside the same box (in the same  
band) will do much?

High power cards make the problem worse.

Any antennas on the APs will make the problem worse (the above was  
using 0dBi antennas)

When the APs drop modulation rate (from 11Mbps to 5.5Mpbs or 2Mbps or  
even 1Mbps), the data traffic is 'on the air' for a longer period of  
time, and thus the chances for this type of "interference collision" go  
up.

Running 3 cards on ch1, 6 and 11 makes the problem worse.  ch6 is  
adjacent to both ch1 and ch11, and the alternate ch performance of a  
Prism2/2.5 chipset, while not given in the spec, is about 20dB better,  
which still isn't enough to degrade range.

So, for a client 'at range', any transmission by either of the two  
other radios will smash its packets at the intended receiver.

There are ways to fix this, but they require additional money.

Jim


On Apr 20, 2004, at 4:58 PM, Mustafa C. Kuscu wrote:

> Is the loss due to interference? If so, why is that, especially when  
> the cards are supposed to operate on non-overlapping channels?
>
> I am wondering if it is because of something like coupling among  
> hardware. Denis's post gives a hint but I am not sure..
>
> Denis's post:    
> http://sisyphus.iocaine.com/pipermail/hostap/2004-April/006210.html
>
> Jim Thompson wrote:
>
>> 40dB of loss in the first meter.
>