[OLSR-users] Double radio mesh and OLSR

Jim Thompson (spam-protected)
Fri Apr 8 19:12:51 CEST 2005

John Clark wrote:

> Henrion Benjamin wrote:
>> Jim Thompson <(spam-protected)> [050408]:
>>> Unfortunately, it is highly unlikely that you will "avoid 
>>> interference" here, since the selectivity of
>>> common 802.11 radios is not high enough to prevent severe ACI.
>>> Now, if you were to move them to different *bands*, then you might 
>>> have something, then again, you might introduce enough hysteresis 
>>> that things really break.   
>> I was thinking to use two different types of antennas, one in Hpol and
>> the other in Vpol. For instance, a slotted waveguide and a simple
>> colinear.
> I have not had time to test things out, but in the near future I'll be 
> setting up a system with
> 3 wifi cards, and attendent antennas. I have however, set up several 
> of these boxes to have
> an active wifi card on one antenna, and one of the cards be a 
> 'monitor' for the traffic on a
> particular channel, and have seen packets on other channels, when I've 
> just had several pairs
> operating on different channels. The antennas are the 'short' 20-25 
> cm. and placed about 10 cm
> appart from each other.
> Since I'm setting up an ad-hoc network, I can 'fix' things by making 
> each network on the respective
> channels a different IBSS, and while packets will be received, they'll 
> be ignored. This does not
> prevent lower bandwith due to collisions, etc, but at least the 
> packets will be filltered out early.

Minimum Adjacent Channel Rejection for an IEEE 802.11b card is 35dBm.   
This is 1-on-6, 6-on-1, 6-on-11
or 11-on-6.    The cards with the best ACR spec have I/F sections with a 
SAW filter.  (Lucent/Agere with
the Hermes MAC, or Prism2/2.5/3.0.)   These tend to have an ACR spec of 
around 41dBm.

Lets say the antennas are separated by 1m, so 40dBm path loss.

Lets also say that the signal leaves one card at 17dBm (50mW, a fairly 
standard/poor card these days),
plus 2.2dBi of antenna gain.

17dBm + 2.2dBi = 19.2dBm EIRP.   We encounter 40dBm of path loss, so the 
signal is at -20.8dBm.  It
is "amplified" by the receiving antenna (2.2dBi), so it enters the 
remote receiver at -18.6dBm.  If we
assume a good card, then we can reject a signal on an adjacent channel 
by 41dBm, so the apparent
signal level to the receiver is -59.6dBm.  

The receiver (operating on, say ch6) sees this signal (from either ch1 
or ch11) as interference.   

Common 'tricks' suggested:

Bigger antennas (more gain):  This actually hurts.   The directivity of 
any antenna will typically be at least 0dBi
in any direction, and often (a lot) more.   The gain components add.

Cross polarization:  Good for no more than a theoretic 20dBm.   If 
employed in our example, the signal level is
now at -79dBm, which is close (but still above) the minimum rx 
sensitivity for an 802.11b receiver.  Restated,
the interference power is still (at least 3dBm) higher than the signal 
of interest.

Operating on "just" ch1 and ch11:    Potentially good for another 20dB 
of isolation.   There is no minimum spec
for adjacent channel rejection in 802.11b (there is in 802.11g/802.11a), 
but lets *assume* that this figure is 55dBm, and that a super-het 
receiver enjoyes 60dBm of adjacent channel rejection.   This yeilds the 
same result
as cross-polarization.  

So, as you can tell by now, If you employ both cross-polarization and 
operation on alternate channels (1 and 11), don't use large antennas, 
and separate the antennas by at least 1m, with super-het cards, you 
stand a chance of reducing the in-channel "interference power" of the 
alternate channel to something within 10dB of the noise floor.  Other 
ideas include using "channel filters" on the radios.

BTW, IEEE minimum adjacent channnel rejection for 54Mbps (11a or 11g) is 
-1dBm, minimum alternate channel rejection @ 54Mbps is 15dBm.

I now return you to the protocol show, already in-progress.


More information about the Olsr-users mailing list