CoLocation

Subject: Re: Linksys 802.11b signal amp
From: Anthony.N.Martin
Date: Thu, 2 Jan 2003 15:19:02 +0000
X-Message-Number: 3


>> The only issue with this is that the RF Front Ends on the 802.11
>> equipment is quite wide. I suspect that there will be some adjacent
>> channel interference meaning that the overall network capacity is not
>> quite double. How much less is not certain.

> Do you have any numbers that could quantify this? I've long been
> interested in just how good (or bad) 802.11b RF front ends are, on
> both the TX and RX sides. What does the spectrum of a typical 802.11b
> card look like -- is there any filtering at all, or do you see a
> typical sin(x)/x spectrum?

Tx modulation envelope is given in the 802.11 spec.
Sidelobes must be suppressed so filtering is required.


I did some co-site measurements on some 2.4GHz devices; although not WLAN
they use common WLAN components and the same RF circuit arrangements. I
needed to answer some basic questions for demonstrating a repeater at
2.4GHz.

I measured the Tx noise level at channels far away from
the carrier.

I calculated (datasheet) values for the 1dB compression
point of the receiver front-end LNA/mixer.

Reciprocal mixing with Rx LO noise is also a performance limiting factor,
but it was figured that the Tx LO was the same thing and with modulation and
PA so the Tx noise would always be worse.

These results both suggested attenuation of around 45dB
would be needed between a near transmitter and a receiver
so as not to degrade its sensitivity to a weak signal

I constructed a test setup using a screened low-level "wanted" signal
transmitter and a high-level "interferer" transmitter mixed at different
levels and fed to a receiver. The low-level signal level was set so that it
was just above receiver threshold and the interferer level increased until
the wanted signal was lost.

Measured results showed 50dB and on some channels 55dB isolation was
required between the interfering transmitter and the receiver to avoid
degraded Rx sensitivity.

This translates into a horizontal separation of dipoles of
the order of 10m.

40dB isolation was easily achieved by stacking 2 dipoles vertically with a
horizontal metal plate between; 50dB is achievable with accurate positioning
and with no reflecting objects nearby; only practical outdoors. Otherwise
you're looking at cavity filters or directional antennas if you must put
them close together.

As the emissions of WLAN are broader-bandwidth, using sufficient channel
separation may be more of an issue.

But as a general rule to derive from this, keep your AP's or antennas at
least 10m apart if you want them to work on different channels at the same
time.

These results were at 50mW; if you're jacking up the Tx power then isolation
must be increased pro-rata. Some cards may be worse if they use lower levels
in the Tx modulation chain or LNA's with lower compression points. When this
transceiver was designed, we were paying attention to these issues; WLAN RF
designers probably rate co-site of little importance.

Also bear in mind the screening effectiveness of a WLAN card at 2.4GHz is
not great and two WLAN cards in the same PC are likely to block one another
irrespective of what kind of antennas you use. The bus-connector end of a
PCMCIA card cannot be closed effectively, and is typically left open,
leaving a nice slot-length for coupling RF in and out of the card. Keep at
least 0.5m physical separation between cards, unless you build additional
screening boxes effective at 2.4GHz.


All this only matters if cards are on *different* channels; in a typical
SOHO using one frequency the desks don't have to be 10m apart!

Ant M1FDE