You are right to be concerned about interference between AP's because of the
effect it can have on throughput.  Two AP's that interfere with each other
will have the roughly the same total throughput capacity as a single AP, and
who wants one for the price of two.  You are also partially correct that the
extent of the interference is related to the distance between AP's, but not
for the reason most people think.  In an AP only environment (let's forget
clients for a minute), two AP's on the same channel will either interfere or
not based on the receiver sensitivity and the received signal strength.
Once two AP's are close enough to interfere, the interference will not be
made worse by moving them closer together.  For example, lets assume we have
two identical AP's on the same channel (AP-1 & AP-2) with a received signal
strength (RSSI) from AP-2 at AP-1 of -80dB.  If the receiver sensitivity of
AP-1 is less than -80dB, say -90dB, AP-2 will interfere with AP-1.  If the
receiver sensitivity is greater than -80dB, say -70dB, AP-2 will not
interfere with AP-1. If the RSSI is greater than the receiver sensitivity,
it doesn't really matter by how much.  Be aware that this is a simplified
account of what actually happens.  Most late-model enterprise type radios
employ some manner of DSP to eliminate some of this interference, but that
doesn't change the yes/no nature of interference, only the signal level at
which interference is a factor.

Unfortunately, our networks don't consist solely of AP's, and network
clients complicate the issue tremendously.  For a given radio, the factors
that will determine the achievable throughput are the transmission rate and
the size of the collision domain.  Unlike the wired networks of old, where
the collision domain was the same for every node on the network segment,
WLAN collision domains are determined by RF signal propagation and are
different for each radio.  Consider again our AP's from above.  If the AP's
are far enough apart so that the coverage areas (region within which a
client can associate with the AP) don't overlap, AP-1's collision domain is
not affected by the presence of AP-2 or it's clients.  If however the
coverage areas do overlap, the clients in the overlapping region are in the
collision domain for both AP's.  So, if there are 30 clients associated to
each AP and coverage areas don't overlap, the collision domain for each AP
is 31 nodes.  If, with the same number of clients, the coverage areas
overlap by 1/3, each collision domain is 41 nodes.  In the latter case,
network performance will be worse than the number of users associated to
each AP would suggest.  This is why moving AP's closer together can
negatively affect throughput.  It's not so much the magnitude of the
interfering signal, but the size of the collision domain.

Finally, you must also consider the size of each client's collision domain.
This is especially true if high-gain directional antennas are employed.
These narrow beam antennas are capable of producing non-overlapping cells
with large amounts of area in close proximity to each other.  Picture the
coverage area of our two AP's as adjacent oversized lanes on a bowling
alley.  It's quite possible to have 15 clients (30mW radios with omni
antennas) in each lane that can only see one AP, but can see every other
client.  In that case the collision domain for each AP is 16 nodes, but for
each client it's 31.  Again, performance will be worse than the number of
users associated to each AP would suggest.

The good news is that for low client densities using applications with
modest bandwidth & QOS requirements, throughput may not be as important as
everyone makes it out to be.  However, may need to consider all of the
factors in a high client density environment, especially if 802.11 phones or
other streaming apps are being used.  My advice is to keep your AP's from
interfering with each other, since that's easily controllable, and the
extent to which you worry about clients is up to you.  I know I didn't
exactly answer your question, but I strongly believe that only you can.
Every design in the WLAN business is a one-off, where performance, expense,
ease of use, ease of administration, etc. are traded off against one
another.  You should be suspect of anyone who says they have "THE" answer.

Chuck Enfield
Sr. Communications Engineer
PSU, Information Technology Services
110 USB 2
UP, PA 16802
Ph. 814.863-8715
Fx. 814.863-9851

-----Original Message-----
From: James Savage [mailto:[EMAIL PROTECTED]
Sent: Thursday, November 13, 2003 4:02 PM
To: [EMAIL PROTECTED]
Subject: Re: [WIRELESS-LAN] AP radio interference


I'd love to use 5g radios but that's not going to happen.  I've considered
preaching the 11a gospel to the other institution on the chance they might
convert  ;+)

...thanks for the response........J

On Thu, 13 Nov 2003, John J. Brassil wrote:

> It works OK - our engineering school recently completed a new wing
> that has a large central atrium similar to the one you describe.  This
> is an extremely dense environment for RF - 84 APs for this single
> wing!  They wanted as much throughput as possible, so our contractor
> designed a picocell network with as many as 3 APs in a single room,
> one on 1, one on 6 and one on 11, dialed to 1mW and in some cases
> futher attenuated with inline 10 and 25 dB resistors (I know resistors
> aren't measured in dB but that's the net effect of the attenuation.)
>
> Unfortuanely, depspite their best efforts, there is still a lot of
> leakage outside the intended coverage areas - not a lot of signal
> strength, but
> single- and low double-digit signal strength from non-primary APs in lots
> of areas, the worst of which is the atrium which has at least 5 devices
> visible  to Netstumbler or AirMagnet in most places on all three channels!
>
> We haven't benchmarked the throughput rates but I would imagine they
> are not full rate (we fix our APs at 11mbps) but my laptop at least
> has been able to hold a connection in there every time I have tried.
> Others' MMV.
>
> The good news is that it all still works.  For the situation you are
> describing, I wouldn't screw around with 4 channels, that's just a
> hack since the frequencies are what they are and playing games with 1
> 5 9 & 11 or some other such silliness doesn't change that.  Keep your
> same channel APs as far apart as possible and do the best you can.
>
> Or buy some 5MHz radios and do it properly. :)
>
> John
>
> John J. Brassil | Network Engineer, Vanderbilt Data/Video Engineering
> voice 615.322.2496 | ICQ 9660375
>
> --On Thursday, November 13, 2003 3:33 PM -0500 James Savage
> <[EMAIL PROTECTED]> wrote:
>
> > Hi,
> >   We're sharing a building with another institution and must also
> > share the airspace.  We've agreed to go to the four non-overlapping
> > channel model instead of three.  Each institution will use two
> > channels.  This is my first shot at full building coverage and
> > foresee channel interference issues as I'm restricted to two instead
> > of three channels.  Presumably, this can be addressed with
> > directional antennas and adjustment of signal strength but I thought
> > I might ask more experienced folks who might have already
> > experimented with channel interference.  If two APs operating on the
> > same channel can 'hear' each other, is it simply a throughput hit or
> > do things just not work at all?  Presumably, there's a relationship
> > to how close the APs are to each other (signal
> > strength-wise)...'closer' means more interference?  Also, the amount
> > of traffic is a factor as well?  This particular building has a
> > central open area surrounded by offices/classrooms with glass
> > windows overlooking the open space. The signal seems to easily
> > penetrate the glass and cover the open space as well.....ie....I
> > have multiple APs operating on the same channel bleeding into the
> > open space.  Is this a show-stopper for the open space or is it
> > possibly a slow but liveable scenario?
> >
> > ....advice or comments are greatly appreciated.
> >
> > ......thanks in advance.........Jamie
> >
> > James Savage                              York University
> > Senior Com. Tech.                         108 Steacie Bldg.
> > [EMAIL PROTECTED]                          4700 Keele Street
> > phone: 416-736-2100 ext.22605             Toronto, Ontario
> > fax: 416-736-5701                         M3J 1P3, CANADA
> >                 /\      /\      /\      /\
> >                /  \    /  \    /  \    /  \
> >                    \  /    \  /    \  /
> >                     \/      \/      \/
> >
> > **********
> > Participation and subscription information for this EDUCAUSE
> > Constituent Group discussion list can be found at
> > http://www.educause.edu/cg/.
>
>
>
>
>
> **********
> Participation and subscription information for this EDUCAUSE
> Constituent Group discussion list can be found at
> http://www.educause.edu/cg/.
>

James Savage                              York University
Senior Com. Tech.                         108 Steacie Bldg.
[EMAIL PROTECTED]                          4700 Keele Street
phone: 416-736-2100 ext.22605             Toronto, Ontario
fax: 416-736-5701                         M3J 1P3, CANADA
                /\      /\      /\      /\
               /  \    /  \    /  \    /  \
                   \  /    \  /    \  /
                    \/      \/      \/

**********
Participation and subscription information for this EDUCAUSE Constituent
Group discussion list can be found at http://www.educause.edu/cg/.

**********
Participation and subscription information for this EDUCAUSE Constituent Group 
discussion list can be found at http://www.educause.edu/cg/.

Reply via email to