I'm sure I said "1,6,11" are *NOT* non-overlapping in the real world. Oh sure, the transmit masks don't overlap, but the receiver architectures aren't good enough.
Kuhl, Vince (DotComm) wrote:
Thanks for the replies. Sorry I am not able to present a clearer picture. Some of our users jump from one radio to another, we are using eap and the logs show the re-auth/re-assoc occur on the new radio. When this occurs the user losses connectivity for their TN3270 session. I am not confident in the original network design and am looking for resources on preventing users from jumping from radio to radio. Should this be done on the client side? Also I understand the "1,6,11" channels are non-overlapping - we attempted to spread the frequencies out as much as possible using all channels. If this is the case, should you only use "1,6,11"
-----Original Message----- From: Jim Thompson [mailto:[EMAIL PROTECTED] Sent: Tuesday, September 21, 2004 11:24 PM To: [EMAIL PROTECTED] Subject: Re: [SOCALWUG] WLAN Design Question
In my (ahem) experience, many "experienced wireless LAN designer(s)" have precious little theoretic base. They just "know what works".
e.g. the oft-quoted (but inaccurate) 3 non-overlapping channels.
Worse are those pretty picture of hexagons with each hexagon representing the (idealized) coverage area of a single AP, with adjacent
hexagons (APs) in different colors (representing different frequencies),
promulgating the "1, 6, 11" arguement, and causing anywhere mild to severe interference, no matter what load is applied to the network.
It causes cognative dissonence when I say it, but the better (dare I say "correct") way to deploy multiple APs is to put them on the (you'll hate this, I know) *same* channel. There are limits to this, of course, see below. At the very least, this makes the "hexagons" somewhat larger.
This allows 802.11's CSMA/CA to work. (If the preamble is detected, CCA gets set and the MAC holds off transmitting for the duration of the packet.)
Move off channel, and all the baseband can 'see' is more noise (due to the in-channel power of the STA (or AP) operating on an adjacent channel).
If the operation of an adjacent (or alternate) channel STA (all APs are STAs, btw) is such that the local receiver can't decode the preamble (which is sent at the lowest modulation) then you're probably safe moving to the next adjacent (or better, alternate) channel. Thus, the distance between the centers of the hexagons should be about the distance at which you can maintain (for 11b) 1Mbps connectivity.
Still worse are the people who attempt to engineer large coverage via high-gain omnis on a set of APs. This actually makes the problem worse (and I can prove it, the math is simple).
jim
Jack Unger wrote:
Get help from an experienced wireless LAN designer. You likely have too many access points located too close together. Under normal-use traffic loads, you will have an unacceptably high level of self-interference (too many packets colliding with each other). The result will be slow throughput.
jack
Kuhl, Vince (DotComm) wrote:
I was looking for some resources on wlan design. I have a 6 story building
with approx 5 radios per floor. All radio's are in the same SSID. The idea
was to provide redundancy for the users but most pc's only get an acceptable
signal from one radio anyway. It seems as if this one large SSID approach is
causing more of an interference problem than anything else. Would it be
better to break the building up into separate SSID's? Any thoughts or
resources would be appreciated.
Thanks
