Non-overlapping channel (1/6/11) isn't always optimal

Just found another case where a non-overlapping channel isn't optimal:

* Router: Hawking HWR54G Rev M with latest firmware (Version 1.0 Release 11)
* PC Card: NETGEAR WG511v1 with latest drivers
* Host: IBM ThinkPad T30 running Windows XP SP2
* Fairly weak signal due to intervening walls.
* Interference from neighbors on channels 1, 6, and 10

After considerable experimentation, I found that I couldn't get a reliable
connection on any of the non-overlapping channels (1, 6, 11), but that I could
get a reliable connection on channel 3.

I've previously reported a case with similar characteristics where I also got
the best results with the router (D-Link DI-624) on channel 3.

--
Best regards, SEE THE FAQ FOR ALT.INTERNET.WIRELESS AT
John Navas <http://en.wikibooks.org/wiki/FAQ_for_alt.internet.wireless>

John Navas <(E-Mail Removed)> wrote:
>Just found another case where a non-overlapping channel isn't optimal:

>* Interference from neighbors on channels 1, 6, and 10

>could get a reliable connection on channel 3.

Well, sure, if there's significant interference from the optimal set
of non-overlapping channels, then go for the non-optimal midpoints.

[POSTED TO alt.internet.wireless - REPLY ON USENET PLEASE]

In <(E-Mail Removed)> on Thu, 30 Mar 2006 14:04:14
-0500, William P.N. Smith <(E-Mail Removed)> wrote:

>John Navas <(E-Mail Removed)> wrote:
>>Just found another case where a non-overlapping channel isn't optimal:
>
>>* Interference from neighbors on channels 1, 6, and 10
>
>>could get a reliable connection on channel 3.
>
>Well, sure, if there's significant interference from the optimal set
>of non-overlapping channels, then go for the non-optimal midpoints.

My point is that 1, 6, and 11 aren't necessarily optimal.

In fact 1, 6, and 11 are only really optimal in terms of minimizing
interference when you control multiple overlapping access points, and even
then they aren't necessarily optimal -- 4 channel spacing (e.g., 1, 4, 8, and
11) will allow 4 access points (instead of just 3) to overlap with minimal
interference.

In an uncontrolled environment an overlapping channel might be optimal, as it
is in my case that started this thread, and in some other cases I know of.

It nonetheless makes sense to recommend first trying the non-overlapping
channels (1, 6, 11) even in an uncontrolled environment because (a) they are
the most likely to be optimal and (b) they are the most "neighbor friendly".
But it nonetheless also makes sense to try other channels, particularly the
midpoints between the non-overlapping channels, when the non-overlapping
channels don't provide optimal results.

--
Best regards, SEE THE FAQ FOR ALT.INTERNET.WIRELESS AT
John Navas <http://en.wikibooks.org/wiki/FAQ_for_alt.internet.wireless>

~ >>Just found another case where a non-overlapping channel isn't optimal:
~ >
~ >>* Interference from neighbors on channels 1, 6, and 10
~ >
~ >>could get a reliable connection on channel 3.
~ >
~ >Well, sure, if there's significant interference from the optimal set
~ >of non-overlapping channels, then go for the non-optimal midpoints.
~
~ My point is that 1, 6, and 11 aren't necessarily optimal.
~
~ In fact 1, 6, and 11 are only really optimal in terms of minimizing
~ interference when you control multiple overlapping access points, and even
~ then they aren't necessarily optimal -- 4 channel spacing (e.g., 1, 4, 8, and
~ 11) will allow 4 access points (instead of just 3) to overlap with minimal
~ interference.

Our test results do not concur (assuming that you control your RF
environment). The interference created by using 1/4/8/11 rather
than 1/6/11 channel spacing will give you less aggregate throughput
throughput the 2.4GHz band. With 4 APs, you are better off having two on
the same channel.

This is because, with multiple BSSIDs on a single channel, the 802.11 MAC
can (potentially) recognize the signals from the other BSSID as being valid
802.11 frames, and can hence do an informed backoff/transmission. With
multiple BSSIDs on adjacent overlapping channels, the other BSSID's
transmissions are just perceived as noise.

http://www.cisco.com/en/US/products/...0802846a2.html


having said that ...

~ In an uncontrolled environment an overlapping channel might be optimal, as it
~ is in my case that started this thread, and in some other cases I know of.
~
~ It nonetheless makes sense to recommend first trying the non-overlapping
~ channels (1, 6, 11) even in an uncontrolled environment because (a) they are
~ the most likely to be optimal and (b) they are the most "neighbor friendly".
~ But it nonetheless also makes sense to try other channels, particularly the
~ midpoints between the non-overlapping channels, when the non-overlapping
~ channels don't provide optimal results.


agreed - if you don't control your RF environment, then do whatever works best.
(Within the bound of the law, of course.)

Aaron

[POSTED TO alt.internet.wireless - REPLY ON USENET PLEASE]

In <(E-Mail Removed)> on Thu, 30 Mar 2006 14:32:48
-0700, Aaron Leonard <(E-Mail Removed)> wrote:

>~ My point is that 1, 6, and 11 aren't necessarily optimal.
>~
>~ In fact 1, 6, and 11 are only really optimal in terms of minimizing
>~ interference when you control multiple overlapping access points, and even
>~ then they aren't necessarily optimal -- 4 channel spacing (e.g., 1, 4, 8, and
>~ 11) will allow 4 access points (instead of just 3) to overlap with minimal
>~ interference.
>
>Our test results do not concur (assuming that you control your RF
>environment). The interference created by using 1/4/8/11 rather
>than 1/6/11 channel spacing will give you less aggregate throughput
>throughput the 2.4GHz band. With 4 APs, you are better off having two on
>the same channel.

That's not my experience, which is consistent with these documents:

4 Simultaneous Channels Okay For 802.11b
<http://www.extremetech.com/article2/0,3973,708876,00.asp>

The Overlapping Channel Problem
<http://www.ja.net/services/publications/factsheets/063-overlapping-channel-problem-v2.pdf>

WiFi was *designed* to support many interfering, overlapping networks,
handling the resultant packet collisions with minimum fuss.

Channel Overlap Calculations for 802.11b Networks
<http://www.cirond.com/whitepapers/FourPoint.pdf>

>This is because, with multiple BSSIDs on a single channel, the 802.11 MAC
>can (potentially) recognize the signals from the other BSSID as being valid
>802.11 frames, and can hence do an informed backoff/transmission. With
>multiple BSSIDs on adjacent overlapping channels, the other BSSID's
>transmissions are just perceived as noise.
>
>http://www.cisco.com/en/US/products/...0802846a2.html

As it notes, that paper is at odds with the other authorities. In addition,
the only supporting test was both limited and highly artificial.

It may be that what works best is different in different types of
environments.

--
Best regards, SEE THE FAQ FOR ALT.INTERNET.WIRELESS AT
John Navas <http://en.wikibooks.org/wiki/FAQ_for_alt.internet.wireless>

On Thu, 30 Mar 2006 21:55:36 GMT, John Navas <(E-Mail Removed)> wrote:

~ [POSTED TO alt.internet.wireless - REPLY ON USENET PLEASE]
~
~ In <(E-Mail Removed)> on Thu, 30 Mar 2006 14:32:48
~ -0700, Aaron Leonard <(E-Mail Removed)> wrote:
~
~ >~ My point is that 1, 6, and 11 aren't necessarily optimal.
~ >~
~ >~ In fact 1, 6, and 11 are only really optimal in terms of minimizing
~ >~ interference when you control multiple overlapping access points, and even
~ >~ then they aren't necessarily optimal -- 4 channel spacing (e.g., 1, 4, 8, and
~ >~ 11) will allow 4 access points (instead of just 3) to overlap with minimal
~ >~ interference.
~ >
~ >Our test results do not concur (assuming that you control your RF
~ >environment). The interference created by using 1/4/8/11 rather
~ >than 1/6/11 channel spacing will give you less aggregate throughput
~ >throughput the 2.4GHz band. With 4 APs, you are better off having two on
~ >the same channel.
~
~ That's not my experience, which is consistent with these documents:
~
~ 4 Simultaneous Channels Okay For 802.11b
~ <http://www.extremetech.com/article2/0,3973,708876,00.asp>
~
~ The Overlapping Channel Problem
~ <http://www.ja.net/services/publications/factsheets/063-overlapping-channel-problem-v2.pdf>
~
~ WiFi was *designed* to support many interfering, overlapping networks,
~ handling the resultant packet collisions with minimum fuss.
~
~ Channel Overlap Calculations for 802.11b Networks
~ <http://www.cirond.com/whitepapers/FourPoint.pdf>
~
~ >This is because, with multiple BSSIDs on a single channel, the 802.11 MAC
~ >can (potentially) recognize the signals from the other BSSID as being valid
~ >802.11 frames, and can hence do an informed backoff/transmission. With
~ >multiple BSSIDs on adjacent overlapping channels, the other BSSID's
~ >transmissions are just perceived as noise.
~ >
~ >http://www.cisco.com/en/US/products/...0802846a2.html
~
~ As it notes, that paper is at odds with the other authorities. In addition,
~ the only supporting test was both limited and highly artificial.
~
~ It may be that what works best is different in different types of
~ environments.

I've glanced at your references, and have this comment: ALL of them
reference 802.11b not 802.11g, and all of them point back to the original
'02 Mitch Burton paper. Note that Burton concedes that the 4-channel
allocation scheme is not well suited to 802.11g OFDM, which uses considerably
more energy at the edges of the channel.

Surely more and better testing using 802.11g is in order.

Regards,

Aaron

[POSTED TO alt.internet.wireless - REPLY ON USENET PLEASE]

In <(E-Mail Removed)> on Fri, 31 Mar 2006 11:01:05
-0700, Aaron Leonard <(E-Mail Removed)> wrote:

>On Thu, 30 Mar 2006 21:55:36 GMT, John Navas <(E-Mail Removed)> wrote:
>
>~ [POSTED TO alt.internet.wireless - REPLY ON USENET PLEASE]
>~
>~ In <(E-Mail Removed)> on Thu, 30 Mar 2006 14:32:48
>~ -0700, Aaron Leonard <(E-Mail Removed)> wrote:
>~
>~ >~ My point is that 1, 6, and 11 aren't necessarily optimal.
>~ >~
>~ >~ In fact 1, 6, and 11 are only really optimal in terms of minimizing
>~ >~ interference when you control multiple overlapping access points, and even
>~ >~ then they aren't necessarily optimal -- 4 channel spacing (e.g., 1, 4, 8, and
>~ >~ 11) will allow 4 access points (instead of just 3) to overlap with minimal
>~ >~ interference.
>~ >
>~ >Our test results do not concur (assuming that you control your RF
>~ >environment). The interference created by using 1/4/8/11 rather
>~ >than 1/6/11 channel spacing will give you less aggregate throughput
>~ >throughput the 2.4GHz band. With 4 APs, you are better off having two on
>~ >the same channel.
>~
>~ That's not my experience, which is consistent with these documents:
>~
>~ 4 Simultaneous Channels Okay For 802.11b
>~ <http://www.extremetech.com/article2/0,3973,708876,00.asp>
>~
>~ The Overlapping Channel Problem
>~ <http://www.ja.net/services/publications/factsheets/063-overlapping-channel-problem-v2.pdf>
>~
>~ WiFi was *designed* to support many interfering, overlapping networks,
>~ handling the resultant packet collisions with minimum fuss.
>~
>~ Channel Overlap Calculations for 802.11b Networks
>~ <http://www.cirond.com/whitepapers/FourPoint.pdf>
>~
>~ >This is because, with multiple BSSIDs on a single channel, the 802.11 MAC
>~ >can (potentially) recognize the signals from the other BSSID as being valid
>~ >802.11 frames, and can hence do an informed backoff/transmission. With
>~ >multiple BSSIDs on adjacent overlapping channels, the other BSSID's
>~ >transmissions are just perceived as noise.
>~ >
>~ >http://www.cisco.com/en/US/products/...0802846a2.html
>~
>~ As it notes, that paper is at odds with the other authorities. In addition,
>~ the only supporting test was both limited and highly artificial.
>~
>~ It may be that what works best is different in different types of
>~ environments.
>
>I've glanced at your references, and have this comment: ALL of them
>reference 802.11b not 802.11g, and all of them point back to the original
>'02 Mitch Burton paper. Note that Burton concedes that the 4-channel
>allocation scheme is not well suited to 802.11g OFDM, which uses considerably
>more energy at the edges of the channel.
>
>Surely more and better testing using 802.11g is in order.

I would agree.

--
Best regards, SEE THE FAQ FOR ALT.INTERNET.WIRELESS AT
John Navas <http://en.wikibooks.org/wiki/FAQ_for_alt.internet.wireless>