Using all 5 Ghz channels available in your regulatory domain is not a choice, it’s a matter of fact. Here’s my experience.
I did a validation survey at one of our remote sites. The survey data showed severe co-channel interference almost everywhere. The site lies in open countryside and has no detectable neighboring APs: all of the APs are our own and they all transmit in the same 4 non-DFS channels as shown in the reading taken at location A on site:
There are at least 6-8 APs transmitting in each channel, forcing up to hundreds of users to share the same contention domain. Measures at different locations C and D are very similar. APs that are 50+ meters and several walls away are still detectable at about -85 dBm.
Another issue showing in the graphs at locations A B C are several APs received at about the same RSSI. Too may APs or not enough attenuation from walls.
This situation originated from several concurring factors:
- the misconception of playing safe with non-DFS channels only, in order to avoid DFS events altogether;
- incomplete information, at the time of the original project, about the building structure and the materials, which have much lower attenuation values than expected;
- the misconception of more APs is better;
- the effect of high ceilings (5m), light wall materials, hole in floor areas;
- incorrect transmit power;
The loath of using DFS channels could be a relic of past negative experiences (perhaps from old equipment particularly bad at managing DFS events), airport proximity (not really) or lack of proper CWNP education.
The idea of playing safe is very appealing to a conservative mindset: spare the discomfort of a DFS channel switch by using the reliable non-DFS channels 36-48. This may have worked 15 years ago when APs were few and precious, not hundreds per site like today.
The problem with this choice is that to avoid the minor, temporary nuisance of switching channels after a DFS event all the clients are permanently stuck in a huge contention domain with highest channel utilization and no airtime available.
The only solution is using all the available 5 Ghz channels in the regulatory domain to design a well spaced channel plan with minimum CCI. Of course there will be DFS events. It’s mostly false positives by client devices, it’s sporadic and it’s manageable.
After redesigning the channels accordingly and lowering transmit power about 3 dB this are the measured values at the same location A B C:
There are still issues with some strong RSSI AP that has CCI from 2 APs below (channel 52 in location C for example). But this is manageable if we
- turn down overall transmit power;
- tune and refine the channel plan;
Our purpose here is to separate contention domains, therefore it’s worth lowering transmit power so that the second channel causing CCI does not exceed 4dB above the noise floor (signal detect). Real devices (laptops, tablets, phones) are less sensitive than the dedicated, proprietary survey device that was used to collect the data. Tone down the transmit power -3 dB and factor in -3 -6 dB offset for the client device, and you have a respectable -9 dB difference that will hopefully minimize CCI.
Let’s also consider that we are surveying when students and faculty are on summer/COVID19 leave, all spaces are empty. Crowds in the lecture halls, labs, canteen and cafeteria would add their attenuation factor.
Regarding AP numerosity, it might make sense to turn off and remove some AP. It’s cheaper to remove a redundant AP than to install a missing one, so more is better still makes some sense (almost).
Regarding 2.4 Ghz it’s such a mess that I don’t even care to paste a graphic here. I turned off 50% of the 2.4 Ghz radios and still got significant CCI. As this site does not have neighboring APs I am strongly inclined to switch from 3 to 4-cannels in the 2.4 Ghz ISM band and improve the situation a tiny bit.
Bottom line: use all available 5 Ghz channels, monitor DFS events and adjust channels as required, lower transmission power, stop worrying and and live happy.