I've setup ZigBee to channel 25 (2475MHz), which should overlap only with sideband lobe of WiFi channel 11, but according to metageek, even this could harm ZigBee network:

Sideband lobes might not carry Wi-Fi data, but they are fully capable of drowning out ZigBee transmissions.

The two channel reuse approach seems to make impossible to not have overlap on the same channel.

I've got APs setup in a triangle - two indoors (hAP ax³ and hAP ax²), and one outdoors (wAP ac). The outdoors one is on full power for 2.4GHz, the indoors ones are on low power for 2.4GHz - only to cover 5GHz incapable devices.

One of indoors AP not-negligbly overlaps with outdoors AP only in not highly used areas - such as vestibule, hallway. So, I've set this one to be on same channel 6 as the outdoors one, the other indoors is left on channel 1.

What would you go with for WiFi?

• 1, and then 6 and 6 for weakly area-overlapping APs, or
• 1, and then 6 and 7 for weakly area-overlapping APs, or
• some overlapping three channel reuse, i.e. 1, 4, 7, or?

And, is there an option to have 10MHz wide WiFi channels for 2.4GHz?

10MHz is not standard chanel width and "random" wireless clients likely won't be able to use it.

Don't go for partially overlapping channels (e.g. 6 &7) because wireless devices, using one of channels, won't be able to detect traffic on the other channel and will heavily interfere with each other. When APs (snd station devices) operate on same channel, they can detect activity and simply pause for a short while (collision detection). Yes, it will reduce overall capacity but this reduction will be less than in case of interference.

When it comes to modern WiFi channels, it is actually possible to operate efficiently in layout 1-5-9 (see 2.4GHz chapter of wikipedia article on WiFi). Due to historical reasons (802.11b was using 22MHz wide channels, US doesn't allow using channel 13 which then allows for use of 1-5-9-13) internet is full of "use 1-6-11" recipes and before using anything else you better check radio surroundings to see if that's even feasible (due to potential adjacent channel interference).
However, using 1-5-9 does make enough guard band towards mentioned ZigBee channel that it should not interfere with it much.
If you find that your neighbours all stick to 1-6-11 and their signals are not too strong inside your house (e.g. signal strengths are below -80dB), then you could use channel 1 on the outdoor AP and stick to 1-5-9 layout.

Q:
why 1/5/9 if for years it always has been said it should be 1/6/11 (as you say, internet is full of it) ?
Just wondering...

A:
Because internet is USA-centric. In USA, use of channels 12 and 13 is discouraged and the remaining frequency bandwidth is most uniformly used in 1-6-11 layout. Plus it allows for proper guard band between channels if one operates 802.11b (does anybody still do it? IMO it should be outright banned by now).
Elsewhere on the Globe, where channels 12 and 13 are not restricted, using 1-5-9-13 gives 33% of performance increase as it allows 4 non-overlapping channels (only guard bands overlap). Of course channel layout 1-5-9 is apllicable also in USA, but until OP identified the problem with WiFi drowning ZigBee nobody in the States saw any need to use frequency band efficiently. But then, drowning others with WiFi isn't seen as a problem by most of WiFi users, bitching about DFS prooves that.

Then what if you use channel 13 and are suddenly faced with devices unable to connect to it ?
Especially IoT stuff where there is little to no possibility to tweak things.

Been there, done that and the main reason I don't use 13...

Then what if you use channel 13 and are suddenly faced with devices unable to connect to it ?
Especially IoT stuff where there is little to no possibility to tweak things.

Been there, done that and the main reason I don't use 13...

When adhearing to 1-5-9-13 nobody forces you to use channel 13, by using 1-5-9 you still have your 3 non-overlapping channels (just as you have with 1-6-11). But be nice to neighbours who don't use brain-dead gadgets which refuse to work with APs with (legally configured) channel 13.

And, BTW, brain-dead gadgets refusing to work on channel 13, are result of USA-centrism of internet.

Don't go for partially overlapping channels (e.g. 6 &7) because wireless devices, using one of channels, won't be able to detect traffic on the other channel and will heavily interfere with each other. When APs (snd station devices) operate on same channel, they can detect activity and simply pause for a short while (collision detection). Yes, it will reduce overall capacity but this reduction will be less than in case of interference.

On this topic, I actually found conflicting information.

Some people claim, that it's better to use partially overlapping channels, because stronger signal wins in areas, where APs overlap with one being significant stronger than the other - i.e. interference from distant AP won't have much impact.

Other people claim, that it's better to use same channel rather, because of bandwidth sharing, that you've mentioned.

When it comes to modern WiFi channels, it is actually possible to operate efficiently in layout 1-5-9 (see 2.4GHz chapter of wikipedia article on WiFi). Due to historical reasons (802.11b was using 22MHz wide channels, ... if one operates 802.11b (does anybody still do it? IMO it should be outright banned by now)

Thanks, this is useful info.

Is it possible to see somewhere in RouterOS, if any connected device uses 802.11b? I'm thinking only of robotic vacuum cleaner, that could be using it - doesn't need much of traffic, only control commands and status reporting, so production could be cost-cutting on chips?

And, if no my/ours device uses 802.11.b, then is it possible to block it - i.e. don't allow guest devices to use this mode?

However, using 1-5-9 does make enough guard band towards mentioned ZigBee channel that it should not interfere with it much.

The WiFi channel 9 (2452Mhz) covers range from 2442 to 2462MHz in OFDM. Won't sideband still cover 2475MHz (ZigBee channel 25)?

Because internet is USA-centric.

Quite true. EIRP limits tables for 5GHz WiFi are quite misleading, too. Unless explicitly specify EU in search query, then google tends to show images for USA tables.

Elsewhere on the Globe, where channels 12 and 13 are not restricted ... channel layout 1-5-9 is apllicable also in USA, but until OP identified the problem with WiFi drowning ZigBee nobody in the States ...

I live in EU, so I could legally/theoretically use channel WiFi 13, and ZigBee channel 26. But, some devices are not manufactured in different variants, and then also the ones sold in EU don't support frequencies disallowed in USA - ZigBee channel 26 and WiFi channels 12 and 13.

So, that's why I choose ZigBee channel 25.

... saw any need to use frequency band efficiently. But then, drowning others with WiFi isn't seen as a problem by most of WiFi users, bitching about DFS prooves that.

Tragedy of the commons.

Regarding co-channel / adjacent-channel ... believe whom ever you want. Personally I don't care about neighbours as long as their wifi signal is at least 30dB lower than mine. But when setting channels for my own APs I keep using 1-5-9-13 scheme. If not for other things, if AP that should be selected as roaming target, is on adjacent chanbel and is thus interfered by currently serving AP, station will less likely detect it as usable roaming target candidate.

However, using 1-5-9 does make enough guard band towards mentioned ZigBee channel that it should not interfere with it much.

The WiFi channel 9 (2452Mhz) covers range from 2442 to 2462MHz in OFDM. Won't sideband still cover 2475MHz (ZigBee channel 25)?

WiFi channels, since 802.11g they use OFDM, actually use 16.25MHz channel width. The rest up to 20MHz is already guard band (1.87MHz on each side) a.k.a. sideband. And modern wifi devices are not that shitty when it comes to spurious transmissions into adjacent frequency space.
So if you use channel 9, then highest frequency used by wifi will be 2460.2MHz (give or take) ... but ZigBee center frequency of 2475MHz is "almost half of universe" away. I don't think any wifi device could receive type rating if it would be transmiting 13MHz into adjacent frequency space (it's more than 2 wifi channels away).

if you can use channel 11 with low power and try to locate ap using channel 11 far from ZigBee devices that will mitigate somewhat the problem

if that is not feasible then sacrifice Wi-Fi leave that frequencyes exclusive for ZigBee

If you are relying on 2,4ghz for Wi-Fi is a fact you dont need High Performance Wi-Fi so put all your Wi-Fi AP on 1 and 6 channels, i am sure you can live with some co-channel interference

Done

captured with metageek tool

with moderate power (AP set to 20dbm)
and
2 metter distance between AP and Spectrum Analizer

sideband slope is not so extended, only below 90dbm yo can see somewhat after 10 minute capture

additional note, 802.11b datarates are disabled