disable WiFi cross-channel interference

[Sebastian Lindner]

Hi,

we're simulating different WiFi networks on different WiFi channels. So far we'd been assuming that there's no cross-channel interference for non-overlapping channels. For example, if we configure 1 network to operate on the 5GHz 80MHz-wide channel with number 42, and another on number 58, we had expected to have no interference. However, our current evaluations show that there's a pretty massive impact.

Is there any way we can disable cross-channel interference on non-overlapping channels? We're using the SpectrumWifiPhy, I'm not sure if YansWifiPhy would be a solution (but then, what do we lose, it seems that YansWifiPhy is a much "simpler" Phy, i.e. in the comments it says that it is a model for 802.11a, but we need to simulate 802.11ac/ax -- I guess this is not a choice for us then?)

[Tommaso Pecorella]

Channels 42 and 58 should have minimal interference (not zero, but not massive).

My first suggestion is to triple check that it's indeed cross-channel interference. Also, try to use ns-3-dev, and in case open an issue (with an example).

About having "perfect" isolation, you can create two channels, one for each network. However, I wouldn't suggest this solution.

[Sebastian Lindner]

Hi Tommaso,

thanks for your reply.

We've circumnavigated this issue by using non-neighboring 80MHz 5GHz channels. The issue is resolved when we do this.

Note that the cross-channel interference would create a hidden node scenario for us, where on the downlink a large number of packets is dropped (because next to the receiving station, many interfering stations on the neighboring channel were operating), whereas on the uplink, everything's fine (because those stations are far away from the access point).

By leaving sufficient frequency isolation, this effect vanishes.

[Tommaso Pecorella]

Even if channels are different (and supposedly non-overlapped), some interference is expected (but only a little).

If the interference is too high, please try using ns-3-dev, and in case report the issue. If it's a bug we need a valid test case.