Long duration Site Survey, Multi Sensor, Congested Environment

Sorry, this post got long. TLDR in bold.

I'm helping a friend with apartment in multi unit, multi floor building.
His unit is 1600 sq ft, one floor

The existing old wifi is not performing well.
The goal is to understand the existing wifi environment including neighbor APs, channels used, signal strength, band utilization (2.4, 5, and 6), channel utilization. Then, ultimately recommend a solution.

I don't have any data yet.

I found these posts helpful for baseline information.

What I'd like to do for diagnostics is deploy multiple OpenWRT routers in the unit as sensors.
Maybe something like RT3200 w/ OpenWRT replaces existing router and provide remote access, 3 ASUS onhub w/ OpenWRT connected by wire for monitoring sensors.
Using those devices since they are available.

I'm leaning toward Kismet on OpenWRT for diagnostics since it has a multi sensor architecture.

Looking for alternatives, feedback and advice.

comments and questions:

#1 Doing a site survey with a phone or laptop gives a good starting picture. However, it is only a snapshot in time and maybe you miss the worst case scenario. That is why I want to do a long duration site survey maybe a week or two.

#2 Can I get channel utilization from Kismet? One of the mentioned posts links to an ARS article where they mention that Ubiquiti and InSSIDer can report on utilization. Seems to me just seeing an AP beaconing on a given channel is inadequate information. If it doesn't have any utilization I wouldn't expect it to impact me.

#3 Is using a router as a site sensor a bad idea vs using a laptop or phone? Could it be that having bigger better antenna will give a higher signal of your and your neighbor's APs and therefore not represent the end use case of phones and laptops.

#4 For heatmap diagrams on congestion, do you need one for every channel? Usually I see heatmaps used for design. Put some APs on a drawing and simulate the coverage. Neighbor APs typically don't seem to be included. In that case the theoretical heat map is the same for every channel in 5GHz, so you don't care. I want to see a heatmap for existing congestion. I guess that means a heatmap for every channel.

#5 Can low dBm neighbor devices impact your performance? If your client sees another SSID/device on the same channel and it has high utilization but something like -90dBm would it still impact your performance.

#6 Any good ideas on post processing the data? Maybe I will load into Graylog or Elastic.

#7 The software that does all the things is probably thousands of dollars. I'm will to go a few hundred if there is something that provides a good benefit.

#8 I haven't seen anything free/cheap that ticks all the boxes. Intuitibits has an analyzer that will allow remote capture from a wlanpi. I also got this to work on an Ubuntu sensor. Haven't tried Netspot yet, but looks like it might help. They have a feature for combining multiple saved captures. These tool seem to just capture pcaps of 802.11 management frames with tcpdump.

This is an interesting question, and my first inclination is to jump in and full throttle do all that stuff you wanna do ... But my second inclination is to say that's probably not at all necessary and that if the system that exists isn't working well, likely it's not because of something that only becomes apparent at certain times of day or days of week or whatever. If you design something that works better at 6Pm on a Tuesday it is likely to work better all the time.

Specifically it'd be good to figure out some data on the symptoms of the current thing. Do you have high packet loss? High latency? Frequent Dissociations? Problems roaming? Problems with bandwidth sharing? It might not be the Wifi itself that's the issue... So maybe you can give idea of what are the symptoms.

1 Like

Modern radios (ath10k wave2 and mt76 wifi6) can sustain background passive scan without impairing main ap function.
eg aruba does this for years.

The tenant reports getting 'knocked off' of their wifi router and having to shut off/on wifi on phone and/or power cycle the router to get it working again.

Agreed that what I'm planning is overkill. There is an element of learning/fun to all this.

There is a shared building internet that s terminated in the unit with ethernet. This upstream connection may be the problem, or part of the problem.

There is also a building wide wifi service but it is reported to perform even worse. Could it be that this system is actively sending dissociations to tenant owned wifi router?

You are looking for WIDS/WIPS.
For "knockedoff" - enable 11w aka pmf "optional" to keep modern (2020+) devices safe.

Your speculation is true, happened before, some hotels paid fines for that.

A while ago I performed a survey with https://github.com/jantman/python-wifi-survey-heatmap for comparing OEM vs Openwrt firmware - steps involved were

  • install software
  • draw floor plan
  • record measurements at several spots (7) on ch1/36/100 (each measurement took ~60secs)
  • generate heat maps

Like you may have guessed this is quite time consuming ...
(*for results check EAP615-Wall; OpenWrt or Stock? - #10 by ed8)
Other Survey apps that you already found are likely less time consuming. An alternative in case of a supported Vacuum Robot is https://github.com/ccoors/Valeronoi


Possibly useful: for long-term observability you can also use monitoring apps:

ps: the setting which brada4 mentioned can be found here:

1 Like