The Wi-Fi plan predicts, the measurement proves. The two rarely line up exactly, and that is not a fault in the plan: a drawing does not know about a metal filing cabinet, a neighbour who put up a new access point last week, or a door that can be closed.
Seeing who else is transmitting
RouterOS can scan with its own radio. Which command you use depends on the Wi-Fi package on the device. With the new wifi package:
/interface wifi scan wifi1 duration=10With the old wireless package:
/interface wireless scan wlan1 duration=10Scanning takes the radio off its channel. Clients on it lose their connection for the duration. So do not do this on the access point you are connected to, and on a live site preferably outside office hours.
What you get back is a list of SSIDs, channels and signal strengths. What matters is not the number of networks but how they spread over the channels. Three neighbours on channel 1 and nobody on 11 is useful information; fifteen networks at -85 dBm are background.
For how busy a channel really is, occupancy matters more than the number of names. On the old package, /interface wireless frequency-monitor wlan1 gives the load per frequency. A channel with two weak neighbours transmitting all day is busier than one with six strong neighbours doing nothing.
Signal and noise
For clients that are already connected, look at the registration table:
/interface wifi registration-table print
/interface wireless registration-table printPer client you see the signal strength and the rate it is talking at. That second number matters more in practice than the first: a client at a high rate has a good link, whatever the dBm says.
The numbers are negative and closer to zero is better. A rough guide:
| Signal | What to expect from it |
|---|---|
| -50 dBm and above | Excellent. You are standing next to the access point. |
| -60 to -50 dBm | Good. Here you get what the hardware was built for. |
| -70 to -60 dBm | Usable. Video works, throughput drops. |
| -75 to -70 dBm | The edge. It works, and it fails the moment a door closes. |
| below -80 dBm | Do not count on it. |
The noise floor is normally around -90 to -95 dBm. The gap between signal and noise is what counts: roughly 25 dB or more gives you a connection worth having. A signal of -65 dBm with a noise floor of -80 is worse than -70 dBm with a noise floor of -95, even though the first number looks better.
If the noise floor is high, something nearby is transmitting that is not Wi-Fi. On 2.4 GHz the usual suspects are a microwave, a wireless camera, bluetooth equipment and a bad power supply.
Walking the building
One measurement at the access point tells you nothing. What you want to know is how it is where people sit, and that means walking.
- Walk the rooms with a phone or laptop and note the signal per spot and, more importantly, whether a video call holds up.
- Mind the edges: the furthest desk, the meeting room behind the concrete wall, the stairwell.
- Close doors that are normally closed. With every door open you are measuring a different building.
- Measure once when it is empty and once when it is full. People are mostly water, and water stops 5 GHz.
- Walk the route someone actually walks and see whether your device moves to the next access point. Roaming is the client's decision, so this is the only thing that tells you.
Use one device throughout. Two phones from different makers give different numbers in the same spot, and then you are comparing measuring equipment rather than the network.
Next to the Wi-Fi plan
In the Wi-Fi plan you drew walls, placed access points and watched the coverage colour in. Put your measurement beside it and you get three kinds of outcome.
- It matches. Then you also know your plan for the next floor can be trusted.
- The measurement is better than the plan. Usually a wall set heavier than it is in reality. Adjust the plan, or next time you hang one access point too many.
- The measurement is worse than the plan. This is where the work is. Look at what is there in reality and not on the drawing: a lift shaft, pipework, a mirror, film on the glass, a steel rack. Add it as an obstacle and see whether the plan lines up then.
The plan cannot know your neighbours. Your scan does. If the channel you picked turns out to be full, adjust it in Wi-Fi channels: the fields frequency 2.4 GHz and frequency 5 GHz are empty by default, which means automatic. On 2.4 GHz you stay on 1, 6 or 11 (2412, 2437 or 2462) and leave the width at 20 MHz; anything wider overlaps your neighbours.
On 5 GHz there is more room, but the DFS channels can move you without warning when radar is detected. If an access point changes channel by itself, read the log before you fix something that is not broken.
The TX power field in Wi-Fi settings is empty by default, which means the maximum allowed for the country you pick. Transmitting harder is not always better: your access point reaches the client, but the client with its small antenna does not reach back, and you add noise for your own neighbouring access point. With several access points close together, turning the power down is often the improvement.
What the tool cannot do here
The configurator does not scan. It runs in your browser and has no radio and no connection to your hardware. The Wi-Fi plan is a simulation of what you draw, and the checks about Wi-Fi look at your configuration, not at the air. Everything in this chapter is measured on the device itself or with an app on your phone.
Read on: Wi-Fi channels, The Wi-Fi plan and Checks about Wi-Fi.