Start with the same conditions
In the evening, more devices may be using your network, neighboring wireless networks, and your provider's equipment. These reasons cannot be separated based on the feeling “after seven it’s worse.” We need a small table of observations, not a dozen settings changed at the same time.
Test latency on one device in one room. Before taking measurements, write down the time, Wi-Fi range and active downloads. On the first pass, do not reboot the router: it is useful to first save the manifestation of the problem.
Ping and jitter test
FreeLoading the check...
Make four observations
| Condition | What to write down |
|---|---|
| During the day, Wi-Fi | Median, p95, trials without response |
| Evening, Wi-Fi | Same indicators and activity of other devices |
| In the evening, cable | Has behavior changed without a radio channel? |
| In the evening, next to the router | Do walls and distance have an effect? |
p95 helps you see the high latencies that are hidden behind the usual median. However, a small set of samples cannot be considered reliable statistics for the entire evening. If the failure is rare, use long magazine.
If cable makes it better
Check the position of the router: it should not be covered with a metal case or hidden near the floor behind furniture. Compare the available ranges in the same place. A higher frequency does not automatically mean a better signal across multiple walls, so choose based on your workroom measurement.
See if the problematic device has connected to a distant point in the mesh system. The network name may be the same, but the radio path may be different. If you change the channel manually, first save the current settings and repeat the test after the change. Randomly changing the channel without comparison can only transfer the problem.
If it's equally bad via cable
Pause large transfers on your devices. It is especially useful to check the outgoing flow: the backup can work unnoticed. Run load lag checkto reproduce a similar situation in a controlled manner.
If your own network is free, and the problem repeats in the evening on several devices and via cable, send a comparison to your provider. Check the time and time zone, not just the maximum number on the chart. Our server can also be a limiting link, so independent measurement according to support instructions is useful.
What such a test will not establish
The browser does not see the occupancy of all neighboring radio channels and does not know the load on the provider's switch. It observes the response of a particular node. The result helps you choose the next step, but does not allow you to automatically blame your neighbors, declare the router faulty, or prove a violation of the terms of the tariff.