26 April - Real World GPS Test #4

Owned by Diarmid Mackenzie
Last updated: Apr 26, 2020
3 min read

Test seems to show that WiFi contributes to GPS instability. But test incomplete because it was ruined by complete lock up of GPS logging for 8 hours!

On my phone a (Motorola G7) I had seen significant noise in GPS measurements when indoors, but not so much outdoors.

I had a suspicion that my indoor WiFi might be a cause.

So at about 11pm local time (21:00 UTC) I put my phone down, in a static location, with WiFi disabled. About 8 hours later, I then enabled WiFI, and left it for another couple of hours.

Unfortuantely, just 2 hours after the start of the experiment, GPS logging stopped completely for 7h58. Enablign WiFi on the phone at 7am local time (05:00 UTC) did not trigger it back into life. It started again just after 9am. Unfortunately I don’t know for sure whether I handled the phone then or not. I know I started the app to send in ocation data at 9:40 local time = 07:40 UTC, but by that time t had been logging again for ~30 mins.

For the 2 hours of data that we do have, location data looks to have been rock solid. That is in strong contrast to the data we had from yesterday when the phone was in a static indoor location with WiFi enabled, and the data jumped about.

The later data (points 23 onwards) is from the morning (after the 8h lock up at point 22), when WiFi was anebaled again. We see much more variability there, althoguh since this was after the end of my experiment, I don’t have detailed records, and may have been moving around with the phone at that time.

I will perform anotehr static test with WiFi enabled again later, for comparison/

 

Key learnings

  • Wi-Fi does appear to have an impact on the noisiness of GPS data. WIth WiFi off, a stable phone appears to have a stable GPS signal. Since we cannot ask our users to disable Wi-Fi, I don’t know whether this is an actionable insight. Before taking any action based on this, we should run more tests, in a variety of situations, and with a number of different phones, to further test this hypothesis.

  • GPS logging lock-ups continue to be a major problem, and can occur even in the most benign conditions (unhandled phone, stationary, with WiFi disabled.

GPS logging lock-ups are both a major problem for production deployment of the app, and also because of the impact they have on any further experiments in this space.

GitHub issue #454 already covers the fact that logging cadence varies a lot, but evidence from this testng & other recent testign shows that we must now prioritize this much higher, as it is a critical defect, and impeding GPS testing.