GPS fix times and possible memory leak

[Tron]

We have a custom board with nrf9160, ncs 1.2.0 and modem fw 1.1.1. We experience strange behaviour with the GPS: As an example,on units with 30 min fix intervals, we experience the following with units in the field: 

1. In the beginning, the unit has fix times of roughly 2-5 secs, with the occational 30 sec fix time
2. When the unit runs and have sent roughly 500 messages, the fix times are increasing, and often passes 1 min
3. After roughly 1500 messges, we see that the GPS takes a fix every second try, and when it takes a fix, the fix times are longer than a minute.
4. When the user resets the unit, we are back to point (square) 1. 

For us, this looks like a memory leak in the GPS modem fw. 

Yours, 

Tron

[Øyvind]

HI Tron, 

I'm on the case and will need to discuss the matter with our R&D team. I will get back to you asap!

Kind regards,
Øyvind

[Tron]

Hi

Modem trace is unfortunately out of the question. The units are on sheep in rural areas of Norway. 

I log the GPS Fix times to the server, or to be more precise, every second second I increase a fix counter, which is transmitted to the server every time the unit tries a GPS fix. I have a timeout on the GPS on 2m30s, so every unsuccessful GPS try gives a value of 75, ie 150 seconds. THis is the only logging, together with the GPS positions, of course. 

Resets are also logged. 

Are these of interest?

[Øyvind]

Tron said:

How long do I have to log for any leaks to be visible?

 I guess this would be as you described in the OP:

1. In the beginning, the unit has fix times of roughly 2-5 secs, with the occational 30 sec fix time
2. When the unit runs and have sent roughly 500 messages, the fix times are increasing, and often passes 1 min
3. After roughly 1500 messges, we see that the GPS takes a fix every second try, and when it takes a fix, the fix times are longer than a minute.
4. When the user resets the unit, we are back to point (square) 1. 

 How long does it take before the user is at point 4?

[Tron]

100 mesages a day, 15 days roughly

[Øyvind]

Then I'm afraid that is how long the modem trace must run in order to see the issue.

Will it be possible for you to run that long on your side?

-Øyvind

[Tron]

Hi again

I will have  to make it run as long as needed. 

In order to run such a long test, I would like to test the modem trace first, if possible. This is to verify that the modem trace is readable, looking at my recent problems with the modem trace. Is it OK to submit a trace for validation that the trace is working?

Tron

[Øyvind]

Yes, of course! Please send a trace for verification.

-Øyvind