timestamp problem

RE: timestamp problem

alan11 — Thu, 13 Apr 2023 01:41:32 GMT

Hi Emil Lenngren

I have the same problem, and it sounds like a great tip.

1.Can you provide simple sample code? For RTC overflow calibration problem

2.How to detect the accuracy of external LFXO crystal oscillator?

Best wishes

RE: timestamp problem

Victor A — Tue, 11 Apr 2023 06:38:19 GMT

If, for example, diff_tick contains ~1.001007 seconds, 1001 ms will be added on the last time,1007 microseconds will be saved and added the next time.

This is the result of RTC COUNTER.That means it actually ran 1.001007s,If discard it, that will lose 1007us

RE: timestamp problem

Emil Lenngren — Mon, 10 Apr 2023 06:44:38 GMT

Because your calculations are wrong. On the last line you add an integer number of milliseconds that have passed since last time. To compensate for rounding errors, the remain_tick variable contains a remainder that is added the next time. But it's an incorrect remainder. If, for example, diff_tick contains ~1.001007 seconds, 1001 ms will be added on the last time, but ~1007 microseconds will be saved and added the next time, resulting in exactly a one millisecond extra drift.

Using an app timer that performs cpu work like this every second is not really battery efficient anyway. Instead, whenever you want to retrieve the time, you can have a function that gets the time directly from the COUNTER register of the RTC, correctly adjusted with how many RTC overflows have happened, plus the synchronisation value that converts it to a Unix timestamp.

Also make sure you configure the project to actually use the LFXO and not the internal oscillator.