Calculating ppm offset between the 32khz and the 16mhz clocks

[Omri Friedman]

I am trying to build a run time test routine that will return the ppm offset between the 32khz and the 16mhz clocks.

Can i use PPI to hook 10 clocks 32khz ticks to capture the value of a 16mhz time?

Can you instrurct me how to do that?

I need to get 20ppm resolution, so i need to count 16mhz ticks in between 10 32khz ticks.

[Omri Friedman]

I have implemented the ppm test code mentioned above, but i get offset of ~4000ppm between the 32768hz and the 16mhz  clocks. 

I read this http://infocenter.nordicsemi.com/index.jsp?topic=%2Fcom.nordic.infocenter.nrf52832.ps.v1.1%2Frtc.html and wondered if that has to do with my results.

I saw in the link 15 to 46 μs delay on task START and +/- 15 μs on "START to COUNTER increment" and +/- 62.5 ns on "COMPARE to COMPARE".

I start the RTC and use CC[0] and CC[1] to trigger capture of TIMER2 (via PPI). then i use the delta to calculate ppm offset.

Which of the above jitters/delays should i expect? just the +/- 62.5 ns?

[Einar Thorsrud]

The way I read your code, you should only be affected by the compare-compare jitter as the timer is started and stopped by compare events via PPI. You could perhaps let the test run for a longer time to get a more exact average.

If the accuracy you have measured is correct, then there could be an issue with your LF oscillator. Are you using a crystal or the internal RC? If you use the internal RC, then  you should calibrate it regularly. If you are using a crystal, then you need to make sure that you have proper load capacitors.