timer drift

[Tong Yu]

Hey guys,

I followed this guide and successfully synchronized the timers from two devices(nRF52 DK). But if I let these two synchronized devices keep running all the time, I guess sooner or later they will be non-synchronized again.

I want to estimate how much time the timer drift after a certain time.

In the guide, the NRF_TIMER2 was the target timer to be synchronized. What I thought is this timer is based on (I am not sure about that) the 64 MHz crystal oscillator (HFXO), and I saw the information on infocenter, the 64 MHz crystal oscillator's ppm is +-60. So I did such calculation, for one month, the time drift could be:
(1/64,000,000-1/(64,000,000*(1+0.000,06 ))60seconds60minutes24hours30days = 2.43 * 10^(-6)
Am I right? What else can affect the timer drift? Could somebody give me some hints?

Many thanks.

[Sigurd]

The NRF_TIMER2 is based on the HFCLK (16 MHz/ PCLK16M) and have a 62.5 ns resolution. The HFCLK is based on the 64 MHz crystal oscillator (HFXO), which is controlled by the 32 MHz external crystal. See the drawing here.

The 8ppm is the additional HFLCK Frequency tolerance needed when using synthesized clock (not recommended). The HFXO Frequency tolerance is 60ppm, this means that you need a 32 MHz external crystal with at least 60ppm for the radio peripheral to work. On the nRF52-DK the 32 MHz external crystal is 10ppm.

[Tong Yu]

OK，thank you very much for the information!