We are currently porting our proprietary stack from the nRF51 to the nRF52. To save on BOM we are using an external 20ppm 32Mhz crystal to synthesize a 32Khz clock (power consumption is not an issue in this application). Clock configuration is as folows:
NRF_CLOCK->LFCLKSRC = CLOCK_LFCLKSRC_SRC_Synth;
We observe quite large clock drift and wonder which level accuracy is obtainable with the Synth clock source. The datasheet states that:
LFCLK can also be synthesized from the HFCLK clock source. The accuracy of LFCLK will then be the accuracy of the HFCLK.
However this is not what we are observing when we look at the radio behaviour. Before embarking on an extensive measurement campain we would like to know if there are any caveats associated with using the Synth source. Or that differences in interrupt latency or radio implementation (we use nRF52 compatability mode) between the nRF51 and nRF52 are the source of our problems.
