nRF52833 LFCLK Accuracy

[aweng]

Hello,

I'm trying to measure the accuracy of the internal low frequency clock on the nRF52833 for hardware verification. I've set it up so that an RTC generates a tick event which is then connected to a PPI which is then connected to a GPIOTE task out to toggle the GPIO pin polarity. The setup is similar to the one mentioned here but I am OK with the 16384 Hz output:

 How to output 32768Hz external crystal LFCLK at a GPIO pin? 

The GPIO pin is then connected to a universal frequency counter to measure the frequency. The datasheet and sdk_config.h settings specify a 500 ppm accuracy, but all of my measurements have exceeded that value and go up to 10,000 ppm (i.e. measuring 16546 Hz).

Is there some inherent inaccuracy with using this type of low frequency clock output connection? I have an external high frequency oscillator and using a similar setup I am able to output very accurately up to 20 ppm. Only the low frequency output sees such a large discrepancy.

I'm also using soft device s140 7.0.1.

Thanks.

[Kazi Afroza Sultana]

Hello,

This inaccuracy might be related to the calibration of 32.768 Hz RC oscillator. Have you calibrated the crystal before measurement? In other way, you can try to synthesize the LFCLK from the HFCLK source to get the proper accuracy? Here is the link you can look at: https://infocenter.nordicsemi.com/topic/ps_nrf52833/clock.html?cp=4_1_0_4_3_1_4#concept_zhp_4vy_2q. 

Thanks.

Best Regards,

Kazi Afroza Sultana

[aweng]

This is the correct answer where the LFCLK needed to be calibrated first. For what it's worth, my hardware verification needed to check the accuracy of the RC LFCLK itself so using the HFCLK to synthesize wouldn't be testing the RC LFCLK. But calibrating it did the trick, thanks!

[Kazi Afroza Sultana]

You are welcome. Great that you have solved the problem.