32.768 kHz XTAL Accuracy

[eiadj]

Hi,

We are in the HW verification stage of a product with an nRF52 series MCU. We were easily able to measure the 32MHz XTAL accuracy using the buffered output feature; however, we are having a hard time doing the same for the 32.768 kHz crystal. There doesn't appear to be a buffered output option; we've tried using a RTC event with a prescaler (400) but the accuracy we measure is very high (5000ppm). An error of that high makes me think that we are not getting an accurate representation of the crystal accuracy.

Our questions are:

• Is it plausible to see an error as high as 5000ppm on the crystals? Is accuracy of the 32.768 kHz clock a common issue?
• Does Nordic have any recommended process or technique for measuring the 32.768 kHz crystal accuracy?

Thanks,

Eiad Jandali

[Kaja]

Hi Eiad, 

measuring the accuracy for the 32.768 kHz crystal is usually not necessary, it's usually enough calculating the load capacitors based on the crystal load capacitance, CL, and using the crystals accuracy. 

Choosing load capacitors: Ccap = CL*2 - 4 pF. 

 There will always be a couple ppm inaccuracy, and the accuracy spec of the crystal will come on top of this. It should be fairly easy to experiment with with the configuration. Also, you might want to look into how accurate the 32.768 crystal actually is and take that into account.

Do you generate an interrupt form the event? The measurement is probably not accurate, you could take out the signal via PPI/GPIOTE, and use a oscilloscope.

Best regards,
Kaja

[Kaja]

Hi Eiad, 

measuring the accuracy for the 32.768 kHz crystal is usually not necessary, it's usually enough calculating the load capacitors based on the crystal load capacitance, CL, and using the crystals accuracy. 

Choosing load capacitors: Ccap = CL*2 - 4 pF. 

 There will always be a couple ppm inaccuracy, and the accuracy spec of the crystal will come on top of this. It should be fairly easy to experiment with with the configuration. Also, you might want to look into how accurate the 32.768 crystal actually is and take that into account.

Do you generate an interrupt form the event? The measurement is probably not accurate, you could take out the signal via PPI/GPIOTE, and use a oscilloscope.

Best regards,
Kaja