32.768KHz crystal calibration

[yingchc]

I use nRF52832 with 32.768KHz crystal, is that possible to make the frequency error smaller than 5 ppm by calibration or doing something?

[Andreas]

Hi,

 

On nRF52 series there is no calibration routine or similar to do this. The base frequency of the crystal is determined by the capacitive loading of the crystal. In theory the crystal load capacitors (C1 and C2) should be C1 = C2 = 2*C_L - C_pin - C_pcb, where C_L is the specified load capacitance for the crystal, C_pin is the nRF52 Pin capacitance, and C_pcb is the pcb trace/pad capacitance. C_pin is 3-4 pF, depending on RF52-variant and C_pcb is ~1pF.

We still recommend that you measure the LFCLK frequency and tune the load caps as needed to adjust the crystal frequency as needed. Probing the crystal directly will load it, so it is better to measure the frequency by toggling a GPIO on RTC tick events, using GPIOTE trough PPI.

The same theory as outlined above applies to the 32 MHz crystal, the only difference is we recommend you measure the frequency accuracy by measuring the frequency accuracy of an unmodulated RF carrier.

 

nRF5340 however has tunable capacitors onboard, meaning such calibration e.g. in production could be made possible.

 

Best regards,

Andreas

[yingchc]

Hi Andreas,

Thanks for the answer, is 5 ppm for 32.768KHz achievable?

Eric

[Andreas]

Hi Eric,

 

It should be achievable to get the base frequency below 5ppm, but the actual frequency will also factor in the frequency tolerance so you would need a <5ppm crystal. Not sure if this is available, it might be necessary to go with a external clock oscillator to get these levels of accuracy, nRF52 allows using external clock signals as source for LFCLK.

 

Best regards,

Andreas