I am designing with nrf52832 and experiencing some BLE connection issues at 0C. The system gets disconnected every few seconds with 32k crystal, but no issue with internal RC oscillator, so the issue should be related to the crystal. I confirmed the crystal meets the datasheet specs and changing the caps help. What's the best way or tool to validate the capacitor value? Right now I just changed the cap and put the system at different temp to check the connections for 30+ minutes (including soak time). Any other less time-consuming methods, e.g. toggling GPIO and measuring the frequency drift?

  • \$\begingroup\$ A 10MHz calibrated frequency counter or another same Fosc to mix that is calibrated to <1ppm. Read xtal load cap specs and double value for each side minus input cap of CMOS and use NPO for 0+/-50 ppm cap temp stability. Trim with varicap or varactor DC pot biased cap or easier, buy a 1ppm TXCO chip or.... Make a lab reference for future measurement beat frequency with programmable PLL from OCXO to mix 32khz * 30 cycles within 1ppm drift accuracy in 1ms. so many solutions. Even faster is an HP Time Interval microwave clocked counter in 1 cycle \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Dec 1 '16 at 23:53

From the nRF52832 data sheet: enter image description here

The easiest is to pulse or toggle a GPIO pin on every TICK interrupt and measure the frequency on your oscilloscope, or even better, compare with a known accurate frequency source. Also, modern PC sound cards can sample 32 kHz, and they are surprisingly accurate, especially when calibrated.

Alternatively, write out the RTC value over the serial port or Bluetooth. Capture the RTC values overnight using a terminal program that has timestamping capability (e.g. Tera Term). After a good night's sleep, your measurement error should be in the 5 ppm region. Don't forget your PC clock drifts too.

Finally, make absolutely sure that you are using two NP0 capacitors on your crystal, and that they conform to the crystal manufacturer's specified value, taking into account the parasitic capacitances of pins and traces.

| improve this answer | |
  • \$\begingroup\$ Remember those USB GPS modules we used with our laptops for navigation before having maps on our phones? If you still have one around, their NMEA strings are often precisely timed to 1s intervals. Or get one on eBay. \$\endgroup\$ – neonzeon Dec 2 '16 at 2:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.