Hi,
I'm having issues with the accuracy of the 32KHz crystal. In summary I can't reliably get a 20ppm crystal to give me a 20PPM frequency.
We’re measuring multiple ways:
1. Using a frequency counter with 1ppm accuracy (but the probe loads the XTAL some, so I recorded the frequency from the dev board and tried to match that measurement with my board – what we ended up with for this crystal (http://www.digikey.com/product-detail/en/9HT10-32.768KDZF-T/887-1504-1-ND/2690372) was 22pF rather than 12pF to tune the frequency to match the dev board.
2. We also measured the timing of the output pulses using a high-speed oscilloscope before and after the capacitor change to validate that the frequency of the pulses was actually adjusted to be more accurate – and it seemed to validate our results on the frequency counter.
So effectively we’ve tuned the XTAL on one of our boards to be the correct frequency, so we have two open items:
1. PCB #1 uses the XTAL in the link above. For now, we seem to have a useable fix by installing 22pF capacitors, but we need to understand why there is a 100ppm frequency offset from expected when using the recommended capacitors before we go into higher volume production.
2. For PCB #2, which uses the smaller XTAL (http://www.digikey.com/product-detail/en/ABS05-32.768KHZ-T/535-11899-1-ND/3508067), we were not able to successfully tune the 32kHz frequency to within 20ppm by changing the capacitance on the board (the higher capacitance dampened the oscillations before reaching the desired frequency), so we need another component or solution with a similar footprint (there is no extra room on the layout to place the larger XTAL components).
There isn’t much to the schematic for the XTAL, but I have attached it.
Attached is also a snippet from our frequency counter test data from testing PCB #1. The “Target” frequency is measured on the CC2650EVM using the same frequency counter and probe:
We’ve tried at least 3 XTALs on our boards and have had issues with all three, so I’m really looking for a root cause here. Are we violating some spec for operational mode, capacitance, equivalent series resistance, etc.? Can you comment on XTAL specs or if you 
have an approved list of part numbers, that would be helpful. We may order more parts to test, but it would save us a lot of time if our efforts were directed to rule out a suspected cause rather than random part selection … and I really have no clue what the root cause is.
