Kinetis K65 RTC crystal selection

[GordyCarlson]

My customer is having issues with the Kinetis K65 RTC resetting itself to 1987 (default values) when they do environmental testing on their end product prototypes for several hours at -40C.  We initially reviewed the VBat supply integrity, and it appears OK.  Our attention is now on the external RTC crystal + circuit he uses. Perhaps it is having oscillation issues at the low temp.

We have reviewed the K65 data sheet and it's crystal requirements/specifications.  And would like to run this by you for additional comment...

See circuit and attached crystal data sheet used for the 32KHz RTC oscillator, EPSON PN MC-306-32.7680K-A0:ROHS.

The crystal is designed for a 12.5pF load capacitance.

Presently we have (2) 18pF caps as shown in schematic for a load capacitance (CL) of 9pF plus any stray capacitance (Cs). (CL = (C42*C41)/(C42+C41) + Cs).

The K66 data sheet lists Cpara = 5pF typical, where Cpara is the parasitical capacitance of XTAL32 and EXTAL32.

Would I use 5pF for Cs, resulting in CL = 9pF + 5pF = 13pF or would it be the actual load capacitance be calculated differently?

I wondering if I should considered adding 2pF using K66 internal capacitance registers or leave as is based on Cpara value.

-Gordy Carlson

Freescale FAE

Upstate NY

[egoodii]

I would remove your external R and Cs in this circuit -- the R is internal, and you should use the internal Cs based entirely on loading to 'nominal accuracy' at 25C.  For my 12.5pF XTAL ABS07-32.768KHZ-1-T (+/-10ppm tol), I chose 20pf (16+4 selections) per pin, but I have virtually NO other PCB parasitics.  Together these hold me to about +/-25ppm overall accuracy (+/-2s/day) over 0 to 50C.  I test accuracy by synchronizing to a 'timeanddate' web site, then checking it the next day.

These changes may keep the oscillator running at -40C, but I would also watch for condensation in that test -- if such is a concern within the product, then you may need to coat this extremely high-impedance area.

[egoodii]

By the way, if you are expecting any significant ACCURACY over extended temperature ranges, like If one is TRULY trying to hit 'about 1s/day' (+/-11.6ppm), IMHO the cleanest option is an external 'compensated oscillator' from SiTime: SIT1552AC-JE-DCC-32.768D.  Takes a little more battery, but insures +/-10ppm 0-70C (extended temp versions available -40+85C, holds +/-13ppm).  Also AVOIDS these very issues of high-impedance/low-signal-level connections on the PCB.

[GordyCarlson]

Thanks for the comments Earl,  I'm going to review these suggestions and see if they resolve or improve the issues with RTC operation at -40c. 

Gordy Carlson