[dsavitsk]

Should C8 and C9 be 20pF instead of 33pF to match the crystal's load capacitance?

[hagtech]

No, they should be 40pF. 

But actually, they are!  Once you add in the capacitances of the traces and internal circuitry, they are pretty close to 40pF (just a hair on the shy side).  From the crystal's point of view, it sees these two capacitors in series, hence 20pF total loading.  This is a common mistake 9/10 people make.  Maybe more.  Shoot, even a world-famous DAC designer once thought I had it wrong...

jh

[dsavitsk]

Thanks Jim, that's very interesting.

In the data sheets for some of these crystals, instead of giving a loading capacitance, simply say "series". What does that mean?

[hagtech]

Depends how the xtal is intended to be used.  You can run them in resonance or anti-resonance modes.  Slightly different frequencies.  In series mode you do not load the xtal with a parallel capacitance.  Pierce is parallel mode only, Colpitts might be able to go either way.

You can operate a typical 20pF xtal in series mode, but the frequency will be lower than specified.  The series type are purposely adjusted in cut to compensate, so that in that mode they do hit the target.  Similarly, using a series type in a Pierce circuit will run fast.  The difference?  I forget, maybe something on the order of a few hundred ppm.

jh

[hagtech]

Here, this covers some of it...

http://www.ieee-uffc.org/freqcontrol/quartz/vig/vigtoc.htm

jh