# ATmega XTAL input capacitance and crystal load capacitance

I need precise low-ppm crystal for my circuit with 8-bit ATMEGA (AT90USB1286). Currently 18pF TXC HC-49 (11.5x5mm) is used. I need also minimize size to reduce potential EMC and PCB. TXC - 7M/B (5x3.2/3.2x2.5nn) seems as potential alternative. But they are 10pF crystals.

Atmega datasheets do not mention XTAL pin capacitance, there is general Capacitance for each I/O Pin as high as 10pF (with exclamation that is not accurate and not dependend on socket TQFP/QFN). Does 10pF value relate also for XTAL pins? And does it implies that the low load capacity crystal cannot be used because of formula $C_x = 2 \times (C_{xtal}-C_{pin})-C_{stray}$? SMD 18pF alternative e.g. Abracon ABM3B (5x3.2mm).

BTW: Does make difference if the crystal has GND pin (4 pins or metal case) or is in 2 pin ceramic socket e.g. Abracon ABM7 (18pF). AFAIR strong EMC recommendation is to connect metal case with GND.

## 1 Answer

Crystal pins are not I/O pins, and 10pF is a guaranteed maximum not a nominal capacitance.

Usually 5pF will suffice for a crystal pin plus stray capacitance with reasonable layout (as short traces as reasonable, possibly over a ground plane). However, the AT90USB1286 datasheet does not recommend load capacitors of less than 12pF so 10pF total would seem to be marginal (you'd get 11pF if the input+stray was 5pF).

If you don't care about the exact frequency, 12pF or even a bit higher would probably be okay with a "10pF crystal", it would just shift the nominal oscillation frequency very slightly. Otherwise you can pick a different crystal.

• I try to achieve accurate (16MHz) clock, i.e. use low-ppm (10ppm) crystals. Maybe capacitance non-matching adds more error than I gain with more accurate crystal, i.e. answer is avoiding 10pF crystal. You're right, recommendation is 12-22pF and probably both conditions should be satisfied.
– TMa
Sep 1, 2014 at 15:44
• Keep in mind that you're not going to be able to predict the capacitance to better than a couple of pF, but choosing to go higher will give a systemic shift. Sep 1, 2014 at 15:45
• P.S. If you're really interested in a guaranteed accurate frequency, I would suggest considering a packaged oscillator rather than a crystal. Sep 1, 2014 at 16:09