Use 8pF 200ohm 10MHz crystal with ATMega88 w/o caps

atmegacrystalmicrocontroller

I'm trying to save board space. I know crystals typically have 20-30 pF ratings, and need 20-ish pF caps to ground.

I found this crystal: http://www.mouser.com/ProductDetail/AVX-Kyocera/CX3225GB10000D0HEQCC/?qs=sGAEpiMZZMsBj6bBr9Q9aSMajaFx5yeOFWolaA6tkgz4rswCxFq0Sg%3d%3d

It's extra-small (2.5×3.2 mm), has an ESR of 200 ohms (which seems high – is that going to be a problem?) and a load capacitance of 8 pF.

I know that in general, you need caps to ground. But I have also read that there's typically 5-10 pF of parasitic capacitance which should be subtracted from the caps you add. So – does an 8 pF crystal need caps at all?

This is for a real product, so I don't want to cut corners if it will hurt reliability.

EDIT: I found two other people asking the same question; I read the answers they got; I still don't know whether it's correct to leave off the caps.
How do I deal with different XTAL pin capacitance when selecting XTAL and load capacitors and How do I choose the correct capacitors to use with my Crystal?

One of the answers said it would be "easier" to design with a crystal with more capacitance, but didn't say why; other answers say that getting within a factor of 2 is probably sufficient. From this, I deduce that the more capacitance the crystal needs, the less you will be messed up by mis-estimating the stray capacitance. So I should pick a crystal that needs 30 pF, put 25 pF on each leg, and I'll probably be good.

Best Answer

This ESE answer contains a good explanation of what having an incorrect load capacitance could do.

Personally I would never do this, because I don't like coming back to pain-in-the-butt problems when they surface much later down the line - I like to ensure they never happen instead.

From this TI Application report we can estimate around 1-3pF of stray capacitance per inch of trace for a PCB with a power and ground plane, Bear in mind that given the high contribution from fringing fields (as the capacitor's gap is larger than its width) you will get material permittivity contributions from your solder resist as well as the FR4 of the PCB. Hence if you are relying on a few stray pF to get you through, a different resist process might have a relatively large impact on this. As you note in your question, this is why crystals with a higher loading requirement are easier to design with - a larger percentage of the load capacitance is in the tightly-controlled precision component you paid for, and less in the vagaries of the PCB production process.

If space is an issue for you, have you considered either using an internal oscillator source or a ceramic resonator (which is a one-component solution) as an alternative? Neither will have the same accuracy or stability as a crystal, but you may find you don't actually need that timing accuracy in your application.