Drive level for the 32kHz crystal too high?

[Quax64]

The drive level of the nRF51822 is specified as 1µW, whereas the EPSON crystal suggested in the BOM for the EK boards PCA10000 and PCA10001 is specified for just 0.5µW (with 1µW being mentioned in brackets plus a note saying "Please contact us if you require 1.0µW max."). Apparently, a too high drive level can affect crystal accuracy, so I wonder whether I should look for a different crystal. I also noticed that NORDIC suggests the crystal version which is optimized for 9pF load capacitors but, according to the schematics, uses 12pF. Is this somehow related, that is, could it be that the crystal can cope with 1µW but higher load capacitors have to be used to compensate for the frequency shift that might be caused by high drive level?

[Asbjørn]

1 uW is maximum according to our product specification. We usually stay with capacitors close to the optimized load, but sometimes we see that we get better accuracy from other values. It depends on the board and the crystal. You can tune it slightly with the capacitor values. This isn't normally a problem as long as you keep within spec.

[Quax64]

Sorry, but have you read my question? The point is that the drive level provided by your product might be too high for the EPSON crystal. My guess was that the crystal can still take it but the nominal frequency would be shifted quite a bit and, therefore, has to be shifted back by load capacitors which are rather high as compared to EPSON's specifications. I am no expert in crystals, so I don't know whether my guess is correct and whether there are hidden downsides of this approach (like higher drift depending on temperature, age, etc.).