How accurate to you want it to be? What is your acceptable drift. Currently you have about 0.017% error.

Firstly, how do you know it's not your software accumulating errors due to interrupt latency etc?
To prove it is the crystal, try another brand crystal and another processor. Is the error the same? Put some load capacitors on and do the test, remove the capacitors and do the test. Any variation? How accurate is your measurement device.


Crystals are normally specified for a ppm error with a specified load capacitance. What are the specs on your crystal? Whilst you're looking at the specs on the crystal - look at the specs on ageing and temperature and calculate how that is going to affect your timekeeping. No matter what you do short of synchronising your device to a atomic clock are you going to have near to zero drift. It all comes down to how much drift is acceptable and how much you're willing to pay.

If you could adjust the frequency of the crystal, what would you use to measure it? How accurate would your measurement device be? But you still have drift due to temp and ageing.

In answer to your question: Yes, the load capacitors on the crystal do affect the frequency. How much? Depends on the crystal. There is a lot of info regarding these issues on the web. Best be doing some searching.


The other question is: how are you measuring the time?