Franc Urbanc wrote:
-------------------------------
Install into system clock oscillator one variable capacitor from 5-22pF or so.
------------------------
Or if you prefer software solution...
Just make a simple uC clock and synchronize it with a reliable plain clock. Leave it running for a couple of days, the longer the better. Note the total number of seconds the thing has been running for and the difference it made.
Let's say it gained 20s/day (not unusual!). This means that every (3600*24)/20=4320 seconds you have to drop one second from the system clock. Remember that the time unit is not a second but an interrupt interval - much less than a second - the clock runs without any noticeable compensation. Don't round X/Y if it is fractional. It means that every X interrupts you have to drop not one but Y of them.
This way 0.2s/day is easily obtainable and even that is far beyond the long term stability of ordinary crystals.