Thank you Jinx. I didn't see this until now, because for some reasons it
scored as spam in Eudora.

Your clock is very precise. But the board is done and I don't feel like
redoing it. Anyway, I am not using the Timer1 counter as the clock itself,
rather, I add a 64bit value (typically 0x0000000100000000 if the 32768
source was perfect) each second (i.e. each 32768 xtal ticks). Later, I
will be able to change that 0x0000000100000000 to a slightly more or less,
to compensate for the non-perfectly-32768Hz frequency of the xtal. Adding
temperature compensation would turn it into an almost atomic clock ;) but
I don't have the time for that now, maybe for the next board.

Cheers,
Mario


At 20.52 2008.05.21, you wrote:
>> Maybe it's important to add that it's a dsPIC (30F5011 to be exact)
>> and that I want it to be precise
>
>Mario, if you want short- and long-term accuracy, may I suggest
>the attached circuit. You can get 1 minute per year stability. By
>my reckoning that's about 1 part in 12,000,000, or keeping a 32k
>crystal stable to 0.0026Hz
>
>In a back-up situation with a 12V SLA, I use 150k/22k as a voltage
>divider to get ~1.7V for the clock driver and a 5V LDO to power the
>CMOS and a 12F675 time-keeper. Shown is part of the circuit for
>a battery-powered remote logger

 

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist