> >> One suggestion I have for keeping an accurate clock is to add a
> >> temperature sensor and an EEPROM to your PIC, measure the
> >> oscillator's error and temperature coefficient, and store it as
> >> a table in EEPROM.  This could be used in conjunction with
> >> services such as ACTS (dialup) to allow the service to be
> >> called less frequently and save long-distance phone charges.
> >> Or it could allow WWVB or other VLF signals to be used in fringe
> >> areas where reception is only good at certain times of day.

> >Neat idea. I'd not really thought about call costs at this stage - no
> >doubt the client will! My only concern is variance between individual
> >RTCs.  If one set of cal data would do for a production run then
fine,
> >but it isn't going to be economical to do individual cals on these
> >units unfortunately. I'll have to get a handful of RTC chips and
> >experiment with this.

> Another thought I had was to measure the average and max and min
> temperatures inbetween time measurements (calls to ACTS or powering up
the
> WWVB receiver) and if they are not too spread out (temperature didn't
vary
> too much) then put an entry into the EEPROM for the average
temperature and
> measured oscillator error.  Thus it would start out with poor
performance
> and through use it would automatically learn the characteristics of
the
> oscillator and improve its performance.

Excellent idea! An intelligent oscillator, I like it very much.

This may be the excuse I needed to invest in the fuzzy logic kit...


Many thanks for the input.  I'll post a summary to the list when the
project is done, and put up some pages with more detail too.



Cheers,

Andy.








.