[time-nuts] Chronometer contest sponsored by IEEE Spectrum

Jeff Mock jeff at mock.com
Thu Nov 29 22:05:42 EST 2007

>> I think that I would start by looking at 32kHz watch crystals, I've 
>> often wondered how good a timebase you can make out of one.  The tempco 
>> is a parabola around 25C with a max slope of something like 0.05 PPM/C, 
>> so they are naturally a pretty good timebase with good aging 
>> characteristics. The crystals are really tiny,  maybe insulating it with 
>> a material that has an interesting heat of fusion along with a micro to 
>> model the physics of the parabolic shape of the crystal performance.
>> jeff
> Maybe an ensemble of watch crystal clocks and a PIC microprocessor  per 
> David Allan paper
> of some years ago. After testing you could assign deferent weighting to 
> the different clocks.
> Bill K7NOM

Yes, an ensemble of 32kHz crystals, I like it a lot.  Perhaps you can 
setup a temperature gradient from hot to cold across a number of 
crystals.  If the tempco for the crystal is really a parabola, each of 
the crystals can be a datapoint along the parabola of crystal 
performance.  With a little work you might be able to precisely 
calculate the maxima of the parabola and precisely set the frequency 
independent of temp and aging.  Ha!


More information about the time-nuts mailing list