[time-nuts] Casio Watches 13 Year Drift in Seattle

[Bob Camp]

Hi

Even back in the 1970’s we used computing counters to read the 32 KHz. You
get 7 digits in about 10 ms these days. It will take you longer to power up the 
module and stabilize it than the frequency reading takes.

The semiconductor process for the watch chips has always been a bit odd. They 
never need anything faster than 32 KHz for a clock. They are after *very* low 
leakage. Back in the old days having it run at 1 V was considered strange. It’s 
pretty common today. 

They use the same tricks to dump an analog function into a digital process as the 
MCU people. It’s not a *great* analog sub system, but it’s good enough for the purpose. 
They may even calibrate the temperature and voltage when the set the frequency. That
would let them get away with a *lot* of slop on those sub systems. 

If they digitize the temperature range from 0 to 50 C, that’s plenty for any normal 
environment the watch will see. A 5 bit ADC would be good enough for that task
(with the proper full scale input). 

The same thing likely applies to the voltage. The circuit is unlikely to function below
half voltage on the battery. Again a few bits of ADC will tell you everything you 
need to know to drive the table. 

I’d bet that they run a sigma delta and are quite happy with information at a “once
a minute” sort of rate. There’s not a lot of “analog” stuff in that case. It would be
pretty easy to dump onto the die. 

Bob

> On Jun 30, 2015, at 4:42 PM, Andy <AI.egrps+tn at gmail.com> wrote:
> 
> Measuring a 32kHz frequency to ~100 PPB accuracy takes some time, even for
> ATE.  Time is money.
> 
> I didn't think the hardware to do the computations and the digital offset
> was any problem.  I thought the temperature and voltage sensors might be,
> since they are analog.  But you can integrate them into a good mixed
> technology IC process.  I just didn't think that was what they are using.
> 
> Andy
> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.