[time-nuts] Frequency Dividers

[Dave Andersen]

Tom Van Baak wrote:
>> I understand that the micros work fine, it just seemed like you would
>> need a second one to discipline the OCXO because of the timing
>> constraints on the divider.
> 
> Correct. But at $2+/- each it's easy to use one uC
> for a divider and another uC for the GPSDO algorithm.
> Yet another for an IRIG option; perhaps another for
> NTP. I think I remember RickH saying that his CNS II
> clocks use up to 5 microprocessors.

Actually, you don't always need to.  Many uCs will provide timer inputs
that you can wiggle in various ways to get the behavior you want in an
interrupt fashion, where the number of cycles it takes to get the
interrupt is predictable.  You can then schedule the other functions
(like poking the osc.) immediately afterwords, and know that they'll
have enough time to complete.

Hard to do with divide by 10;  easy to do with divide by 1000 if you
just want 10Khz on down output from your divider.

If you want to be more clever, you can get some of them to act as
dividers in hardware for the divide by 10 or 100.  Gets a little more
tricky programming the timer/counter on the uC, but it's doable.  I've
got some starter C/asm code for the atmel chips if anyone wants it as a
starting point.  (It's not complete;  I got distracted trying to debug a
broken LCD display...)

  -Dave