[time-nuts] Lady Heather's Tbolt oscillator auto-tune function

[Bob Camp]

Hi

The pps sync is done by resetting the counter that generates the PPS. At a 1 ppm frequency 
offset, it could take 500,000 seconds to steer it in with the OCXO. It unlikely people would wait 
for over a week for the PPS to line up ….

Bob

> On Sep 13, 2016, at 5:58 PM, Scott Stobbe <scott.j.stobbe at gmail.com> wrote:
> 
> Interesting discussion about startup. At startup the phase error of the
> synthesized PPS is +- 0.5 s. Is this coarsely set to the nearest ocxo cycle
> once gps time is established (would make sense to do it this way), or is
> the half second recovered steering the ocxo?
> 
> On Tuesday, 13 September 2016, Charles Steinmetz <csteinmetz at yandex.com>
> wrote:
> 
>> Mark wrote:
>> 
>> I just ran a tbolt (which has been off for a couple of months) and logged
>>> the state for a couple of hours...  and then remembered something about the
>>> initial DAC value setting that I had figured out long ago... it has little
>>> to nothing to do with oscillator disciplining.    The tbolt drives the GPS
>>> from the 10 MHz ocxo.  If the ocxo is too far off freq it can't track
>>> satellites.   The initial dac setting is used to speed up acquisition of
>>> satellites and not to speed up the OCXO disciplining loop lock.
>>> 
>> 
>> Well...  by doing the one, it also does the other.
>> 
>> As soon as a satellite is acquired (after a couple of minutes), the DAC
>>> voltage jumps and the disciplining starts.  A few seconds later when more
>>> sats are tracked, it gets underway in earnest (and by then the OCXO is warm
>>> enought to be within 0.1 Hz).  After 1 hour the box temperature has
>>> stabilized and the freq is within a couple of milli Hz.  After two hours
>>> the oscillator has settled down to the point where the DAC curve goes into
>>> "wandering around" instead of following  a smooth decay compensating for
>>> the oscillator warm-up.  The attached image show the first hour of the
>>> process.
>>> 
>> 
>> If you look carefully at the first 3-4 minutes, you'll see it does exactly
>> what I described.  The DAC reference is 0.510v, and the scale is
>> 5000uV/division (=5mV/division).  According to the paramaters, the initial
>> DAC voltage (INIT) = 0.499v.  I assume this was previously stored as the
>> DAC value after the Tbolt had fully stabilized, some time in the past.
>> 
>> Sure enough, the DAC voltage starts at just about 0.499v (it looks like
>> 0.494v on the graph), and when the second satellite is acquired it jumps
>> very quickly to 0.529v -- an overshoot of some 55% -- before settling back
>> to ~0.518v, at which time it appears to be on frequency within 1e-8 or so.
>> From that point disciplining continues as the crystal warms up.
>> 
>> If one accepted my suggestion, the initial DAC voltage would be set to
>> ~0.518v for this oscillator.  In that case, it should be within a few
>> millivolts of the voltage required when the second satellite is acquired
>> and the huge step with its 55% overshoot should be avoided.
>> 
>> I would be very interested to see the result of another dead cold start of
>> this same Tbolt, with INIT set to 0.518v.  Of course, the time at which the
>> second satellite is acquired (hence, the temperature of the crystal when
>> discipline begins, and thus, the exact DAC voltage required for a stepless
>> transition, will be a bit different from one start to the next, so it won't
>> be perfect.  But it will be a hell of a lot better than starting from
>> 0.499v).
>> 
>> Now -- does what happens during the first five minutes really make any
>> difference, given that no time-nut is going to do serious work with a GPSDO
>> for at least several hours after a cold start?  No, probably not.  But we
>> are time-nuts, after all, aren't we?
>> 
>> Best regards,
>> 
>> Charles
>> 
>> 
>> _______________________________________________
>> time-nuts mailing list -- time-nuts at febo.com
>> To unsubscribe, go to https://www.febo.com/cgi-bin/m
>> ailman/listinfo/time-nuts
>> and follow the instructions there.
>> 
> _______________________________________________
> 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.