Hi

There is both short term and long term noise on the GPS PPS signal
out of a GPS module. Some of the short term noise can be reduced
by software using the “sawtooth correction” out. That assumes the
module you are using provides this output. The uBlox M8T is one
of many that provide this information.

Even with correction, you are still stuck with “many nanoseconds” of
noise on the GPS PPS. A good OCXO is much more stable than
this. A filter in the locking circuit is needed to reduce this noise.

The lock circuit can be either a PLL (phase locked loop) or a FLL
(frequency locked loop). If it is a FLL, the time output of the GPSDO
will drift. If you want time to stay locked, you need to use a PLL.

Even with a PLL long term accuracy is dependent on an integrator
in your lock circuit. That is what watches the long term problems
and corrects for them.

Working out the timing / corner frequencies of the proportional and
integrator portions of the loop is highly dependent on your oscillator
and the goals you intend to achieve. A reasonable starting point
for the proportional portion is in the 10 to 100 second range. The
integrator usually is 10 to 100X longer.

Getting all this working involves comparing it to “something better”.
Often in a basement project, finding something better becomes
a major effort.

Fun

Bob

On 1/15/2021 8:30 PM, paul swed wrote:

Seems to me you already have a temperature sensor in the OCXO.

If you temperature soak it overnight and monitor the frequency as power
is applied, you should see the starting freq of the oscillator.  The
heater will come on immediately and start it warming up, but the
oscillator should stabilize in less than a 100 ms or so, and the heater
should not be a major factor.  Do this at several temps and you should
get a pretty decent picture.

I suspect the fly in the ointment in this approach is that the oven temp
is usually picked to be around the flattest spot in the temp/freq curve.
You may not see much resolution.

You can also monitor oven current which I guess should translate to a
temperature error term, though I'm not sure off hand how you'd calibrate
that.

Oz (in DFW)

Hi

Ok, what do you know / have a pretty good guess at?

1. There’s an SC (or similar)  cut crystal in there

2. That crystal has a curve that looks something like the “0” or “1” curve in figure 7 of:

https://coloradocrystal.com/applications/ https://coloradocrystal.com/applications/

3. It probably is not a true SC so those curves are only approximately correct.

Since the frequency of the oscillator is low, there are a few possibilities.

1. It was originally on the lower turn and now is on the upper turn. ( = it got hotter)

2. It was in the vicinity of the turns and now is colder.

Without knowing where it started, there isn’t much way to work out which
one is correct. Also without being able to repair the circuit / adjust the
temperature the information isn’t of a lot of use.

Simple answer it so watch eBay for the next couple months and grab a
few replacement OCXO’s when they next come up for sale.

Bob