On 4/19/2016 7:34 AM, Rob wrote:
rickman wrote:
What happens now is that the DAC in the GPSDO steps up and down every
couple of seconds, and the oscillators wobble around the correct frequency
in the 1E-9 area (with of course an average that is very close, more like
1E-11), and this results in funny interference patterns. What is
happening is clear when you put the 10 MHz outputs of two independent
boxes on the scope in X-Y mode.

I don't know what they are using for an oscillator, but if you have
control over it, can you reduce the corner frequency of the LPF on the
control loop? It sounds like the control loop is hunting to me. But at
10-9 I suppose it could be ambient thermal drift too. Yes, I think a
rubidium GPSDO would do the job.

I think the problem is that our GPSDO has a 16-bit DAC and it is
dithering the digital value to obtain the correct frequency. So when
the correct DAC value would be 32000.2 it will do 32000 for 8 seconds
then 32001 for 2 seconds, obtaining a long-term average that is quite
good, but a wobble with 10-second period as well.

Are you referring to the voltage used to control the VCXO? Not sure
what parts you can tweak, but I don't see why the dithering can't be
sped up (say 1 kHz) and low pass filtered to produce an actual value of
32000.2.


These are old Datum 9390 units, we also have some Trimble Thunderbolts
that should be better.

I remember a couple/three years ago Symmetricom came out with a chip
scale atomic clock that can sync to a 1 pps. "Two orders of magnitude
better accuracy than oven-controlled crystal oscillators". Only $1,000.
Might do the job. They likely package this in a box level product
that will do what you want. Check out this one...

I would prefer a box that works from GPS and outputs the 1PPS.
The Datum and Trimble are in that category.

http://www.microsemi.com/products/ti...-2750#overview

They were bought by Microsemi it seems.

I Have seen and considered these before. That is indeed a GPS referenced
rubidium standard. Of course we always prefer stuff that is either
cheap or available as surplus (like the Datum and Trimble) :-)

One man's "cheap" is another man's "precious". You'd have to use a
number for me to know what you consider "cheap".

You seem to be buying boxes rather than building stuff, but if you have
access to the innards of the system above and can up the sample rate, $1
worth of components can low pass filter the output to give better
accuracy of the control signal.

--

Rick