[time-nuts] Allan deviation -> finally some data!!!

[Tom Van Baak]

> Since I have only one readily accessible spare OCXO at the moment and I 
> want to get the procedure right before I take anything else apart, I 
> will feed the output of the 10 MHz reference from the 5370 into the 
> START channel, and my DUT into the STOP channel and select SEP instead 
> of START COM.

If you have only one oscillator, yes, this is what you
have to do. But as soon as you have two, you can
leave the 5370's own oscillator out of the picture and
compare the two external oscillators using the 5370
as a simple TIC.

> I believe that once I have the setup correct, I will rapidly find that I 
> need dividers on both OCXOs... 10 MHz will be too fast and I will be 
> skipping periods rapidly.

Dividers are nice, but often not really necessary.

Even if the two 10 MHz OCXO differ in frequency by
as much as 1e-9 that's only 1 ns per second of drift.
That gives you about 100 clean samples before a
cycle wrap.

Two solutions. 1) simply adjust them closer than 1e-9;
this should be easy, or 2) take care of cycle wrap in
software. I have code or tools to do this if you need.
It is the most reliable solution -- because you don't
have to touch, adjust, or lock either of the OCXO.

> The easiest from there will probably be to hook up the GPS into the 
> START channel, and a divided down 10 kHz or 1 kHz on the STOP channel.

Yes. But don't wait for a divider. The other advantage
of processing cycle wrap in software is that you can
collect clean, long-term, OCXO data with just this:
    5370 function TI, N=1
    START: GPS 1PPS (0.5 VDC, 50R)
    STOP: OCXO 10 MHz (0.0 VAC, 50R)

> Maybe I can close the loop using the PC, driving the EFC input of the 
> OCXO from the serial port. That would allow me to conveniently adjust 
> the loop parameters.

Don't close the loop. It will surely introduce humps
in the ADEV plots. Closing a loop means you are
measuring the performance of a disciplined oscillator
instead of a stand-alone oscillator.

/tvb