[time-nuts] OCXO Advice

[Tom Van Baak]

Bob Q,

Both of your UCT 108663 oscillators look ok, although the blue plot 
seems 2x to 3x worse than the red plot. Neither is strongly correlated 
with temperature. It appears that you're using a GPSDO as a testbed so 
some of the noise and strange wiggles seen in your plot may be coming 
from the GPSDO board and not the oscillator itself. In that case a 
better OCXO might not help.

Each vertical grid line of your plot is 0.5 mHz / 10 MHz = 5e-11. Your 
red plot matches mine pretty well:

http://leapsecond.com/pages/uct108663/uct-01.png

BTW, a California eBay seller for UCT 108663 is:

https://www.ebay.com/itm/202863421344

Suggestions:

1) Swap the OCXO's to see if the noise follows the OCXO or follows the 
GPSDO board. I mention this because the 2nd hour of the blue plot looks 
like GPSDO wander, not typical OCXO noise.

2) Maybe try each OCXO in an isolated test for a day. That is, remove 
the GPSDO from the equation -- place the OCXO in an open but undisturbed 
no air flow location, power it from a clean stable power supply, and 
float or ground the EFC. Measure with your Rb (LPRO) as the reference. 
The results of these baseline tests will suggest if the instability that 
you see is due to the OCXO or due to the GPSDO boards you're playing with.

3) Consider using both your phase comparators on one OCXO at the same 
time to verify the consistency of your measurement system. For example, 
the unusual red glitch at T+6 hours looks more like a measurement issue 
than something a OCXO would do.

In general, when you have two GPSDO boards, two OCXO, and two counters, 
it's not likely they are all equal. So these multiple tests allow you to 
gain confidence in your parts and to pinpoint the weak element(s) in 
your setup.

/tvb