I have an HP 5334B with option 010, the good crystal. One channel is measuring the 10 MHz from an HP Z3801A.. The other channel is measuring the output of another HP 5334B. I've set the gate time to 100 seconds which gets answers like 1.00000007122E+07 and 9.9999997860E+06. That's more digits than will fit on the display.
If you assume that the Z3801A is accurate (and stand on your head) you can compute the frequency of the OCXO inside the 5334B.
The Z3801A has a poor antenna. I was expecting to watch a combination of short term glitches as the Z3801A recovered from holdover or poor satellite conditions and a long term drift as the OCXO aged.
My lab is not temperature controlled, so things also wander as the temperature changes. The other 5334B should be stable when the Z3801A glitches. If the measured frequency of both change, it's probably the OCXO in the measuring system that has changed.
Here is an interesting event. The graph is 3 days wide. Odd/even days get different colors. The Y offsets have been fudged to fit better on a graph. I'm interested in stability rather than accuracy.
It looks like the OCXO in the measuring 5334B jumped by 0.2 PPB, then 18 hours later, it jumped back. There is also a similar jump 9 hours later that only lasted 10 minutes.
The freq-out from the measuring 5334B goes to a TICC that is watching the PPS from the Z3801A and the PPS from a Sure SKG16B demo board.
The graphing software has to correct for the error in the reference frequency the TICC is using. Without that, you get a diagonal line. The slope tells you the offset but you can't see the small wiggles. The big sloped section in the middle is what happens when that correction is not accurate. It confirms that the frequency of the OCXO changed. If you look closely at hour 57, you can see a small step. That's the 10 minute jump.
Here is another pair of graphs zoomed in to look at that short event.
Here is another pair of graphs for another event..
Here is another pair of graphs for an event that lasted a half hour.
The slope of the Z3801A on the freq graph is due to the temperature change on the OCXO in the measuring system. That change also produces the curve in the TICC graph.
Here is another pair of graphs for an event that lasted an hour.
HP 5334B#1---\
^ ^ |
10 MHz | | |
HP 5334B#2 ------------/ | |
| | 10 MHz
10 MHz | |
Z3801A ------------------/ |
| |
|PPS via RS232 |
| v
\-------->FatPPS------------->TICC
^
|
PPS |
Sure, SKG16B---------------------/