The Z3801A GPS-Disciplined Oscillator - Stability
This page has data on the stability of my two HP Z3801A oscillators. I compared them to my HP 5065A rubidium frequency standard (denoted as "N8UR-3"), which was simultaneously compared to the LORAN station in Dana, Indiana. I measured each oscillator for 48 hours so that any daily pattern could be seen. I also compared the two Z3801As against each other.
The measurement technique used digital dividers to reduce both the 5065a's 5MHz output, and the Z3801A's 10MHz output down to a 1 pulse per second signal. This was fed into an HP5370B time interval counter and GPIB software was used to log the time difference. The 5065A also provided the reference input to the counter.
The 5370B has a resolution of 20ps, and its internal jitter is about 60ps RMS. To be safe, I assume that the measurement system is stable and accurate to 100ps, or 1x10-10 over 1 second. Since the fractional stability increases over longer averaging times ("tau"), at 100 seconds averaging (i.e., 1x10-12) the system noise becomes low enough to provide accurate results. Consequently, I recommend that you ignore the data below for tau of less than 100 seconds. (I'm planning to build the equipment necessary to get high-resolution results at averaging times of 1 second or lower, but that's on the "not-quite-finished-yet" list.)
Here are the results, combining the stability data from the three runs into a single chart:
It appears that N8UR-1 is quite a bit more stable than N8UR-2. This makes sense if you look at the individual unit data below, which shows that N8UR-1's adjusted phase chart has half the peak-to-peak noise of N8UR-2's, or of the comparison of N8UR-1 to N8UR-2.