[time-nuts] Disciplining Rubidium
Tom Van Baak
tvb at LeapSecond.com
Thu Apr 24 23:47:06 EDT 2008
> Hi Antonio, Tom,
> recently I had to unplug my PRS10 Rb from GPS for about 3 days. It drifted a
> couple 100ns in that time frame.
> When I plugged the GPS 1PPS back in, I saw a significant frequency error of
> a couple of parts to the E-10 while the PRS10 was shifting the 1PPS back onto
> Pretty bad for a Rubidium I thought.
> For such a large phase error I would have expected the PRS10 to just reset
> the 1PPS rather than drift it.
But why expect a reset in this case? This would then mean a
sudden frequency change of 100 ns / 1 s, or 1e-7! So you
have to ask what's worse: a sudden jump in the e-7 range
or a gradual change in the 1e-10 range. Or, should it jump
the 1 PPS and drift the frequency? Or keep the 1PPS and
frequency in phase and drift them both? Or jump them both?
The answer is -- there is no right answer. It all depends on
the design specs of the unit, or the needs of your application.
Sometimes a jam sync is in order, sometimes an elevated but
bounded drfit rate.
If you're lucky, the GPSDO gives you an option whereby you
can program your threshold; your "expectation". But internally
every GPSDO has to make a decision about when to jump vs.
when to drift. Whether it's hardcoded or programmable is the
See section 5.1.1 of the Trimble Thunderbolt manual for a
good example of how this can be done.
More information about the time-nuts