[time-nuts] GPSDO recovery from holdover

Magnus Danielson magnus at rubidium.dyndns.org
Sat Dec 1 18:01:04 UTC 2012

On 12/01/2012 03:10 AM, Hal Murray wrote:
> Does anybody know what happens in a TBolt or Z3801?  (or any other boxes?)
> Suppose your system goes into holdover for long enough to be interesting.
> Suppose for discussion that the clock drifts so that the PPS if off by a
> mircosecond.
> I can see two ways to recover.  One is to jump the 10 MHz clock by 10 cycles.
>   The other is to adjust the frequency so that the PPS slews back to on-time.
> The first approach gives you a second with the wrong number of cycles.  The
> second approach has your clock frequency off for a while with a trade off
> between how far off and how long it's off.
> Are there any other approaches?

Jumping should be used when time to come within frequency and time 
constrains is too large. When jumping, it might be better to squelch the 
output during the jump, and doing that for long enough such that you 
definitly misses a PPS pulse. I choose to set a time-error limit, beyond 
which squelching and jump is used rather than just using the normal 
linear track-in process. By regulating the time-error limit the balance 
can be set depending on the equipment using the signals.
Naturally, the track-in rate depends on the PLL loop bandwidth, but 
there is a limit to high large you can and will set it. FLL track-in 
help a lot, but again, the rate of change can upset downstream equipment 
so taking the "hit" of a jump can become the best option.


More information about the time-nuts mailing list