[time-nuts] Frequency comparison

Charles P. Steinmetz charles_steinmetz at lavabit.com
Mon Feb 8 23:50:57 UTC 2010

Joop wrote:

>Yes, I noticed these things take time ;-)
>This is what I logged after everthing was powered on for over 24 hours
>And this is with the LPRO is cold started. About 5 minutes after lock.
>The GPS was already running for about an hour I have to guess.
>It is less flat at the last two hours than what I was hoping for.

There is a reason that precision oscillators are often specified for 
stability "after running continuously for a period of [x]," where "x" 
is generally from one week to 90 days, sometimes even longer.  They 
are not really made to be turned on and off, even if you let them 
settle for a matter of hours.  Even with a GPSDO, the performance for 
small tau isn't really at its best until the unit has been on for 
weeks to months.  What range of tau is "small" depends on how you 
tune your disciplining dynamics.  I have found that the Tbolts in my 
limited very sample (2 units) work best with TC = 500 to 2000 seconds 
(bear in mind that I don't care about the PPS output, just the 10 MHz 
frequency), so for me "small" tau are those less than about 
1000.  Note that the improvement in small-tau stability probably 
won't appear in an ADEV plot that uses the received GPS time for 
comparison, because the oscillator is more stable than the GPS 
receiver at small tau -- so you're essentially measuring the GPS, not 
the oscillator.

It is best to leave your stable quartz and Rubidium oscillators 
running continuously, preferably on a UPS in case of power failure, 
for best results.

Best regards,


More information about the time-nuts mailing list