[time-nuts] GPSDO Sigma Tau at large observation times

[Tom Van Baak]

Hello again Estanislao,

Nice set of questions. The short answer is that for a GPSDO,
- The left side of the ADEV plot will show the difference between the quality of the local oscillator, in your case TCXO vs. OCXO vs. Rb. 
- The middle of the ADEV plot will show the difference in the choice of time-constant or the environmental conditions during the test.
- The right side of the ADEV will show the quality of the GPS timing receiver.

Also,
- The left side of the ADEV plot may expose limitations in your measurement equipment or the local low-noise reference that you're using.
- The right side of the ADEV plot may expose limitations in your local frequency reference. A cesium or H-maser helps here. If you use one GPSDO to test another GPSDO the very right side of the ADEV plot is misleading.

See also the examples here: http://leapsecond.com/pages/gpsdo/

Now, for your specific question:

> My question is, 
> when you are in the second regime, you are basically measuring the 
> ability of each clock to follow each other (is like there is just one 
> clock!, with the worst one setting the Adev value),

Right. For long tau, a GPSDO is not so much an independent clock as just a time-transfer system. As such, ADEV is not an ideal or fair statistic. For long-term performance consider using TDEV instead or just the 1- or 2-sigma RMS of the phase (time) error. This will better reveal how well the PLL is working.

> does then the 
> sigma-tau becomes meaningless for GPSDOs at time scales larger than the 
> TC of the oscillator being locked to GPS?

Yes and no.

Yes -- it is often meaningless because a GPSDO is really just a PLL. ADEV was sort of intended to compare clocks. In the long-term the phase error of a PLL is bounded by some rms value and so ever longer tau doesn't change how the GPSDO works, but ADEV will keep showing a line dropping like a rock at -1 slope. So ADEV is kind of misleading at this point. Consider using TDEV instead of ADEV. Also make sure your frequency reference is better than your GPS receiver for the tau you plot. Use low-noise quartz for short-term; use Cesium or H-maser for long-term.

No -- it is still meaningful because the ADEV you see out at 10^4 or 10^5 or 10^6 becomes very much a function of the quality of your GPS timing receiver (and antenna and environment, etc). If you were to expand your experiments to use half a dozen different GPS timing receivers you would observe a significant difference in rms timing noise, all of which translates to the actual ADEV value as it crosses, say, tau 10^5 and 10^6 s.

/tvb

----- Original Message ----- 
From: "Estanislao Aguayo" <eaguayo at thinksrs.com>
To: "Discussion of precise time and frequency measurement" <time-nuts at febo.com>
Sent: Wednesday, October 05, 2016 8:57 AM
Subject: [time-nuts] GPSDO Sigma Tau at large observation times


> Hello Time-Nuts,
> 
> My name is Tani Aguayo, I'm a newbie in this exciting precision timing 
> world, and so I have lots of questions. So here I go, first, let me 
> describe my experimental set-up for you.  I've been measuring the Adev 
> of several GPSDOs against each other, and once they are locked with a 
> suitable TC, the Adev plots show two regimes of operation (See 
> attached), the first one, at short time scales, has a lot to do with the 
> oscillator being disciplined characteristics, temperature, pressure.. 
> and there is a second one, where the Adev just drops linearly (in a 
> log-log scale). Each oscillator gets to the second regime earlier or 
> later, depending on the GPS PLL TC. In  the attached plot, the TCs of 
> the units compared, have been carefully selected depending on the 
> individual oscillator behavior. My thinking was that once in this second 
> regime, every unit is following the same clock, with the jitter of the 
> worst clock being compared, and since the sigma-tau makes no sense to me 
> for a single clock, I need some help with this concept. My question is, 
> when you are in the second regime, you are basically measuring the 
> ability of each clock to follow each other (is like there is just one 
> clock!, with the worst one setting the Adev value), does then the 
> sigma-tau becomes meaningless for GPSDOs at time scales larger than the 
> TC of the oscillator being locked to GPS?
> 
> I just wanted to thank the active members of this list, I feel like I'm 
> learning a lot with the different discussions and I wanted to say thank you.
> 
> - Tani Aguayo
> 
> 
> 
>


--------------------------------------------------------------------------------


> _______________________________________________
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
> and follow the instructions there.