[time-nuts] Serial-Ethernet

Tom Van Baak tvb at LeapSecond.com
Thu Mar 17 10:00:35 EDT 2016


Hi Joe,

The "TI" value reported through SCPI by a HP GPSDO is the current "time interval"; the phase error within the closed loop. So this is PLL data, not clock data. As such, double-difference ADEV statistics selected in that plot of yours are somewhere between misleading and invalid.

The TI value should wiggle around by many ns to tens of ns. What's interesting are the peak-peak or RMS value. What you can do is plot the TI value as a time series, or summarize it by looking at RMS, or even RMS over time. There may even be specialized PLL statistics to apply in this case, or FFT. Sometimes TDEV or tau*ADEV(tau) is useful to expose periodic variations.

But the clue is that your ADEV (MDEV, etc.) plots all just head straight down into the -13's and -14's. If you ran it longer it would continue going into the -15's and -16's and pretty soon it would look like you have the best frequency standard in the world! That's what happens when ADEV is used on (bounded) PLL data instead of (unbounded) clock data. If not wrong, at least it's misleading, especially to newcomers to T&F metrology.

As to why it's included in Z38xx? Or Heather as well? There was a time when the authors of both those tools misunderstood what it means to measure a GPSDO. They confused actual measured performance of the 10 MHz BNC frequency output compared to an independent frequency standard with the self-measured internal phase locking performance of the PLL. This mistake rarely happens with Stable32 or Timelab.

However there is one case, with LH and a TBolt, where its ADEV graph is valid. If you put the TBolt in the state where GPS is active but disciplining is disabled, then the TI values really do represent a valid measurement between an independent reference (GPS) and a clock (the now free-running OCXO). In this case the ADEV plots in LH are perfectly valid. IMHO, this feature should be in every GPSDO, but I think it's unique to the TBolt.

/tvb

----- Original Message ----- 
From: "Joseph Gray" <jgray at zianet.com>
To: "Discussion of precise time and frequency measurement" <time-nuts at febo.com>
Sent: Wednesday, March 16, 2016 8:00 PM
Subject: Re: [time-nuts] Serial-Ethernet


Yes, I should have realized that it wasn't comparing the GPSDO against
anything. So, what is the point of this graph in Z38xx?




More information about the time-nuts mailing list