[time-nuts] GPSDO three cornered comparisons

Tom Van Baak tvb at LeapSecond.com
Mon Oct 22 07:05:28 EDT 2012

I very much agree with this. It works because OCXO thermal effects you're looking at are in a domain where GPS is able to provide an adequate reference against which to measure.

Where this scheme doesn't work is with short-term OCXO noise or thermal transients (GPS isn't a stable enough reference at short tau), or long-term GPS accuracy (OCXO isn't a stable enough reference at long tau). That's the beauty of a GPSDO - you essentially get the best of both worlds.

Lucky for you, the temperature effects you speak of (especially diurnal) occur a little past the cross-over point. Otherwise I'm not sure you'd be able to measure it your way.

Another way of saying it -- you can use the OCXO to measure short-term tempco of the GPS receiver, or you can use GPS to measure long-term tempco in the OCXO. But if the tempco is anywhere near the cross-over point you can't be sure if it's OCXO or GPS. To solve that you do it the old-fashioned way: with a thermometer and a 5065A/cesium/maser reference.


----- Original Message ----- 
From: "Said Jackson" <saidjack at aol.com>
To: "Tom Van Baak" <tvb at leapsecond.com>; "Discussion of precise time and frequency measurement" <time-nuts at febo.com>
Cc: "Discussion of precise time and frequency measurement" <time-nuts at febo.com>
Sent: Sunday, October 21, 2012 9:27 PM
Subject: Re: [time-nuts] GPSDO three cornered comparisons


This is true for the phase data and calculating short term stability, but not necessarily for measuring the tempco of the OCXO..

If we assume most GPSDOs have very stable DACs and DAC references (I have measured typical DAC units to be at least 10x more stable than the best DOCXOs over temperature) then the changes in DAC voltage over time can give an extremely accurate absolute measurement of the OCXOs thermal performance.

The loop will cause random chaotic small corrections to be made as it tracks the GPS phase. All we have to do is filter these small AC type corrections out of the reported EFC voltage with a reasonable time constant, say 30 minutes, then we are left with a very good representation of how the OCXO reacts to ambient changes. Knowing the sensitivity of the EFC input allows calculation of the changes in frequency that the OCXO is generating over temperature and time.

The CSAC GPSDO makes this even easier as it is steered digitally in parts per trillion and gives the steering info in ppt as well, and also has two very accurate high resolution temp sensors on board.

Most of the time the resulting curve will have a very strong diurnal correlation to ambient temperature.

Knowing the ambient temperature changes around the OCXO then allows us to easily calculate the thermal stability of the OCXO (again assuming the Dac voltage tempco is 10x or more smaller then the OCXO thermal stability).

For example, assume the diurnal voltage changes are +/-60 microvolt (typical DOCXO) and we know the EFC sensitivity is 8Hz/V and the temperature changes +/-5 Degrees C then:

8Hz/V * 0.00012V / 10C = about 1E-011 per Degree C

So that OCXO would have a thermal stability of 0.6ppb peak to peak over a typical temp range of 0C to 60C which is quite good.

All that is needed to make this measurement is a frequency counter to establish the EFC sensitivity of the OCXO once.. And that counter is doing a relative measurement, so its inaccurate internal reference can be used..


On Oct 21, 2012, at 17:39, "Tom Van Baak" <tvb at LeapSecond.com> wrote:

> David,
> This is problematic, since the Z38xx software is not really measuring the output of the GPSDO. Instead it's taking the internally reported PLL time error measurements from the disciplining loop and pretending they are a measure of real performance as measured against a real frequency standard.
> However, this PLL data can be used to see how well the loop is working; that in itself can be interesting. For example, differences in an ADEV plot of the PLL error term can be used to reveal the time constant used by the PLL. So one can still use the ADEV-style calculation on the phase error of a closed PLL; but don't confuse this with the actual performance of the GPSDO as a time/frequency standard. Does this make sense, or shall I explain more?
> Or, if you have some raw data you can send me off-line I'll take a look at it.
> Thanks,
> /tvb
> ----- Original Message ----- 
> From: "David Hooke" <dhooke at gmail.com>
> To: <time-nuts at febo.com>
> Sent: Sunday, October 21, 2012 4:44 PM
> Subject: Re: [time-nuts] GPSDO three cornered comparisons
>> Tom,
>> The ADEVs I quoted are from the Z83xx software, so I assume it's 
>> comparing time from the GPS system to it's own oscillator's output, in 
>> the same way that LH does. I've been monitoring the basic stuff you 
>> suggested.
>> It's because I don't have a known reference to compare the units with 
>> that I want to use the 3-corner method to start sorting the wheat from 
>> the chaff.
>> david
>>>> I have an HP 58503A and a Symmetricom 58503B which are behaving quite
>>>> differently to one another using a common antenna. Using Z38xx, there's
>>>> almost an order of magnitude difference between the reported ADEVs at
>>>> 40k seconds (HP:7e-13, Sym:5e-12). Both have 10811 oscillators and are
>>> Where are you getting the "reported ADEVs"? What counter, what reference, what software?
>>> Did you happen to periodically log any SCPI data during the runs, so that you can compare SV counts, signal strengths, TI averages, and so on? In general it might be good to have a quick look at this mundane data first.
>>> /tvb
> _______________________________________________
> 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.

More information about the time-nuts mailing list