[time-nuts] Re: GPS Control Loop

[Lux, Jim]

On 4/26/22 1:31 PM, ASSI wrote:
> On Montag, 25. April 2022 18:27:01 CEST André Balsa wrote:
>> A PDF of Shera's article can be found here (many thanks to whomever is
>> hosting this file):
>>
>> https://www.qsl.net/n9zia/wireless/QST_GPS.pdf
>>
>> To me, there is no doubt Shera's original design inspired all the following
>> DIY GPSDO designs in one way or another. Also this remark:
>>
>> "Figure 6 also suggests that two major causes of frequency
>> instability—temperature shift and aging—could be predicted and largely
>> eliminated by tracking the performance of the VCXO for a while to estimate
>> the aging parameters and by measuring the ambient temperature. The
>> predicted corrections could be applied to the VCXO independently of the
>> PLL, which might allow much longer loop filtering time constants to be
>> used, further reducing GPS jitter. Although this scheme would be ultimately
>> limited by sources of crystal frequency instability that are random and
>> inherently unpredictable, it might be interesting to explore."
> Establishing aging parameters for a modern non-ovenized crystal is a fools
> errand in my experience, at least if you keep the system operational for a
> long enough period of time.  If you don't, then you'll need to learn the aging
> parameters anew or you'll at least have to wait out the retrace before re-
> using data from the previous run.  When the initial retrace / aging transient
> has subsided, a linear model is good enough for short timescales (out to
> several days), but the actual logarithmic aging behaviour ensures that the
> slope gets very small.  I have some systems that are going into their fifth
> year of mostly uninterrupted, self-ovenized operation and aging induced
> frequency drift is swamped by other influences at the timescales of a
> reasonably imaginable control loop, although it is still visible on (much)
> longer timescales of course (currently drifting at about 100…200ppb/year).
> Feed-forward compensation of temperature fluctuation does work reasonably
> well, but you can expect only about one order of magnitude performance
> improvement from doing that, maybe two if you manage to get a really close
> coupling of the sensor to the actual crystal temperature. IIRC, some TCXO used
> to have a second quartz platelet with a special cut to act as a temperature
> sensor.  It's also possible to interrogate the crystal temperature by exciting
> multiple harmonics and looking at their frequency difference, but I don't know
> if any commercial applications employ that effect.
>

That's the MCXOÂ  - uses third harmonic and fundamental to measure the 
temperature.  Q-Tech sells them. Or, more properly, has them in their 
catalog and may be happy to quote a price and delivery. The datasheet 
revisions are >5 years ago, except for the Space version which was 
updated a couple years ago.  They're fairly good temp stability (a few 
ppb over  -40 to +90) and lower power than a OCXO of comparable 
performance.  The fancy temperature compensation doesn't say anything 
about aging, of course.  It's unclear whether cycling the temperature up 
and down at room temp makes an oscillator age faster than one held at a 
constant higher oven temp.

https://q-tech.com/products/mcxo-products/