[time-nuts] time-nuts Digest, Vol 49, Issue 4

[Tom Clark, K3IO]

   Jim Palfreyman noted


I took my recently acquired Thunderbolt, along with my 5370B, to the
observatory on the weekend and plotted its residual against the NR Active
Hydrogen Maser. See the attached graph (ignore the "line of best fit"
heading - 0.00 was simply the starting point).

Note that I didn't let the Thunderbolt do a survey, I simply plugged in the
local GPS coordinates which I later verified were about 5m to the east of
the actual Thunderbolt antenna.

The y axis is residual in nanoseconds and the x axis is Modified Julian
Date.

I find the residuals fascinating. Note the identical repeating residuals in
a 24 hour cycle.

Is this due to thermal or constellation issues?

The local GPS (not a GPSDO but an early version Totally Accurate Clock)
showed a roughly (very roughly) similar pattern when averaged with 5 minute
averages to hide the awful 100nsec sawtooth.



   Not at all unexpected. I recommend you look at 3 papers on our archive
   at [1]http://gpstime.com/:
   In "Low-cost, High Accuracy GPS Timing" you will  see the
   state-of-the-art circa 2000. In slide 16 you see the same type of
   diurnal bumps, which are due to the residual ionosphere. The GPS
   satellite broadcast message contains a gross ionospheric correction
   number based on sunspot count and 2.7 GHz solar flux. But the typical
   single-frequency receiver has ZERO correction for a day-night
   ionospheric cycle. At the GPS frequency (1575.42 MHz) the total effect
   of the ionosphere in the zenith is about 5 meters; the effect at lower
   elevation angles is even more -- say 7 meters or so. And the day-night
   variation (max late afternoon local time so GPS runs slow, min early
   morning so GPS is fast) typically is in the range of 5 meters = 15 nsec
   or so.
   In "Critical Evaluation of the Motorola M12+ ..." take a look at Figure
   7 (in the PDF) or slides 17/18 (in PPT). You will see the diurnal
   ionosphere causing ~10-20 nsec of variability and even a 50 nsec
   excursion due to a large solar "incident".
   The 2007 version of my "Timing for VLBI " tutorial shows some more
   modern data, including a description of what causes the "awful
   sawtooth"  and how it can be removed in hardware. FYI -- the 100 nsec
   sawtooth you see was characteristic of the early 6- and 8-channel
   Motorola Oncore receivers that I used in my original TACs and Rick used
   in the first generation of CNS Clocks. When Motorola brought out their
   new M12-series, the magnitude dropped to ~24 nsec. And it could always
   be removed in software (even with the older receivers); a binary
   message from the receiver tells the user the size of the error
   (quantized in 1 nsec steps) of the next 1PPS tick.
   Rick's TAC32 software can log the corrected 1PPS value from a low-cost
   (HP53131/132) counter, so the "ugly sawtooth" is never a problem. Of
   course. averaging over tens of minutes wipes out the effect too -- just
   look at the 1995 results from Onsala in the papers.
   Regards, Tom

References

   1. http://gpstime.com/