[time-nuts] Thunderbolt "good performance" example

[Stewart Cobb]

When one starts trying something new, it can be hard to tell whether one's
results are good or bad. That knowledge comes from experience, which is
exactly what one is lacking when trying something new.

One solution is to look for results that others have obtained. Time-nuts is
pretty good for this sort of thing. The attached plot is a small
contribution to this knowledge base. This is an example of Thunderbolt
performance  with a good antenna installation.

This screenshot from Lady Heather shows roughly a year of satellite signal
strength data and the most recent 9.5 days of phase and DAC data. The exact
latitude and longitude are not shown, but the lab is located in Silicon
Valley, California, at about latitude +37 and longitude -122.

The circular plot in the upper right is the signal strength data. By
default, LH "fills in" gaps in the signal strength data. I used the command
S-A-D to disable this fill and show only actual data. The satellite orbits
repeat exactly in sidereal time, which means that they drift slightly in
ordinary time. Over a year, this drift is noticeable and helps to fill in
this plot. Some satellites are stronger than others, and the plot only
shows the most recent data for each pixel, so a lot of stronger signals get
overwritten by weaker ones.

The antenna is a very good survey-grade choke-ring antenna located on the
roof of the lab. A choke-ring antenna has a sharp rolloff in gain below
about 20 degrees elevation, which helps reduce multipath errors. You can
see this in the "bullseye" shape of the data. You can see the shadow of a
single tree in the data (azimuth 240 degrees, up to about 30 degrees
elevation) but other than that the sky is clear. The usual hole to the
north appears, caused by the orbital inclination of the satellite
constellation.  The strongest signals are at C/No of 50 to 51 dB/Hz, and
the weakest trackable signals are below 30.

Others have recommended setting the elevation mask to 30 degrees, to reduce
multipath errors. While this works, it means the receiver will not even
attempt to track satellites below 30 degrees. I achieved approximately the
same result by setting the elevation mask to 5 degrees and the signal level
mask to 6 AMU (roughly 40 dB/Hz). With these settings, the receiver will
track all satellites above 5 degrees but will not use them in the solution
unless they are stronger than 6 AMU. With my antenna, the plot shows this
signal level corresponds to about 32 degrees elevation. You may need to
choose a different AMU mask for your antenna installation.

The horizontal plot along the bottom of the screen shows the most recent
9.5 days of timing data. The yellow line shows the diurnal temperature
variation in a lab which has no HVAC but benefits from Silicon Valley's
temperate climate. This particular Thunderbolt was built with the later,
low resolution temperature sensor.

The purple line shows that the Thunderbolt's own estimate of its timing
error is within 5 ns of zero. This is optimistic, because the Thunderbolt
is affected by ionosphere and troposphere changes that it cannot measure,
but it is the error input that its timing control loop uses because it's
all it knows.

The green line shows the tuning DAC response to the measured errors. The
plot legend says the span over 9.5 days is about 0.725 mV. The DAC
calibration (top left) shows that the OCXO tunes 3.717 Hz/Volt, or about
370 ppb/V. Multiplying these, we find that the OCXO wander due to aging and
temperature variations over 9.5 days is about 270 ppt, which is not too
bad. This Thunderbolt was continuously powered for most of a decade before
this plot was taken, which helps explain the low aging. From several areas
of the plot, we can estimate a short-term change of about 75 uV/C,
corresponding to a tempco of about 27 ppt/C for the OCXO. Again, not too
bad.

The control loop time constant was set to 300 seconds. For best frequency
performance with this unit, one could set it longer. For best PPS timing
performance, this seems to be optimum for this unit.

Hope you find this useful.

Cheers!
--Stu
-------------- next part --------------
A non-text attachment was scrubbed...
Name: LabMainTbolt-2021-04-29.png
Type: image/png
Size: 87301 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20210724/6fdb5c76/attachment.png>