Check system timing accuracy (using astrometry of GNSS
Repeat the Process
This is an outline of how to use several online software tools made available by
Bill Gray (see MPML message
#32538) allowing the overall
timing accuracy of a telescope system to be determined by measuring
the position of Global Navigation Satellite Systems (GNSS) such as the US GPS
and Russian GLONASS
The orbits of these communications satellites are very well known, from
where they were soon after launch, some dating back to the early 1990's to where
they will be up to about 24 hours into the future and to an
accuracy of just a few centimetres.
As these satellites are generally moving at
apparent speeds of 2,000 - 2,400 arc seconds per minute, or 30 - 40 arc seconds
per second, they can be used as a very sensitive gauge of the absolute timing
accuracy of an imaging system, with all the delays and errors
introduced by the camera shutter, imaging software, operating system, timing source and even
observatory topocentric coordinates (see non-timing errors)
combining to affect the final measurement.
The positional accuracy in RA & Dec can be expected to be of the order of 1
arc-second or better but a timing error of 1 second in the clock will manifest
itself as a very obvious 30-40 arc-second discrepancy with the ephemeris, so
providing a very sensitive measure of timing accuracy.
Configure capture software
The software that captures images and stamps the time of exposure within the
FITS headers may need configuring to allow extra precision in the recorded time.
Maxim DL is used at Great Shefford and by default it only records the time of
exposure to the nearest second. However, by entering a non-zero value in the
Camera Control, Expose tab, Options, Measure Shutter Latency... option, Maxim
saves the exposure times to a precision of 0.01 seconds (see also
Of course the
recorded time may not be accurate but at least it will be saved to two
decimal places. However, the GNSS tools described here will allow you to determine how
accurate the times really are.
Pointing the telescope
first problem for the observer is to point the telescope at one of these very
fast moving objects.
Use the following two links to select a satellite and then generate an ephemeris for
- Choose a target from a list of GNSS satellites visible at your site (gps_find)
- Generate an ephemeris for the selected satellite (gps_eph)
The GNSS satellites are generally 12th to 14th magnitude, moving at 30 - 40
arc-seconds per second. Ideally exposure times should be short enough to limit trailing so that the entire
trail fits within the measurement annulus during measurement. This should allow automatic
and accurate centroiding of the target, rather than trying to measure trail ends.
At Great Shefford, with a 2.2"/pixel scale and a 3-pixel radius aperture in
Astrometrica, a 0.3 second exposure
allows the resulting 10-12 arc-second trails to be easily measured.
short exposures there may not be enough reference stars recorded
to achieve a good astrometric measurement, so check
that there are catalogue stars distributed across the field and that the target
has reference stars on all sides (in Astrometrica tick the Reference stars option within the Images/Select
Markings... menu option).
Here is an example showing a typical short trail from a 0.3 second exposure:
The trail is short enough for the measurement annulus in Astrometrica to
determine a good centroid:
Repeat the process
Depending on how quickly your system can take consecutive images you may need
to reposition the telescope a number of times to get enough images to determine
a statistically relevant result. These GNSS satellites pass through the 18 arc-minute field of view of the telescope at Great Shefford
Observatory in about 30 seconds and with a gap of 1 - 3 seconds between
exposures about 10 measureable images can be obtained from one telescope
It is suggested to obtain at least 10 positions, preferably more and from a
number of different fields of view and in different directions in the sky, using
more than one GNSS satellite.
The next step is to measure astrometric positions of the target satellite.
The date and time needs to have enough precision in the astrometry output file
so that the accuracy of the method can be fully utilised. In Astrometrica this means making sure that
in the CCD tab of the Program settings the "Time in File Header" precision is
set to 0.1 seconds or better.
The resulting astrometry, either in the MPC's 80
column format or in the new ADES format can then be copy/pasted into the link
below to work out how far off the measurements for each observation are in along-track seconds
of time and cross-track arc-seconds:
Here is a sample of the resulting output from gps_ast, with most of the detail lines
removed for clarity:
|Current time = 2021 Aug 6
Earth rotation parameter file date 2021 Aug 5
2003-010A C2020 02 29.84487712 11 53.44 +47 03 23.3 13.3 G J95 xresid
0.547435" along 0.0586419s G21 2003-010A
2003-010A C2020 02 29.84491812 12 03.53 +47 02 34.1 13.3 G J95 xresid
0.411983" along 0.0296576s G21 2003-010A
...70 lines removed...
2003-010A C2020 02 29.85459512 49 13.99 +43 23 30.7 13.0 G J95 xresid
0.730248" along -0.0056586s G21 2003-010A
2003-010A C2020 02 29.85462812 49 21.00 +43 22 42.6 13.1 G J95 xresid
0.087513" along -0.0330421s G21 2003-010A
74 observations found
Avg cross-track : -0.018644 +/- 0.338065"
Avg along-track (timing): -0.0050475 +/- 0.0382892 seconds
Negative along-track errors mean your clock was 'ahead' of the actual
i.e., the times reported in the astrometry are later than the positions
of the GPS satellites would indicate.
From a number of similar determinations the average along-track (timing) value
indicated there was about 0.07 seconds of
latency in the overall system at Great Shefford.
This value was then configured in Maxim/DL (in
Control, Expose tab, Options, Measure Shutter Latency... option) so that every
subsequent image taken has the 0.07 second system latency automatically applied
to the time recorded in the FITS header. In the gps_ast output
above, this adjustment has already been applied, reducing the overall
timing latency to -0.005 ± 0.038 seconds.
Combining measurements made on
three different nights and plotting timing errors horizontally in seconds of
time and cross-track
errors vertically in arc-seconds, the random distribution of measurements can be seen here:
This method should be used to periodically check timing accuracy to make sure
any change to the overall telescope system that may affect timing (e.g. hardware or operating system
upgrades) is detected and can then be taken into account.
When initially obtaining readings using this method in 2017 it was noticed that
there were systematic trends in the cross track residuals the further from the
meridian a satellite was imaged. Investigation eventually uncovered that the
coordinates of the observatory issued by the Minor Planet Center back in 2002
had not correctly included the adjustment of the
to the height above mean sea level. The geoid height at Great Shefford is 47.6
meters and the effect of the incorrect coordinates on the cross track residuals can be seen in this figure
where the GNSS satellites plotted in green were observed to the west of the
meridian and those in orange observed to the east of the meridian.
(Satellites are identified in the figure with the 3-character
After the observatory coordinates were corrected, both sets of points are now
properly centred around 0 arc seconds on the cross track (Y) axis: