[time-nuts] Precision GPS timing
doug at niceties.com
Fri Oct 7 02:09:28 EDT 2005
I wanted to comment on a few items in the most recent digest mail.
Regarding GPS antennas, of course there can be multipath issues which are
typically reduced with choke-ring or large groundplane, and site-dependant
multipath issues can still occur. Of course there are many other issues
such as elevation mask and accuracy of coordinates assigned to the antenna
if your receiver is operating in some 0D/timing mode.
Regarding antenna filters, I'd trust whatever Dr Clark says but my simple
understanding is that filters can add delay which may not be flat across
frequencies and maybe have tempco issues.
Regarding single- vs dual-frequency, the first enhancement typically
achieved by dual-frequency is removing the daily ionosphere signal (rather
than just modeling it which might only achive around 50% success). However,
dual-frequency receivers often have some other feature such as
anti-spoofable P-code for authorized government users and/or
geodetic-quality carrier phase (where the receiver is driven by the
frequency from your own clock).
That carrier-phase method involves fancy post-processing of multiple
stations and results in a very precise difference between your clock and
some reference clock (typically an H-maser or better timescale) at the
processing epoch (typically every five minutes or so). The most specialized
receivers can also be calibrated and get an initial time sync so that you
are comparing time rather than just frequency - also you don't need to be
using an external Time Interval Counter. To be clear, those receivers are
typically >$10K and often use ~$5K antenna, and have some PC constantly
logging data to RINEX files (for post-processing).
If you can view or print a .ps file, an example of such dual-frequency
carrier-phase method measuring my 5071A high-performance cesium clock near
the end of August can be found at the location below (note that the chart
excludes a phase offset which was almost 50ns). The chart is dominated by
my cesium's typical wandering around +/- couple nanoseconds; the method
itself is definitely sub-nanosecond. In other words chart would look 10x
better if I had an H-maser (or could somehow steer my clock to reduce the
More information about the time-nuts