[time-nuts] GPS ADEV?
Tom Van Baak
tvb at leapsecond.com
Sat Apr 7 06:11:37 EDT 2007
I like the questions you ask. Here's an updated plot:
Note that the tau 1 second Allan deviation is essentially
the standard deviation of the sawtooth: in the graph the
VP is about 55 ns, CNS about 15 ns, CNS2 about 3 ns.
(CNS is an M12+; the CNS2 adds sawtooth correction).
The plot was made from about 12 hours of 1PPS data
for each of the three Motorola receivers.
Note also that for each decade the relative stability gets
10x better (since the slope is -1; white noise) and that
the ratio among these three receivers stays the same
regardless of the tau; the distance between the lines is
pretty much fixed (about 6 dB). This directly relates to
their relative jitter (sawtooth error).
Now, at much larger tau I would expect the three lines
to merge as something else becomes the limiting factor,
but that's well over tau 10^6 seconds.
I also overlaid the earlier longer-term CNS2 data (pink)
to the plot; the same -1 slope continues to a week.
Note the TIC resolution (150 ps for a 53132A) is not the
limiting factor for any of the plots here. However, if one
were to use a TIC with resolution of several ns instead
of sub-ns you'd see the CNS2 plot (only) shifted up a bit.
For contrast I added a (well-engineered) Datum 2000 to
the plot (black line). It was measured with a TSC 5110A
at 5 MHz rather than a hp 53132A at 1PPS.
What GPS receiver boards do you have handy, so that
you could repro this plot? It looks OK to me but having
a "second source" would make me feel better.
More information about the time-nuts