26-Feb-2009
Below are the results of a comparison between a Trimble Thunderbolt GPSDO and an atomic reference under laboratory conditions. The TBolt was set to factory defaults and was allowed to self-survey for half a day. The default time constant (100 s) was used.
About 750,000 seconds (~200 hours or ~8.5 days) of raw 10 MHz phase difference data was collected. The phase, frequency, and stability plots below were created from this data set.
In each of the following plots the frequency averaging is increased by 10.
For all averaging times (including 1 and 10 below) the frequency is almost always within ±1×10-10.
Frequency error (1 second raw data):
Frequency error (10 second averages):
For averaging times of 100 seconds, the frequency is
almost always within ±4×10-11.
Frequency error (100 second averages):
For averaging times of 1000 seconds, the frequency stability is
usually within ±5×10-12.
Frequency error (1000 second averages):
Frequency error (1000 second averages), same as above with 10× scale:
The following plots were made from one second samples of 10 MHz
Thunderbolt GPSDO output against a 10 MHz reference standard.
Relative frequency and phase offset were removed prior to plotting.
These plots give a good idea of the excellent phase stability of the GPSDO;
always within ±10 ns of the mean.
It is also evident that 10, 100, even 1000 second averaging
does not significantly improve time stability --
meaning that the noise processes responsible for the wander are
either not white noise
or require averaging times much longer than thousands
or even tens of thousands of seconds.
Phase error (1PPS raw data):
Phase error (10 second averages):
Phase error (100 second averages):
Phase error (1000 second averages):
Phase error (3600 second moving average):
We can see why for very long-term comparison against GPS,
only one average point per day is sometimes used.
Below is the Allan deviation plot (overlapping, many tau) for the data set.
Stability starts in the low -12's
(expected for this OCXO),
reaches a high of 1×10-11 near 100 seconds
(expected for the default TC)
and approaches the mid -14's at one day averaging time.
ADEV (Allan deviation, overlapping, many tau) plot:
Since error bars are not shown,
it's probably best to ignore points much beyond about
1×105 in these stability plots.
TDEV (time deviation) plot:
The one-day periodicity in the phase plots and
the tiny bump in the ADEV plot near tau 86400 seconds
suggest further analysis.
Autocorrelation plot (note strong, narrow peaks at 24 hour intervals):
Phase Plots
Stability Plots
Notes to Self
Return to LeapSecond.com home page.
Comments/questions to tvb.