[time-nuts] TBolt and Z3801A performance data

John Ackermann N8UR jra at febo.com
Sat Aug 9 20:27:59 UTC 2008


I just put the results of some tests of 2 "Time-Nuts Special"
Thunderbolts, as well as 2 Z3801As, at
http://www.febo.com/pages/gpsdo_comparison/

I learned an interesting (and important) lesson in doing these
measurement.  I initially measured the pairs of GPSDO against each other
(e.g., one TBolt as "reference" and the other TBolt as "DUT").  In
theory, if the two oscillators are identical, and if their noise is
uncorrelated, the results of the pair can be used to deduce the results
of the individual units.

However, in this case doing so gave a very optimistic view -- the TBolts
were better than the pair of Z3801As!  Another set of measurements
comparing the GPSDOs against an independent reference revealed that the
first measurement was a lie.

I guess you can think of it like this.  Picture two OCXOs that both age
at the same rate and in the same direction.  Because they drift
together, measuring their relative phase hides their actual drift and
makes them look better than they are.  On the other hand, if they were
drifting in opposite directions, they would look worse than they are.
An identical aging trend is a "correlation" even though it's external to
the things we usually think of.

John




More information about the Time-nuts_lists.febo.com mailing list