[ntp:questions] Re: RMS error calculation

Dale Worley worley at dragon.ariadne.com
Thu Oct 30 18:07:32 UTC 2003


ptrojane at mion.elka.pw.edu.pl (Piotr Trojanek) writes:
> Yes, true. But I want to estimate (predict) errors introduced by this
> clock model, expecially those caused by "other badness". How should I
> compute them from Allan deviation plots?

A non-trivial problem.  If the clock errors are pure random noise, the
clock's deviation is a random walk, and the average error increases
proportionally to the *square root* of the time the clock has been
free-running.  On the other hand, if the major source of error is the
temperature sensitivity of the clock, and the temperature has a
noticeable daily cycle, the error patterns will be more complex but
more predictable.

Dale




More information about the questions mailing list