[time-nuts] Disciplining Rubidium

Tom Van Baak tvb at LeapSecond.com
Thu Apr 24 00:31:54 EDT 2008

>I have truly enjoyed 'reading the mail' on this group.
> However, I need some help or a 'refresher' on the lingo.
> I am a Clinical Cardiac Electrophysiologist but in a bygone millennium, I
> received a BEE and a MSEE from Georgia Tech before I went to Medical School.
> 'tau'?
> Thanks,
> Joe

> 'tau'?


Short answer: tau is the measurement duration.

Long answer:

The accuracy, or resolution, of a frequency measurement
depends very much on the duration, or "averaging time",
of the measurement interval.

Also, the stability of an oscillator typically depends on the
duration over which the measurement is made. For example,
some oscillators are noisy from second to second, but look
much more stable when averaged day to day. Others may
be very stable minute to minute but drift over the span of
hours or days (averaging doesn't help, it makes it worse).

So the Greek letter t, or tau, refers to this averaging time.
You see if most often in all those log-log Allan deviation
plots (tau is the x-axis).

You'll also often hear the words "short-term" or "long-term".
There's no fixed rule about what's short and what's long. It's
mostly a way to acknowledge that clock behavior at small
averaging times may be quite different than behavior at
large averaging times

I bet the same is true for you with EKG's. You might learn
one thing by looking at a minute of data (short-term); and
learn quite something else by looking at trends in data taken
each day for the span of a week, or even once a year over
the span of several decades. So it is with oscillators: they
have short-term wiggles and long-term trends.

See also the Allan deviation of my heart beat, for tau from
10^0 to 10^3 seconds, near the beginning of:



More information about the time-nuts mailing list