[time-nuts] Cesium DOES Drift, not

Tom Van Baak tvb at leapsecond.com
Mon Apr 18 20:21:11 EDT 2005

> The FTS4060 time interval is following the following equation (it takes 
> about a month to get this equation):
> y = -1.2594x2 + 236.37x - 10318
> where Y is in ns and X is the Day Of the Year.  The first term is the 
> fractional frequency stability, i.e. drift rate and is
> 1.14E-14 per day which is pretty good.

Do be careful here. Excel will blindly report
equations with 5 significant digits no matter
what the data looks like.

Here's something to try: break up your data
into three 10-day segments and see how well
the x2 term of the equations agree.

Or convert phase to frequency and then plot
30 days of frequency. If you have real drift it
should be clear from this plot.

DougH, JohnA, and I have Stable32 which makes
this a snap if you want to send any of us the raw
phase data.

> The HP-Agilent 5071A is specified at <1E-14 per day.
> Note that the fractional frequency stability of a good lab grade crystal 
> standard is about 1E-10 per day, so Cesium is 10,000 times better, but 
> still has drift. 

All frequency standards have frequency instabilities.
Hydrogen masers, Quartz, and Rubidium have drift,
but Cesium standards are generally considered to
have zero drift. That's one reason UTC is based on
that technology.

And note that fractional frequency [in]stability is not
the same thing as frequency drift. I can go into this
in more detail if you wish.

> This explains a lot about why setting the C field near 1E-14 is 
> difficult, the frequency is changing all the time.

If you make frequency plots in addition to phase plots
you will see dramatically why setting the C-field of
your 4060 to 1e-14 is hopeless. Frequency plots will
graphically show frequency instability (the width of
the line) and frequency drift (the slope of the line).


> Now Having Fun,
> Brooke Clarke, N6GCE

More information about the time-nuts mailing list