[time-nuts] Pretty ADEV/Tau plots

John Ackermann N8UR jra at febo.com
Wed Mar 14 17:51:15 UTC 2012


On 3/14/2012 1:35 PM, Chris Howard wrote:
>
>
>
> These pretty ADEV/Tau plots, do people have an automated
> system to produce these things? How much work is involved?
> How many samples are taken? Sample for a month, omputer crunching
> for weeks?
>
> I have no feel for what the process is like.
>
> I have two oscillators and a Racal 1992 counter. If I were
> to hook up the computer to the counter would I have the minimum
> amount of "stuff" needed?

The stuff on my pages at febo.com come primarily from two sources:

1.  Screenshots from the TSC analyzer's display (actually, I trick the 
TSC into thinking it's network printing to a Postscript device, then I 
grab the incoming bitstream and convert it to .png -- not a pretty process.

2.  Phase or frequency data from the TSC or other counters manually 
massaged using a *nix WYSIWIG graphing tool called Grace.  In a few 
cases, I've also munged a way to automatically generate Grace plots 
every X minutes from live data.  That, also, is not pretty.

However, lately I've been using John Miles' TimeLab software (even 
though it's Windows...) because it is so damn easy to capture data from 
lots of counter types, and display multiple runs and various plot types. 
  It just takes all the work out of it.

Length of capture depends on what you're trying to do.  Generally, you 
want a minimum data length of X times the longest tau, where X can range 
from 3 (but huge error bars) on out to perhaps 8 - 10 for good 
reliability.  So if you want to plot stability out to 100K seconds, 
you're going to be collecting data for at least a week or two.

I've had PPS measurement setups where I took data perhaps every 10 
minutes for several months.  That starts to be an exercise in "design 
for reliability"!

John





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