[time-nuts] HP 5730 data polling technique?

Magnus Danielson magnus at rubidium.dyndns.org
Mon Sep 7 19:38:27 EDT 2015


Hi,

On 09/07/2015 02:23 AM, Tom Van Baak wrote:
>>> Finally analyzing the data with Tom Van Baak's adev1 (
>>> http://www.leapsecond.com/tools/adev1.htm)
>>>
>>> [paul at localhost Documents] $ ./adev1 1 < ../datafile.txt
>
> Hi Paul,
>
> There's also the adev4 and adev5 tools too. But my main suggestion is to use John's TimeLab instead of a command line tool. Go to http://www.ke5fx.com/timelab/readme.htm to download.
>
> It turns out your raw data is very corrupted. If you view the phase and frequency data graphically it will be apparent. This is why a self-contained graphical tool like TimeLab is ideal. Until you get a clean phase difference series, the ADEV numbers produced by a command line tool (or any tool for that matter) are sort of meaningless.
>
> I've sent you a bunch of sample plots off-list.

Crash-and-burn a number of times with TimeLab is really a good way to 
learn a number of things. Eventually you get a good clue already on the 
ADEV and when you see the phase/freq plot you just go "ah, I knew it!".

Sometimes swapping between wrapped (w) and unwrapped (p) phase provides 
a good clue, but once you unwrap your phase correctly, swapping between 
phase (p) and frequency (f) views helps. Allan deviation (a) and time 
deviation (t) should be useful. Sometimes a modified allan (m) or 
Hadamard (h) deviation helps.

Experiment and measure!

Cheers,
Magnus


More information about the time-nuts mailing list