[time-nuts] Allan deviation

Magnus Danielson magnus at rubidium.dyndns.org
Mon Apr 28 16:50:35 EDT 2014


On 04/28/2014 10:32 PM, Tom Van Baak wrote:
> Can one you NTP timing experts kindly help this gentleman? I've had some email with him.
> Attached is an ADEV plot I made from his loopstats raw data. It seems to agree with his, though it doesn't seem to me that an ADEV plot is an appropriate statistic for this data (a single RMS number is sufficient).
> But I'm no NTP expert, so please take it from here. His previous emails are:
>> Hello! I built a graph Allan Deviation. Correctly built them? How to know the accuracy of time on them?
> ...
>> Sorry, could not speak English. Built in Gnu Octave.
>> Program took away http://www.vanheusden.com/misc/blog/2009-01-06-ad.php. Here the data:
> ...
>> A big thank you!
> Remember to keep it time-nuts related, not a general purpose NTP support thread.

Which loopstats data was used? time-error or frequency error?

The trouble with using in-loop data like this is that it will not 
disclose directly how stable the source is against an independent 
reference, but how stable the locked oscillator is against the reference.

That is, noise from the previous reference and noise on the link will be 
added on top of the tracked-in noise of the locked oscillator.

Knowing it is NTP loop data helps.

Most importantly, what is it that you want to illustrate in these 
graphs? Produced stability?


More information about the time-nuts mailing list