[time-nuts] Method for comparing oscillators

Magnus Danielson magnus at rubidium.dyndns.org
Thu Aug 6 13:37:50 UTC 2009


Steve Rooke wrote:
> 2009/8/7 Magnus Danielson <magnus at rubidium.dyndns.org>:
>> Steve Rooke wrote:
>>> 2009/8/6 Magnus Danielson <magnus at rubidium.dyndns.org>:
>>>> Ulrich Bangert wrote:
>>> ...
>>>>> Well, stability over time is what exacly is displayed in a
>>>>> tau-sigma-diagram
>>>>> of an oscillator. Since only a few words before he is saying that he is
>>>>> NOT
>>>>> intersted into Allan Deviation plots, then he is perhaps interested into
>>>>> something else?
>>>> Yes. Sigma-Tau plots of the Allan Deviation fame (with friends) addresses
>>>> the instability of the noise part of things. For crystal oscillators and
>>>> other non-atomic oscillators "linear" factors in frequency drift is not
>>>> best
>>>> specified, described or measured using that method, which was invented
>>>> purely to be able to handle the phase noise side of things, not the slow
>>>> frequency drift.
>>> For these sorts of measurements on drifting oscillators would it not
>>> be prudent to use the Hadamard Deviation?
>> Hadamard Deviation does not fully cancel the non-stable drift.
> 
> From the flow of the discussion I gathered, perhaps incorrectly, that
> we were talking about the fairly linear drift in crystal oscillators
> for which HD handles. If there is a non-stable drift, it's going to be
> an interesting exercise to be able to process that out of the data
> easily. Agreed if the drift can be characterized to fit some form
> function but I would call that a stable drift anyway. I also had the
> impression, and please correct me if I'm wrong, that the use of the
> additional adjacent data point in HD calculations resolved the issues
> with drift at any point in the data stream.

As been investigated before, linear drift is an approximation and better 
models have been found to match the data better, and also been shown to 
connect to the various "aging" mechanisms known to exist even if 
production techniques can reduce many of them to low levels.

HDEV does a better, but not perfect, job at canceling the non-constant 
drift. If you want ADEV or similar, then using methods described will 
get you better results.

I found a nice little graphical tool that let's you pull data in, match 
against one of several models and then see the produced result. Nice 
tool to get started. Can't recall the name right now.

Cheers,
Magnus




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