[time-nuts] Noise types and Allan Deviation

[Bruce Griffiths]

Tom Van Baak wrote:
>> A table indicating the dependence of the ADEV of these phase noise types
>> on the measurement system bandwidth would also be helpful.
>>     
>
> Bruce,
>
> I've never bothered with the details of measurement system
> bandwidth as you describe. A couple of reasons.
>
> 1) Almost none of the thousand ADEV plots I've ever seen in
> commercial and academic literature ever mentions it. That's
> my first clue it might be a "Bruce thing" ;-)
>
> 2) It looks like it's a pain to compute, and if it doesn't end up
> making any difference then it's not worth the bother.
>
> 3) If you take NEQ BW into account you can see when it
> makes a difference and, as important, when it doesn't.
>
> Here's a nice example from last week:
>
> http://www.leapsecond.com/pages/tbolt-tc/tbolt-tbolt-neq-1.gif
>
> It's the ADEV of two non-optimized TBolt's against each other
> on a TSC 5120A (which allows you a selection of NEQ BW)
>
> The good news for you is that there are data points where the
> measurement bandwidth makes a difference. See tau 0.01
> and 0.02 for example.
>
> The better news is that for the rest of the plot, all the way from
> tau 0.1 to 100 000 seconds, it makes no difference at all, not
> even one pixel.
>
> /tvb
>
>
>   
Tom

The tau range for which the measured ADEV etc are relatively independent
of measurement system bandwidth will depend on when the phase noise
processes for which ADEV etc are bandwidth independent become dominant.

It should be true for larger tau when flicker phase and more divergent
phase noise mechanisms dominate, but the lower limit will vary depending
on the characteristics of the sources involved.

One can either measure the phase noise at very low offsets and calculate
MDEV etc to identify when the various noise processes dominate and thus
infer the minimum vale of tau for which ADEV, MDEV etc are independent
of the measurement system bandwidth, or vary the measurement system
bandwidth to see if the measurements change.

The tau range for which ADEV etc are sensibly independent of measurement
system bandwidth should differ significantly when comparing a pair of
hydrogen masers or a pair of low noise OCXOs (OSA8607, FTS1200 etc).

The lower limit for an 8607 or a good FTS1200 is probably in the 1 to
10s range.
That is around the value of tau for which ADEV vs tau first approximates
a horizontal straight line or reaches a minimum.

Bruce