[time-nuts] Limitations of Allan Variance applied to frequencydivided signal?

[Magnus Danielson]

On 05/13/2011 06:05 PM, Bob Camp wrote:
> Hi
>
> For AVAR you want a time record not a frequency measure. Your time stamps
> are a direct phase estimate. They are what you would use directly for the
> AVAR calculation. If they are faster than your shortest tau, all is well.
> Divide, mix down, what ever, just stamp faster than the shortest tau.

You can use frequency measures.... but there is a number of quirks 
hiding in there which can make a frequency-based analysis biased. Using 
time-stamps avoid those quirks, but naturally you can fluke those too...

For instance, use of averaging can be a bad idea. It can be used, but it 
needs to be blended in not to bias the AVAR measures.

Cheers,
Magnus