[time-nuts] Timepod Phase Noise Measurements and 3 corner hat

[Magnus Danielson]

John,

On 08/31/2015 12:37 AM, John Miles wrote:
>> The is a publication by the NPL that talks about a similar technique, but
>> comes to a somewhat different conclusion.
>>
>> http://publications.npl.co.uk/npl_web/pdf/mgpg68.pdf
>>
>> Bottom line: it holds that the phase noise of all three sources can be
>> determined if they are taken two at a time; one as ref and the other as
>> the DUT.
>
> Yes, that's the three-cornered hat technique.  The document is a bit dated in that respect -- it was published in 2004, just as Timing Solutions was starting to work on cross-correlated direct digital measurements.
>
> The dual-reference method always converges on the DUT noise if set up properly, but it doesn't give you any insight into the two independent sources being used as references.  You have to swap the DUT with each of the references and repeat the measurement if you want to characterize all three sources, while the N-cornered hat returns separated variances for all sources at once (at least ideally.)

Indeed.

On that note, did you look more closely on the NIST analysis of 
cross-correlation and a possible cancellation and thus overly optimistic 
results? Did it have any consequence on your code? What did you take 
away from it?

Cheers,
Magnus