[time-nuts] Effect of sampling intervals on ADEV measurements

[Robert Darby]

I compared an Austron 1250A to an FTS 1050A, both 5 MHz quartz frequency 
standards. I beat both against a 5Hz offset using a Riley DMTD device to 
provide a 1e6 time resolution increase.  There was about a 1.15e-10 
frequency difference between the two oscillators (two weeks on, it's 
about 5.6s-11)

The two oscillators measure about  6e-13 ADEV from 8 to 100 seconds, 
assuming the phase difference at the the ZCDs is between .2 and .11 s.

When the time interval, as measured by the counter, drops below .11s  
two distinct slopes become apparent on the Timelab "Original Phase 
Window"; one from .2s to .11s and the other from .11s to 0s.  As the 
phase wraps the cycle repeats.

I have thought about this a fair bit and the only thing that makes much 
sense is that with small phase differences I get 5 samples per second 
but as the phase difference lengthens, the TIC can no longer deliver 5 
sps.  It has to drop to 2.5 samples per second. If I'm not mistaken, I 
also see a similar, but less pronounced effect using PicTicII's as shown 
in Riley's article.  TimeLab sets the sampling time based on monitoring 
the initial input from the TIC and I assume a change in sampling rate 
will affect the slope. Does this make any sense or I am I barking up the 
wrong tree here?

Anyone using time tagging instead of TICs?  Any serious pitfalls there?

Thanks,
Bob Darby













-------------- next part --------------
A non-text attachment was scrubbed...
Name: FrequencyChange.JPG
Type: image/jpeg
Size: 15541 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20140824/30896788/attachment.jpe>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Rate Change.JPG
Type: image/jpeg
Size: 25171 bytes
Desc: not available
URL: <http://febo.com/pipermail/time-nuts_lists.febo.com/attachments/20140824/30896788/attachment-0001.jpe>