[time-nuts] Phase shifter circuit for DTMD
Magnus Danielson
magnus at rubidium.dyndns.org
Sat Jul 25 00:14:32 UTC 2009
Stephan Sandenbergh wrote:
> Hi,
> Thanks for the replies. I agree. One is interested in the timing of beat
> notes.
>
> But, I'm slightly confused now. It might be a language problem on my side.
>
> Quoting Howe, Allan and Barnes, 1981, "...adjust the phase so that the two
> beat frequencies are nominally in phase; this adjustment sets up a nice
> condition that the noise of the common oscillator tends to cancel..."
>
> I can see a few issues here:
>
> 1) if the beat frequencies are in phase, there will be a very small time
> interval between their zero crossings. This might be difficult to measure
> with accuracy.
No, it is not a particular issue, unless you have problem with
isolation. The method to workaround isolation problems that I described
is just one possible approach out of many, including designing it out of
being an issue.
What they are after is that as coincidence in time makes signals from a
DTMD perspective correlate better in the regard as supressing the
transfer oscillator phase stability. If you keep the coincidence time to
say within 100 us much of the low-frequency noise of the transfer
osicllator beyond 100 us correlate quite well, so cross-correlation will
perform well to supress it where as instabilities with shorter times
will not correlate and thus polute the measurement. As time-separation
increases, more integrate time-noise of the transfer oscillator will be
exposed and induced into the measurement results, for which only
time-averaging will help.
> 2) I agree that reference oscillator noise will cancel to some extend
> because the measurements are done closer to the same time, which makes the
> reference oscillator noise better correlated between the start and stop
> edges.
Correct.
> 3) Also, one would like to compare both clkA and clkB at the same time. Not
> one at t=0 and the other at t=1day to exaggerate a little. I'm not planning
> to measure atomic standards, or the best OCXO available, so I doubt this
> will bother me.
The closer together they are, the better cross-correlation gain can be
achieved. On the other hand, tbe closer the edges is and thus
cross-correlation losses may also be found.
> Other than the above, I agree: it is better to have a greater phase offset
> between the beat notes.
>
> Does my thinking my sense at all?
Sure thing.
Cheers,
M'agnus
More information about the time-nuts
mailing list