[time-nuts] New topics (was Re: He is a Time-Nut Troublemaker....)
Magnus Danielson
magnus at rubidium.dyndns.org
Tue Dec 23 17:25:44 UTC 2008
John Ackermann N8UR skrev:
> Magnus Danielson wrote:
>
>> My intent is to get some stuff done in the lab during the vacation.
>> (Desperatly trying to get some more on-topic discussions going).
>
> Here are two questions that have been running around my head:
>
> 1. Following on from the discussion last week about trying to
> synchronize multiple oscillators to improve phase noise, I've wondered
> about a simpler tack: take, for example, two 5 MHz atomic standards and
> mix their outputs together, using the 10 MHz result to drive a time
> scale. Assuming the standards were of relatively equal quality, would
> this provide a better time scale than using one of the standards alone?
I have been looking at mutual lock systems for say two 10 MHz
oscillators. This mutual lock system behaves much as a PLL circuit (as
contrary to the other mutual lock methods described previous) such that
a phase detector detects the phase-difference between the oscillators, a
loop filter filters this difference and then it is feed back with
opposite signs to the respective oscillator. I envision that a common
EFC input exists to the system. Such a system is kind of interesting in
that within the PLL bandwidth, the two oscillators are locked and you
see the average of frequency errors etc. Well above the PLL bandwidth
the two oscillators is independent of each other and just above the PLL
bandwidth they are essentially independent. As the two oscillators is
phase-locked in, their outputs can be combined and noise reduction can
be achived, which is the significant effect above the PLL bandwidth.
It is interesting to notice that the PLL operates on the difference
between the oscillators, where as the EFC input is acting as a common
mode action, which in the perfect world is not influenced by the
diffrential PLL. Realities is however such that the EFC is not linear,
and the need to regulate differently also ensures that the oscillator
gain constant never really match, and therefore will also the
diffrential PLL "leak-in" on the common mode behaviour.
I have not seen any reference to this type of diffrential PLL action to
mutual locking. I've only seen discussions on mutual locking involving
injection locking.
This diffrential locking technique could be applied to atomic standards,
but then naturally require much improved solution than simple
oscillators. The diffrential locking technique does not magically solve
issues that is typically common mode, such as temperature dependence. It
can however even out individual properties like noise and systematic
drift to some extent. It essentially runs the oscillators as a common
constellation and attempts to achieve the average improvements of those
oscillators in an interlocked fashion. In its simplicity it will do
unweighed averaging. It is fairly easy to do weighed averaging by
individualizing the feedback gain to the respective oscillators. Further
refinements would individualize the proportional and integrate feedback
terms, but as always, the simplicity forms a limit.
I have intended to make some runs just for fun to see how this behaves
in real life.
> Happy holidays!
I wish you all happy holidays!
Cheers,
Magnus
More information about the time-nuts
mailing list