[time-nuts] Checking accuracy of Rubidium standards

[Bruce Griffiths]

Alan Melia wrote:
> This is an interesting thread again.....it may be similar to ones that have
> been discussed, but one or two furthur questions occur to me. I have a
> Montronics sytem that does comparisons by the multiply and mix process, and
> I find (also common to more modern Kethly systems) that the limitation is
> around a part in 10^10 where the noise on the phase output makes it not
> really usable (without a lot of averaging) being around or in excess of 90
> degreees even with a couple of very good OXCOs. How does the 10G comparision
> avoid this problem with standard multipliers? I doubt you can go all that
> way with low-noise multipliers and have any useful signal left, or have I
> missed something. At present I use a phase meter (lock in amps can be quite
> good) at the MHz range and datalog the phase drift for several hours. I have
> determined that setting "on the nose" is not necessary (for my
> applications). It is more useful to know how far a source is "off".
> How does the mix down compare with the seemingly more popular "mix down and
> timestamp" I understand from previous threads that this has more potential
> but might it also be as good even using simpler circuits that the NIST
> system.
>
> Thanks for all your efforts inthe background John..... great reading
> material !
>
> Alan G3NYK
>
>   
Alan

The principal limitation with dual mixer systems is the relatively large
phase shift tempco of the mixers (~ 10ps/C for 10MHz inputs).
If one uses a high end sound card (or equivalent ADC) to timestamp the
beat frequency zero crossings a resolution of better than 1E-12/Tau is
possible when the thermal environment permits.
However a stable low noise offset source frequency is required.
It is also essential to ensure there is sufficient isolation between the
sources being compared to avoid injection locking.

A low end sound card can also be used together with a Collin's style
bandpass limiter/slope amplifier to achieve similar performance if the
slope gain is large enough.

Bruce