[time-nuts] Question on crystal jumps

[Bruce Griffiths]

Steve Rooke wrote:
> Bruce,
>
> 2008/10/24 Bruce Griffiths <bruce.griffiths at xtra.co.nz>:
>   
>> You've missed a vital point:
>> There is no RF output from the rubidium cell itself at all.
>> It doesn't oscillate as does an active hydrogen maser.
>>     
>
> Thanks for putting me right on this, I obviously need to study this
> more. So as I see it, a CPU measures the output of the rubidium lamp
> and tunes a signal source which is used to excite the lamp to give the
> highest output. The output of the signal source is used to discipline
> a 10MHz ocxo and the output of this becomes our reference standard.
> Well, we are at the mercy of the xtal oscillator, if it jumps, so does
> the output, straight away, and it does not matter how good our high Q
> rb standard is as there will be a jump which should be corrected by
> the feedback circuit, IE. the output of the ru lamp photo detector
> will go down and the cpu will have to correct it causing the oxco to
> be corrected again. How long this takes is down to the time constand
> of the feedback mechanism.
>
> OK, even given this setup, would it not be possible to remove the ocxo
> from the loop and just use the cpu to control a non-xtal'd signal
> source rb exciter. The output of this signal source could then be
> divided down to 10MHz and filtered. I appreciate that using a rb
> exciter that is not disciplined would not be as stable but with good
> design this can be mitigated and, once the signal generator is on
> frequency it should be possible to keep it that way with the optical
> feedback to the cpu without a lot of noise from the cpu hunting for
> the photo detector peak. After all, in the ocxo design, this is also
> being varied by the cpu and can be kept stable so why not the rb
> exciter oscillator as well.
>
> 73 - Steve
>   
Steve

The change in light transmission is very small ~ 0.1% so an AC phase
sensitive detection technique is used.
The probe signal frequency is modulated at a relatively low frequency
(usually < 200Hz).
When the interrogating frequency is tuned to the resonance the
transmitted light flux has a modulation component at twice the
modulation frequency and none at the modulation frequency.
When the frequency is slightly off the transmitted light flux has a
frequency component at the modulation frequency with a phase that is
indicative of whether the frequency is high or low.
For small frequency errors the amplitude of the light flux component at
the modulation frequency ids proportional to the frequency error.
Thus a synchronous detector can be detect the transmitted light flux
component at the modulation frequency and steer the frequency to the
rubidium hyperfine transition frequency.

The signal to noise ratio isnt that high so a low noise oscillator with
good short term stability is required.
Typically a crystal oscillator is used because very few other
oscillators have adequate short term stability.

Superconducting resonator and cryogenic sapphire stabilised oscillators
are exceptions.
A conventional LC oscillator, is too unstable to be useful.

Bruce