[time-nuts] Question on crystal jumps
sar10538 at gmail.com
Fri Oct 24 02:52:42 UTC 2008
2008/10/24 Bruce Griffiths <bruce.griffiths at xtra.co.nz>:
> The rubidium standard is passive it doesnt oscillate, it merely acts as
> a high Q filter whose resonance can be probed with an external source.
> Thus you need to replace the crystal oscillator with another low noise
> source to interrogate the rubidium resonance.
So, do I take it you are suggesting that the xtal oscillator is used
to excite the rubidium standard? I can't see how that works, can you
please comment further.
> To be useful the bandpass filter would need a bandwidth of a few
> milliHertz or less.
Well, if the xtals we use are capable of resonating sufficiently well
to provide us with a freq standard that is acceptable for our uses,
IE. within a few miliHertz, I can only see that using multiple xtals
can only improve on the stability of a singe xtal in an oscillator
circuit. Remember that the output of these GPSDOs, and as I understand
it rubidium standards, are really the output of the xtal oscillator.
The centre frequency of that xtal oscillator is relatively slowly
varied to be correct by it's difference to the GPS signal, or rubidum
stage, via some form of comparator driving a low pass filter and onto
the EFC. Hence it is the xtal oscillator which is the signal source we
> The crystals in the lattice filter will also experience similar jumps.
But it is highly probable that they will not all jump at the same time
for reasons I have given in a previous posting. It's similar to what
other people have described about using three or more xtal oscillators
and having a vote on what the frequency should be. Trying to control
multiple oscillators, and implement the voting arrangement, would be
quite complex and has a high probability of being prone to other
effects. A lattice filter is very simple in comparison, you have the
same voting scheme of multiple xtals (and you can have many more than
three) but none of the complexity of the oscillator arrangement. You
just have to make sure that it is driven with the reference standard
and all the stages could be manually tuned for peak after a burn in.
As the xtals change over time the level may drop but the frequency
would remain as stable as the input reference. Of course, if you used
just a very simple lattice, it's possible for all the xtals to drift
in one direction which would make the filter less able to suppress
noise sidebands but it would need to have some math done to
investigate the implications of this. Also a series parallel design
would probably circumvent this.
Sorry, just thinking out loud.
73 - Steve
Steve Rooke - ZL3TUV & G8KVD
Omnium finis imminet
More information about the time-nuts