[time-nuts] Switching oscillators
magnus at rubidium.dyndns.org
Sun Dec 9 22:25:56 UTC 2012
On 12/09/2012 10:53 PM, Joseph Gray wrote:
> The finished project will be placed where I won't have regular access
> to it, so I think I have justification for wanting this complication.
> Also, I want to monitor which oscillator is in use, so my application
> will know.
First reaction would be to lock the TCXO to the OCXO. It's not that hard
to make a PLL which will start acting only when the OCXO is "ready"
(oven detection could be part of it).
Another approach than direct PLL locking could be to use injection
locking, but designing for it could be challenging.
Switching a clock is a bit sensitive, but one solution to achieve it is
to lock the OCXO to the TCXO when the OCXO comes around, and only when
that PLL has lock, then switch between them and just let the lock state
fade out on the OCXO control, as you kill the control loop.
A few ideas. I don't like switching of clocks, so I would avoid that if
possible, even if precaution to make sure they are locked to be
phase-aligned. I would go for the lock-up of the TCXO. Make sure that
the lock-in does not upset any of the measurements, possibly by making
the lock-in a little "weak" so the maximum rate of change isn't too great.
More information about the time-nuts