[time-nuts] Question on disciplining oscillators

[Javier]

Hello all,

I'm working on discipling a FRS-C Rb oscillator using a M12T GPS and a 
AT91SAM7S64 micrcoprocesador (ARM architecture) using the same concepts 
described in the famous Brooks Shera article. I've two questions: For 
timekeeping, it is better to turn on in the GPS both tropospheric and 
ionospheric models, only one of them, or none. The other question is 
about the loop filter. Currently I'm playing with a PI filter, with two 
switched time constants (one for relatively fast first lock and once 
locked, a slower one). I would like to know, in your experience, what it 
the time constant that would be optimal for this kind of oscillator.

Thanks. Best regards to all,

Javier, EA1CRB