[time-nuts] Big Picture View

[Keith E. Brandt, M.D.]

John,
That's a good start towards answering my question. Given my set-up of 
GPS referenced NTP server, how would I integrate a frequency 
reference into the system?

Keith

 >    As a relative newbie to the who time-nut scene, I'm still working on
> >    the big picture of how all the hardware fits together. I've
> >    successfully put together a FreeBSD box with a Garmin 18 LVC as an NTP
> >    server and am getting consistent microsecond accuracy. I've been
> >    looking at the next step of adding a stable oscillator to the system.
> >    I'm not seeing how to integrate this into the system.
> >    The way I understand it, the oscillator is there to provide a very
> >    stable frequency reference, regardless of the type used --> VXCO,
> >    rubidium, cesium, etc. What I'm not seeing is how you link the
> >    frequency reference to UTC. For NTP, I'm guessing that you'd have to
> >    sync a PPS pulse with a known time source such as GPS. Am I on the
> >    right track? Pointers to references appreciated.
> >    73, Keith
>
>Hi Keith --
>
>Frequency isn't linked to UTC; phase is.  What you're doing with an
>external oscillator is simply providing the computer with a stable clock
>frequency.  The PPS signal from the GPS provides the phase information
>that tells NTP when the second marker is.  The stable clock frequency
>just makes it easier for NTP to hold the second marker at the proper point.
>
>Even if the reference frequency is inaccurate (and that may be the
>question your asking), NTP is still better off as it can correct for a
>frequency offset more easily than it can a clock that is erratic.
>
>Hope this helps.
>
>73,
>John