[time-nuts] Low-Cost Rubidium Performance
Charles P. Steinmetz
charles_steinmetz at lavabit.com
Fri Feb 10 02:24:42 UTC 2012
>It's a huge difference what comes in the device and what environment
>we put it in. A temperature dependence, is a systematic effect on the device.
>To some degree we can control temperature, we can predict
>temperature and deal with it. We can handle it as an engineering
>concept and do steering loops etc. It's pretty systematic to me.
True, but while we can arrange to run the oscillator in a
constant-temperature environment to see how it performs without
temperature variations, we can never operate the oscillator without
any internal frequency variation, which would be necessary to fully
characterize its dynamic temperature dependence. The best we can do
is measure both temperature and frequency (the latter containing both
"internal" and "external" errors), and attempt to correlate the
temperature effects by curve fitting. This gets more and more
difficult as the device's response to temperature exhibits second,
third, and higher order effects (hysteresis, nonlinearity, etc.,
etc.). So the best we can hope to end up with is an educated guess,
and one person's guess will be different from the next person's (and
from his or her own next run). In practice, we are unlikely to even
attempt to correct for anything but first-order effects.
Think about recording an orchestra in a concert hall near a busy
street. You can make a semi-independent recording of the traffic
noise (just like measuring the temperature in our oscillator
example), but the complicated manner in which the traffic noise
enters the hall and reverberates inside it makes it impossible to
subtract the traffic noise from the original recording. What you end
up with has more to do with the choices you made in modeling the
noise ingress than it does with the traffic noise itself. That is
why any tinkering one does with the data must be exactingly
explained, and accompanied by the raw data so others can see for themselves.
Better to build the isotemp chamber and get clean data, if you want
to know how your oscillator behaves independent of temperature. Just
don't think that you can count on it for the same stability once you
take it out of the chamber.
More information about the time-nuts