[time-nuts] Know thy components
James Flynn
james.flynn at csun.edu
Fri Aug 7 09:16:30 EDT 2015
Hal Murray <hmurray at ...> writes:
> > Also the carbon composition resistor drifted with time at elevated
> > temperature.
>
> I assume the drift is slow enough so that any any control loop will
easily
> track it.
Actually, "ageing" is a better word than "drift". After the major change
in resistance due to temperature, the change in resistance was a slow
exponential, taking several weeks to level off. The change in the first
day or so was pretty dramatic, causing an error of about 2E-9 per day in
the oscillator. Since my application was looking at the source signal
over the course of a day, the error was appreciable. Even with a GPSDO,
the window has to be in the range of 10000s for the level of precision I
am shooting for.
After the oven stabilized, the changes due to temperature were quicker
and on the order of 2e-10 in a few minutes. This was unacceptable for my
application due to the averaging times. Even the average diurnal change
in ambient temperature was visible in the error plot.
>
> What is the drift like at non-elevated temperatures?
I have not studied this, but I would think that, like other ageing
processes, it would be less.
>
> What is the drift like for good resistors?
Also not studied yet.
> If I'm using 2 resistors as a divider, how well do they track? Does
it
> matter if they have same or different values?
IF you assume they have the same change in percent of their value per
degree C, it would not make any difference, regardless of their relative
sizes. However, I cannot write that "IF" big enough. Even resistors out
of the same batch may not change in the same way. I would not rely on
them tracking. Also if other resistors are included in a configuration
that is not a divider, all bets are off.
>
> The problem I see with resistors (or voltage reference or ???)
drifting is
> that it gets tangled up with the crystal drifting. ...
Exactly. And exactly why I have changed the resistors to ones I have
tested to be less than -1ppm/deg C.
More information about the time-nuts
mailing list