[ntp:questions] Re: Help needed with excessive drift
david at djwhome.demon.co.uk
Fri May 6 21:03:32 UTC 2005
In article <f6qdna0Gb7sSU-ffRVn-sA at comcast.com>,
Richard B. Gilbert <rgilbert88 at comcast.net> wrote:
> It is sometimes possible to tune the crystal oscillator by connecting a
> trimmer capacitor across the crystal and adjusting the capacitor while
> measuring the frequency of the oscillator. Doing so will probably void
Generally, if the clock is out by that much, it is assumed that it
is also unstable, and loading it with a trimming capacitor is likely
to make it even more unstable. If the clock is stable, but so far out,
most operating systems have a facility to bulk correct (tickadj),
typically in steps of 100ppm. That's true of Linux, and SunOS at least.
Besides an out of specification crystal, the other common cause for
the frequency correction hitting the end stop is that something else
is trying to discipline the clock, for example the "hwclock -a" command
mentioned before. Many Unixes (e.g. SunOS and SCO Unix) will try and
align the software clock with the hardware clock unless you explicitly
disable the feature (requires a kernel rebuild on SCO).
Normally this results in steps, but a smaller correction might result
in the frequency hitting the end stop without associated steps.
More information about the questions