[ntp:questions] Re: long-term oscillator frequency
maarten at kittensandcats.net
Thu Aug 7 20:48:14 UTC 2003
Harza wrote in message ...
>Can NTP adjust long-term oscillator frequency if the hardware supports
>that feature? Can it also introduce a temporary change in oscillator
>frequency to adjust to UTC? Or even introduce small time gaps when
>frequency cannot be
>changed at all? Making it is possible that NTP hosts do not have a
>frequency close to true UTC clock?
So what would be the frequency of a true UTC clock?
AIUI, the OS clock is kept by incrementing a counter at regular
intervals, and what can be twiddled is not the regular intervals
but the counter increment.
So if your oscillator that was sold as 100 Hz turns out to be
101 Hz, NTP might change the increment from 1000 to 990, and
most of the error would go away. (I think the terms are that
in this hypothetical case, time would be precise in five digits
behind the decimal point, but accurate only in two.)
More information about the questions