[ntp:questions] About use of PPS in NTP sync

Rob nomail at example.com
Mon Jun 16 07:33:10 UTC 2014


jthulard at gmail.com <jthulard at gmail.com> wrote:
> Hello,
> I'm looking for a way to speed up the ntp convergence of a system which would be restarted after several days being off. Does the use of PPS improve this convergence time ?
> This is local configuration, with one LAN and one NTP server, with about 30 NTP clients.

PPS makes the final convergence a lot faster, because it is usually being
run at a high sample rate (once every 16 seconds).

However, there will still be errors.  This is because ntpd when starting
does not yet enable the PPS, it first wants to be in sync with other
sources (like NMEA or NTP).

So what happens when you restart ntpd (or the system) is: it first considers
the other sources and starts syncing to them, and only when this has been
successful it re-enables the PPS and starts syncing to that.  As the
external sources are never as accurate as PPS is, this will mean the clock
first drifts away from correct, and then drifts back to the correct time.
This also causes a miscalculation of the frequency offset, which also has
to be corrected again afterwards.

It could be argued that instead it should do no syncing during the evaluation
of the other sources, and when after a few minutes it has detected that
they are OK, it should enable the PPS and sync to that.



More information about the questions mailing list