[ntp:questions] Re: PPS ATOM clock set to falseticker

David L. Mills mills at udel.edu
Wed Aug 27 13:37:26 UTC 2003


Ulrich,

You remind me it's time to remind this list again.

The various versions of NTP all do the same thing in the Byzantine
selection algorithm. See the architecture briefing at the NTP project
page linked from www.ntp.org for reference in the following. The
algorithm must find a majority clique of truchimers in order to set the
system clock. This means the correctness intervals of a majority of
sources must overlap. What usually happens is that the GPS serial
timecode is far enough from the PPS signal that the intervals do not
overlap. The recommended fix is to use the calibration feature to
suppress what little evil differences exist between the serial port and
PPS interrupts.

See the reference clock drivers page in the documentation for the
calibration feature. While this feature is in the current development
version, the release version is so old that I have no idea what it does.

Dave



More information about the questions mailing list