[ntp:questions] Re: Frequency and leapseconds!

David Woolley david at djwhome.demon.co.uk
Fri Mar 11 07:12:19 UTC 2005


In article <mailman.10.1110506028.576.questions at lists.ntp.isc.org>,
erik corell <cii99ec7 at i.lth.se> wrote:

[ BROKEN THREADING - NO REFERENCES HEADER  - missing In-Reply-To on
  mailing list interface, or public reply to private message ]

[ not sure how this now relates to leap seconds ]

> I have really tried to look into this issue, and I think that the phase =
> is adjusted

This is all in published source code of for Linux, FreeBSD and ntpd.  The
details depend on the OS, and whether or not the kernel PLL is enabled, but
basically some phase correction is applied on every tick.

However, as a short answer to your problems what you need to do is:

- not simulate catastrophic failures of the timing crystal - if you get
  a 373 ppm step in the actual crystal frequency, throw the mother board
  away, and discard all measurements that used it;

- not override the poll interval to less than the compiled in default
  minimum value, but instead let it find its own level - the best frequency
  stability will occur when it is towards the high end;

- not worry about ntpd's idea of the frequency - for most machines, the
  frequency used by the TSC won't vary by more than one or two ppm, 
  except as a result of large steps imposed by power management.



More information about the questions mailing list