Re: [gpsd-users] Help with extracting time information from GPS to NTPD

[Dave Hart]

On Wed, Nov 9, 2011 at 03:38, Jordan <address@hidden> wrote:
> Thanks!  I downloaded the ntp source and recompiled from scratch with the
> SHM driver.  NTPD can now see the GPS time and is syncing to it.  Now I'm
> noticing another issue where the reported time from the GPS is ~24 seconds
> ahead of the public time servers.  I did a quick google search and saw this
> matches up with Loran-C time vs UTC.  Is it possible the GPS is reporting in
> Loran-C time?  If so, can I compensate for this in NTPD?

The easy answer is yes, you can calibrate out any constant offset from
a ntpd reference clock such as SHM using fudge time1 and/or fudge
time2.  In your case you probably want something close to:

fudge 127.127.28.X time1 -24

That would presumably need additional adjustment with each additional
leap second inserted or deleted in UTC.  Moreover, it means other gpsd
clients will still have incorrect times.  It seems very odd a GPS
receiver would be reporting the LORAN C timescale, though it is a
fixed number of seconds off from the GPS timescale (and both are a
changing but whole number of seconds off from UTC).  My belief is that
it is rare for a GPS to provide calendar-style dates in anything but
UTC.

http://leapsecond.com/java/gpsclock.htm is a useful reference here.

Cheers,
Dave Hart