[LinuxPPS] NMEA Refclock in ntp-dev does not work

[William S. Brasher]

On Mon, 19 Oct 2009, Rodolfo Giometti wrote:

> On Sat, Oct 17, 2009 at 07:01:12AM +0200, Udo van den Heuvel wrote:
> > On 2009-10-16 20:01, William S. Brasher wrote:
> > > 
> > > The NMEA refclock, refclock_nmea.c, from ntp-dev-4.2.5p233-RC does not 
> > > work for me.  NTP refuses to use it as a time source and rejects it and 
> > > PPS.
> > 
> > That is correct.
> > After three years the ntp folks succeeded in not incoroorating the
> > needed patch to refclock NMEA but to do their own changes to PPSAPI.
> 
> Which changes they applyed to the PPSAPI?
> 
> Ciao,
> 
> Rodolfo
> 

I don't know of any changes to the PPSAPI itself, except for the following 
comment at the top of refclock_nmea.c in ntp-dev-4.2.5p235-RC: 

 * Converted to common PPSAPI code, separate PPS fudge time1
 * from serial timecode fudge time2.
 *              Dave Hart July 1, 2009



I have had "configuration issues" with using ntp-dev-4.2.5 on 
linux-2.4.33-NANO, an elderly 2.4 kernel with the PPSkit patch from years 
ago.  Right now the 2.4 box will sync to GPS_NMEA but rejects PPS.

The 2.6.30 and 2.6.31 boxes reject both GPS_NMEA and PPS.  The GPS_NMEA on 
both appears way behind the pps pulse, "looking like" refclock_nmea 
doesn't see the pps signal.  It really looks like the nmea patch has not 
been installed.


All boxes, linux-2.4, linux-2.6.30, and linux-2.6.31 all are happy with 
the stable version of ntp, ntp-4.2.4p7, and sync with GPS_NMEA and PPS.


Since I haven't used ntp-dev-4.2.5 in quite a while, I can not rule out a 
build or configuration error.  I can say that it does not work wih pps
on any of my boxes.