Re: [gpsd-users] PPS from kernel into gpsd

[Dave Hart]

On Wed, Apr 4, 2012 at 18:27, Tomalak Geret'kal <address@hidden> wrote:
> On 04/04/2012 19:05, Alexander Carver wrote:
>>
>> On 4/4/2012 11:01, Tomalak Geret'kal wrote:
>>>
>>> [..]
>>>
>>>
>>> Now I want to use this PPS to steer ntpd in SHM(1), along with the
>>> NMEA-derived time that's already serving me well in SHM(0).
>>>
>>> According to http://catb.org/gpsd/gpsd.html [1], gpsd can use the RFC
>>> 2783 kernel PPS API "for extra accuracy". Can I use it as the sole
>>> source of PPS? How? Do I add /dev/pps0 to gpsd's device list? [..]
>>
>>
>> Skip gpsd and use ntpd's ATOM driver (refclock 22) to directly read in
>> /dev/pps0 into ntpd.
>
>
> Ah.. that appears to make much more sense.
>
> It appears not to be entirely working for me yet:
>
> # ntpq -p
>     remote           refid      st t when poll reach   delay   offset
>  jitter
> ==============================================================================
> *SHM(0)          .SHM.            0 l   17   16  377    0.000   -6.959
>  74.995
>  PPS(1)          .PPS.            0 l    -   16    0    0.000    0.000
> 0.008
>  localhost       .INIT.          16 l    -   16    0    0.000    0.000
> 0.000
>
> But I suppose that's a question for another list now.

PPS(1) contradicts /dev/pps0.  Either provide a /dev/pps1 link, or
change the refclock server line to 127.127.22.0.  Also use "prefer" on
the 127.127.28.0 line.

Cheers,
Dave Hart