[ntp:questions] Panic stop captured

A C agcarver+ntp at acarver.net
Sat Dec 15 20:15:06 UTC 2012


On 12/15/2012 02:43, Rob wrote:
> A C <agcarver+ntp at acarver.net> wrote:
>> Yes, the SHM is configured currently as the preferred peer and PPS is
>> the system peer usually.  SHM is coming from gpsd and PPS is coming
>> directly from a serial port (separate from the one used by gpsd).
>
> I take it that you then use the first SHM provided by gpsd, the one
> that is calculated from the serial messages from the GPS receiver?
> It is quite usual that it wobbles by several hundred milliseconds
> on some receivers.
>
> The second SHM provided by gpsd is the PPS one, but in reality it is
> not independent of the first one, because of a limitation of SHM.
> It is not possible to convey PPS information via SHM, only absolute
> time information.  So gpsd works around this by adding the current
> system time to the relative time provided by the PPS signal.  As
> this is a circular definition, gpsd attempts to keep sanity by only
> putting valid time into the second segment when the system time is
> within some defined offset from the time received in the GPS serial
> messages.

I only use the first SHM from gpsd (the time stamp).  I do not use the 
PPS SHM from gpsd.  I use the ATOM refclock (22) and send the PPS signal 
directly to ntpd via DCD on a second serial port.  Yes, the first SHM 
does wander around a bit but for the most part it stays on one side of 
zero (on purpose, I have been tweaking the fudge time to try and keep it 
single sided).



More information about the questions mailing list