[gpsd-users] Sudden shift in time?

[Alexander Carver]

Does anyone have an idea why the time reported to ntpd from gpsd (via 
SHM(0) ) would suddenly be off by more than half a second?

I'm running tests with ntpd on one system so I'm constantly watching the 
peer status.  This system has seven internet peers, a PPS input (not 
from gpsd) and the SHM(0) input from gpsd.  The SHM(0) clock had been 
reporting time to within about +/-50 ms of the clock tick and then out 
of the blue it was reporting -680 ms (+/-10ms).  The jitter reported was 
of the same magnitude (opposite sign) for the very first cycle where the 
large offset appeared.  It settled back to a more normal jitter of about 
10ms

I didn't do anything else, I just watched it to see what would happen. 
On another machine I query this instance of gpsd's satellite info and 
all looked good there.  Ten satellites were visible, eight used in the 
fix, HDOP was 1.2.

The offset managed to fix itself with no intervention from me.  It was a 
sudden step correction again just like when it went haywire.  Again the 
jitter reported was huge for the first cycle after the step change 
occurred and then settled back to a normal value.

The system clock did not change during this process.  The SHM(0) clock 
was labeled a false ticker and kicked out of the calculations.  The rest 
of the internet clocks and the PPS signal did not change their offsets 
or show any other unusual behavior so it was confined to just gpsd.

I'm just not sure what would cause such a sudden, huge offset.  Most of 
the time the offset wanders around in small increments but stays within 
the +/-50 ms window.