[ntp:questions] Why does GPS time diverge from system time?

Charles Elliott elliott.ch at verizon.net
Fri Apr 13 19:01:22 UTC 2012


I just installed NTPD version ntp-4.2.7p265-win-x86-bin from Dave Hart's web
site.  I am using a SiRF BU-353 (USB) GPS device.  The time2 fudge factor is
0.372258.  NTPD is always synced to the GPS device since all the remote
clocks are marked noselect.  From clockstats.20120413, the first entry after
the ntpd restart indicated a difference between the GPS and system time of
0:00:00.450 seconds.  Here is the clockstats line:

 

56030 48614.450 127.127.20.2
$GPRMC,133014.000,A,3959.5981,N,07507.5565,W,0.51,40.43,130412,,,A*4F

 

The last clockstats line indicates an offset between GPS and system time of
0:00:00.658 seconds:

 

56030 56152.658 127.127.20.2
$GPRMC,153552.000,A,3959.5979,N,07507.5549,W,0.20,74.34,130412,,,A*46

 

I have copied all the clockstats lines since the restart below, where all
duplicate offsets have been deleted.  You can see that the difference
between GPS and system time has been increasing almost monotonically.   The
average offset since the restart from the offsets computed by NTPD and
output in the loopstats file is -0.054193 (ms), while the stdDev is
1.507155.  The maximum and minimum offsets are about +5 and -5 ms,
respectively, while the overwhelming majority of offsets are between +- 3
ms.

 

Why is the difference between GPS and system time diverging?  Shouldn't they
be converging?

 

Also, why is the delay for all remote clocks always 0.977 in this version of
NTPD, where as normally it is about 0.230 ms?  In addition, why is the most
common computed jitter equal to 0.977? 

 

 

Charles Elliott

 

For peerstats file see pastebin.com.  Search for Peerstats.  You may be able
to find it here: http://pastebin.com/E5XktXSR.

 



More information about the questions mailing list