[LinuxPPS] LinuxPPS Digest, Vol 29, Issue 4

[Hal V. Engel]

On Thursday 11 September 2008 09:44:44 am david wrote:
> Thank you Hal for your help,
>
> I patch ntp with the file nmea.patch (link in wiki) and the diff with
> original give this :
snip

I do not know much about this patch since I don't have a NMEA GPS.  I have to 
assume that using the one linked to in the wiki is OK.

>
> AFTER, I build Makefile like this :
>
> ./configure --disable-all-clocks --disable-parse-cl
>                               ocks --enable-NMEA --enable-LOCAL-CLOCK
> --prefix=/usr

This is OK.

>
> and make.
>
> I run ntpd like this : /usr/src/ntp-4.2.2p4/ntpd/ntpd -c /etc/ntp.conf
>
> NOTE: If I check timex with a grep output I show that :
>
> ./configure --disable-all-clocks --disable-parse-cl
>                               ocks --enable-NMEA --enable-LOCAL-CLOCK
> --prefix=/usr | grep timex
> checking timex.h usability... no
> checking timex.h presence... no
> checking for timex.h... no
> checking for sys/timex.h... yes
> checking for __adjtimex... yes
> configure: WARNING: *** But clock_settime() would be better (if we had
> it) ***
> checking if sys/timex.h has STA_FLL... yes
>
> Is it correct or not ?

Yes this looks OK.  You should also check to see if NMEA is OK during the 
configure.  But looking at your output from ntpq -pn it looks like the NMEA 
driver is working and that it is seeing PPS pulses since your reach is 377.  
This will only happen if the driver is seeing PPS pulses.  

Also have you unloaded the ktimer module?  You need to do this to make sure 
that the NMEA driver is seeing the right PPS pulses.

The issue I think is that your offset (10.107 milliseconds) and jitter (3.417 
milliseconds) are very high for a machine with a refclock.  But it is normal 
for these to be high when the ntpd daemon is first started.  If there is valid 
data in the ntp.drift file it will in most cases take a few hours for the 
clock to stabilize and reach offset and jitter values that are expected from a 
GPS disciplined ntpd (IE. < 10 microseconds most of the time).  And much 
longer if there is no drift data like would be the case with a new 
installation.   How long after starting ntpd did you wait before getting these 
values?  Have you let ntpd run for an extended period (like 24 hours or more) 
to see if these numbers get better?

Also what do you get when you run ntptime?  Do you get a status that includes 
NANO?  You should see something like this:

$ ntptime
ntp_gettime() returns code 0 (OK)
  time cc73d8f3.ff20748c  Thu, Sep 11 2008 10:50:43.996, (.996589562),
  maximum error 6738 us, estimated error 1 us
ntp_adjtime() returns code 0 (OK)
  modes 0x0 (),
  offset -2.509 us, frequency 87.862 ppm, interval 1 s,
  maximum error 6738 us, estimated error 1 us,
  status 0x2001 (PLL,NANO),
  time constant 4, precision 0.001 us, tolerance 500 ppm,

Hal


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://ml.enneenne.com/pipermail/linuxpps/attachments/20080911/9fdd3093/attachment.htm