[ntp:questions] Re: Some anomalies

Vaidotas J. dtroit at ktu.lt
Tue Dec 7 12:08:36 UTC 2004


Hello,

I think that You haven't analyzed my link with graphs, where every step is beeing described ;)
Also I found the cure already - i had disabled event polling option on my refclocks driver, and after enabling it - the time variation stabilized to about -10 : 10 microseconds (on notebook), and on my WS the offset is now almost always 0 usecs (-1 : 3 the largest values).


On Sat, 04 Dec 2004 08:37:18 +0100
mike <michael.no.spam.cook at wanadoo.fr> wrote:

> You say your platform is a laptop. Have you disabled the power saving
> features??? These cause clock frequencies to be changed, which throws 
> ntp. I think it's a FAQ.
> 
> 
> 
> Richard B. Gilbert wrote:
> > Vaidotas J. wrote:
> > 
> >> Hello,
> >>
> >> recently my office decided to create some QoS monitoring system and 
> >> the main goal of it's correct results on some tests is precise time 
> >> synchronization. So we bought few Trimble Acutime 2000 SyncKits and 
> >> I've conected them at first to my WS and then to project notebook. 
> >> It;s a little bit pitty, that (as far as i know), that Trimbles 
> >> Palisade ref clock driver doesn't use ports B output also no PPS usage 
> >> too ;(
> >> But anyway, the synchronization is received and so on... After some 
> >> time, i wrote this veeeery small c program to get the offset value 
> >> from ntp_adjtime() function for one hour (calling it every 1 second). 
> >> Here it is:
> >> #include <sys/timex.h>
> >> #include <sys/time.h>
> >> #include <stdio.h>
> >> int main(){
> >> int i;
> >> struct timeval ts;
> >> struct timex *adjtm; //ntp_adjtime pointeris
> >> adjtm = (struct timex *)malloc(sizeof(struct timex));
> >> for (i=0;i<=3600;i++) {
> >> ntp_adjtime(adjtm);
> >> printf("%i\t%i\n",i, (unsigned int) adjtm->offset);
> >> sleep(1);
> >> } //foras
> >> }
> >> And i graphed some results enabling/disabling some kernel functions:
> >> http://orka.litnet.lt/~dtroit/
> >> (gnuplot was used as a ploting engine, with just simple input x,y from 
> >> file, generated by this smaaaal c program of mine:)
> >>
> >> So in one graph we see that for sometime offset was 0, in some of the 
> >> graphs (that are wildely absurd) the offset is taken just right after 
> >> the ntpd is started and other graphs are commented.
> >> The last graph I think is quite good and stable (that last jump is a 
> >> mystery to me), but such offset with direct startum0 connected to PC 
> >> isn't satisfieing. So maybe any ideas/responses to these graphs or 
> >> some experience using Trimble GPS Acutime? What's the offset with Your 
> >> Trimble Acutime and the settings of the system that runs it and other 
> >> things. So if any ideas please help me to get the smallest ofset :) 
> >> Because rigth now i've no more ideas :)
> >>
> >> PS> The notebook is:
> >> model name      : Intel(R) Pentium(R) 4 Mobile CPU 1.70GHz
> >> System parameters:
> >> Linux ippm 2.6.9 #4 Fri Nov 26 16:39:29 EET 2004 i686 GNU/Linux
> >> Debian 3.1 distr.
> >> System has no other processes than ntpd, apache, sshd and the GPS is 
> >> connected via UST-RS232 converter (no comm port on this notebook).
> >>
> >> Sincerely,
> >> Vaidotas J.
> >> LitNET,
> >> Network engineer,
> >> http://www.litnet.lt
> >> dtroit at ktu.lt
> >>  
> >>
> > My first guess is that you are having problems receiving the satellite 
> > signals.  Further, I'd say that the times when you show a long run of 
> > zero offset, you are getting a solid signal from four satellites.  Four 
> > satellites are necessary and sufficient to solve for latitude, 
> > longitude, height, and time.
> > 
> > I'm not familiar with your Trimble hardware or the driver for it but; in 
> > principle, if you know your position accurately you need only one 
> > satellite to get the time.
> > 
> > Things to check:
> > 1.  Does your antenna have a clear view of the entire sky?  If not, find 
> > a better location for it.
> > 2.  Use nptq -p and check the "reach" field for your reference clock.  
> > If it is not 377 you have a problem!
> > 2.  Does your driver allow you to ignore satellites close to the 
> > horizon?  If so, it may help to enable this feature.
> > 
> > Another possibility is that both Linux and Windows have a reputation for 
> > losing interrupts!  Losing clock ticks will do horrible things to your 
> > timekeeping!!!!
> _______________________________________________
> questions mailing list
> questions at lists.ntp.isc.org
> https://lists.ntp.isc.org/mailman/listinfo/questions



More information about the questions mailing list