[LinuxPPS] ntptime status - relaunch of that issue

Tue Feb 24 08:58:57 CET 2009

Hello again,

once more I have doublechecked my setup procedure.
Although I have followed the setup instructions precisely, still the
jitter / offset problem appears!
I played around with different hardware and realised, that then faster a
machine is, then less the offset / jitter is.
For example on an amd64 X2 5000+ jitter/offset is about 0,100 - 0,010
+/- while with the Pentium M processor the offset/jitter rises up far
beyond 1,000...

I think, the Pentium M (1700MHz) should be way enough for a ntp server?

Can anyone send me a kernel .config file, if possible for a pps patched
2.6.28  kernel  and write down their ntp ./configure parameters...

Also, when building the ntp server, which timex.h or linuxpps.h files do
I have to put in which (sub)-directory under /usr/include?

Thank you all once more for your help.

regards,

Felix

Hal V. Engel schrieb:
> On Monday 16 February 2009 11:26:44 pm Felix Joussein wrote:
>   
>> Hello list,
>>
>> as I am really blocked out right now in my work, I re-launch my
>> question/issue regarding the ntptime status codes...
>>
>> Additionally to these facts mentioned below, I have made the following
>> observations during my comparison of the "should be" state and the "is"
>> state, apart from the status codes which prove, to me, that my setup is
>> messy:
>>
>> So additionally to the ntptime output already mentioned, the interval
>> stands still, always on 1s. On my ppsKit machines running on linux 2.4
>> the interval is increasing up to 256, also the offset and jitter from
>> ntpq-p -c rl for the pps source should not bounce more then from 0,001
>> to 0,010.
>> In my case it bounces from around -0,150 to 0,150.
>>     
>
> I don't see this on my machine.  After startup it takes about 1/2 hour to 
> stabilize and once it is stable my jitter is typically around 1 or 2 
> microseconds and my offsets are almost aways <20 microseconds.  So there is 
> definitely an issue on your configuration or hardware if your jitter is as 
> high as 150 microseconds.  
>
> What the cause is I don't know.  But looking at your configure command for ntp 
> I do see some differences from what I used and what is recommended in the 
> wiki.  Specifically I have not used ac_cv_var_tick=no, ac_cv_var_tickadj=no, 
> --enable-accurate-adjtime or --enable-linuxcaps.  I don't know what these do 
> or how they affect how ntp interacts with LinuxPPS.  I think most LinuxPPS 
> users are not using any of these configure settings.
>
> I think that the PPSKit patches implemented an in kernel time keeping consumer 
> (IE. time_pps_kcbind()).  LinuxPPS does not since this functionality is 
> optional.  This is the likely explanation for the interval always being 1s and 
> the missing PPS related status codes.   There have been some discussions here 
> about this but no one has stepped forward to implement this functionality and 
> this is very likely not a trivial undertaking. 
>
>   
>> I am for 100% sure, it is not my pps source, as I used on the one hand a
>>  real atom clock and in the lab a garmin 18lvc.
>> Both sources work as expected under my old Linux 2.4 setup.
>>
>> Also the rest of the hardware is not the problem... neither on a Pentium
>> M 1.7GHz nor on an amd64, same on both machines...
>>
>> Here is my previous post.
>> Please help me to get this work!
>>
>>
>> many regards,
>>
>> Felix
>>
>>
>> I have running ntp-4.2.4p4, built with the following options:
>> ./configure CFLAGS='$(CFLAGS)' \
>>                 ac_cv_var_tick=no ac_cv_var_tickadj=no \
>>                 --prefix=/usr \
>>                 --enable-all-clocks --enable-parse-clocks --enable-SHM \
>>                 --enable-ATOM --enable-accurate-adjtime --with-crypto \
>>                 --sysconfdir=/etc \
>>                 --enable-debugging \
>>                 --with-sntp=no \
>>                 --enable-linuxcaps \
>>                 --disable-dependency-tracking
>>
>>
>> My Kernel ist 2.6.28 with the corresponding pps patch.
>>
>> The output of ntptime is:
>>
>> ntp_gettime() returns code 0 (OK)
>>   time cd3455ac.081b5418  Wed, Feb  4 2009 18:58:04.031, (.031667534),
>>   maximum error 13322 us, estimated error 52 us, TAI offset 34
>> ntp_adjtime() returns code 0 (OK)
>>   modes 0x0 (),
>>   offset -11487.545 us, frequency -25.923 ppm, interval 1 s,
>>   maximum error 13322 us, estimated error 52 us,
>>   status 0x2001 (PLL,NANO),
>>   time constant 4, precision 0.001 us, tolerance 500 ppm,
>>
>> I'm concerned about the status 0x2001 (PLL,NANO) because when I use a
>> 2.4.33.2 kernel with the ppskit, same ntp version, same configure
>> options, ntptime gives the following output:
>>
>> ntp_gettime() returns code 0 (OK)
>>   time cd34564c.6f9324e4  Wed, Feb  4 2009 18:00:44.435, (.435839520),
>>   maximum error 529 us, estimated error 1 us, TAI offset 134514656
>> ntp_adjtime() returns code 0 (OK)
>>   modes 0x0 (),
>>   offset 2.688 us, frequency -108.801 ppm, interval 256 s,
>>   maximum error 529 us, estimated error 1 us,
>>   status 0x2107 (PLL,PPSFREQ,PPSTIME,PPSSIGNAL,NANO),
>>   time constant 4, precision 0.448 us, tolerance 496 ppm,
>>   pps frequency -108.801 ppm, stability 0.012 ppm, jitter 2.063 us,
>>   intervals 14986, jitter exceeded 6916, stability exceeded 258, errors 16.
>>
>> The  status 0x2107 (PLL,PPSFREQ,PPSTIME,PPSSIGNAL,NANO) is different.
>>
>> What makes the difference, at which part of the howto from
>> http://wiki.enneenne.com/index.php/LinuxPPS_installation
>> might I have not followed correctly?
>>
>> Thank you for your help,
>>
>> Felix
>>
>>
>>
>> _______________________________________________
>> LinuxPPS mailing list
>> LinuxPPS at ml.enneenne.com
>> http://ml.enneenne.com/cgi-bin/mailman/listinfo/linuxpps
>> Wiki: http://wiki.enneenne.com/index.php/LinuxPPS_support
>>     
>
>
> _______________________________________________
> LinuxPPS mailing list
> LinuxPPS at ml.enneenne.com
> http://ml.enneenne.com/cgi-bin/mailman/listinfo/linuxpps
> Wiki: http://wiki.enneenne.com/index.php/LinuxPPS_support
>
>
>   