[ntp:questions] Re: Public NTP server as Falsetick while preferred clock is GPS time

Roland roland.heise at gmx.de
Wed Nov 26 11:55:05 UTC 2003


roland.heise at gmx.de (Roland) wrote in message news:<9a27ce45.0311250036.3c7c3de2 at posting.google.com>...
> I am looking for some support to identify and resolve the issue of my
> configuration. I built an NTP time server by using the GPS NMEA
> protocol sentences from a Trimble SV-6 module and the 1PPS-signal of
> the same box. Due to the transfer rate of 4,800 baud I stretched the
> 1PPS-signal from 1..5us to about 110ms by using a CMOS timer chip 555
> to ensure that the signal can be converted to RS232 (using Maxim-IC
> MAX232) and detected by the serial port (UART 16550A). The PC system
> runs on a pentium III, 450 MHz with Linux kernel-2.4.21 and the
> corresponding kernel patch of the PPSkit 2.1.2. The NTP version is
> 4.2.0.
> Beside the ref-clock of NMEA and PPS a third ref-clock is used from a
> public NTP time server (ntp1.ptb.de). The relevant part of the
> ntp.conf is as followed:
>    server 127.127.20.0 mode 2 prefer minpoll 4 maxpoll 4
>    fudge 127.127.20.0 stratum 0
>    pps /dev/pps0 assert hardpps # PPS device
>    server 127.127.22.0 minpoll 4 maxpoll 4
>    fudge 127.127.22.0 flag3 1 # enable PPS API
>    server ntp1.ptb.de  # public NTP time server

Hello again
Learning from another conversation I understood that the PPS input is
probably triggered by the wrong edge. Thus I change the ntp.conf:
   server 127.127.20.0 mode 2 prefer
   fudge 127.127.20.0 stratum 0 flag2 1 flag3 1
   # pps /dev/pps0 assert hardpps
   # server 127.127.22.0
   # fudge 127.127.22.0 flag3 1
   server ntp1.ptb.de

I am not sure whether removing the pps device from the ntp.conf is
correct. However, that is what I read at
http://www.eecis.udel.edu/~mills/ntp/html/drivers/driver20.html
"If the Operating System supports the PPSAPI, RFC-2783, it will be
used."
Maybe, someone can comment on this subject.

The ntpq -p command shows that the offset for GPS_NMEA is about -1.2ms
and for the remote NTP time server about 2.2ms. I hope I can tune it a
little bit with the fudge parameter of time1.

Thanks and regards,
Roland.



More information about the questions mailing list