[ntp:questions] Re: PPS source not always being chosen?

David L. Mills mills at udel.edu
Mon Aug 15 18:40:25 UTC 2005


John,

See the description of the clustering algorithm in the architecture 
briefing on the NTP project page. Consider the offsets between each 
truechimer and the others and toss out the hombre with the largest 
offset relative to the others. The PPS signal was tossed out, leading to 
exactly what you observe.

If you really want the most precise time, don't allow so many cows in 
the herd. Look at the timecode jitter, which is huge compared to the 
others. Also, it shows a large offset relative to the PPS signal. If 
only the timecode and PPS were allowed in your herd, there would be no 
majority clique and the client would not synchronize to either source. 
If you can't fix the jitter and offset, tinker the minimum dispersion 
something larger than 10 ms.

I haven't looked at the HP driver, but the large offset discrepancy 
could be the result of not using the median filter properly. You want as 
many offset samples from the clock as possible, say at one-second 
intervals, before sweeping the median filter at poll times.

Dave

John Ackermann N8UR wrote:

> I'm seeing a strange quirk in the selection process of my NTP servers
> that use refclocks that provide both a serial data timestamp as well as
> a PPS signal.  I mentioned this some time ago, and with the most recent
> 4.2.0a-20050720 snapshot (running on FreeBSD 5.4), I'm still seeing the
> same thing.
> 
> Attached below are the results of ntpq -p showing that the PPS source,
> while apparently far better in jitter and offset, is being deselected in
> preference to the serial timestamp source.  That just doesn't seem
> right.  Is there something in my configuration (sample attached below)
> that might be causing this?
> 
> Any suggestions on how to solve this would be greatly appreciated.
> 
> Thanks,
> 
> John
> ----
> 
> jra at flob:~$ ssh tick ntpq -p
>      remote           refid      st t when poll reach   delay   offset 
> jitter
> ==============================================================================
> *GPS_HP(0)       .GPS.            0 l    4   64  377    0.000    2.582  
> 4.843
> -PPS(0)          .PPSG.           0 l   16   64  377    0.000   -0.003  
> 0.004
> -tock.febo.com   .GPS.            1 u   16   64  377    0.481    0.003  
> 0.069
> +toe.febo.com    .PPSW.           1 u   17   64  377    0.460    3.111  
> 0.098
> -time-A.timefreq .ACTS.           1 u   24   64  377   49.374    1.269  
> 1.193
> +tock.usno.navy. .USNO.           1 u    6   64  377   51.010    3.595  
> 1.227
> +ntp2.usno.navy. .USNO.           1 u   51   64  377   51.232    3.525  
> 1.578
> 
> 
> # /etc/ntp.conf, configuration for ntpd
> 
> # ntpd will use syslog() if logfile is not defined
> logfile /var/log/ntpd
> 
> driftfile /etc/ntp.drift
> statsdir /var/log/ntpstats/
> 
> statistics loopstats peerstats clockstats
> filegen loopstats file loopstats type day enable
> filegen peerstats file peerstats type day enable
> filegen clockstats file clockstats type day enable
> 
> enable pps
> 
> # Z3801A
> server 127.127.26.0 ttl 1 prefer
> fudge 127.127.26.0 time1 -0.963
> fudge 127.127.26.0 refid GPS
> 
> # PPS
> server 127.127.22.0
> fudge 127.127.22.0 refid PPSG
> fudge 127.127.22.0 flag3 1
> 
> #server tick.febo.com iburst
> server tock.febo.com iburst
> server toe.febo.com iburst
> server time-c.timefreq.bldrdoc.gov iburst
> server ntp1.usno.navy.mil iburst
> server ntp2.usno.navy.mil iburst
> 
> 
> _______________________________________________
> questions mailing list
> questions at lists.ntp.isc.org
> https://lists.ntp.isc.org/mailman/listinfo/questions
> 




More information about the questions mailing list