[ntp:questions] Assistance with PPS on Windows

Martin Burnicki martin.burnicki at meinberg.de
Wed May 11 17:12:52 UTC 2011


Hi Ken,

sorry for the late reply. I'm currently on a business trip.

Ken Link wrote:
> Hi Martin,
> 
> We had considered using the time adjustment service from the driver
> package at one point, but it seemed to conflict with NTP's
> configuration at the time. If I disable NTP entirely and only use
> MbgMon and it's time adjustment service, things look great and the
> clock is synced to +/- 0.003s from the IRIG signal.

Hm, 3 *milliseconds* is worse than it should be. Where do you see this
value?

> However I can't do 
> that for two reasons:
> 
> 1) I would like to share this highly accurate time as an NTP stratum 1
> peer (as well as verify it's accuracy by running NTP with multiple
> peers configured), but when I run NTP and the MbgMon time adjustment
> service at the same time, NTP tries adjusting the clock and causes a
> drift of a few hundred msec and high jitter because the two services
> are fighting over the clock.

Hm, the configuration we suggest under Windows is to let the Meinberg time
adjustment service discipline the Windows system time, and run ntpd in a
configuration with local clock at stratum 0 only, so ntpd doesn't touch the
system time but makes the synchronized system time available on the
network.

> 2) In the case where the IRIG signal fails or is inaccurate (e.g.
> Arbiter clock becomes unlocked from GPS signal) there is no
> redundancy, validation of correct time/clock discipline, or other time
> server to connect to unless I am running NTP, which goes back to point
> #1.
> 
> In a perfect world I could run MbgMon's time adjustment service and
> NTP together (with multiple peers), NTP won't touch the clock until
> the IRIG signal is determined to be bad (loss of IRIG signal, loss of
> GPS signal at the clock, or local clock is rejected when compared to
> other peers), and other peers can rely on this computer for accurate
> and stable time via NTP.
> 
> Because we couldn't figure that out, we decided to give PPS a try. Is
> there a way to configure NTP to play nice with MbgMon?

See above. I must admit there is no possibility for redundancy for ntpd in
the configuration mentioned above. As soon as you configure one or more
additional time sources for ntpd ntpd would start to adjust the system time
and thus work against the time adjustment service from the driver package.

So maybe what you are trying is helpful in your case. Anyway, I thought I'd
like to hear about the nature of the problems you have encountered.

Regards,

Martin
--
Martin Burnicki

Meinberg Funkuhren
Bad Pyrmont
Germany




More information about the questions mailing list