[ntp:questions] Interesting results with PPS

Mauro Fiacco Mauro.Fiacco at ipaccess.com
Thu Feb 8 09:01:25 UTC 2007


Isn't a good thing that if two clocks disagree they are discarded? It's
excellent! The local clock has PPS, therefore chances are that it is
accurate (based on historical data). The good frequency stability
values, provided by PPS, implies that time accuracy should be maintained
even if several samples are discarded. 

Assuming that the probability of jitter for different packets is
uncorrelated, only packets providing similar error values are used.
While large discrepancies are thrown out. Even for two packets, if they
agree on the measured error, the probability is that they are correct;
if they disagree it is not possible to take a decision and it is better
relay on the local clock.

Also, in my experience, I never seen samples being discarded... clock
hopping is the usual behaviour... But when it comes to NTP, experience is
often not enough!


> Mauro Fiacco wrote:
> > The only difference is that NTP server #3 employs two servers for
> > time synchronisation, while NTP server #4 employs only one server. 

> > I think the only difference is the working of the clustering algorithm.
> > The use of two NTP servers improves stability and accuracy when compared
> > to only one servers.

> _Never_ use two sources!

> You should always use either 1 or at least 4, having just two means that 
> if those two should ever disagree about the correct time, your client 
> will disregard both.

> Terje
Mauro Fiacco 

URL: www.ipaccess.com  

More information about the questions mailing list