[ntp:questions] Tiime Sync solution

David Woolley david at djwhome.demon.co.uk
Tue May 31 20:58:59 UTC 2005

In article <1838664.rKQ7aZY5Jb at altfrangg.fortytwo.ch>,
Adrian 'Dagurashibanipal' von Bidder <grazdan at fortytwo.ch> wrote:
> > At 1:55 PM +0800 2005-05-30, Eric Liu wrote:
> >=20
> >>    I need to implement time synchronization on LAN. And there should be 1ms

I suspect that this doesn't actually mean *time* synchronisation (i.e. both
frequency and phase matching UTC, and what NTP is designed for) but really
about measuring small time differences, i.e. the common free running
local clock case that seems the most common (mis)use of NTP.

> >>  accuracy.

> +  2 u   85 1024  377   20.503    1.772   0.42=

This is not 1.772 ms offset from true.  It is 1.772 +/- 10.252 +/- 
dispersion offset from the immediate upstream server.  One needs root
delay and root dispersion to work out the actual error bounds.

More information about the questions mailing list