[time-nuts] NTP stratum 0

M. Warner Losh imp at bsdimp.com
Sat Jan 14 14:19:51 EST 2006

In message: <000a01c6193a$03346200$0500a8c0 at darius.domain.actdsltmp>
            "Bill Hawkins" <bill at iaxs.net> writes:
: Perhaps it needs to be better than a millisecond. I remember an
: NTP accuracy of 100 microseconds, after enough round-trip time
: statistics were gathered. Perhaps the net is noisier now, and
: more uncertain. That certainly characterizes the messages carried
: by the 'net.

My rough and ready rule was based on the 'people noticing and
complaining' standard, which is fairly easy to meet...  I base this on
one of our main ntp servers that gets its time from a machine that's
connected to a GPS server, as well as a few ohters and routinely see
bigger offsets:

     remote           refid      st t when poll reach   delay   offset  jitter
+EnsembleAGPS.ti .GPS.            1 u  160 1024  377    1.200   -4.389   0.072
+tick.usnogps.na .USNO.           1 u  223 1024  377   98.290    2.130   0.646
*india.colorado. .ACTS.           1 u  395 1024  376   61.283    0.514   4.185

: In any event, the noise from the frequency standards should be at
: least a factor of ten smaller than the filtered noise limit of the
: Internet, IMHO.

It should be about a thousand times better...


More information about the time-nuts mailing list