[ntp:questions] USNO Clock Fast?

Brian Garrett mgy1912 at cox.net
Mon Nov 6 21:45:48 UTC 2006


"Richard B. Gilbert" <rgilbert88 at comcast.net> wrote in message
news:rPGdnSzleqQbPNLYnZ2dnUVZ_ridnZ2d at comcast.com...
> Daniel Norton wrote:
>
> > Why does USNO show several seconds faster than time.gov (and other
> > references)?
> >
>
> It is not, now, off by "several seconds".  I'm inclined to doubt that it
> ever was!   I think you are misinterpreting something.  Perhaps you are
> reading milliseconds as seconds?
>
He's not misinterpreting anything.  The animated clocks and the static
display have indeed been several seconds fast lately as compared to
time.gov, WWV, etc.  (I haven't checked lately; they may have fixed it by
now.)
I just assumed this was due to their having deprecated these older Web-based
services in favor of the time.gov site, which is a joint project with NIST.

>  From here, tick.usno.navy.mil is within 898 microseconds of my own
> stratum 1 (GPS reference clock) server which is amazingly good.  Usually
> internet servers are off by anywhere from two to ten milliseconds!  Note
> that the servers themselves are within a few microseconds or even a few
> nanoseconds of the correct time but asymmetric network delays usually
> introduce an offset of several milliseconds.
>
> FWIW, the "offset" in the ntpq -p banner is in milliseconds!

I've no doubt that USNO is spot-on to at least ten decimal places
internally, but once their internet-based time transmissions leave
Washington they're subject to getting stuck in traffic, running out of gas
and carjacking just like the rest of us.


Brian Garrett





More information about the questions mailing list