Re: Date/Time accuracy (was Re: First TOM1 image under Rob's Software)

[droege@snapmail.us]

This technology must be really getting cheap.  I bought a
radio/alarm/colck that automatically sets itself.  I also have my watch
which not only sets itself at 2 am, but also has a high quality crystal.

At first the alarm clock was 9 seconds slower than my watch.  Over about a
week, it has slowly corrected itself so that the time matches my watch to
within a fraction of a second.  So it must be computing the rate of drift
of it's (cheap) crystal and making a correction over time.

OK, this does still not fix the variation with distance that NTP attempts
to fix.  But (I think) GPS does a good job of this and when I compare my
watch to GPS, they turn over at the same time as near as I can tell, to
about 0.1 second.

Tom Droege


> As an example, I am synching my NTP server to three different sources,
> each a different distance away in terms of ping times, and NTP is doing
> all the fancy stuff that Chris talked about to create a solution, some
> of which you can see here (if it copies and pastes ok):
>
> ntpq> pe
>       remote           refid      st t when poll reach   delay   offset
> jitter
> ========================================================================
> ======
> +bastille.visi.c 128.101.101.101  3 u   46 1024  377    1.069    1.256
>   3.334
> *nss.nts.umn.edu 160.94.54.248    2 u  929 1024  377   23.730   -4.168
>   2.548
> +ntp-1.gw.uiuc.e 128.174.38.133   2 u  838 1024  375   91.928   38.156
> 41.659
>
> I then synch all my machines to my local master using the same
> technology.
>
> Nice stuff.
>
> Michael
>
>
>
>