Alan Melia wrote:
Hi Uwe, is it not possible to allow for the delay down the connecting wire
to the NTP sever. In the days I played TCP/IL over packet the PING facility
gave the round-trip delay for paths through regenerators, and packet nodes.
At least that software might be useful for keeping my home network all in
sync
NTP is not a simple time fetch protocol, it is a clock synchronisation
protocol that polls the master clock at regular intervals.
It does figure out what the RTT is and processes out this error (over
a period of time).
For non-local time servers the error with a 1024s polling interval
seems to be better than 2ms. For a local time server accuracies of
200us with a 64s poll are the norm.
See
http://www.eecis.udel.edu/~mills/database/brief/perf/perf.pdf
- Stewart G3YSX
Cheers de Alan G3NYK
[email protected]
----- Original Message -----
From: "jannsen" <[email protected]>
To: <[email protected]>
Sent: 12 February 2004 22:18
Subject: Re: LF: Timing GPS
Hi Stewart es All,
I do some PC clock synchronizations via internet using the program
"Time Watch", downloaded from http://www.blue-series.de. the NTP server
transmitting the synchronization signals must be in your vicinity.
it is not usefull telling us that their timer is <1uS accurate
(in my case dirived from GPS receivings) if there is a long wire to the
server. it`s something else if you would know the phase fault on the line
and can correct it in some way.
reading the PC clock in mSec from the window "Time Decoder" of SpecLab
of one PC with the clock (motherboard and sound card) permanent
synchronized by the standard time signal station DCF77 and at the
same time reading the mSec from the screen of the other PC just
clocking by the signals from the internet server, I believe I see
a difference, DCF77 in front.
be that as it may, we should reach an agreement about what kind of
synchronization we will make use of.
|
