----- Original Message -----
From: "Dr. David Kirkby" <drkirkby@ntlworld.com>
To: <amps@contesting.com>
Sent: Thursday, October 24, 2002 7:17 AM
Subject: [Amps] Re: [Moon-Net] Looking for GPS-module to keep PC-clock
correct.
> Joe Taylor wrote:
> >
> > Dave / W6TE wrote:
> >
> > > It's my understanding that PC clock stability drifts too much to be
useful
> > > for some programs that use synchronization between stations (JT44). So
> > > Dimention4, I believe, is not useful for these applications. I found
new
> > > Motorola GPS boards on EBAY for $20.00 they interface with the RS-232
port
> > > using available freeware. I have purchased two of the boards but have
not
> > > finished the interfacing as of yet. Another source of precise clocking
is
> > > the recent availability of the HP Z3801 disciplined GPS clocks. They
are
> > > available for around $250 and are complete with internal double ovened
> > > oscillator. It needs an antenna and a 50 VDC power supply to run.
> > >
> > > Dave / W6TE
> PC clock stability is inherently poor - a lot of it due to the fact there
are
> large temperature variations over time. I don't know what JT44 is, and
hence
> what timing accuracy is required to use it. However, its easy to achieve
100 ms
> accuracy using Network Time Protocol (NTP) on a computer as a software
solution.
> NTP makes uses of accurate clocks (some of which are caesium based) via
the
> internet. The round-trip delay in getting the information via the internet
is
> estimated fairly accurately. Several NTP servers are then used to give
improved
> accuracy, with NTP taking into accout the jitter on the data from
different
> clocks and attaching more weight to those that have the lowest jitter.
I've only
> experience of using NTP on Unix systems, but I know PC solutions exist.
>
> This may not be sufficiently accurate, but its free and easy to use.
>
> Incidentally, recently I had a need to look at the data sheet on a clock
chip
> used in my computer (Sun Ultra 60) and found that if you know the average
drift
> per month, you can actually re-program the clock chip to run slower or
faster.
> Personally I could not be bothered to do this, since NTP does the
adjustments
> continuously and I suspect there is nothing you could do about it if a PC
is
> switched on/off during the week, as the temperature would change too much.
> --
> Dr. David Kirkby,
> Senior Research Fellow,
> Department of Medical Physics,
> University College London,
> 11-20 Capper St, London, WC1E 6JA.
> Tel: 020 7679 6408 Fax: 020 7679 6269
> Internal telephone: ext 46408
> e-mail davek@medphys.ucl.ac.uk
NTP is indeed a good solution. On unix the program is run as a service and
called 'ntpd' or 'xntpd' for Windows users, Windows-2000 and Windows-XP
have the client-only side of NTP built in to the operating system, M$ don't
publicise this though.
So, if you have an always on connection (xDSL or whatever) then you can
configure NTP on your 2000/XP box as follows:
a) log in as administrator
b) start a command prompt
c) issue the following commands:
net stop w32time ; it may be stopped already
net time /setsntp:<ip address of clock>
net start w32time ; restart the service
d) go to Start->Settings->Control-Panel->Admin Tools->Services
locate the "Windows Time Service" and ensure its set to "Started
Automatic"
You can set a single clock by ip address with the command:
net time/setsntp:1.2.3.4
or multiple clocks with a space-seperarted, quoted list:
set time/setsntp:"1.2.3.4 3.4.5.6 5.6.7.8"
If you have a Win-NT4 machine you can install the W32TIME service from
Resource Kit #4 CD.
Information on NTP, time servers, public domain software for other operating
systems, etc. etc. can be found at:
http://www.eecis.udel.edu/~mills/ntp/
Don't run NTP from a dialup account unless you want to remain online
forever...
Mike G8TIC/M0VHF
|