[ntp:questions] Problem with system clock refresh

David J Taylor david-taylor at blueyonder.not-this-bit.nor-this-part.co.uk.invalid
Tue Nov 24 09:27:11 UTC 2009


"JuanFran" <juanfranciscojara at gmail.com> wrote in message 
news:6a4e421a-318b-4fa1-9105-6ef77a861ec8 at p19g2000vbq.googlegroups.com...
[]
> Hi! My system should be accurate to within approximately 10
> milliseconds .I think forcing upgrade all computers in my LAN at the
> beginning should be enough. Thank you very much for the anwser!

Be aware that the timing resolution on Windows is in the order of 10 
milliseconds, so you are asking a lot.  Having a local stratum-1 server 
(GPS-driven), and using a LAN polling interval of 64 seconds (or even 16 
seconds - maxpoll 4), this should /just/ be possible with Windows XP in a 
reasonably temperature controlled environment, and with no highly variable 
CPU loads or other factors which induce timing changes.  A better way 
which two of us have suggested might be to distribute an RS-232-level NMEA 
and PPS signal to /all/ the PCs (hope they have serial ports!) and run 
them /all/ as stratum-1 NTP devices.

Be aware that PC clocks can easily have an error of 100 parts per million 
(ppm) - that's about 8 seconds per day, so a 10 millisecond error could 
build up in about a minute and a half.  Syncing once at the beginning is 
not enough.  NTP allows for PC clock errors five times that - up to 500 
ppm.

Cheers,
David 




More information about the questions mailing list