[time-nuts] Can anyone help?

Poul-Henning Kamp phk at phk.freebsd.dk
Sat Oct 28 21:39:54 UTC 2006


In message <4543C6CF.6070902 at cox.net>, Didier Juges writes:

>Poul-Henning Kamp may weigh in with his valuable expertise, but I 
>believe the ntp standard offers the capability for the clients (your 
>computer's operating system) to compensate for the time it takes for the 
>ntp info to get to you. This assumes that the delay from you to the ntp 
>server is the same as the delay from the ntp server to you, which is not 
>always a correct assumption, but seems to work pretty good in most cases 
>(within the range John cited).

Phk's NTP precision rules of thumb:

	Don't trust the statistics from NTPD, they are not
	independent and not calibrated.

	Precision will almost always be better than 20msec

	Getting precision better than 10msec puts demands on path
	symmetry which not all technologies can lift (cable and
	ADSL in particular).  Calibration or at least verification
	is required.

	Getting _guaranteed_ precision better than 1msec is only
	possible with OCXO or better, or with stable temperature
	on an unloaded LAN with high polling frequency.

	Getting _guaranteed_ precision better than 1msec requires
	special hardware and a lot of calibration.  Since a GPS
	costs only $300 so it's not generally worth it.

Poul-Henning

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.




More information about the Time-nuts_lists.febo.com mailing list