> You aren't going to get nanosecond accuracy out of a ntp server running over > an ethernet[1]. On the other hand, sub-ms isn't hard with a good OS and/or > good software. > > It's fairly easy to get a reasonable sanity check on a (s)ntp server. Just > setup a known good ntp system and have it monitor the DUT. The reference > implementation for ntp (http://www.ntp.org/) has lots of support for > collecting data. > > The key step in making a PC keep good time is tweaking the clock frequency. > This is the software equivalent of the EFC on an oscillator. ntpd calls it > drift. You can use it as a thermometer.
I don't think the oscillator quality in the typical PC is good enough to ever get nanoseconds, even with tons of tweaks and temperature compensation, etc. The short term variability/phase noise is too high. Microseconds, I think you could do. > > ---------- > > 1) There is a group working on that level of accuracy. It takes special > hardware that can put a time-stamp on a packet as it leaves or arrives. > http://en.wikipedia.org/wiki/Precision_Time_Protocol > > More than just a casual group. You can buy IEEE 1588 products from, among others, Symmetricom. _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
