Eugen COCA wrote: > For sure, FreeBSD does not have PPS_SYNC enabled by default. > > NTP developers say that the ntpd service does not need PPS_SYNC > enabled as you activate PPS in ntp.conf (enable pps). But, as we may > see in the graphs, there is an enormous difference between the two > configurations. > > All my three servers are "web enabled" and the configurations are > visible if you put the mouse over the "hardware" word in the page. > > The main problem (as I think) with your configuration is that the GPS > receiver does not see the whole sky. I have 2 GPS receivers (one of > them connected with ntp1 and the other with ntp2 and ntp3 - YES, it > works) mounted on the top of the roof (100% clear sky). Your error is > almost the same at almost the same time. Look at the graphs ! > > You may test all your servers with only one GPS receiver, at the same > time. I made a test and it works - one GPS receiver and three servers.
I'm hoping that one of the NTP developers will come back on this - my understanding now is that there is a "software PPS", treated just like any other sync source, and a "kernel PPS" which is treated differently and rather than be part of a phase-locked loop simply forces correct time. As I've not looked at the sources, it is this latter "kernel PPS" I am unsure about. Your PCs are all somewhat better specified than my old 133MHz/48MB Pentium, but the timekeeping is similar as you say. The experimenter in me want to get down from 10us jitter to 1us jitter, but I really don't have the time to do this. I have just 50% sky coverage in the present arrangement (I added a photo to my Web page), so I will see how well that continues to work. As I run Windows on everything else here, there seems little point in splitting the GPS receiver feed, but it sounds an excellent idea. Cheers, David _______________________________________________ questions mailing list [email protected] https://lists.ntp.isc.org/mailman/listinfo/questions
