On Tue, Feb 1, 2011 at 5:50 PM, Mike S <mi...@flatsurface.com> wrote:
> Chris, I've seen similar when using an Atom (PPS) refclock linked to another > refclock (using the "prefer" keyword) which delivers time via a serial > character stream. The PPS is, as it should be, very accurate. But the serial > (Trimble Thunderbolt, in this case) would bounce around +- quite a few ms. > When they got too far apart, NTP would consider the pair bad (or maybe only > the PPS dropped out, I can't recall). > > I don't know if the monolithic Oncore driver pays any attention to this, but > my issue was corrected by adding a "tos mindist 0.015" line to > /etc/ntp.conf. That allows the two refclocks to be separated by up to 15 ms > before NTP considers it a problem. > I suspect now the PPS is failing somehow. I can see the serial data appear in the log every second. But mixed in there is a error message that reads "ONCORE[0]: ONCORE: oncore_get_timestamp, error serial pps". I'll likey grep the code for the error string and see what it's doing. I just bught an HP "universal counter" and when UPS delivers it I'll have the abilty to measure the PPS at sub uS level. I could have a "plumbing" pproblem as the UT+ is 3rd floor and the computer is 1st floor with 75 feet of RS232 over cat-5 between. -- ===== Chris Albertson Redondo Beach, California _______________________________________________ questions mailing list questions@lists.ntp.org http://lists.ntp.org/listinfo/questions