Ry said the following on 04/15/2006 02:14 AM: > I'm 1462.5 km (4.8 ms) from the WWVB transmitters, which is well inside > the coverage area maps on the NIST site. I figure WWVB is my best > (only?) option for a refclock at this location. Is sub-milisecond > accuracy possible using a WWVB refclock? The NIST site says 0.1 to 15 > ms phase error, which is two orders of magnitude. I see ~5 ms average > offsets when I sync with nearby stratum-1 servers via NTP. Would WWVB > perform any better?
I can't answer your question about current availability, but I can show you some WWVB results. I have an old Spectracom 8170 and the accompanying large ferrite antenna is mounted on my roof. I'm about 1770km from the station. At http://www.febo.com/time-freq/ntp/stats/index.html I have a set of plots derived from the peerstats file of one of my NTP servers, databox.febo.com. It shows the offset of my other internal NTP boxes, as well as three external stratum 1s, versus the system time (ie, the offset field from peerstats). databox itself is a stratum 1 machine, using a shared memory PPS driver to sync to an HP Z3801A GPS disciplined oscillator. So the offset of the other servers is being measured against a fairly stable baseline, and if you're sneaky you can back out any wander in databox's time by looking at common-mode wiggles in the plots. (Unfortunately, at the moment I'm doing some reconfiguration and two of the servers that should be stratum 1 are actually running at stratum 2, and one of them has a very strange wander pattern over 5ms or so that makes the rest of the plot harder to read than it should be; all the other machines are typically grouped well within 1ms.) In general, toe.febo.com, which is hooked to the Spectracom 8170, holds within several hundred microseconds of the other clocks. However, every now and then, perhaps once every couple of weeks, there is a downward excursion of perhaps 20ms and a recovery that takes a couple of days to fully complete. I suspect this occurs when the 8170 loses lock and its 10MHz oscillator goes into free-run mode for a little while. It may be that a receiver alignment would reduce the magnitude of this effect, but my longer term plans are to use an external reference oscillator. So, other than that occasional anomaly, the WWVB receiver provides very good time, usually well within 1ms. By the way -- I also have some plots of WWVB received signal strength at http://www.febo.com/time-freq/wwvb/sig-strength/index.html, comparing the performance of a couple of antennas, and showing the strength change from day to night. Having said all that, I think you may have problems getting decent reception if your WWVB antenna is in a machine room near the bottom of an urban canyon, but it's worth a shot. Another thing you might consider is one of the CDMA refclocks that get their time from the cellular network. It won't be as traceable as using an NIST or USNO source, but you shouldn't have trouble getting a signal. John _______________________________________________ questions mailing list [email protected] https://lists.ntp.isc.org/mailman/listinfo/questions
