[time-nuts] GPSDO and holdover
Tom Van Baak
tvb at LeapSecond.com
Sat Apr 26 17:09:43 EDT 2014
>> I have noticed skipped 1PPS on the Adafruit GPS also.
> I've always assumed this could happen but as a result of RF signal loss not
> a glitch in the gps. So I've started recording event timestamp deltas
To be clear, we are not talking about a system-wide GPS problem here; the satellites are fine. All my other GPS receivers are fine. It's just one particular GPS receiver board that Bob, and now Bob and I, are questioning.
> using the Linux kernel PPS interface. I read assert events (e.g.
> 1398449188.001000242#1741672) and compute timestamp and event deltas. If
> the t delta is < .9 something horrible must have happened and if it's > 1
> some didn't happen assuming the event count delta is always 1.
> I wonder if this is a reasonable approach or if I'm being lazily
> optimistic. I just started (and I haven't added a join with the valid fix
> indicator yet) but I've had two missing pulses in the last 24 hours.
Sure, that's reasonable. I'm using a 53132 TIC to compare my house atomic 1PPS (start) against the Adafruit GPS 1PPS (stop) and so no timestamping is even necessary: the readings themselves tell you if there is a missed pulse. For example, you get TI readings like 0.0000000nn for hours or days and then once in a while you get a 1.0000000nn or 2.0000000nn, indicating a missed pulse.
After Bob Stewart's first mention of the Adafruit Ultimate GPS board, I dug out mine and collected 30 days of data to see if I could duplicate or understand his measurements. As usual, a simple and quick test has turned into something more complicated and perplexing. I'll share some results maybe next week. The test is now multiple GPS boards, antennas, and counters.
More information about the time-nuts