[time-nuts] WWVB Clocks don't sync anymore (revisited)
turner at ussc.com
Tue Mar 19 16:55:17 EDT 2013
A few weeks ago I posted a question/comment about some of my WWVB-based
"Atomic" clocks no longer setting themselves properly. These two clocks,
SkyScan #86716, would show the symbol indicating that they had set
themselves, but their time was drifting away from UTC. Interestingly,
they *would* set themselves exactly once upon installation of the
battery, but never again.
Since that time, I've done a bit of digging around.
The first suspicion was that, perhaps, the NIST had fudged a bit in the
WWVB timecode recently, so I manually decoded a few frames and analyzed
them: Nothing suspicious there.
The next question was if the addition of the BPSK somehow skewed the
timing of the TRF's AGC/threshold - but logically, this didn't make
sense since the clock *did* set itself exactly ONCE - and it wouldn't
have been able to do this at all were this the case. Out of curiosity I
poked around on the board and found the trace containing the time code
and found that despite the BPSK, its timing was exactly as it should
have been: No surprise there.
This left the clock itself, so I did what any other Time Nut would do:
I built a WWVB simulator.
Initially, I set it to a 2010 date - a time that I knew that the clock
worked properly. I had two clocks: One that I'd just reset by pulling
and replacing the battery while the other had been "stuck" for a few
weeks, not resetting itself nightly as it should. I put both of these in
the coupling loops from my WWVB simulator and over the next few days,
the recently re-set clock happily synchronized itself while the other
one with the 2013 date was still "stuck." I then reset that clock and
it, too, behaved itself from then on.
I then reset the clock on the simulator to a February 2013 date and
time. Initially, both clocks reset themselves to the current time and
date at their next midnight, but after that, they got "stuck", never
resetting themselves at "night" again.
So, it appears to be a problem with "Broken Sand" (e.g. a silicon problem).
For the morbidly curious, I have documented my efforts here:
The initial testing
The testing with the WWVB simulator
More information about the time-nuts