[time-nuts] Some More questions
Tom Van Baak
tvb at leapsecond.com
Wed Jan 18 09:11:22 EST 2006
> The timing accuracy is ± 100 nanosecond (1 sigma) and is available
> only when valid position fixes are being reported. Repeatability
> checks of 10 sets of 100 one second samples taken
> over a period of 20 minutes showed an average variation of
> approximately 100 nanoseconds (not allowing for SA)."
> I'm trying to determine what to make of these specs in terms of the
> accuracy of the 1PPS signal as a timing signal. Not necessarily
> accuracy as relates to GMT but to the actual 1.000xxx pps signal. 1E-7
> doesn't seem all that good for GPS. Can I assume that the actual
> timing accuracy is better? Jitter?
The timing accuracy is probably close to 1e-7 (100 ns)
but you have to ask yourself over what timeframe. With
a little averaging things improve. 100 ns over a day is
about 1e-12, which is pretty standard for low-cost GPS
> I'd like to use this signal as a gating signal for one of my frequency
> counters. Would this be better or worse than the un-ovenized TCXO?
> This is an 80s vintage HP counter, sorry I don't have the model handy.
Over the short-term (e.g., 1 or 10 second gating) the
GPS 1 PPS will be worse than a TCXO I would think.
> I should be able to improve the receiver performance some by ovenizing
> it. I wonder if it would be worth the effort?
You can check this with a hair dryer. Measure the effect
of a 10 C rise and then extrapolate back to 0.1 or 0.01C
to see what the result of making an oven would buy you.
> If I divided this down to 1 PPS, would it be a better timing reference
> than the GPS receiver? Is there a better method of calibrating this
> oscillator than beating it against WWV, assuming the usual compliment
> of electronic lab equipment but no special timekeeping instruments?
Again, it depends on what sort of timing reference you
need. But unless you have a full-blown GPSDO, I think
it might be best to use your GPS 1 PPS as a way to
calibrate your xtal's rather than using the raw 1 PPS
as the gate to your counter.
Yes, using a GPS 1 PPS is often a much more convenient
way to calibrate than beating against WWV. All you need
is two TI measurements, separated by some interval. And
the best part is you usually don't even need to divide down
your xtal to 1 Hz.
Put your counter in TI mode; send the GPS 1 PPS to the
start channel; send the xtal RF (e.g., 1 MHz) to the stop
channel; make a measurement of the phase (if your counter
can average, even better). Then wait a while.
Say a minute later you make one more such measurement.
If you had 20 ns phase difference the first time and 32 ns
the second time your frequency error is (32 ns - 20 ns) / 60 s,
which is 12 ns / 60s = 2e-10.
The noise level will depend on the counter, the xtal, your
GPS 1 PPS source -- and so you may have to wait longer
than one minute to get clean results. But as you watch
the counter you will get a feel for the nose level. In this
example you're able to measure down to parts in 10^10th
in just a minute, which I think beats using WWV.
> Next, I really want one of David Forbes' CRT clocks! I sent Dave an
> email asking if he's planning on making any more but I haven't heard
> back. (you lurking here, Dave?) Anyone have one in good condition
> he'd like to part with? Prefer the clear case.
More information about the time-nuts