One additional point: Many ordinary PC sound cards do not have proper
input filtering, so anything as fast as a PPS pulse will cause severe
aliasing, even after it has been voltage-clamped to a level that will
not overload the ADC. (Thus my mention of LP filtering under pulse
The sound cards that do have adequate input filtering generally use
rapid-cutoff filter alignments, which can ring like crazy -- very much
what you do not want when you are trying to use a pulse for timing.
So: Best to slow the pulse rise- and fall-times down so you are not
banging on the input filter (if present), or causing aliasing.
If you could, would you see if you can take the pps out of your
lucent box and use it to sync the pc sound card clock via continuous
calibration for using SL for fx measurements. Maybe by lucent box
pps output is bad.
The PPS directly out of most GPSDOs is too fast for sound cards. But
once you stretch it, there is another problem. Assuming 5v logic, and
knowing that the sound card is AC coupled, the leading edge of the
(assumed positive-going) pulse will produce a 5v spike that decays over
a time period that depends on the post-capacitor DC resistance inside
the sound card. It might decay fully (back to 0v) during the pulse, or
less than that.
The falling edge of the pulse will generate a negative-going spike with
a peak negative voltage of [decayed voltage after leading edge minus
5v]. This could be anywhere from just slightly negative to the full
logic supply (if the post-capacitor voltage decays fully before the
falling edge occurs). The same would be true for 3.3v logic,
substituting "3.3" for "5" above.
These peak voltages are way outside the input range of most sound cards,
so the positive spike is almost certainly going to overload the ADC,
with unpredictable (and, very likely, unfortunate) results.
At the very least, you will need some sort of sound-card oscilloscope
program so you can see what the sound card is actually capturing. Then
you can start shaping your trigger pulses (with voltage clamping and
possibly LP filtering) to give predictable and stable results.
time-nuts mailing list -- firstname.lastname@example.org
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.