And, it turns out that it's not that hard at all. I did this, with 140k points per capture. There was one failed waveform in just over 200 segments. That one waveform does not show any evidence as to the trigger's stated conditions being met, despite the fact that the trigger fired.
Yes because it rearmed.
I'm not sure I follow. It rearmed after it fired. Firing causes the capture. Because the capture contains the point in time that the trigger fired, and because the time span of the display (which defines the time span of the capture) is enough to capture the preconditions of the trigger, it follows that the captured waveform would, in a perfect world, contain the data for both the trigger precondition (rising past 3.5V somewhere between 1.2 and 1.4ms before the trigger fires, which means a voltage value below 3.5V *must* be present somewhere prior to, and relatively close to, the precondition point seen by the triggering mechanism) and the trigger activation itself (rising past 4.5V at the point when the trigger fires).
A segment is a full capture, and what I was referring to is where one of the segments shows an improper capture.
Use Single and it should capture it at any timebase setting.
I did this, with a 14 megasample capture buffer, captured with peak detect, and finally managed to stop myself from pressing the "single" button after seeing a problem capture.
I've attached a screenshot of the end result, as well as a 7-zip archive with the CSV of the data from -1.5ms to 0ms.
And the analysis of the data shows the same absence of a valid initial condition, with 4.33V being the minimum voltage seen between -1.5ms and the trigger (zero) time:
kevin@ubuntu:/tmp$ cat FAIL1GS-SINGLE.csv | awk -F, 'BEGIN {min = 100.0; mintime = 0;} { if ($1 >= -1.5E-03 && $1 <= 0 && $2 < min) { min = $2; mintime = $1; } } END {print mintime, min}'
-1.062040E-04 +4.330000E+00