This makes me wonder what folks do when they have no idea what kind of a glitch or anomaly they are looking for. Do they use single sweep repeatedly until they spot something and then zoom/expand to investigate and go from there to set up an appropriate trigger?
Depends on the anomaly Many (most?) scopes have a glitch, runt, or width trigger for finding pulses that are the "wrong" level or duration. There are also timeout triggers for when signals get "stuck" at a certain level, and masks or zone triggers are also useful for finding signals that diverge too far from expected values.
I have a video that talks about the different trigger types (and how they are implemented on one of our oscilloscopes)
(Note that he Batronix demo board manual actually uses this scope - the MXO - for most of their examples)
Nice video, as always.
Even most basic scopes nowadays have pretty much what would be "very advanced set of triggers" just 10 years ago....
It is quite easy to look at nonmonotonic edges or risetimes, pulse widths that are less or more than they should be, runts, dropouts, setup times violations etc etc..
You setup triggers for it and then wait. On scopes with all time running history you can also go back and shuffle through violations etc..
But question was about how do you find something that you don't already know how to search for.
In most scopes you can only use persistence (waiting to catch something on the screen that will give you some clues what to look for in more specific terms), or mask test. With mask test you have also on some scopes capability to save all violations for later (in history) or do some other stuff with it.
As an example, on Siglent and LeCroy scopes you also have deep memory measurements where you can setup several measurements simultaneously and use statistics and histicons to look at distribution of certain parameters.
This is very powerful technique to verify parameters of signals.
For instance you measure pulse period, P-P voltage, risetime, etc.
And let scope run for some time. You end up with statistics for millions of periods with statistical distribution right there.
And if distribution is tight and shape is Gausian (meaning stohastic) you have nice stable signal.
If distribution is multimodal (more than one peak in histogram), then you have some discrete states of the sistem influencing that parameter.
If you measure pulse period (or frequency, same difference, 1/t) and you get distribution that looks like bathtub, that means that signal is frequency modulated, and specifically with sinusoidal modulation...
So with good analytical tools, it is not about staring into the screen waiting for something to happen. It is about configuring the scope to do the hunting for you, and automating tedious work.
And then having an insight that this, this and this parameter seems to be "wondering around" you can tailor triggers to catch it "red handed" and try to correlate anomaly with other parts of the system to try and catch what causes it.
Advantage of this approach is that you quickly start not only to know "something" is wrong, but you start to analyze, understand and quantify signal qualities from the start.
More advanced scopes have SignalScan/Wavescan type of tools. Those tools practically automatically do some steps for you, isolating certain parts of signal with characteristics that seem to be statistical outliers, for operators consideration... These tools might assist you to create needed triggers and such.
Nice timesaver, but using proper steps you can actually achieve most of that manually, if scope has the rest of basic capabilities.