Capturing Transients and Single-Shot Events

Section Overview

The chapter so far has read signals that repeat; this section captures the ones that do not — the glitch, the sag, the one-time spike — by making the scope wait for them (triggering-capturing-a-stable-waveform). Single-shot capture is the foundation. Single mode arms the scope once and freezes the first qualifying event, and with the trigger position set inside the record, the pre-trigger view preserves the lead-up as well as the aftermath (the-oscilloscope-as-the-diagnostic-instrument). The trigger is set where only the fault can reach. A level placed outside the signal's healthy swing — below a rail's normal minimum, above a line's normal peak — turns the trigger into a tripwire that fires only when the abnormal happens. Persistence makes the rare visible. Persistence overlays many sweeps into one accumulated picture, so a deviation that appears once in thousands of cycles shows as a ghost behind the normal trace instead of a flicker the eye misses. Peak-detect keeps narrow events alive. At slow timebases a glitch narrower than the sample spacing can fall between kept samples and vanish; peak detect stores each interval's highest and lowest point so the spike survives — while averaging does the opposite and is kept away from a glitch hunt. And the captured record is read for cause. The pre-trigger lead-up, the zoom into the stored trace, and measurements on the frozen waveform turn one capture into a diagnosis (reading-and-measuring-waveforms). Arm the tripwire, accumulate the ghosts, keep the spikes alive, and read the record — and the faults that only happen sometimes stop being invisible.

Why This Matters

The steady faults were found by earlier chapters; the ones that remain are intermittent, brief, or one-time, and they are found by capture, not by watching (triggering-capturing-a-stable-waveform). This matters because the eye cannot catch a one-shot: a millisecond sag at power-up or a spike that strikes once is gone before any observer reacts, and only an armed capture holds it (the-oscilloscope-as-the-diagnostic-instrument). This matters because rare events hide inside healthy signals: a bus that glitches once a minute looks perfect on a triggered sweep, and persistence is what makes the one bad cycle in a million visible. It matters because acquisition settings decide what survives: a narrow glitch can be lost by the scope itself at slow timebases, so peak-detect versus averaging is the difference between catching the fault and erasing it. It matters because the tripwire trigger saves hours: a level set just past the healthy envelope lets the scope stand watch unattended, so the technician stops babysitting the screen (measuring-rail-voltage-ripple-and-noise). And it matters because the lead-up is the diagnosis: a captured shutdown shows what the rail and the enable did just before it, and that pre-trigger record is often the only witness to the cause. Capture the event once, with its lead-up, and an intermittent fault becomes ordinary evidence.

Required Prerequisites

Triggering — Capturing a Stable Waveform — Section 8.2 taught single mode, trigger level and slope, and the pre-trigger view; this section turns them from controls into a hunting method for events that may strike only once.
The Oscilloscope as the Diagnostic Instrument — Section 8.1 established acquisition: the sampled record stored around the trigger point, whose sampling and storage behaviour decides what a capture can and cannot hold.

Recommended Consumables

A compensated 10x probe with a short ground path — to deliver honest edges to a capture that cannot be repeated (reading-and-measuring-waveforms)
A notebook or worksheet — to record each trigger setup and what it did or did not catch
A switched load such as a resistor on clip leads — to provoke rail sags and load transients on demand (measuring-rail-voltage-ripple-and-noise)
An insulated prodding tool — to provoke an intermittent by flexing or tapping while the scope stands watch
A camera or screenshot workflow — to preserve captured one-time events before they are overwritten

Recommended Practice Hardware

A digital storage oscilloscope with single mode, persistence, and peak-detect — to practise every capture method in this section (the-oscilloscope-as-the-diagnostic-instrument)
A board whose power-up can be repeated — to practise single-shot captures of start-up behaviour
A repeating signal with an occasional induced anomaly — to hunt ghosts with persistence
A source of narrow pulses — to watch sample mode lose a glitch and peak-detect save it
A power rail with a switchable load — to practise tripwire triggers on sags and recoveries (measuring-rail-voltage-ripple-and-noise)
A notebook of capture setups — to build a library of trigger recipes for events you expect to hunt again

Real-World Applications

Capture technique is how intermittent complaints become closed repairs. A technician chasing a device that reboots randomly sets a tripwire trigger below the rail's healthy minimum in single mode and comes back to a captured sag — with the pre-trigger showing the load step that caused it (measuring-rail-voltage-ripple-and-noise). A repairer told a bus "sometimes corrupts" puts infinite persistence on the data line and watches a runt pulse build up as a ghost behind thousands of clean cycles (triggering-capturing-a-stable-waveform). Someone verifying a power-up sequence arms a single-shot on the enable line and reads the stored record to check that the rails rose in order (the-oscilloscope-as-the-diagnostic-instrument). A technician hunting a spike on a slow timebase switches from sample to peak-detect and the microsecond glitch that was falling between samples appears on every sweep. And a repairer with a captured shutdown zooms the frozen record, measures the sag's depth and duration with cursors, and matches it against the supply's specification (reading-and-measuring-waveforms). The failures this prevents: shipping back an "unreproducible" fault the scope could have caught overnight, polishing away a glitch with averaging, and losing a one-time capture because no one saved it.

Common Challenges

The event cannot be described before it is seen. A trigger needs a level and slope, but the fault's shape is unknown until the first capture — the difficulty is arming for the abnormal in general, which is done by setting the tripwire just outside the healthy envelope rather than guessing the fault's exact form (triggering-capturing-a-stable-waveform).
The scope's own acquisition can erase the evidence. Sample mode at slow timebases drops what falls between kept samples, and averaging buries the outlier by design — the difficulty is that these losses are silent, so the mode must be chosen for the hunt before trusting an empty screen.
One capture is easy to waste. A single-shot event may not repeat for hours — or ever — the difficulty is that a mis-set vertical, a clipped trace, or an unsaved screen throws away evidence that cannot be recalled, so the setup is rehearsed on a healthy signal first.

Safety Notes

Risk Level: Medium. Transient hunting keeps a probe on a live circuit for long, unattended stretches, and provoking intermittents means flexing and tapping powered boards — the discipline of live work applies for the whole watch.

A standing watch is still live work — secure the probe, and provoke faults with an insulated tool

A capture session leaves the probe connected to a powered board while attention is on the screen or elsewhere, so clip or brace the probe securely — never balance a live tip where it can slip and short adjacent pins — and route the lead so it cannot be snagged during the wait (the-oscilloscope-as-the-diagnostic-instrument). The grounding discipline is unchanged: probe ground only to verified circuit ground, an isolation transformer on any mains-referenced circuit under test — never on the scope — and the isolated board still treated as live, because the transformer removes the earth path, not the voltage (triggering-capturing-a-stable-waveform). When provoking an intermittent, flex and tap with an insulated tool, not a finger, keep one hand clear, and switch loads with insulated clip leads rather than bare wire ends. An unattended watch on a faulty board is unattended powered operation of faulty equipment — current-limit or fuse the supply where possible, keep the area clear of flammables, and do not leave a suspected power-stage fault running overnight in an unoccupied space. The takeaway: capture work stretches live-board time from minutes to hours — set the connections up to be safe without supervision before arming the trigger.

Professional Tips Before Starting

Rehearse the capture on a healthy signal. A one-time event gives no second chance — set the vertical, timebase, and trigger on normal behaviour first, so the real event lands in a ready frame (reading-and-measuring-waveforms).
Put the tripwire just past the envelope. A level inside the healthy swing fires constantly; far outside it may never fire — set it just beyond the worst normal excursion.
Save before you touch. A captured single-shot is overwritten by the next arm — screenshot or store the record before adjusting anything.

Capturing the Unrepeatable — Single-Shot, Persistence, and Peak-Detect

Recap and Frame

Sections 8.1 through 8.4 built the skills for signals that repeat; this section aims them at the faults that do not, and the difference is that the scope, not the technician, does the watching (the-oscilloscope-as-the-diagnostic-instrument). The toolkit is small and specific. Single mode freezes one event, the pre-trigger preserves its lead-up, persistence accumulates rare deviations into visibility, and peak-detect keeps narrow spikes alive through slow acquisitions (triggering-capturing-a-stable-waveform). The trigger becomes a tripwire. For steady signals the trigger was set inside the swing to lock the picture; for fault hunting it is set outside the healthy envelope, so it fires only when the abnormal happens. The acquisition mode is part of the hunt. What the record keeps — every extreme, or a polished average — is a setting, and the wrong one erases the evidence before it is ever displayed. The capture is evidence to be read. A stored record is zoomed, measured, and read backwards from the trigger into its lead-up, the same measurement discipline as before applied to a frozen trace (reading-and-measuring-waveforms). And rehearsal protects the one chance. Because the event may never repeat, the whole setup is proven on healthy behaviour before the real wait begins. Hold the frame — tripwire trigger, right acquisition, accumulated ghosts, and a record read for cause — and the unrepeatable becomes capturable.

Arming a Single-Shot Capture

The basic capture is single mode with a deliberate frame around the expected event (triggering-capturing-a-stable-waveform). Single mode freezes the first qualifying event. Armed once, the scope waits, captures the first trigger that meets the condition, and stops — the event is held on screen instead of being overwritten by the next sweep. The vertical and timebase are set before arming. The channel is ranged so the expected event fits without clipping and the timebase chosen so the event's duration spans a useful part of the screen, both set against normal behaviour first, because a mis-framed capture cannot be retaken (reading-and-measuring-waveforms). The trigger position splits the record. Placing the trigger point mid-record gives the capture both a lead-up and an aftermath, and the position is shifted toward more pre-trigger when the cause is the question, or more post-trigger when the consequence is. The slope points at the event's first move. A sag is caught on a falling slope, a spike on a rising one — the trigger is set on the edge the fault must make first, not on the recovery. A blank screen while armed means no event yet. In single mode the screen waits dark or frozen until the condition is met, so patience — not fiddling — is the discipline once armed (the-oscilloscope-as-the-diagnostic-instrument). Armed once and frozen, framed before arming, split around the trigger, sloped toward the fault's first move, and left alone to wait — and the single-shot capture is set. Frame the event before it exists, and the scope catches it the only time it happens.

The Tripwire — Triggering Only on the Abnormal

A capture is only as good as its trigger condition, and for fault hunting the condition is placed where healthy behaviour can never reach (triggering-capturing-a-stable-waveform). The healthy envelope is measured first. A few minutes watching normal behaviour establishes the signal's real limits — the lowest the rail dips under load, the highest the line swings — and the trigger is set just beyond them (measuring-rail-voltage-ripple-and-noise). Below the floor catches sags. A trigger level set under the rail's worst normal minimum, on a falling slope, fires only when the rail drops abnormally — a dropout, a brownout, a load fault — and ignores every healthy cycle. Above the ceiling catches spikes. A level set over the highest normal peak, rising slope, fires only on an overshoot, a transient, or an injected disturbance. Normal mode turns the tripwire into a counter. With the same tripwire level in normal mode instead of single, the scope draws only when the abnormal recurs, so how often the screen flashes is a live measure of how often the fault strikes. Too tight rings false, too loose misses. A tripwire inside the envelope triggers on healthy noise and a tripwire far outside may never fire, so the level is walked outward from the envelope until false triggers just stop. Envelope measured, floor and ceiling staked out, slope aimed at the fault's first move, normal mode for counting, and the level walked outward to just past the noise — and the tripwire is set. Put the trigger where only the fault can go, and every capture is a fault.

Persistence — Making the Rare Visible

Some faults are too frequent to wait for one by one and too rare to see on a refreshing sweep, and persistence is the display mode that catches them (the-oscilloscope-as-the-diagnostic-instrument). Persistence overlays sweeps instead of replacing them. Persistence keeps previous traces on screen as new ones draw over them, for a set decay time or indefinitely, so the display becomes an accumulation of everything the signal has done rather than only its latest cycle. A rare deviation becomes a ghost. One bad cycle in thousands leaves a faint trace — a runt pulse, a shifted edge, a sag — hanging behind the bright normal waveform, so the anomaly announces itself without a trigger ever being aimed at it. Infinite persistence is the overnight net. Set to accumulate without decay, the scope keeps every excursion it acquired during the watch, and a clean picture after hours is strong evidence — though not proof, because a scope is blind for the dead time between acquisitions, so a rare event can strike unseen, and the conclusive watch is a tripwire trigger armed in single or normal mode. Persistence accumulates only what the acquisition kept. A watch for narrow glitches is run with peak-detect on, so the ghosts the display collects are events the acquisition actually held onto. The ghost then aims the tripwire. Persistence shows what the anomaly looks like and where it goes, and that shape tells you where to set the single-shot trigger level to capture one cleanly for measurement (triggering-capturing-a-stable-waveform). It pairs with a stable trigger. Persistence accumulates around the triggered, overlaid waveform, so the trigger discipline from Section 8.2 — a locked, stable trace — is what makes the ghosts legible deviations rather than smear. Sweeps overlaid, deviations as ghosts, infinite persistence as the overnight net, ghosts aiming the tripwire, and a stable trigger underneath — and persistence turns rarity into visibility. Let the screen remember every sweep, and the one bad cycle has nowhere to hide.

Peak-Detect and Averaging — What the Acquisition Keeps

Whether a narrow event survives to be displayed is decided by the acquisition mode, set before the hunt (the-oscilloscope-as-the-diagnostic-instrument). Slow timebases stretch the sample spacing. A digital scope covering seconds of screen time keeps far fewer points per second than at fast timebases, so the effective spacing between kept samples grows long. A narrow glitch can fall between the kept points. A glitch — a brief, abnormal excursion such as a narrow spike, a runt pulse, or a momentary dropout — that is narrower than the kept-sample spacing can come and go entirely between points, so ordinary sample mode draws a clean trace over a signal that is not clean. Peak-detect keeps the extremes. Peak detect has the scope record the highest and lowest value seen in each sample interval instead of a single point, so a spike narrower than the interval still leaves its mark on the trace — the standard mode for glitch hunting at slow timebases. Averaging does the opposite. Averaging mode combines many sweeps to suppress random noise, which polishes a repeating waveform beautifully — and buries a rare outlier by design, so it is the right mode for measuring a steady signal and the wrong mode for hunting a fault. The trade is visual roughness. Peak-detect traces look thicker and noisier because they carry every extreme; that roughness is the evidence surviving, not a defect of the mode (reading-and-measuring-waveforms). Long spacing at slow timebases, glitches lost between points, peak-detect keeping each interval's extremes, averaging reserved for steady measurement, and roughness read as evidence — and the acquisition keeps what the hunt needs. Choose what the record keeps before the hunt, because what it drops is gone.

From Capture to Cause — Reading the Record

A capture is not a diagnosis; the stored record has to be read, and it is read with the measurement discipline already in hand (reading-and-measuring-waveforms). Save first. A single-shot record is overwritten by the next arming, so the screen is stored or photographed before anything else is touched. The pre-trigger lead-up is read for cause. The part of the record before the trigger shows what the signal — and, on a second channel, its neighbours — did in the moments before the event, which is where the cause usually sits (triggering-capturing-a-stable-waveform). Zoom works on the stored record. The captured waveform can be expanded horizontally and vertically after the fact, so the event is examined at magnifications that were not chosen in advance. Cursors put numbers on the event. Depth and duration of a sag, height of a spike, time from cause to effect — measured on the frozen trace and compared against specification, the same comparison habit as on live signals. A second channel doubles the story. Capturing the suspect line and a related signal — the enable, the load, the clock — in the same record turns "the rail sagged" into "the rail sagged when the motor switched," which is a cause. And the event is tied to its provocation. Noting what was happening when the trigger fired — power-up, a flexed board, a switched load — completes the chain from symptom to mechanism. Saved first, lead-up read, zoomed after the fact, measured with cursors, correlated across channels, and tied to its provocation — and the record gives up its cause. Treat the capture as evidence to be read, and one event closes the case.

Common Mistakes

Hunting a glitch in averaging mode. Averaging suppresses outliers by design, erasing the fault being hunted — use peak-detect for the hunt, and averaging only for measuring steady signals (the-oscilloscope-as-the-diagnostic-instrument).
Setting the tripwire inside the healthy envelope. A level healthy cycles can reach fires constantly and buries the real event — measure normal limits first and set the level just beyond them (measuring-rail-voltage-ripple-and-noise).
Arming with an unrehearsed frame. A clipped vertical or a wrong timebase wastes a capture that may never repeat — set and prove the frame on healthy behaviour before arming.
Re-arming before saving. The next arm overwrites the record — store or photograph the capture before touching the controls.
Trusting a clean trace at a slow timebase. Sample mode can draw a clean line over a glitching signal — a slow-timebase all-clear counts only in peak-detect.

Troubleshooting Guidance

Transient capture comes down to a rehearsed frame, a tripwire trigger, the right acquisition, and a record read for cause. If the fault is one-time or rare: arm single mode with the trigger level just outside the healthy envelope and the trigger position mid-record (triggering-capturing-a-stable-waveform). If you do not know what the fault looks like: run persistence — infinite for an overnight watch — and let the ghost show its shape before aiming a trigger at it. If a suspected glitch never appears at slow timebases: switch from sample to peak-detect before concluding it is not there. If the tripwire keeps firing on healthy signal: walk the level outward until false triggers just stop — the envelope was wider than measured (measuring-rail-voltage-ripple-and-noise). If the capture is blurry or clipped: the frame was not rehearsed — reset vertical and timebase on normal behaviour and re-arm. If the event is captured but the cause is not visible: shift the trigger position toward more pre-trigger and capture again, or add the suspected cause on a second channel. If the screen stays blank for hours in single mode: that is data — the event did not cross the tripwire — so widen the watch or move to the next suspect node (the-oscilloscope-as-the-diagnostic-instrument). If a captured record must be compared to spec: zoom and measure the stored trace with cursors, exactly as on a live waveform (reading-and-measuring-waveforms). The throughline: rehearse the frame, place the tripwire where only the fault can go, keep what the record needs, and read the capture as evidence.

Verification & Testing Methods

Confirm your captures are deliberate, not lucky:

[ ] I rehearsed the frame — vertical range, timebase, and trigger position — on healthy behaviour before arming, so the one-time event landed unclipped and well-placed in the record.
[ ] I set the tripwire deliberately: healthy envelope measured first, level just beyond it, slope aimed at the fault's first move, and I walked the level out until false triggers stopped.
[ ] I used persistence — including infinite persistence for a standing watch — to make a rare deviation visible as a ghost, and used its shape to aim the single-shot trigger.
[ ] I hunted narrow events in peak detect, treated a clean sample-mode trace at a slow timebase as unproven, and kept averaging for steady measurements only, so no glitch was erased by the acquisition.
[ ] I saved each capture before re-arming, read the pre-trigger lead-up for the cause, measured the stored record with cursors, and noted what provoked the event — with the probe secured and grounded to verified circuit ground for the whole watch.

Then try the practice exercises below — hands-on capturing; scenarios differ from the quiz.

Practice Exercises

Capture a power-up single-shot (5 minutes, hands-on). Rehearse the frame on a running rail, then arm single mode with the trigger position mid-record and power the board up — read the stored record's pre-trigger lead-up and zoom into the rise (triggering-capturing-a-stable-waveform).
Set a tripwire on a rail (5 minutes, hands-on). Watch a loaded rail long enough to know its healthy minimum, set a falling-slope trigger just below it, then switch a load in with clip leads and confirm the scope captures the sag — measure its depth and duration with cursors (measuring-rail-voltage-ripple-and-noise).
Ghost-hunt with persistence (5 minutes, hands-on). Put infinite persistence on a repeating signal, provoke an occasional anomaly — a flexed connection tapped with an insulated tool — and watch the deviation accumulate as a ghost behind the normal trace; note how its shape tells you where a tripwire would go.
Watch peak-detect save a glitch (3 minutes, hands-on). At a slow timebase, feed a narrow pulse and compare sample mode against peak-detect, watching the spike vanish in one and survive in the other (reading-and-measuring-waveforms).

These core steps — arming a single-shot, setting the tripwire, accumulating with persistence, keeping glitches alive with peak-detect, and reading the record — are tested in the Chapter Quiz at the end of this chapter, where a score of 80% is required to continue.

Key Takeaways

A single-shot capture is framed before it exists — vertical, timebase, and trigger position rehearsed on healthy behaviour — because a one-time event gives no second chance, and the record is saved before re-arming (reading-and-measuring-waveforms).
The fault-hunting trigger is a tripwire: the healthy envelope is measured first and the level set just beyond it, on the slope of the fault's first move, so the scope fires only on the abnormal — and in normal mode the firing rate counts how often the fault strikes (triggering-capturing-a-stable-waveform).
Persistence overlays many sweeps into one accumulated picture — infinite persistence for a standing watch — so a rare deviation shows as a ghost behind the normal trace, and the ghost's shape aims the single-shot trigger.
At slow timebases a glitch narrower than the kept-sample spacing can vanish in ordinary sample mode; peak detect keeps each interval's highest and lowest point so the spike survives, while averaging suppresses outliers by design and is kept away from a glitch hunt.
A capture becomes a diagnosis when the record is read: the pre-trigger lead-up for the cause, zoom and cursors on the frozen trace for the numbers, a second channel for the correlation, and a note of what provoked the event.

Skills Learned

You can now arm a single-shot capture — trigger level, slope, and pre-trigger position — for an event you have not yet seen.
You can now set trigger conditions outside the healthy swing so the scope fires only on the abnormal event.
You can now use persistence to make a rare deviation visible as a ghost behind the normal trace.
You can now use peak-detect so narrow glitches survive slow timebases, and keep averaging away from a glitch hunt.
You can now read a captured record — pre-trigger lead-up, zoom, and measurement — to move from the event to its cause.

Glossary Additions

glitch — a brief, abnormal excursion on a signal — a narrow spike, a runt pulse, a momentary sag or dropout — that departs from the waveform's healthy pattern and usually strikes rarely or irregularly. Glitches matter because a single one can corrupt a bus transaction, reset a processor, or trip a protection circuit, yet they are easy to miss: the eye cannot catch a one-time event, a refreshing sweep overwrites it, and at slow timebases a glitch narrower than the kept-sample spacing can fall between samples and vanish from an ordinary acquisition entirely. Glitches are hunted with capture technique — a tripwire trigger set just outside the healthy envelope, persistence to accumulate rare deviations into view, and peak-detect acquisition so narrow events survive to be displayed.
persistence — an oscilloscope display mode that keeps previous traces on screen as new sweeps draw over them, for a set decay time or indefinitely, so the display accumulates everything the signal has done rather than showing only its latest cycle. A deviation that appears once in thousands of cycles — a runt pulse, a shifted edge, a sag — leaves a faint ghost behind the bright normal waveform, announcing an anomaly no single sweep would reveal, and the ghost's shape and position then tell the technician where to set a single-shot trigger to capture one cleanly. Infinite persistence turns the scope into a standing watch: left accumulating for hours, it keeps every excursion the acquisition captured, and a clean picture after a long watch is strong evidence — though not proof, since a scope is blind for the dead time between acquisitions — that the fault never visited that node.
peak detect — an oscilloscope acquisition mode in which the scope records the highest and lowest value seen during each sample interval instead of a single point, so brief events survive slow acquisitions. At slow timebases the effective spacing between kept samples grows long, and a spike or dropout narrower than that spacing can come and go between points, letting ordinary sample mode draw a clean trace over a signal that is not clean; peak detect preserves each interval's extremes, so the narrow event still marks the display. The mode's characteristically thicker, noisier-looking trace is the evidence surviving rather than a defect, and peak detect is the standard acquisition for glitch hunting — the opposite of averaging, which suppresses rare outliers by design and belongs only on steady-signal measurement.

Suggested Next Sections

Must read next:

Diagnosing with the Oscilloscope — Section 8.6 brings the chapter together: triggering, probing, measurement, and capture combined into a method that takes a fault from symptom to the waveform that reveals it.

Recommended:

Reading and Measuring Waveforms — the measurement discipline applied to every stored record this section captures.
Measuring Rail Voltage, Ripple, and Noise — the rail-measurement context where sags, dropouts, and load transients — this section's favourite quarry — were first characterised.