Surge Damage in Power Supplies: Reading the Evidence

Q: How do I trace a surge failure?

Classify the signature (melted, cracked, degraded), match it to likely events (inrush, lightning, load dump), and confirm with the operating history. The signature narrows the causes; the history confirms the one.

Surge damage leaves evidence before it leaves a part: a melted die, a cracked package, or a slow leakage drift. Reading the signature and tracing it to the event — inrush, lightning coupling, or load dump — is what turns a failed part into a protection upgrade. This guide covers the signatures, the tracing, and the fixes.

(Last modified date: August 31, 2026)

Failure-Analysis Method — Signature First, History Second, Waveform Third:
Classify the visual signature (melted, cracked, degraded), then match it to the likely event (inrush, lightning, load dump), then confirm with the operating history and the real surge waveform. The signature narrows the causes; the history and waveform confirm the one.

Surge Damage Signatures: Melted, Cracked, Degraded

Three signatures cover most surge failures, and each points in a different direction:

A melted or cratered die usually means a single overcurrent event that exceeded the device’s surge capability.
A cracked package or lifted bond can follow either a single violent event or repeated smaller surges that fatigued the connections.
A degraded part — higher leakage, softer breakdown, drifting characteristics — is the signature of repeated stress that never quite killed the part outright.

The inspection is done safely: de-energize the supply and discharge the capacitors before handling, and treat the disassembly as live-work until verified. Classify the signature before the circuit theory starts, and read the visual evidence before any electrical testing — the melted die, cracked package, or lifted bond are evidence that testing can destroy. Photograph the part in place, remove it, and only then measure the electrical characteristics the signature predicts. The degradation signature is the one that hides: a part with drifting leakage passes the functional test at the bench and fails in the field, which is why the returned part is measured against its datasheet limits rather than against “it still works.”

Tracing the Event: Inrush, Lightning, Load Dump

Each signature maps to likely events, not certainties:

Power-on inrush from capacitor charging is the most common surge source; its signature is often a shorted rectifier or bridge die after repeated hot-plug cycles.
Lightning coupling arrives as a fast, high-energy current pulse; its damage can be cratering or package cracking, and it usually appears after a storm, not after routine operation.
Load dump in automotive systems is a voltage spike when the battery disconnects under alternator load; its signature is a breakdown in the input protection or rectifier.

The event history is the second half of the trace: what happened before the failure, how many events, and at what operating condition. Collect the history from the same people who saw the event — the operator who heard the pop, the log that records the hot-plug count, the weather report that matches the lightning strike. The trace is a small investigation, and the evidence list — signature, history, waveform — is its record.

IFSM vs the Real Event Waveform

The surge-rating check compares the event to the device’s capability: define the event’s shape, width, and peak, read the device’s IFSM at the matching width and starting temperature, and add margin for repeated events. A part that failed during power-on inrush after months of hot-plug cycles was likely failing the count, not the single event; a part that failed after a lightning strike was likely outside the rating’s time frame. The waveform comparison separates the two.

The comparison is where the trace becomes quantitative: the inrush waveform from a current probe, the lightning surge characterized by its 8/20 µs shape, and the load-dump pulse from the automotive standard each have a width and a peak, and each is checked against the part’s surge capability at that width. The headline number becomes a set of three checks. The three-event check also sets the protection design: the layer that covers the widest event — usually the lightning-coupled one — cannot be skipped, because the other events are handled by the same chain.

Protection Upgrades: TVS, NTC, Fusing

The fix follows the traced cause:

If the event is voltage-driven, a TVS or varistor across the line clamps the transient before it reaches the rectifier, sized by the three-voltage method (stand-off for the rail, breakdown clearing normal transients, clamping below the protected limit).
If the event is inrush-current-driven, an NTC thermistor or soft-start circuit limits the charging current and shrinks the surge requirement on the rectifier.
If the event is a downstream fault, the fuse or breaker must open before the rectifier’s capability is exceeded.

The three layers work together, and each is sized with margin, not to the edge: the TVS clamps below the rectifier’s blocking capability, the NTC limits inrush below the surge rating with room for repeated events, and the fuse opens before the rectifier’s single-event capability is reached. Each layer protects the one behind it, and the chain is only as strong as its weakest sizing.

Preventing Recurrence

Prevention is the loop that closes the diagnosis: the trace identifies the event, the upgrade addresses it, and a re-test with the event reproduced confirms the fix. The re-test runs the worst case — maximum input, discharged capacitors, repeated cycles for the hot-plug case — and verifies that the rectifier’s surge and thermal margins hold. If the numbers still do not close, the honest next step is a higher surge class or a redesigned protection chain, not a different brand of the same part.

The re-test record is part of the fix: the waveform, the surge count, the case temperature, and the pass criterion are written down so the same event can be reproduced by the next engineer. A surge diagnosis that is not documented as a test procedure is a story, not a fix. The record also includes the part’s final state — replacement or redesign — so the field history stays consistent. The diagnosis is complete when the fix, the test, and the record all exist.

Engineering note: the signature classification and event tracing follow failure-analysis practice as documented in JEDEC JEP122 — signature first, history second, waveform comparison third. The signatures are possible causes, not certainties, and the protection upgrades are sized by the traced event.

Frequently Asked Questions

What does a melted rectifier die mean?

Usually a single overcurrent event that exceeded the device’s surge capability — an inrush beyond IFSM or a lightning-coupled pulse. The event history confirms the cause.

How do I trace a surge failure?

Classify the signature (melted, cracked, degraded), then match it to the likely events — inrush, lightning, or load dump — and confirm with the operating history. The signature narrows the causes; the history confirms the one.

Why does a part fail after repeated hot-plug cycles?

Each inrush ages the die and connections even inside the single-event rating; the failure is the count, not one event. Add margin for repeated duty or add an NTC/soft-start.

What protects against lightning-coupled surges?

A TVS or varistor across the line, sized by the three-voltage method, with the surge waveform matched to the real event. The clamp keeps the transient off the rectifier.

Should I replace the part or the protection?

Diagnose first: if the part failed outside its ratings at the real event, the protection chain is the fix. Only when the part fails inside its envelope is the part itself the suspect.