Welding and induction heating are the harshest IGBT duties outside the motor drive, because both throw high-current pulses at the device faster and harder than a steady load ever does. A welder draws a current spike at every arc strike; an induction heater switches at high frequency against an inductive coil. Both punish the IGBT at the same two limits — the rate of voltage change dv/dt and the rate of current change dI/dt — and both produce failure signatures that a smoke test can misread. This article compares the load profiles, explains the two rate limits, decodes the common failure signatures, and gives the repetition-and-cooldown selection logic that keeps the part alive.
Pulse Profiles of Welding vs Induction vs Motor Drives
The three IGBT duties differ in their pulse profiles, and the difference sets the protection and selection. A motor drive repeats a steady switching cycle at a fixed frequency, with a relatively predictable current. A welding inverter draws a high-current pulse at every arc strike, with a sharp rise and a variable duty as the arc strikes and restrikes. An induction heater switches continuously at a high frequency into an inductive coil, so its current is a steady train of fast edges rather than discrete strikes.
The consequence is three different stress signatures. The motor drive stresses the IGBT thermally and by switching loss at a steady rate; the welder stresses it by repeated high-current inrush with cooldown gaps; the induction heater stresses it by continuous high-frequency edges where the switching loss dominates. A part selected for one profile can fail on another, which is why the profile is the first thing to name. The IGBT guide and the 650 V appliance article frame the application families; this article is the pulse-profile comparison that the selection relies on.
dv/dt and dI/dt: The Two Limits That Kill IGBTs in These Jobs
The two rate limits are the crux of welding and induction failure. The voltage slew rate dv/dt is how fast the collector-emitter voltage changes during switching; a high dv/dt is needed for low switching loss but stresses the structure and couples into the gate. The current slew rate dI/dt is how fast the collector current changes; a high dI/dt in an inductive load can push the current beyond the part’s capability during the transition.
In a welder, the arc strike produces a sharp current rise with a high dI/dt; in an induction heater, the continuous high-frequency switching keeps dv/dt elevated. When a design pushes either rate past the datasheet limit, the failure is not gradual heat but a fast, destructive event — the junction is overwhelmed in a single edge rather than cooked over time. The high-dv/dt driver article covers the drive and layout that control these rates, and the motor-drive IGBT article shows the frequency context where the rates are managed.
The design response to the rate limits is a mix of the right drive, the right gate resistance to control dv/dt, and the right snubber or layout to control dI/dt. Pushing the rates for efficiency without respecting the limit is exactly how the fast failures happen.
Where the Smoke Test Misleads: Common Failure Signatures
The smoke test — powering the circuit and watching what burns — is the last resort, and in these duties it misleads because the fast rate failures leave little time for heat to accumulate. A part killed by a single high-dv/dt event may fail instantly with no prior warmth, and the smoke test attributes the failure to whatever was powered at that moment rather than to the rate limit. The signature of a rate failure is different from a thermal one: no gradual heat, a single catastrophic event, and often a shorted junction.
The failure signatures worth recognizing: a shorted collector-emitter is the fast rate or overcurrent signature; a gate that fails open or shows leakage is the dv/dt-coupling signature; a part that runs hot before dying is the thermal or repetition signature. Reading the signature instead of the smoke tells a technician which limit was exceeded, and the rectifier failure-modes guide frames the same evidence-reading discipline that separates one signature from another. A simple rule sharpens the read: if the part failed instantly with no prior heat, suspect a rate or overcurrent limit rather than a thermal one, and check the drive and the strike waveform before blaming the device, because in a welder the waveform, not the silicon, is the more common offender. This is the field check that separates a real IGBT failure from a driven one.
A worked welder example puts the repetition math in numbers. A spot-welder IGBT carries a 120 A pulse for 2 ms at every strike, in a 60 °C cabinet, with the strikes spaced 500 ms apart — a 4 ms-period duty of 2 ms on, about 0.4%. The peak check: the part lists a pulse current of 300 A at 1 ms and 200 A at 10 ms, so the 120 A at 2 ms is inside the pulse limit with margin. The thermal check: the saturation drop of about 2.5 V at 120 A gives a conduction loss of 0.3 kW per pulse; over the 0.4% duty that averages to about 1.2 W, which the junction sheds easily. The part is comfortable at this repetition. Now speed the strikes to 100 ms spacing — duty rises to 2%, the average jumps to 6 W, and the thermal check tightens; at 50 ms spacing the same part approaches its limit and the cooldown becomes the binding constraint. The example shows the two checks working together: the peak kept the part in the pulse region, and the duty decides whether the average stays cool.
The three duties compare in one table:
| Duty | Pulse shape | Dominant stress | Selection emphasis |
|---|---|---|---|
| Welding | High-current strikes, variable duty | dI/dt and repetition | Pulse rating + cooldown |
| Induction | Continuous fast edges | dv/dt and switching loss | Frequency corner |
| Motor drive | Steady switching cycle | Steady switching loss | Frequency + thermal |
The table is the profile comparison in one view: welding picks the part by its pulse and repetition headroom, induction by its frequency corner, and the motor drive by its steady thermal balance. A design that names the dominant stress before opening the catalog is already most of the way to the right part, because the three emphases point to three different rating columns.
Selecting for Repetition: Current, Time, and Cooldown
The welding duty is defined by repetition: the arc strikes, the current pulses, and the cooldown gaps between. The selection must account for the peak current per strike, the duration of each pulse, and the cooldown that lets the junction reset. A part rated for a single high-current pulse is not automatically safe at a high repetition rate, because the junction accumulates heat across the strikes when the cooldown is too short.
The selection math is the duty-cycle logic from the SOA reading: the peak must stay inside the pulse limit for the pulse duration, and the average power over the full strike-and-cooldown cycle must stay inside the DC limit at the ambient. A welder that strikes and cools slowly lets the junction reset; a welder that strikes rapidly without cooldown turns a safe peak into an overheated average. The thermal reality worksheet and the loss method apply the same duty math, and the IGBT family spans the current classes sized for the repetition.
Field Notes: Diagnosing IGBT Failure in a Welder
The repair-side diagnosis closes the article with field practice. When a welder IGBT fails, the diagnosis runs in order: first, confirm the power stage is discharged and safe; second, test the IGBT’s junctions with the meter to identify a short, open, or gate failure; third, read the failure signature to name the limit that was exceeded — a short suggests overcurrent or a rate event, a gate failure suggests dv/dt coupling or a drive problem; fourth, check the cooldown duty to see whether the repetition outran the junction reset.
The diagnosis feeds the repair: fixing the IGBT without addressing the limit that killed it guarantees a repeat failure. A welder that keeps losing IGBTs to dv/dt needs a gate-resistance or drive fix, not a bigger part; one that loses them to repetition needs a cooldown or thermal-path fix. The rectifier failure diagnosis and the overheating evidence method are the two field companions that make the signature reading reliable, and the IGBT families supply the replacement parts once the real cause is named. The closing rule for these harsh duties is that the IGBT is not failing randomly — it is failing on a specific, nameable limit, and the limit is almost always one of four: a pulse peak too high, a repetition cycle too fast, a dv/dt or dI/dt rate pushed past the part, or a thermal path too weak. Naming the limit before replacing the part is what turns a recurring repair bill into a one-time fix, and it is the same discipline the welding and induction cases demand from every engineer who services them.