Transistor Module Overcurrent Shutdown Control: Techniques That Actually Protect Your Devices

Overcurrent is the fastest way to kill a transistor module. A short circuit on the load side can push current to ten times the rated value in under a microsecond. If the driver does not react fast enough, the device enters its short-circuit safe operating area, the junction temperature spikes, and the silicon melts — sometimes explosively.

The challenge is not detecting overcurrent. The challenge is reacting fast enough while avoiding false trips that kill your uptime. Every millisecond of delay in the shutdown path burns more energy into the device. Every false trip shuts down a perfectly healthy system. Getting this right means understanding the detection window, the shutdown speed, and the recovery behavior as an interconnected system — not as isolated features.

How Fast Do You Actually Need to React

Most datasheets specify a short-circuit withstand time of 5 to 10 microseconds for IGBT modules. For SiC MOSFETs, it is even shorter — often 2 to 3 microseconds. This is the total time from fault onset to the point where the device fails irreversibly.

But the 5 to 10 microseconds is not your shutdown budget. It is your total budget, and it includes the fault detection time, the driver response time, and the transistor turn-off time. If detection takes 2 microseconds and turn-off takes 1 microsecond, you have only 2 to 7 microseconds left for the driver to actually act.

This means the detection circuit must trigger within 1 to 2 microseconds of the fault. Any slower, and you are eating into the device's survival margin. In practice, most well-designed systems aim for a total shutdown time under 3 microseconds — giving the transistor a comfortable safety margin even under worst-case temperature and voltage conditions.

Desaturation Detection: The Most Common Method

The go-to technique for overcurrent detection in transistor modules is desaturation (desat) detection. It works by monitoring the collector-emitter voltage during the on-state.

When a transistor is fully on, the VCE is low — typically 1.5 to 2.5 volts for an IGBT. When a short circuit occurs, the current rises fast but the transistor cannot saturate, so VCE climbs to the bus voltage. The desat circuit watches for this voltage rise and triggers shutdown when VCE exceeds a threshold — usually 7 to 9 volts for IGBTs.

Setting the Threshold Correctly

The desat threshold is a trade-off between sensitivity and noise immunity. Set it too low — say 5 volts — and normal switching transients or inductive kickback from the load will trigger false shutdowns. Set it too high — say 12 volts — and the device spends too long in the high-voltage, high-current region before the fault is detected.

A practical starting point is 7 volts for IGBT modules and 4 to 5 volts for SiC MOSFETs (since SiC devices have a lower on-state voltage and the desat window is smaller). But you must verify this on your actual hardware with an oscilloscope. Inject a known overcurrent condition and measure the VCE waveform. The threshold should sit well above the normal on-state voltage but well below the bus voltage.

The Blanking Time Problem

Right after turn-on, VCE is naturally high for a few hundred nanoseconds while the transistor exits its active region and enters saturation. If the desat circuit is active during this window, it will see a normal voltage and interpret it as a fault. This is why every desat circuit includes a blanking time — a short window (typically 2 to 5 microseconds) after turn-on during which the desat comparator is disabled.

The blanking time must be long enough to cover the entire turn-on transient. If it is too short, you get false trips on every switching cycle. If it is too long, you lose precious detection time during an actual fault. Measure the turn-on transient on your specific module under worst-case conditions (high temperature, high bus voltage) and set the blanking time to 1.5 times that measured value.

Current Sensing Alternatives That Avoid the Desat Trap

Desat detection is simple and cheap, but it has a fundamental weakness: it detects overcurrent indirectly, by watching voltage. This means it reacts slowly — the voltage has to rise before it triggers. For faster protection, you need to sense current directly.

Shunt Resistor Sensing

A low-value shunt resistor in series with the emitter or source gives you a direct current measurement. The voltage across the shunt is proportional to current, and you can set a hard threshold with no blanking time required.

The problem is power loss. A 1 milliohm shunt at 500 amperes dissipates 250 watts continuously. This is unacceptable in most applications. You can reduce the resistance to 0.1 milliohm, but then the signal voltage is only 50 millivolts at 500 amperes — tiny, noisy, and hard to measure accurately.

Shunt sensing works best in low-current, high-precision applications like battery management or small motor drives. For high-power transistor modules, it is usually impractical.

Hall Effect Current Sensors

A Hall effect sensor gives you galvanically isolated current measurement with no power loss in the sensing element. The bandwidth of modern Hall sensors reaches 100 kilohertz or more, which is fast enough to catch overcurrent transients in the microsecond range.

The downside is latency. Even the fastest Hall sensors have a response time of 1 to 2 microseconds. Combined with the driver shutdown time, this puts you close to the edge of the transistor's short-circuit withstand time. Hall sensors are best used as a secondary protection layer — they catch faults that desat detection misses, like slow overcurrent ramps that never push VCE high enough to trigger desat.

Current Transformer for AC Applications

In AC motor drives and grid-tied inverters, a current transformer on the output phase gives you fast, isolated current sensing with excellent bandwidth. The transformer output is a scaled replica of the phase current, and you can detect overcurrent in under 500 nanoseconds.

The limitation is that current transformers do not work for DC. They also saturate under large fault currents, which distorts the waveform and can delay detection. Use them in combination with desat detection, not as a replacement.

Shutdown Speed: The Driver Side Matters More Than You Think

Detecting the fault is only half the job. The driver must turn the transistor off fast enough to limit the fault energy.

Active Clamping vs. Soft Turn-Off

When desat detection triggers, the driver must pull the gate low immediately. A standard driver with a passive gate resistor turns off the transistor at a speed determined by the resistor value. This might be fast enough for IGBTs, but for SiC MOSFETs, even a few hundred nanoseconds of slow turn-off during a fault can be fatal.

An active Miller clamp forces the gate voltage below the threshold with high current — 1 to 5 amperes — overriding any noise and ensuring the fastest possible turn-off. This is not optional for SiC modules. It is mandatory.

A soft turn-off — deliberately slowing down the gate discharge to reduce dV/dt — is the opposite of what you want during a fault. Soft turn-off is for normal operation to reduce EMI. During overcurrent, you want the hardest, fastest turn-off the driver can deliver.

Negative Gate Drive During Fault

For IGBT modules, applying a strong negative voltage (-8 to -15 volts) during fault shutdown speeds up turn-off dramatically. The negative voltage actively pulls carriers out of the base, collapsing the stored charge faster than a zero-voltage turn-off.

For SiC MOSFETs, the negative rail is equally important. A -5 volt gate drive during turn-off ensures the device fully turns off even if the Miller capacitance injects noise during the transient. Without a negative rail, the gate can float up during the fault, causing the transistor to re-turn on — a condition called fault re-ignition, which is almost always destructive.

Recovery Behavior: What Happens After the Shutdown

Shutting down is easy. Recovering safely is hard.

Auto-Retry vs. Latch-Off

After an overcurrent fault, you have two choices: auto-retry or latch-off. Auto-retry means the driver attempts to turn the transistor back on after a delay — typically 10 to 50 milliseconds. If the fault is gone, the system recovers automatically. If the fault is still there, the driver trips again.

Latch-off means the driver stays off until the controller explicitly resets it. This is safer because it prevents repeated fault energy pulses from destroying the device. But it requires the controller to diagnose the fault condition before allowing restart.

For industrial motor drives, auto-retry with a current limit is common. The driver retries, but limits the initial gate drive voltage to 50 percent. If the current stays below the threshold, it ramps up to full drive. This lets the system recover from transient faults (like a momentary mechanical jam) while protecting against sustained short circuits.

Soft Restart After Fault

When the driver retries after a fault, do not slam the gate with full voltage. The transistor just survived a high-energy event. Its junction temperature is elevated, and its safe operating area is reduced.

A soft restart ramps the gate drive voltage up over a few microseconds. This limits the inrush current and gives the device time to cool slightly before it sees full stress. Most modern gate driver ICs have a built-in soft restart feature — enable it.

False Trip Prevention: Keeping Your Uptime High

The best overcurrent protection is useless if it shuts down your system every time a motor starts or a capacitor charges.

Differentiating Between Inrush and Fault

Motor starting currents can be 5 to 8 times the rated current for a few hundred milliseconds. Capacitor charging on the DC bus can produce similar current spikes. These are not faults — they are normal operation.

Use a time-delay circuit on the desat output. If the overcurrent condition persists for less than 5 microseconds, ignore it. If it persists beyond 5 microseconds, trigger shutdown. This filters out inrush transients while still catching real short circuits.

The delay must be shorter than the transistor's short-circuit withstand time. For IGBTs with a 10 microsecond withstand time, a 5 microsecond delay leaves 5 microseconds for shutdown — which is tight but workable. For SiC modules with a 3 microsecond withstand time, you cannot afford a long delay. In that case, rely on current sensing (which reacts faster than desat) rather than desat with delay.

Temperature-Compensated Thresholds

The desat threshold should shift with temperature. At high junction temperatures, the on-state voltage rises, which means the desat window shrinks. A fixed threshold that works at 25 degrees might cause false trips at 125 degrees.

Some advanced drivers include temperature-compensated desat thresholds. If yours does not, you can add an external circuit that adjusts the threshold based on a temperature sensor mounted on the module baseplate. This is not trivial, but it is the difference between a system that trips on hot summer days and one that runs reliably year-round.

PCB Layout for Overcurrent Protection

The detection circuit is only as fast as the signal path between the transistor and the driver.

Keep the desat sense trace as short as possible — under 10 millimeters from the collector or drain pin to the driver input. Long traces add inductance that slows down the voltage rise seen by the comparator, delaying detection.

Route the sense trace away from high-current power traces. Capacitive coupling from a switching node into the desat sense line can inject false voltage spikes that mimic a desat event. Use a grounded guard trace alongside the sense line if the distance exceeds 5 millimeters.

The decoupling capacitors on the driver supply must be sized for the fault condition, not just normal operation. During a fault, the driver draws maximum current while trying to turn off the transistor. If the supply sags, the gate discharge current drops, turn-off slows, and the fault energy increases. This is a vicious cycle that ends with a destroyed module.