Test of surge withstand capability of discrete semiconductors

Discrete Semiconductor Surge Capability Testing: Methods That Separate Survivors from Failures

Surge capability is the parameter nobody thinks about until something blows up. A voltage spike hits the bus, the diode clamps, the MOSFET absorbs the energy, and everything keeps running. Or it does not. The device punches through, the board catches fire, and the field failure report lands on your desk with a single line: "unknown cause."

That unknown cause is almost always a surge event that exceeded the device's capability. The datasheet gave you a number. You ignored it because it did not fit your simulation model. Now you are debugging a destroyed converter at two in the morning.

Surge testing is not optional. It is the difference between a design that survives real-world transients and one that works perfectly on the bench and dies in the field.

What Surge Capability Actually Means

Surge capability is not a single number. It is a family of parameters that describe how much energy a device can absorb before it fails. The most common ones are non-repetitive surge current, repetitive surge current, single-pulse avalanche energy, and I squared t rating.

Non-repetitive surge current is the peak current the device can handle once. Think of it as the worst-case lightning strike or load dump event. The device sees this current for a defined pulse width, usually 8.3 milliseconds for a half-sine wave or 10 milliseconds for a rectangular pulse, and it must survive without degradation.

Repetitive surge current is the current the device can handle over and over. This matters for applications with frequent transients, like automotive alternator load dumps or motor drive inductive kickback. The repetition rate might be once per second or once per minute, and the device must survive thousands of cycles without parameter drift.

Single-pulse avalanche energy is the total energy the device can absorb during an avalanche event. This is measured in millijoules or joules, and it tells you how much inductive energy the device can clamp before the junction temperature exceeds the damage threshold.

I squared t rating is the thermal limit. It defines the maximum integral of current squared over time that the device can survive. This number is derived from the junction thermal mass and the maximum allowable junction temperature. Exceed it and the silicon melts, not gradually, but instantly.

Core Surge Test Methods

Non-Repetitive Surge Current Test: The 8.3ms Half-Sine Pulse

The most common surge test for diodes and thyristors uses an 8.3 millisecond half-sine current pulse. This waveform simulates a single surge event like a load dump or a short-circuit fault.

The test setup uses a capacitor bank charged to a defined voltage, discharged through the device under test with a series inductor that shapes the current into a half-sine wave. The peak current is set by the capacitor voltage and the inductance value. The pulse width is set by the LC time constant, which for an 8.3 millisecond half-sine requires a specific L and C combination.

You apply one pulse. Then you measure the device. For a diode, you check the forward voltage at a low test current. If VF has shifted by more than 5 to 10 percent, the device has degraded and should be rejected. For a thyristor, you check the holding current and the leakage current. Any shift indicates junction damage.

The critical detail is that this is a destructive test if you get the energy wrong. If the pulse energy exceeds the device rating, the junction melts and the test is over. You do not get a reading. You get a short circuit. This is why you start with a low energy pulse and step up to the rated value, not the other way around.

Repetitive Surge Current: The Endurance Test

Repetitive surge testing is where most devices fail in qualification. The single pulse test is easy. The device sees one hit and survives. But 10,000 pulses at 80 percent of the rated surge current will find weaknesses that a single pulse never reveals.

The test applies pulses at a defined repetition rate, typically one pulse per second or one pulse per minute, for a defined number of cycles. The standard counts are 100, 1000, and 10000 pulses depending on the application class.

Between pulses, the device must cool down enough that the junction temperature returns to near ambient. If the repetition rate is too fast, the junction heats up cumulatively, and you are testing thermal fatigue, not surge capability. The two are different failure modes, and mixing them up gives you meaningless data.

After the pulse sequence, you run a full parameter check. Forward voltage, leakage current, breakdown voltage, switching times if applicable. Any parameter that has shifted beyond the datasheet limits means the device failed. A shift of 3 percent in VF might look small, but it indicates micro-damage in the junction that will grow over time in the field.

Avalanche Energy Test: The Real Story for MOSFETs and IGBTs

For power MOSFETs and IGBTs, surge capability is about avalanche energy, not surge current. When an inductive load is switched off, the energy stored in the inductance has nowhere to go. It forces the device into avalanche breakdown, and the device must absorb that energy without destroying the junction.

The standard test uses a clamped inductive switching circuit. You charge an inductor to a defined current, then turn off the switch and let the inductive energy drive the device into avalanche. The energy is calculated as one-half L times I squared.

The key measurement is the avalanche energy at failure, Eav. This is the maximum energy the device can absorb before the junction temperature reaches the critical damage point, typically around 300 to 400 degrees Celsius for silicon.

The test is destructive by nature. You increase the inductance or the current step by step until the device fails. The last pulse that the device survives is your Eav. Compare it against the datasheet rating. If it is below the spec, reject the lot.

For qualification testing, you do not test to failure. You apply a pulse at 80 percent of the rated avalanche energy, 1000 times, and then check for parameter drift. The device must show no measurable degradation after 1000 avalanche events. This proves it can handle repeated inductive kickback in the field.

Test Circuit Design: Where Most Engineers Get It Wrong

Inductor Selection and Energy Control

The inductor in your surge test circuit is not just a component. It defines the entire test. The wrong inductor gives you the wrong waveform, the wrong energy, and the wrong result.

For an 8.3 millisecond half-sine pulse, the inductance must be chosen so that the current decays to zero in exactly 8.3 milliseconds. The formula is L equals V times t divided by I peak, where V is the capacitor voltage, t is the pulse width, and I peak is the target surge current.

If your inductance is too low, the pulse is too short and the energy is too small. You are under-testing the device and getting a false pass. If your inductance is too high, the pulse is too long and the energy is too large. You are over-testing and destroying good devices.

Measure the actual inductance with an LCR meter before you start. Do not trust the nominal value. Inductors have tolerances of 10 to 20 percent, and that tolerance directly translates into energy tolerance.

Capacitor Bank Sizing and Voltage Control

The capacitor bank stores the energy that becomes the surge pulse. The energy is one-half C times V squared. For a target energy of 10 joules at 100 volts, you need 2000 microfarads. At 200 volts, you need only 500 microfarads.

The capacitor voltage must be controlled precisely. A 5 percent error in voltage creates a 10 percent error in energy, because energy scales with V squared. Use a regulated supply to charge the capacitors, not a raw bench supply with ripple.

The capacitor ESR matters too. High ESR limits the peak current and rounds off the pulse edges. For fast surge pulses, use film capacitors or ceramic capacitors with low ESR. Electrolytic capacitors have too much ESR for precise surge testing.

Current Sensing and Peak Capture

You need to know the exact peak current of every pulse. A current transformer with insufficient bandwidth will miss the true peak, especially if the pulse has fast rise time. For an 8.3 millisecond half-sine, the rise time is roughly 4 milliseconds, which means the frequency content goes up to about 250 kilohertz. Use a current transformer with at least 1 megahertz bandwidth.

A Rogowski coil works well here because it has no core saturation and no DC offset. Place it as close to the device as possible to minimize loop inductance. The voltage output of the Rogowski coil is proportional to di/dt, so you need to integrate the signal to get current. Use a passive integrator with a time constant matched to the pulse width, or a digital integrator in your oscilloscope.

Temperature Effects on Surge Capability

Surge capability drops at high junction temperature. The reason is simple: the junction is already hot, so the avalanche energy needed to push it to the damage threshold is smaller. A device that can absorb 50 millijoules at 25 degrees might only handle 30 millijoules at 125 degrees.

This means your room-temperature surge test is optimistic. If the device will operate at high temperature in the field, you need to test at temperature. The standard practice is to run the surge test at the maximum rated junction temperature, or at least correct the room-temperature number using the temperature derating curve from the datasheet.

For automotive applications, the test temperature is often 150 degrees Celsius junction temperature, because that is what the device sees under the hood in summer. A surge test at room temperature for an automotive part is almost useless.

Cold temperature has the opposite effect. Surge capability increases at low temperature because the junction can absorb more energy before reaching the damage threshold. But cold temperature also makes the device more brittle, so mechanical stress from the surge current pulse can crack the bond wires. This is a failure mode that shows up only at low temperature, and it will surprise you if you never test there.

Qualification vs Production Testing

Qualification testing and production testing serve different purposes, and they use different methods.

Qualification testing is about proving the design can survive the worst case. You test to destruction or near-destruction. You run thousands of pulses. You test at temperature extremes. You measure every parameter before and after. This takes hours per device and you only do it on sample lots, not every part.

Production testing is about sorting good parts from bad parts. You apply a single pulse at 80 to 90 percent of the rated surge energy. You measure forward voltage before and after. If VF shifts by more than 5 percent, the part fails. This takes seconds per device and you can run it on every part that ships.

Do not skip production surge testing because you passed qualification. A weak bond wire that survived 1000 pulses in qualification might fail on the 50th pulse in the field. Production testing catches those marginal parts before they leave your facility.

Practical Tips That Prevent False Failures

Always discharge the capacitor bank completely between pulses. Residual charge adds to the next pulse and creates an unpredictable energy level. Use a bleed resistor across the capacitors with a time constant under 1 second.

Check the device temperature before each pulse. If the junction has not cooled to within 10 degrees of ambient, the next pulse will start from a higher baseline temperature, and your energy calculation is wrong. The energy rating assumes a cold start. A hot start changes the physics.

Verify your pulse shape on every test run. Connect the oscilloscope to the current probe and confirm the half-sine waveform. If the pulse looks distorted, clipped, or has ringing, stop and fix the circuit before testing any devices. A distorted pulse delivers energy differently than a clean half-sine, and your pass/fail decision is based on a waveform that does not match the standard.

Ground everything properly. Surge currents create large magnetic fields that induce voltages in nearby loops. If your measurement ground loop picks up induced voltage, your current reading will be wrong. Keep the current sense loop as small as possible and use twisted pair for the sense leads.