Key points for testing the breakdown voltage limit of discrete semiconductors

Breakdown Voltage Limit Testing for Discrete Semiconductors

Breakdown voltage testing is the one test every engineer dreads. Not because it is hard to set up, but because it is the test that can destroy your device in milliseconds if you get it wrong. One wrong setting, one missed current limit, and a perfectly good sample turns into scrap.

For discrete semiconductors — diodes, MOSFETs, IGBTs, BJTs, thyristors — the breakdown voltage (Vbr, BVdss, BVces, etc.) is the hard limit. Cross it, and the device either degrades permanently or fails outright. This test defines that limit. It also reveals weaknesses that no other test can catch.

This article covers the actual testing methods, the parameters that control accuracy, and the mistakes that turn good data into garbage.

Why Breakdown Testing Is Different From Every Other Test

Most semiconductor tests are gentle. You apply a small signal, measure the response, move on. Breakdown testing is not gentle. You push the device to the edge of destruction and measure how close it gets.

The purpose is not to break the device. The purpose is to find the exact voltage where it starts to conduct heavily in reverse, then verify it stays within spec. The difference between a device that breaks down at 600V and one that breaks down at 450V is the difference between a design that works and one that catches fire.

This test also exposes latent defects. A device with a microscopic crystal defect might pass every other test and fail breakdown at 70% of its rated voltage. That defect would not show up in forward voltage or leakage current tests. Breakdown testing is the stress test that finds it.

For automotive applications under AEC-Q101, breakdown testing is mandatory. For industrial applications under IEC 60747, it is required. For military under MIL-STD-750, it is even more stringent. There is no skipping this test.

Understanding What Actually Happens During Breakdown

Before you run the test, you need to understand the physics. Two mechanisms dominate: avalanche breakdown and Zener breakdown. They look similar on a curve but behave very differently under stress.

Avalanche Versus Zener: Why It Matters for Your Test

Avalanche breakdown happens when carriers gain enough kinetic energy to knock electrons loose from the lattice. This is a chain reaction. One carrier frees two, those two free four, and current explodes. This dominates in devices rated above about 6V.

Zener breakdown happens through quantum tunneling. The electric field is so strong that electrons tunnel directly through the bandgap. This dominates below about 5V. The breakdown is sharper, more predictable, and less destructive.

For your test, this matters because avalanche devices are more sensitive to temperature and current ramp rate. Zener devices are more stable but can still fail if you overshoot the current limit.

The datasheet will tell you which mechanism applies. If it does not, assume avalanche and test conservatively.

Soft Breakdown Versus Hard Breakdown: The Distinction That Saves Devices

Soft breakdown means the device enters high conduction gradually. The current rises smoothly as voltage increases past the knee. The device can survive this if you limit the current.

Hard breakdown means the device snaps into full conduction instantly. There is no knee, no warning. The current jumps from microamps to amps in nanoseconds. If your current limiter does not react fast enough, the device is dead.

Most modern discrete devices are designed for soft breakdown. But process variations can turn a soft device into a hard one. You will not know until you test. That is why current limiting is not optional. It is survival.

Setting Up the Test Correctly

The setup is simple. The execution is where things go wrong.

Voltage Ramp Rate Controls Everything

The speed at which you ramp the voltage changes the measured breakdown voltage. Faster ramp rates give higher apparent breakdown voltages. Slower ramp rates give lower values. This is not a measurement error. It is real physics.

At fast ramp rates, the device does not have time to heat up. At slow ramp rates, localized heating reduces the breakdown voltage. The standard ramp rate is typically 10V/s to 100V/s depending on the device type and the applicable standard. Check IEC 60747 or JEDEC JESD22 for the exact rate.

If you ramp too fast, your data looks better than reality. The device will break down at a lower voltage in the field where the ramp is slower. If you ramp too slow, you might fail good devices that would have passed at the standard rate.

Use a programmable power supply with adjustable ramp rate. Do not use a manual knob. Manual means inconsistent, and inconsistent means useless data.

Current Limiting: The One Setting That Prevents Destruction

Set the current limit before you apply any voltage. This is not a suggestion. This is the rule.

The typical test current is 1mA to 100µA for diodes, and 250µA to 1mA for transistors, depending on the rated voltage. Check the datasheet or the applicable standard for the exact value. The current limit must be low enough to prevent damage but high enough to give a clean, measurable reading.

If the device hits breakdown and the current limiter does not engage within microseconds, you have destroyed the sample. Use a supply with fast current limiting, ideally under 1µs response time. Bench supplies with slow limiting are not acceptable for this test.

Also, set the compliance voltage. This is the maximum voltage the supply will output even if the current limit is not reached. Set it 10% to 20% above the expected breakdown voltage. This prevents the supply from hunting or oscillating near the breakdown point.

What the Data Should Look Like

A good breakdown test produces a clean curve. Voltage on the x-axis, current on the y-axis, log scale for current. Below breakdown, the current is flat and near zero. At breakdown, the current rises sharply. The knee of the curve is your breakdown voltage.

If the curve is noisy, your measurement has interference. Shield your cables. Use a guarded fixture. Keep the test leads short.

If the curve shows a gradual rise instead of a sharp knee, the device may have a soft leakage path. This could indicate contamination or a packaging defect. Flag it and test another sample.

If the device does not break down at all up to your compliance voltage, you have either a very good device or a wrong test setup. Verify your connections, then verify your supply is actually reaching the set voltage. A broken probe will give you a false pass.

Run at least five samples per lot. Breakdown voltage has part-to-part variation. One sample tells you nothing. Five samples tell you whether your lot is in spec.

Common Mistakes That Invalidate Breakdown Data

The most frequent mistake is testing at the wrong temperature. Breakdown voltage has a positive temperature coefficient for most devices. A device that breaks down at 600V at 25°C might break down at 620V at 125°C. If you test at room temperature and ship the device to a customer who operates at high temperature, you have a mismatch.

Test at the worst-case temperature. For most silicon devices, that is 25°C for the minimum breakdown voltage. For some applications, you also need to test at the maximum operating temperature to verify the device does not exceed its absolute maximum rating.

The second mistake is not preconditioning the device. Apply a forward bias pulse before the breakdown test. This removes any trapped charge in the junction. Without preconditioning, your first breakdown reading can be off by 5% to 10%.

The third mistake is ignoring humidity. Moisture on the device surface creates leakage paths that distort the breakdown measurement. Test in a dry environment or bake the devices at 125°C for 24 hours before testing. This drives off moisture and gives you a clean reading.

The fourth mistake is using the wrong fixture. A socket with poor contact adds resistance. That resistance drops voltage before it reaches the device. Your measured breakdown voltage is lower than the actual value. Use a dedicated test fixture with Kelvin connections. If you must use a socket, verify contact resistance before every test session.

Writing a Breakdown Test Spec That Engineers Will Actually Follow

Start with the device identification. Part number, lot number, package type, and date code. This sounds obvious, but skipped identification is the number one cause of mixed-up test data.

Define the test conditions explicitly. Voltage ramp rate: 50V/s. Test current limit: 100µA. Compliance voltage: 700V for a 600V-rated device. Temperature: 25°C ± 2°C. Preconditioning: 10mA forward pulse for 1ms. Number of samples: 5 minimum.

Write the pass/fail criteria in absolute terms. Breakdown voltage must be greater than or equal to the minimum specified value. Not "close to." Not "approximately." Greater than or equal to.

Record the raw data. Voltage at 10µA, voltage at 100µA, and the full I-V curve if possible. The voltage at a specific current is more repeatable than the knee voltage, which can shift depending on how you draw the line.

Store the test report with the lot traceability. When a field failure comes back six months later, you will need to pull that report and compare it against the returned device. If you did not record it, you cannot compare it.

Test fixtures degrade over time. Socket contacts wear. Cable insulation cracks. Calibrate your setup quarterly. Verify the current limit response time annually. A fixture that passed calibration last year may not pass this year.

Breakdown testing is not complicated. It is unforgiving. Every parameter matters. Every step has a reason. Skip a step, and the data means nothing.