A diode that switches off too slowly kills your efficiency. A transistor that lingers in the linear region during turn-off burns up your heatsink. Recovery time is the parameter that separates a fast switching device from a slow one, and testing it wrong gives you data that looks good on paper but falls apart in the field.

Recovery time testing for discrete semiconductors covers diodes, BJTs, MOSFETs, IGBTs, and thyristors. Each device has its own recovery mechanism, its own test method, and its own failure modes. Getting the test right means understanding what recovery actually looks like inside the device, not just reading a number off a curve tracer.

This specification walks through the testing methods, the conditions that affect the results, and the setup errors that turn valid data into noise.

What Recovery Time Actually Measures

Recovery time is not a single event. It is a sequence of events that happen when a device transitions from conducting to blocking, or from blocking to conducting. Each event has its own time constant, and each one matters for a different reason.

For diodes, reverse recovery time (trr) is the big one. When you reverse-bias a conducting diode, it does not stop conducting instantly. Stored charge in the junction keeps current flowing in the reverse direction for a short time. That reverse current spike causes ringing, EMI, and extra loss in any circuit with a complementary switch.

For BJTs, storage time (ts) dominates. The base region stores minority carriers during saturation. When you try to turn the device off, those carriers have to recombine or be swept out before the collector current can fall. Until that happens, the device is still on, and your switching waveform has a long tail.

For MOSFETs and IGBTs, the picture is more complex. You have gate charge removal, drain current fall time, and voltage rise time. The turn-off delay, the current fall interval, and the voltage rise interval each contribute to the total switching loss.

Testing recovery time means measuring these intervals accurately under conditions that match your real application.

Reverse Recovery in Diodes: Why It Destroys Efficiency

When a diode switches from forward conduction to reverse blocking, the stored charge in the drift region creates a reverse current pulse. The duration of that pulse is trr. The peak of that pulse is Irr. The product of the two, multiplied by the reverse voltage, is the energy lost every switching cycle.

In a hard-switched converter running at 100kHz, that energy loss adds up fast. A diode with trr of 50ns might dissipate 2W in switching loss alone. A diode with trr of 15ns dissipates 0.6W. The difference is not in the datasheet footnotes. It is in your thermal design.

The shape of the recovery current matters as much as the duration. A soft recovery diode ramps the current down gradually. A snappy recovery diode drops it fast but creates a sharp current spike that excites parasitic inductance. Neither shape is universally better. It depends on your circuit topology.

To test this properly, you need a test circuit that replicates the actual commutation conditions. A simple resistor-inductor load with a fast switch gives you a clean current commutation. Measure the reverse current waveform with a current probe or a shunt resistor and a high-bandwidth oscilloscope. Extract trr, Irr, and the recovery charge Qrr from the waveform.

Storage Time in BJTs: The Tail That Kills Switching Speed

A BJT driven into deep saturation stores a large amount of charge in the base. When you remove the base drive, that charge has to go somewhere. It recombines slowly, and during that time, the collector current stays high. This is storage time, and it is the reason BJTs are slower than MOSFETs in most switching applications.

The storage time depends on how deep the saturation is. A BJT with a forced beta of 10 stores more charge than one with a forced beta of 50. If your drive circuit overdrives the base, you are making the storage time worse on purpose.

Test storage time by driving the base with a current pulse that puts the device into saturation, then removing the drive abruptly. Measure the time from drive removal to the point where collector current falls to 10% of its on-state value. That is your storage time.

Run the test at multiple base drive levels. The storage time at light saturation is much shorter than at deep saturation. Use the worst-case value for your design, not the typical value.

Test Circuit and Measurement Requirements

The test circuit is not a detail. It is the test. A bad circuit gives you bad data no matter how good your oscilloscope is.

Building a Commutation Test Circuit That Works

For diode reverse recovery, the standard test circuit uses a pulse generator to drive a switch that commutates current from the diode under test to a commutating capacitor or a second diode. The key is to make the current commutation fast and repeatable.

Use a low-inductance layout. Long traces add inductance that slows the current fall and distorts the recovery waveform. Keep the loop area between the diode, the switch, and the commutating capacitor as small as possible. A loop area of more than 1cm squared will add enough inductance to change your trr measurement by 20% or more.

For BJT storage time, the test circuit is simpler. A pulse generator drives the base through a resistor. A collector load resistor sets the on-state current. The key is to make the turn-off edge fast. A slow edge lets the base charge bleed off gradually, which makes the storage time look shorter than it really is. Use a totem-pole driver or a dedicated BJT driver IC to get a fast, clean turn-off edge.

For MOSFET and IGBT turn-off, you need a gate driver that can sink current fast. The gate charge removal speed determines how quickly the device leaves the linear region. A weak gate driver makes the device look slower than it is. Use a driver with at least 2A peak sink current for power MOSFETs and 5A or more for IGBTs.

Oscilloscope and Probe Selection for Recovery Measurements

Bandwidth is not optional. For a diode with trr of 20ns, you need at least 200MHz bandwidth to see the waveform accurately. A 50MHz scope will smear the recovery current into a blurry blob and give you a trr value that is 30% too long.

Use a current probe for diode reverse recovery. A shunt resistor works if the current is high enough, but at low current the voltage across the shunt is too small to measure accurately. A Hall-effect current probe with 100MHz bandwidth or better gives you a clean waveform without adding series resistance to the circuit.

For voltage measurements on MOSFETs and IGBTs, use a high-voltage differential probe. Do not use a single-ended probe with a long ground lead. The ground lead inductance creates ringing that looks like overshoot but is actually a measurement artifact. A differential probe with short leads eliminates this problem.

Probe compensation matters. An under-compensated probe rounds the edges of your waveform. An over-compensated probe adds ringing. Compensate every probe before every test session.

Temperature and Bias Dependence of Recovery Time

Recovery time is not a fixed number. It changes with temperature, bias current, and voltage. If you test at one condition and design for another, your timing margins are wrong.

How Temperature Shifts Recovery Parameters

For diodes, reverse recovery time increases with temperature. The stored charge increases because carrier lifetime increases at higher temperature. A diode with trr of 30ns at 25°C might have trr of 50ns at 125°C. That 20ns difference changes your switching loss calculation significantly.

For BJTs, storage time increases dramatically with temperature. At high temperature, carrier recombination slows down, and the stored charge takes longer to clear. A BJT that turns off in 80ns at 25°C might take 200ns at 125°C. If your circuit has a dead-time budget of 100ns, the device works at room temperature and fails at operating temperature.

For MOSFETs, the gate charge removal time increases slightly with temperature because the threshold voltage drops and the gate driver has to sink more charge to reach the off state. The effect is smaller than for BJTs but still measurable.

Test at the maximum operating temperature for your application. If the device operates at 125°C, test at 125°C. Do not test at 25°C and assume the numbers scale linearly. They do not.

Effect of Forward Current on Diode Recovery

The higher the forward current before turn-off, the more charge is stored in the diode, and the longer the recovery time. This is not a linear relationship. Doubling the forward current can increase trr by 50% or more.

Test at the actual operating current, not at a low current that gives you a nice number. If your diode carries 10A in the application, test it at 10A. A trr measurement at 1A tells you nothing about the device at 10A.

The same principle applies to BJT storage time. Higher collector current means deeper saturation, more stored charge, longer storage time. Test at the current your circuit actually uses.

Defining Pass/Fail Criteria and Sample Size

A recovery time test without clear pass/fail criteria is just a measurement exercise. It tells you what happened. It does not tell you whether the device is acceptable.

Setting Realistic Specification Limits

The datasheet gives you a maximum trr or ts. Use that as your pass/fail limit. But add a margin. A device that measures 48ns against a 50ns maximum is passing by the letter of the spec but has no margin for process variation or temperature shift.

For production testing, set the fail limit at 80% of the datasheet maximum. A device with trr of 40ns against a 50ns maximum passes. A device with trr of 46ns fails. This gives you a guard band that catches drifting process lots before they ship.

For characterization testing, record the actual value and compare it against the datasheet range. Do not bin the data into pass/fail. You need the distribution to understand your process.

How Many Samples Do You Actually Need

One sample tells you nothing about recovery time variation. Five samples gives you a rough idea. Ten samples gives you a distribution you can trust.

For production testing, sample five devices per lot minimum. For characterization, sample at least ten. Plot the distribution. If the spread is wide, your process has variation that your circuit must tolerate. If the spread is tight, you can design closer to the limit.

Record the lot number, the test temperature, the bias current, the test voltage, and the waveform for every sample. When a field failure comes back, you need to compare it against the exact device that shipped. Without lot traceability, your test data is worthless.

Common Setup Errors That Corrupt Recovery Measurements

The most frequent error is using too much probe capacitance. A standard 10x probe adds 10pF to 15pF of capacitance to your circuit. For a fast diode, that capacitance slows down the current commutation and makes trr look longer than it is. Use a low-capacitance probe or a current probe instead.

The second error is ignoring the commutation speed. If your test switch is slow, the current does not transfer cleanly from the diode to the commutating path. The diode sees a slow current fall instead of a fast one, and the recovery waveform is distorted. Use a switch with a rise time at least ten times faster than the expected trr.

The third error is measuring at the wrong point in the waveform. Recovery time has multiple definitions. Some standards measure from the zero crossing of current to the point where reverse current falls to 25% of Irr. Others measure to 10%. The definition changes the number by 10% to 20%. Pick one definition and stick with it. Document it in your test specification.

The fourth error is not accounting for ringing. Parasitic inductance and capacitance in the test circuit create ringing that overlaps with the recovery waveform. If you measure trr from a ringing waveform, your number includes the ringing period, not just the recovery time. Dampen the circuit with a small series resistor if ringing is a problem. A 1Ω to 5Ω resistor in series with the diode can kill ringing without significantly affecting the measurement.

Writing a Recovery Time Test Spec That Your Team Will Follow

Start with the device identification. Part number, lot number, package, date code. Then define the test conditions explicitly. Forward current, reverse voltage, commutation speed, temperature, gate drive strength. Write these as numbers, not as ranges.

Define the measurement method. Which probe, which oscilloscope bandwidth, which definition of trr or ts. Specify the pass/fail limit as a number with units. No ambiguity.

Record the raw waveform for every sample. A number without a waveform is unverifiable. Store the waveforms with the test report. When someone questions a result six months later, you pull the waveform and settle it in five minutes.

Recovery time testing catches the failures that DC tests miss. A device can have perfect forward voltage, perfect leakage, perfect gain, and still fail in switching because its recovery time is out of spec. Do not skip this test. Do not rush it. The device you catch today is the field return you avoid tomorrow.