AEC-Q101 decides whether a discrete semiconductor is fit for a car. The chamber runs its two hardest environmental stresses, the reverse-biased damp soak that hunts surface leakage and the power cycling a self-heating part puts itself through. A discrete is graded to a junction temperature, not an ambient band, so the box has to hold its conditions steady while the device drives its own silicon hot.
AEC-Q101 is the stress-qualification standard the Automotive Electronics Council wrote for discrete semiconductors, the single transistors, diodes, rectifiers and power MOSFETs that a car leans on outside its integrated circuits. These parts switch the current to a motor, rectify an alternator, protect a rail, and carry the power a controller only commands. The standard puts them through a battery of accelerated stresses to prove they will last the life of the vehicle. A large share of that battery runs inside a temperature humidity chamber. What the chamber asks of a discrete is unlike what it asks of a chip, because a discrete is built to hold off voltage and to carry power. Both of those traits change the test.
An integrated circuit packs millions of low-voltage transistors that each sip a trickle of current. A discrete is one device doing one job at the scale the job needs, often blocking hundreds of volts or passing tens of amps. It might be a MOSFET switching a motor phase, a diode steering current through a converter, a rectifier feeding the battery, or an insulated-gate bipolar transistor carrying the heavy load of a traction drive. That difference reaches into every part of the qualification. A power discrete turns a real fraction of what it switches into heat, so its own dissipation, not the oven around it, sets how hot the silicon runs. It blocks voltage across a single junction, so a reverse field sits at the edge of that junction whenever the part is off. It lives or dies on its package, since the path that carries heat out of the die is the same path that fatigues when the die heats and cools.
The chamber work splits along those lines. The damp tests bias the part in reverse, the way it sits blocking in a circuit, and watch for the leakage that humidity breeds at a high-field surface. The life tests let the part heat itself, cycling its own junction the way a switching load would, and watch the bonds and the die-attach tire. The standard ties the result to a junction temperature, the hottest the silicon is rated to reach. Those traits send different stresses into the chamber, which together make a discrete’s qualification its own.
The signature humidity test for a discrete is H3TRB, high humidity high temperature reverse bias. The part sits at eighty-five degrees and eighty-five percent relative humidity for a thousand hours. While it sits there it is held in reverse, blocking a voltage set at eighty percent of its rated blocking voltage, capped at a hundred volts of direct current. The device is measured electrically before the soak and after it. The change between the two readings is the verdict. The reads cover what the part is built to do. Reverse leakage and breakdown voltage tell whether the blocking junction still holds. For a transistor the gate threshold and the on-resistance tell whether the channel still behaves. A part passes when those numbers stay inside their limits across the thousand hours. It fails when the soak has pushed any of them past what the datasheet allows. The reverse bias is the heart of why this test belongs to discretes and reads differently from the forward-biased soak a chip would see.
A discrete spends the bulk of its working life off, holding a voltage back. When it blocks, the full reverse field falls across the junction and crowds at the junction’s edge, the termination, where the depletion region reaches the surface of the silicon. That surface is the weak ground. Moisture that has worked through the moulding compound settles on the passivation over the termination. The high field there drives it to do its damage, pulling ions across the surface and corroding the metal that rings the junction. A part that blocked cleanly when dry starts to pass a small reverse current that climbs as the surface degrades. The forward-biased soak a chip endures hunts corrosion at the metal between conductors. The reverse-biased soak a discrete endures hunts it at the one place a power device cannot afford it, the high-field edge that has to hold the voltage off. The damage is electrochemical in both. The deeper account of how water and a field corrode a semiconductor surface belongs to the humidity-bias method itself, but the geometry that H3TRB stresses, a single blocking junction under a reverse field, is the discrete’s own.
The chamber’s task under H3TRB is exact and unglamorous. It has to hold eighty-five and eighty-five flat for the full thousand hours, never letting a cool surface inside condense a film onto a part that is sitting at hundreds of volts of stored potential nearby. It has to carry that reverse bias to every device through feedthroughs that stay sealed and insulated in the damp. The verdict means something only when every part met the same moisture and the same field for the same long stretch.
A chip is soaked while it works. A discrete is soaked while it blocks.
A humidity test wants the part wet. A power part, given any current, wants to be dry, because it heats itself and drives the damp off its own surface. That tension runs underneath the discrete’s qualification. Under H3TRB the reverse bias is set low and the leakage small, so the part dissipates almost nothing and stays at the chamber’s eighty-five degrees, wet as intended. The moment a test asks the part to carry real current, the picture flips. The silicon turns watts into heat. The junction climbs above the air around it, by an amount that depends on the device’s power and the thermal resistance of its package. The chamber sets the ambient the standard names. The part sets its own junction on top, and the gap between the two can run to tens of degrees.
This is why the standard pins a discrete to a junction temperature. The number that matters is the hottest the silicon reaches, the chamber air plus the device’s self-heating. That ceiling governs whether the part survives the life tests that follow. It also shapes what a humidity test must guard against. If a part under a damp soak were allowed to draw enough current to warm its own surface, it would dry the film the test depends on, so the soak would lie about how the part will fare. The conditions are chosen to keep the part wet where wetness is the point and to let it heat itself only where the heat is the point. The bias plan and the chamber work together to keep those two regimes apart.
An integrated circuit carries an ambient grade, a single number that names the air it has to endure. A discrete is rated a different way, to a maximum junction temperature, the hottest its silicon may run before its ratings no longer hold. A common figure is a hundred and seventy-five degrees for modern automotive power silicon, with a hundred and fifty for older parts and higher still for wide-bandgap devices. The distinction is not bookkeeping. An ambient grade describes the world around the part. A junction limit describes the part itself under load, which is the honest measure for a device that makes its own heat. A discrete qualified to a high junction temperature has been shown to survive there powered. The chamber that proves it has to control the ambient precisely while the device adds its own rise on top, so the silicon reaches exactly the junction the test intends and no more.
Two life tests lean on that self-heating, intermittent operating life and power temperature cycling. In both, the device is switched on until its junction reaches a set temperature, then switched off to cool, over and over for thousands of cycles. The swing is the junction’s own, driven by the current the part carries. It works the package from the inside. The bond wires expand against the die they sit on and lift at their feet over the cycles. The die-attach solder under the silicon flexes as the die heats and the baseplate lags. It cracks from the edges inward, raising the thermal resistance and making the next cycle hotter still. These are the wear-out modes a power discrete dies of in service. The only way to provoke them in a lab is to let the part heat itself the way it will in the car. The swing can run tens of degrees on the junction, and the count reaches from thousands of cycles into the hundreds of thousands, since a switching load works the part that way for the life of the car. A feedback in the wear makes it bite. As the die-attach cracks inward its thermal resistance rises, so the same current heats the junction a degree higher on the next cycle, which strains the interface a little more and cracks it a little further. A device that starts comfortably inside its rating can walk itself out of one. The test exists to find the parts that will.
The chamber holds the surrounding conditions while this happens. It keeps the ambient steady so the junction swing is the device’s own and repeatable. It carries the heavy switched current in and out through fixtures that stay sound across tens of thousands of cycles. A fixture that warms up or a contact that drifts would change the swing and blur the result, so the box and its wiring have to be as durable as the part they judge.
Not every AEC-Q101 stress wants water. Two of the toughest are dry and electrical, run at temperature with no humidity. High temperature reverse bias holds the part at its rated high temperature, blocking a reverse voltage, for a thousand hours. It hunts the slow drift of the blocking junction and its termination under field and heat. High temperature gate bias does the same for a transistor’s gate, holding a bias across the gate dielectric at temperature to find instability there. For these the chamber is a precise oven, holding a flat high temperature for weeks while the device carries its bias, with the same demand for uniformity the damp tests make. A part buried in a loaded rack that runs a few degrees cool logs an easier test than its neighbours. The verdict only holds when every device saw the same heat.
The newest discretes a car carries are wide-bandgap, silicon carbide and gallium nitride. They push every stress in the qualification harder. A silicon carbide MOSFET blocks six hundred or twelve hundred volts where a silicon part of the same job blocked a fraction of that, so its high temperature reverse bias runs at voltages that turn the chamber’s insulation and feedthroughs into a real design problem. It runs hotter, rated to a junction of a hundred and seventy-five degrees and sometimes beyond, so the ovens have to reach and hold further than silicon ever asked. Its gate sits on a thin dielectric whose stability under bias and heat is a watched concern, which puts extra weight on high temperature gate bias. Because these parts are young, with far less field history than silicon gathered over decades, the qualification leans harder on the chamber to stand in for years no fleet has yet driven. A high-voltage humidity-bias test, run at hundreds of volts in the damp, has grown into part of how the toughest of these devices are proven. It asks the chamber for all that H3TRB does at a far higher potential.
Alongside the powered swings, the part faces the environmental kind, the chamber driving it between cold and hot with no current flowing. AEC-Q101 cycles a discrete a thousand times from minus fifty to a hundred and fifty degrees, after a preconditioning step, to work the mismatch between the silicon, the leadframe and the moulding compound. A large power die makes this harder than a small chip does, because the bigger the die the more its expansion differs from the metal under it, so the more strain the die-attach and the bonds take at each turn. The detailed fatigue physics of thermal cycling belong to the cycling method. The point for a discrete is the size of what is being cycled and the heavy thermal interfaces a power part carries, which is why the cycle count and the span are written hard into the standard. The cycling follows a preconditioning step, the moisture-and-reflow sequence a surface-mount part goes through, whose detail belongs to the preconditioning method. What the cycling then adds is the slow mechanical fatigue, felt first in the solder under a large die and at the feet of the bond wires, the same interfaces the powered cycling tires but driven here by the chamber alone.
A discrete qualification carries high voltage where a chip qualification rarely does. That voltage has to enter and leave a chamber full of moisture. Under H3TRB a part might block ninety or a hundred volts in eighty-five percent humidity. Under high temperature reverse bias a part might block several hundred. Every one of those volts reaches the device through a feedthrough in the chamber wall. A feedthrough that lets the voltage track across a damp surface, or arc to the grounded shell, fails the fixture and risks the run. So the chamber for power discretes is built around insulated, sealed high-voltage feedthroughs, with creepage distances sized for the damp and bias boards that keep the live nodes clear of condensation. Creepage is the distance a current would have to crawl across a surface to short two nodes. Humidity shrinks the safe distance, since a damp surface carries charge a dry one would not. A chamber meant for low-voltage chips can let this pass. A chamber meant for power discretes has to design for it everywhere a live node and a grounded wall sit close. None of this touches the part directly. It is still the difference between a chamber that can qualify a power discrete and one that cannot.
Moisture does not attack the silicon directly. It travels. Water vapour permeates the moulding compound over the days of a soak, following the resin and the lead frame, then gathers at the interfaces inside, the die surface and the passivation over the termination. There it meets the reverse field and goes to work. The journey takes days, since the water has to permeate millimetres of resin and creep along the lead frame before it ever reaches the die. That is one reason the soak runs a thousand hours, far longer than a quick check would. A part with a thin or a flawed passivation, a contaminated surface, a microscopic crack in the mould, gives the water a faster road and shows its weakness sooner as a climbing reverse leakage. The chamber cannot change where the water goes. By holding the humidity high and flat it guarantees the water gets its full chance, so a part that would have failed in five years of damp service fails within the six weeks of the soak.
A discrete is not qualified on one good part. AEC-Q101 draws devices from more than one production lot, so a flaw rooted in the process has somewhere to surface across batches. The standard sets a sample size and an acceptance criterion of zero failures for the stress tests. A single part that fails its post-soak read sends the lot back. The governance is the spine of the result. It is the reason the chamber cannot be the weak link, since a soak that wandered or a feedthrough that wept could fail a sound device and sink a qualification the silicon would have passed. The campaign runs long, the thousand-hour soaks alone filling six weeks each, so a discrete in qualification can occupy a chamber for months between its first read and its last. The qualification is not a one-time event either. A change to the wafer process, a new die size, a move to a different package or assembly site can each send the part back into the chamber to earn its rating again, since any of them alters how the device blocks voltage or sheds its heat.
The reads that decide a discrete are electrical, taken before and after each stress and compared. The clearest signal for a blocking device is reverse leakage, the small current the part passes when it should pass none. A device whose reverse leakage has climbed across H3TRB has told its story even if it still blocks, because the climb is the surface degradation at the termination caught partway. For a transistor the gate threshold and the on-resistance carry the same kind of news, a drift that a healthy part holds flat and a stressed one lets wander. The chamber’s contribution to that reading is the credibility of the stress behind it. A leakage that climbed under a clean, flat soak points at the part. A leakage read after a soak that sagged or condensed points nowhere, since a drift in the device can no longer be told apart from a fault in the box. A power discrete adds one more reason to trust the box. Its leakage and its thresholds move with temperature, so a reading taken while the part is warm differs from one taken cold. The standard fixes the conditions of the read for that reason. The chamber has to bring the part to the same state each time, or the comparison the verdict rests on is weighing two different things.
A thousand hours is about six weeks. A car is meant to run for fifteen years. The qualification bridges the gap with acceleration, the principle that harder stress ages a part faster along the same paths. Heat and humidity in a reverse field drive the surface degradation H3TRB hunts far quicker than a mild, unbiased life would. The models that tie the soak to field years rest on the chamber having delivered the condition without a break. An hour where the humidity dropped or the temperature sagged is an hour the model never got. A claim of fifteen years built on a soak full of gaps is a claim the test never earned.
What AEC-Q101 asks of a chamber is honesty toward a device that fights back. It has to hold eighty-five and eighty-five flat for a thousand hours while a part sits blocking a hundred volts, with no surface anywhere falling to the dew point. It has to hold a high temperature steady while a part blocks its rated voltage dry. It has to carry heavy switched current in and out while a device swings its own junction tens of thousands of times, and high blocking voltage in and out without a track or an arc. A chamber that does all of that lets a maker rate a discrete to a junction temperature and stand behind it through the life of a car. The rating is a promise about how hot and how stressed the part can run. The chamber is what holds the promise to the truth.
H3TRB is high humidity high temperature reverse bias, the signature damp test for a discrete. The part sits at 85 C and 85 percent humidity for 1000 hours while held in reverse at 80 percent of its rated blocking voltage, capped at 100 V. The bias is reverse because a discrete spends its working life blocking voltage. The reverse field crowds at the junction termination on the silicon surface, exactly where humidity drives leakage and corrosion. A forward-biased test would stress a different place.
An integrated circuit carries an ambient grade, a number that names the air temperature it must endure. A discrete is rated to a maximum junction temperature, the hottest its silicon may run, commonly 150 or 175 C and higher for wide-bandgap parts. A discrete makes real heat of its own, so the honest measure of how hard it can be driven is its junction limit, the hottest its silicon runs under load.
Temperature cycling swings the entire part with the chamber air, with no current flowing. Power cycling makes the device heat itself by switching it on and off, so its junction swings from the inside while the ambient stays put. That internal swing is what fatigues the bond wires and the die-attach in service. Only self-heating reproduces it, which is why AEC-Q101 uses intermittent operating life and power temperature cycling.
After preconditioning, the standard cycles a discrete 1000 times from minus 50 to 150 C. The wide span and the high count work the mismatch between the silicon, the leadframe and the moulding compound. A large power die strains its interfaces more than a small chip, so the die-attach and the bonds are the first to tire.
A discrete is tested while it blocks voltage, ninety or a hundred volts under H3TRB and several hundred under high temperature reverse bias. That voltage must enter a humid chamber without tracking across a damp surface or arcing to the shell. The chamber needs sealed, insulated high-voltage feedthroughs and creepage distances sized for the damp, so the voltage reaches the device. The fixture is never the thing that fails.
Envsin reliability and environmental test chambers for automotive discrete-semiconductor qualification.