Power Cycling And Thermal Cycling For SiC Power Modules

Heat from within

Its heat is made, not given.

What SiC power modules face

A power module is the block that switches the current in a converter or a drive, a set of power chips bonded onto a substrate inside one case, and silicon carbide is the material that lets those chips run faster and hotter than silicon ever did. The cold-hot testing that qualifies such a module asks whether the package around the chips can take the swings that performance brings.

The question presses harder for SiC than for the silicon it replaces. A module that runs hotter and switches faster strains the joints that hold it together more, so the test that proves those joints is the gate the module must pass before it is trusted in a drive or an inverter.

Two ways to make the heat

The thing that sets a power module apart from an ordinary part under test is where its heat comes from, and there are two answers, which is the whole of this subject. The first is the heat a chamber gives: a passive temperature cycle warms and chills the entire module from the outside, the air rising and falling around it until the whole body has followed, a trial the general cycling methods set out and a chamber knows how to run. The second is the heat the module makes for itself. A power module exists to carry current, and a switching die that carries current loses a little of it as heat at the junction, so when a test passes real load current through the device the chip warms itself from within, its junction climbing far above the case beneath it while the coolant holds the baseplate cool, and the gap between the hot junction and the cool case opens a steep gradient no external chamber can make. That difference is not a detail; it is the point. In service a power module is heated by its own work, switching on and off, warming as the load rises and cooling as it falls, so the swing it meets in its working life is a junction swing driven from inside, not a body swing driven from outside, and the test that reproduces it must drive current rather than blow air. The rig that does so is not a chamber at all but a bench: a source to push the current, a gate drive to switch the device, a cold plate to hold the case, and a way to read the junction's temperature as it rises and falls. The two trials are not rivals but partners, the chamber swinging the whole body to find what the slow outside change loosens, the power bench swinging the junction against a held case to find what the steep inside gradient tears, and a power module bound for hard service is asked to survive both because its life holds both kinds of heat at once.

Hotter than silicon

Silicon carbide runs hotter than silicon, and the package must take the heat.

Why SiC pushes the package

Silicon carbide is a wide-bandgap material, which is a plain way of saying it keeps working where silicon gives up. It holds off higher voltages in a thinner layer, switches faster with less loss, and stays a semiconductor at junction temperatures well past the limits silicon was held to. It is the material the power world reached for when silicon had run out of room.

Those gains come with a cost paid by the package. A SiC die is smaller than a silicon one of the same rating, so the same power leaves through less area, the heat flux is higher, and the junction swings further and faster than a silicon module's would.

The pace adds to the burden. A SiC module switches faster and is pushed nearer its high temperature ceiling, so its junction not only swings wider but sits hotter between swings, and the joints age under a harder duty than a silicon part ever set them.

So the weak point moves. The chip is no longer the part likeliest to fail; the joints that carry its heat and current away are, and the reliability of a SiC module is set by how well its packaging survives the swings the chip can now make.

The junction's own swing

In power cycling the junction's own swing is the stressor.

Where a power module fails

The first to go are often the bond wires. Thick aluminum wires or ribbons carry the current from the chip to the terminals, anchored to the die at points that heat and cool with every cycle, and the repeated strain lifts a wire from its pad or cracks it at the heel where it bends, a failure the general cycling story names and a power module meets writ large.

Beneath the chip the die-attach tires. The layer that bonds the die to the substrate sees the steepest gradient of all, hot above and cool below, and the solder or sinter there fatigues and delaminates until the heat can no longer leave the chip, which drives the junction hotter still.

The ceramic substrate that isolates the chips has its own way of failing. Its bonded copper can lift at the edges as metal and ceramic pull against each other with heat, and the insulating layer that keeps the high voltage off the baseplate can craze, so the part that gives the module its isolation is itself on trial.

Lower down, the wide solder layers fail in their own way. The joint under the substrate or the baseplate spans a large area, so a small mismatch in expansion becomes a large movement at its corners, and a passive whole-body cycle works that layer hardest of all.

The answer to the hottest joints is a tougher bond. SiC modules turn to silver sinter under the die and to copper wires or ribbons in place of aluminum, materials that hold where ordinary solder and soft wire would tire, because the chip's heat now asks more of them than a silicon part ever did.

Reading the junction

A power-cycling test must know the junction temperature it is swinging, and it cannot reach a probe to the die. So it reads the temperature through the device itself, using an electrical property that shifts with heat, the forward voltage of a junction at a small sense current, as a thermometer the chip keeps for itself.

From that reading the bench sets the swing. It drives current until the junction has climbed the wanted amount above the case, holds, then lets it cool, watching the electrical thermometer the whole way so the swing it applies is the swing it meant to.

Big swing, few cycles

The life of a power module under cycling turns on the size of the swing. A small junction swing can be borne a great many times; a large one is borne far fewer, and the cycles a module survives fall steeply as the swing it is given grows wider.

A module's power-cycling capability is read as that relationship, the count it reaches against the swing it is held to, and because SiC invites larger and faster swings, its packaging must be built to reach the required count under a harder swing than a silicon module faced.

Short cycles, long cycles

Even within power cycling there are two tempers, set by how long each pulse of current lasts. A short cycle, a few seconds of heating and cooling, drives the swing into the chip and the joints nearest it before the heat has had time to soak down to the base.

A long cycle, minutes to a side, lets the heat reach all the way through the stack, so the whole solder column down to the baseplate warms and cools together and is worked as one body.

The two find different failures. The short cycle hunts the bond wires and the die-attach that feel the fast inner swing; the long cycle hunts the larger solders that only a soaked, whole-stack swing can strain, so a thorough program runs both lengths rather than one.

Self-heating is the point

The heat born at the die, not piped in from the air, is the whole difference.

Both tests, different truths

Passive cycling and power cycling are not the same trial told twice. The chamber's whole-body swing works the large outer solders hardest and finds what a slow uniform change loosens; the power bench's junction swing works the die-near joints hardest and finds what a steep inner gradient tears.

So a SiC module bound for hard service meets both. The pair covers the joints from the baseplate up to the bond on the chip, and a module that passed only one would have shown only half of what its life will ask of it.

What the chamber does here

For a power module the chamber does only half the job. It can swing the whole body between a cold and a hot, but it cannot make the junction climb above the case, since it heats from outside, so the steep inner gradient that power cycling needs lies beyond it.

That inner trial belongs to the power bench, a different rig with its current source, its gate drive, its cold plate, and its electrical reading of the junction, run alongside the chamber rather than inside it.

So the chamber owns the body's endurance and the bench owns the junction's, the two dividing the module's joints between them, and a lab qualifying SiC keeps both and reports each.

Sinter for the hot die

A sintered silver layer holds the hot die where solder would tire.

Pass means the package keeps up with the chip

Passing both trials is a statement that the packaging can keep pace with the chip it holds. The SiC die can switch hotter and harder than silicon, and the joints have shown they will not fatigue loose before the chip itself would fail.

That is the assurance a drive or an inverter maker leans on, that the module will carry its current through years of heating and cooling without a wire lifting or a solder layer opening under the swing it will see across the life of the drive.

The chip leads, the joints must follow

Silicon carbide moved the frontier of what a power chip can do, and in doing so it moved the burden onto the package, so the cold-hot and power-cycling tests are where a SiC module earns its trust, at the joints rather than the chip.

A module proven through both swings, the body's and the junction's, is one whose packaging has caught up to the silicon carbide inside it.