Every watt an IGBT burns has to go somewhere: it turns into heat, the heat pushes junction temperature up, and the rising temperature quietly degrades the very parameters you selected the part for. Loss is not just an efficiency line item — it is the starting point of most reliability problems. Here is how the losses break down, and which levers actually move them.
Three Losses, Three Different Beasts
Conduction loss is VCE(sat) × I while the switch is on. It scales with current and with the device's saturation voltage — silicon physics and die size set the floor. Switching loss comes from the transition edges: during turn-on and turn-off, voltage and current overlap for hundreds of nanoseconds, and their product is pure loss. Higher switching frequency multiplies it directly. Driving loss is the energy the gate circuit burns charging and discharging the gate capacitance — small per pulse, but proportional to frequency and gate charge, and non-trivial in high-frequency designs.
What Loss Does to the Device
Loss becomes heat; heat raises junction temperature; temperature drags VCE(sat) and switching times in the wrong direction, which increases loss again — the loop only needs a weak thermal design to become runaway (we covered that failure mode separately). Even short of failure: hotter parts must be derated, so the same module delivers less current in a hot enclosure. In efficiency-critical applications — EV drivetrains, PV inverters — the loss budget is also the revenue budget. If you want the thermal side in numbers, our thermal resistance walkthrough and the piece on how temperature dictates module power cover the details.
The Levers That Actually Reduce Loss
Pick the right device for the frequency. Low-frequency, high-current jobs should optimize VCE(sat); high-frequency jobs should optimize switching energy. Using a low-loss conduction part at 40 kHz — or a ultra-fast part at 2 kHz where it costs conduction performance — both waste money. When the frequency is high enough that IGBT tail currents dominate, that is where SiC MOSFETs cut high-frequency losses.
Tune the gate drive. Gate resistance is the cheapest loss knob you own: larger RG slows edges and reduces EMI but widens the voltage-current overlap; smaller RG does the opposite. The optimum sits where switching loss and EMI trade evenly. Measuring it properly takes a double-pulse setup — see how we characterize dynamic performance with DPT.
Design the thermal path honestly. A heatsink sized for typical load instead of worst case forces derating later. interface materials, mounting flatness and airflow all add resistance between junction and air; every bit of it translates into junction temperature, and temperature translates back into loss.
The Same Rules Look Different per Application
In EV motor drives, load swings constantly and switching loss usually leads; in PV inverters the DC bus is steady and the frequency fixed, so conduction and switching losses sit in a predictable balance that light conditions modulate; in industrial drives, big machines run at low frequency where conduction loss dominates, while small high-speed drives balance both. Same equations, different weights — which is why we always start selection from the application's duty cycle, not the datasheet's headline numbers.
If you would rather buy the integration than tune the knobs: a well-designed IPM optimizes the switch, the driver and the protection together, which is why many teams land on IPM over discrete builds for motor drives. And for platforms migrating to wide-bandgap at the same time, our hybrid SiC IPM keeps the footprint while cutting the loss.
