By ATN Editorial Team
The traction inverter is the component that turns a battery into a moving vehicle. It sits between the high-voltage pack and the motor, converting direct current into the three-phase alternating current that spins the rotor, and modulating that current thousands of times a second to deliver exactly the torque the driver is asking for. In a modern electric vehicle it is one of the largest sources of non-motor power loss in the high-voltage powertrain, which is why it has become the most contested component on the bill of materials. This explainer covers what a traction inverter actually does, how the power module, gate driver and DC-link capacitor work together, why silicon carbide has displaced silicon IGBTs in 800-volt platforms, and where the technology is heading next.
What A Traction Inverter Actually Does
An electric vehicle stores energy as DC and uses it as AC. Virtually every modern EV traction motor is an AC machine — usually a permanent-magnet synchronous machine (PMSM), sometimes an externally excited synchronous machine (EESM) or an induction motor, with switched and synchronous reluctance machines a small minority. To rotate that motor, the inverter must synthesise a smoothly varying three-phase voltage waveform whose frequency, phase and amplitude command the motor’s torque and speed in real time.
The architecture is a three-phase, two-level voltage-source inverter. Six switches are arranged in three half-bridges, one per motor phase. Each half-bridge has an upper and lower switch; turning the upper one on connects the phase to the positive DC bus, turning the lower one on connects it to the negative bus. A pulse-width modulation (PWM) controller — most often space-vector PWM in modern drives — switches each pair on and off at a carrier frequency, typically 8 to 15 kHz on a silicon IGBT inverter and 20 to 40 kHz on a SiC MOSFET inverter. The motor inductance filters the resulting phase voltage so the current the motor sees approximates a sine wave at the desired electrical frequency.
The control loop sits on top of this. A field-oriented control algorithm reads rotor position from a resolver or magnetic sensor, decomposes the measured phase currents into the d-axis (flux-producing) and q-axis (torque-producing) components, and adjusts the voltage commands so the motor produces exactly the torque the vehicle controller is asking for. Everything else in the inverter — the power module, the gate driver, the DC-link capacitor — exists to execute those commands without losing too much energy or overheating.
Inside The Power Module, DC-Link And Gate Driver
Open a traction inverter and three things dominate the bill of materials. The power module holds the six semiconductor switches and their anti-parallel diodes on a ceramic substrate, bonded to a copper baseplate or pin-fin cold plate. The DC-link capacitor sits across the battery terminals to smooth the ripple current that the switching action would otherwise pump back into the pack. The gate driver translates 3.3-volt logic signals from the controller into the 15-to-20-volt gate pulses that turn the switches on and off, with galvanic isolation between the low-voltage control side and the high-voltage power side.
On older IGBT inverters the DC-link capacitor was typically the largest passive component by volume and one of the most expensive; on newer SiC designs the power module, cooling structure and integrated busbars have closed the gap, but the capacitor remains a major cost and packaging constraint. It is sized to handle the worst-case ripple current at peak motor load — typically in the order of 100 amps RMS or more on a 150 kW unit — and must also store enough energy to keep the bus voltage within bounds during transient events. Film capacitors dominate because they tolerate the temperature and ripple current better than electrolytics, but they are bulky. Reducing DC-link size is one of the main reasons inverter designers care so much about switching frequency: faster switching shortens the time the current has to slosh around, lowering the ripple the capacitor has to absorb. Recent gate-driver designs from NXP and others have started integrating the active-discharge function — which safely bleeds the capacitor down on key-off or crash — directly into the gate-driver IC, removing external components and lowering the bill of materials.
The gate driver is the unsung hero. It has to switch tens of amps of gate current in tens of nanoseconds, ride out the dv/dt that the power module throws back across the isolation barrier, monitor the switch for desaturation faults, and shut everything down in microseconds if something goes wrong. Modern automotive-grade gate drivers from suppliers including Infineon, NXP and Texas Instruments increasingly offer dynamic gate-strength control on premium parts: they vary the drive current during the switching event to slow the turn-on edge (cutting EMI) without sacrificing turn-off speed (preserving efficiency).
IGBT, SiC MOSFET And The Hybrid Switches Between Them
The biggest engineering decision in any traction inverter is the choice of switch. For two decades that meant a silicon insulated-gate bipolar transistor (IGBT) with a silicon free-wheeling diode. IGBTs are cheap, robust, and have low conduction loss at high current, which is why they dominate 400-volt mainstream electric vehicles. Their weakness is switching loss: the device’s tail current at turn-off and reverse-recovery charge in the diode cap the practical switching frequency at around 15 kHz, and a meaningful fraction of the energy that flows through the inverter ends up as heat in the module.
Silicon carbide (SiC) MOSFETs change the trade-off. SiC’s wider bandgap (3.26 eV versus 1.12 eV for silicon) gives a critical electric field around ten times higher, which means a 1,200-volt SiC device can be thinner, lower in on-resistance and faster to switch than an equivalent silicon IGBT. In a traction-inverter context, peak efficiency rises from a typical 96 to 97 per cent on silicon to 98 to 99 per cent on SiC. The bigger gain is not at peak load but at the partial loads that dominate real driving — city traffic, motorway cruise, regenerative braking — where SiC’s lower switching losses and lack of an IGBT-style tail current can cut inverter losses by 60 per cent or more. BorgWarner has quoted 40-to-70 per cent inverter-loss reductions versus a silicon baseline on its 800-volt Viper SiC inverter for a German OEM, with the headline range depending on which point of the drive cycle is measured (the reduction is largest at part-load cruise, smallest at peak-current acceleration). Reported system-level WLTP range gains from SiC on an 800-volt vehicle typically land in the 2-to-5 per cent band, with Hyundai having publicly quoted a roughly 5 per cent figure for the Ioniq 5 SiC front-inverter upgrade. The exact gain depends on duty cycle, vehicle mass, baseline inverter efficiency and the rest of the powertrain.
SiC’s commercial weakness is cost. Substrate, epitaxy and yield economics still put a SiC die at several times the price of a silicon IGBT of the same current rating. That has pushed two interesting compromises into production. The first is a split architecture, where the rear-axle inverter on a dual-motor electric vehicle uses SiC for range gain and the front-axle inverter uses IGBT for cost — Hyundai’s E-GMP platform began with broadly this approach — SiC rear inverter on launch, IGBT front — with a later announced SiC upgrade for the front inverter on all-wheel-drive variants. The second is the hybrid switch: Tesla has filed a hybrid traction-inverter patent (WO 2026/010828-A1) pairing SiC MOSFETs and silicon IGBTs in the same module with a coordinated switching protocol — the IGBT taking peak-current bursts where its lower saturation voltage wins, and the SiC carrying the cruise load at high efficiency. The same Si/SiC fusion concept has been demonstrated in prototype hardware, notably Infineon’s HybridPACK Drive demonstrator combining CoolSiC MOSFETs and EDT3 IGBTs under a common gate driver.
Why 800-Volt Architectures Need SiC
The shift to 800-volt battery architectures is the single biggest reason SiC has become the preferred choice in high-performance traction inverters. At 400 volts a silicon IGBT is sized comfortably for the bus; at 800 volts the engineer has a choice between a 1,200-volt SiC MOSFET or two 600-volt silicon IGBTs in series, and on losses, parts count and complexity the SiC route is usually the cleaner answer. Doubling the bus voltage halves the current for a given power, which roughly quarters the conduction losses in the cabling and the busbars and lets the engineer downsize the contactors, fuses and DC-link capacitor. The trade-off is that motor and module insulation must now handle higher dv/dt, and the inverter must manage faster switching edges without exciting motor-bearing currents.
Programmes that have gone 800-volt with SiC inverters include the Porsche Taycan, Audi Q6 e-tron, Hyundai Ioniq 5 and Ioniq 6, Kia EV6, Genesis GV60, Lucid Air and the latest Polestar architectures. STMicroelectronics has been Tesla’s main SiC supplier for the Model 3 traction inverter. Wolfspeed has supplied power devices to Lucid and Volkswagen Group, and in November 2025 launched its 1,200-volt YM six-pack power-module family specifically aimed at traction-inverter integrators. Onsemi’s EliteSiC module powers Hyundai Motor Group’s high-performance E-GMP variants and NIO’s 900-volt platform on the ES9 flagship.
Switching Frequency, Dead Time And Motor Losses
Raising the switching frequency cuts the harmonic content the inverter dumps into the motor. Lower harmonics mean lower copper losses, lower iron losses, less rotor heating and less acoustic noise. The catch is that switching losses in the inverter scale approximately linearly with frequency under fixed conditions, so silicon IGBTs hit a wall at 10-to-15 kHz where the inverter heats up faster than the motor cools down. SiC’s lower switching energy unlocks 20-to-40 kHz operation in production traction applications, and research designs have demonstrated air-cooled SiC inverters running at higher frequencies still. The bargain is essentially that SiC shifts the loss budget. Some loss leaves the inverter (where it is hard to cool); some moves into the motor as a cleaner-waveform reduction in iron and copper losses; some ends up as EMI that the filter has to absorb. The net is a meaningful system-level efficiency gain, not loss elimination.
Dead time is the small interval between the upper switch turning off and the lower switch turning on, inserted to prevent both being on at once and shorting the bus. It is one of the few inverter parameters that creates motor-side distortion entirely from the inverter’s design choices. Too much dead time produces a 5-to-7th harmonic on the output current and a measurable rise in motor THD. Representative dead-time values fall around 1.5-to-2 microseconds on IGBT and 0.3-to-0.5 microseconds on SiC, though actual settings vary with current, package inductance and control strategy and some SiC inverters still run conservatively above 1 microsecond. A 2026 IET Power Electronics paper by Reiter (Infineon) showed that optimising dead time across IGBT, SiC MOSFET and Si/SiC fusion modules can reduce motor THD by 2-to-3 percentage points and, depending on the operating point, recover a fraction of a per cent on system efficiency.
Thermal Management And The 200°C Junction Ceiling
SiC’s thermal conductivity is roughly three times that of silicon — 370 to 450 W/m·K against silicon’s 150 W/m·K — and some SiC devices are rated for junction temperatures approaching 200°C, against silicon IGBT’s typical 175°C ceiling, although continuous operation near the SiC ceiling is uncommon in production designs and carries reliability penalties. In production inverters almost everyone still chooses liquid cooling, because the same coolant loop is doing useful work elsewhere in the powertrain and because pin-fin direct-cooled baseplates extract heat far faster than air ever can. The change SiC enables is not the elimination of liquid cooling but the shrinkage of it: smaller cold plates, lower pump flow, less coolant volume, and module designs that put pin-fins directly under the die without an intermediate baseplate. Sintered silver die-attach and silicon-nitride substrates have largely replaced soldered die-attach and aluminium-oxide substrates in SiC modules, because they hold up better under the larger thermal cycling that high-power-density operation imposes.
Power density has climbed accordingly. Bosch quotes a system-level power density of around 60 kW per litre on its fourth-generation 800-volt SiC drive, and Oak Ridge National Laboratory has demonstrated a research-grade 100 kW SiC traction inverter packaged at over 100 kW per litre.
EMI, dv/dt And Motor Bearing Currents
The same fast switching that makes SiC efficient creates a secondary problem. Voltage edges with a dv/dt typically in the 5-to-10 kV per microsecond range couple capacitively through the motor’s parasitic capacitances to the rotor and out through the bearings, producing high-frequency bearing currents that pit the raceways and shorten bearing life. They also drive conducted EMI back up the DC bus and radiate it from the motor cables, which has to be filtered to meet automotive electromagnetic compatibility limits.
The mitigations are well understood but not free. Common-mode chokes on the inverter output absorb the high-frequency content but add mass and cost. Shielded motor cables with low-impedance grounding cut the radiated path. Shaft-grounding brushes or insulated bearings break the conductive path through the motor itself. On the controller side, dynamic gate-strength schemes deliberately slow the turn-on edge while keeping turn-off fast, which trades a fraction of efficiency for a useful drop in EMI. The trade-off honesty matters: SiC is faster than silicon, but the faster switching is only fully useful if the inverter designer has spent the budget to manage the EMI it creates.
Cost, Supply Chains And The Si/SiC Compromise
Cost is still the topic engineers and procurement teams argue about. On a per-watt basis a 1,200-volt SiC MOSFET costs roughly three to four times an equivalent silicon IGBT, and the gap closes more slowly than the industry expected when 200-millimetre SiC substrates were promised at parity by mid-decade. The supply base is consolidating: STMicroelectronics, Wolfspeed, Infineon, Onsemi, Rohm and Bosch dominate SiC die supply, with multi-billion-dollar long-term agreements binding most OEMs and tier-one inverter integrators to one or two of them. Vitesco has signed two such deals — a $1.9 billion ten-year agreement with Onsemi and a separate $1 billion deal with Rohm — to lock in capacity for its 800-volt SiC inverter platform.
The Si/SiC fusion approach exists to bridge that gap. By putting silicon IGBT and SiC MOSFET die on the same substrate, sharing a gate driver and trading the switch role with engine load, an inverter can capture most of the cruise-load efficiency of a pure-SiC unit at a substantial bill-of-materials saving. The penalty is a slightly worse peak-load efficiency and additional control complexity, neither of which is fatal for cost-driven mainstream segments. Expect the architecture to spread into volume B and C segment electric vehicles over the next two model cycles.
What’s Next: GaN, Integration And Diamond
Gallium nitride (GaN) is the obvious next candidate. GaN HEMTs switch faster than SiC and have already taken the on-board-charger and DC-DC market, where the HiPower 5.0 demonstrator showed a 22 kW GaN on-board charger in 4 litres. In traction inverters GaN faces several constraints: its current automotive-qualified voltage rating tops out around 650 to 900 volts (which suits 400-volt vehicles better than 800-volt), packaging parasitics and thermal density become harder to manage at traction power levels, and the automotive supply chain is years behind SiC. GaN is likely to enter traction in cost-sensitive 400-volt platforms and in 48-volt mild-hybrid applications well before it challenges SiC at the high end.
The two other directions worth watching are integration and ultra-wide-bandgap materials. Integration means folding the inverter, motor and reduction gear into a single housing — Hyundai Mobis’ PE system is a clear example, and Bosch, ZF, Magna, BorgWarner and Vitesco all have integrated drive units on offer. The packaging gains are real, but the thermal coupling between motor and inverter has to be managed carefully, and serviceability suffers. Ultra-wide-bandgap candidates like diamond and gallium oxide offer theoretical figures of merit several times better than SiC, but production-grade automotive devices are still a decade away at minimum.
Further Reading From The ATN Powertrain Library
- Wolfspeed 1,200V SiC Six-Pack: EV Traction Inverter Modules
- BorgWarner Integrates STMicroelectronics SiC in Viper Power Module for Volvo
- Onsemi EliteSiC Powers the NIO 900V EV Platform
- Hyundai Mobis 160 kW Integrated PE System
- HiPower 5.0: 22 kW GaN On-Board Charger In 4 Litres
- What Is A High Power Density Motor? An Engineer’s Guide
- What Is Variable Turbine Geometry (VTG)? An Engineer’s Guide
For more power electronics technology news, click the link.
