EV Power Semiconductor Application Selection Matrix

Mapping EV Power Semiconductors: Which Device for Which Application

The cost of SiC power modules installed in a single inverter is two to three times higher than that of conventional Si-based modules. Yet, EV manufacturers are accelerating their shift to SiC because it can be recouped through improved overall system efficiency and size. However, it's not simply a matter of "using SiC." EV power conversion systems are divided into multiple applications, each with different required characteristics. Some areas demand thorough reduction of switching losses, while others prioritize short-circuit withstand capability. Without a "selection map" that dictates which device to assign to which application, achieving an optimal configuration—both technically and in terms of procurement cost—remains out of reach.

Deconstructing the EV Power Conversion System: Five "Battlegrounds"

First, let's grasp the overall picture. An EV's power conversion system can be broadly divided into five applications: the main inverter (motor drive), the DC-DC converter (auxiliary power supply), the on-board charger (OBC), the fast charger (infrastructure side), and the DC-AC inverter (for V2H or external power supply). These differ significantly in voltage levels, current, switching frequency, and operating cycles, meaning selection criteria that apply to one might not align with others, even when broadly categorized as "EV applications."

The main inverter handles high currents at 400V to 800V bus voltages and is constantly exposed to the risk of instantaneous short circuits. OBCs and DC-DC converters operate at relatively lower power but require miniaturization through high-frequency switching. Fast charging infrastructure is closer to the grid and demands power quality and long lifespan. Discussing these five applications with the same perspective is the biggest cause of misjudged selections.

Keeping these four axes in mind allows us to pose specific questions like, "What trade-offs do we make for which characteristics?" Next, we delve into the core trade-off between short-circuit withstand capability and on-resistance that underlies these questions.

"What Defines 3μs?"—The True Nature of Short-Circuit Withstand Time as a Selection Criterion

When selecting SiC MOSFETs, a parameter often found on datasheets but easily overlooked is short-circuit withstand time (SCWT). This refers to the time from the occurrence of a load short circuit until the device fails. In other words, it is the "grace period" before the protection circuit activates.

For Microchip's SiC MOSFETs (700V/1200V rated), the SCWT under specific conditions is typically listed as 3μs on the datasheet. Whether this 3μs figure resonates with you can actually be a deciding factor in the design. Gate driver's DESAT (desaturation) detection—a method that monitors VDS in the on-state to detect overcurrent—requires detection delay and blanking time to operate. If the total exceeds 3μs, the device may be destroyed before it can be protected.

Further complicating matters is that SCWT is not a fixed value. The capability tends to decrease with higher drain applied voltage, higher gate voltage, and lower junction temperature. Conversely, under high-temperature operating conditions, RDSon increases, suppressing saturation current, which works to improve the withstand capability. For high-voltage, high-current applications like main inverters, evaluating these worst-case conditions comprehensively is fundamental to the design.

SiC, Si, and GaN—What Are the Differences and Where Do They Compete?

The categorization of "SiC for high efficiency" and "GaN for high frequency" is correct, but it's insufficient for application-specific selection. Let's delve into the specific characteristic differences.

SiC's greatest strength lies in its high dielectric breakdown strength due to its wide bandgap (3.26eV). This allows for a thinner drift layer compared to Si devices of the same voltage rating, significantly reducing on-resistance. In comparisons of 1200V rated devices, the on-resistance of SiC reaches a fraction of Si's theoretical limit. The reason SiC is chosen for high-voltage, high-current applications like main inverters is its ability to reduce both switching and conduction losses.

On the other hand, SiC has structural challenges. Due to its small die size and high current density, temperature rise during a short circuit is more rapid than in Si. This is why protection circuit response must be faster, and engineers accustomed to Si designs might enter a danger zone if they apply the same design margins to SiC.

GaN is even more suited for high frequencies, but the maturity of its vertical structure currently lags behind SiC, with limited product availability for high-voltage ratings exceeding 1200V. While it shows advantages in applications with switching frequencies of several hundred kilohertz or higher, such as OBC PFC stages and auxiliary power supplies, its application in main inverters remains limited. Si technologies, exemplified by Toshiba's triple-gate IGBT, are also continuously evolving, and there's no reason to completely discard Si in applications where cost advantage is crucial.

This graph illustrates the typical breakdown voltage ranges for each device. The 1200-1700V band is almost exclusively occupied by SiC as a choice, while the 650V band presents a competitive landscape among Si, SiC, and GaN. By first determining the application and voltage range, the number of competing devices is naturally narrowed down.

Reading the Ron x SCWT Trade-off: Differences in Manufacturer Approaches

Technically, the most challenging aspect of selecting SiC MOSFETs is the trade-off between low on-resistance (Ron) and high short-circuit withstand capability (SCWT). Physically, reducing Ron increases channel current density, leading to greater energy concentration during a short circuit, which in turn tends to shorten SCWT. This relationship is rooted in material physics, and structural ingenuity is required to circumvent it.

Mitsubishi Electric addresses this challenge by introducing a p-type protective layer in its trench-type SiC MOSFETs. Mitsubishi's approach structurally protects areas where current tends to concentrate during a short circuit. Rohm employs a unique device structure in its 4th-generation SiC MOSFETs to achieve both low RonA and high short-circuit withstand capability. Both companies share the goal of breaking the premise that "Ron and SCWT are inseparable trade-offs," but their detailed approaches differ.

In practical selection, instead of directly using the SCWT values from catalogs, checking the values under actual operating conditions—bus voltage, gate voltage, and junction temperature—provides a crucial basis for decision-making. Since condition definitions vary by manufacturer, a simple numerical comparison can lead to a false sense of security.

Protection Circuits: Consider Them as a "Set"—Design Sensitivity of DESAT and Blanking Time

Selecting the device alone is not enough. To unlock the full potential of SiC MOSFETs, the protection circuit design of the gate driver must be considered as a set. The DESAT (desaturation) function, in particular, plays a central role in short-circuit protection in main inverters.

The operating principle of DESAT is simple: it detects when VDS in the on-state becomes higher than during normal operation, judges it as a short circuit (overcurrent), and turns off the gate. However, setting the design parameters determines the success of the protection. Setting the DESAT trigger threshold (VDESAT) too low increases false detections, while setting it too high slows down the response. DESAT current (IDESAT) and short-circuit blanking time—the time to ignore VDS spikes immediately after switching turn-on—also have trade-offs.

If the blanking time is too long, it takes time from the occurrence of a short circuit to detection, and for devices with a 3μs SCWT, protection may not be achieved in time. Conversely, if it's too short, switching noise may be misjudged as a short circuit, leading to malfunction. These settings need to be optimized according to the SCWT of the SiC MOSFET being used, making the order of "decide on the device, then select the gate driver" crucial.

Evaluating the device and gate driver together—this perspective also holds significance in procurement. Beyond comparing individual device specifications, whether they have been evaluated as a recommended set with the gate driver directly relates to operational reliability in actual equipment.

Voltage Class x Application: Drawing a "Selection Matrix"

Integrating the discussions so far, the selection of EV power semiconductors can be organized as a matrix based on "voltage class" and "application characteristic requirements."

This matrix is not static. As 800V buses become more prevalent, device selection will become more refined, and if GaN's high-voltage capabilities reach practical application, the competitive landscape will change. The selection map needs to be updated in conjunction with the technology roadmap.

What to Check Next: Questions Beyond This Article

What has been clarified by reading this far is the general framework of "what to use for which application." However, in actual design, procurement, and business decisions, deeper questions await.

Regarding short-circuit withstand capability, as indicated by Microchip's public figure of 3μs, confirming "under which conditions this value applies" is the priority. Manufacturers like Rohm and Mitsubishi Electric, who have achieved higher withstand capability through structural improvements, may have different actual performance compared to those who have not, even with the same numerical value. How to read datasheets and confirm evaluation conditions are the starting points for technical evaluation.

From a procurement perspective, the concentration risk in the SiC supply chain and the timing of the transition to 8-inch wafers will affect supply stability over multiple years. While there is a view that selecting manufacturers with a complete SiC portfolio from 650V to 1700V, such as onsemi, can diversify risk, it also creates a different risk of single-vendor dependency. Analyzing the structure of supply risk from the perspective of wafer procurement will complement decision-making.

For GaN, applications that leverage its low Ron and high-frequency characteristics are increasingly emerging, primarily for OBCs. Movements such as Renesas' acquisition of Transphorm can be interpreted as the eve of GaN's full integration into the EV supply chain.