Foundations for Power Semiconductor Supplier Evaluation

How to Select Suppliers: A Structured Approach to Evaluation

When a design team preparing to launch mass production of EV inverters needed to select SiC MOSFET suppliers, they encountered an initial hurdle: "We don't know which datasheets to compare." Even when lining up specification tables side-by-side, different measurement conditions render the numbers meaningless. Reliability data from each company uses different formats, making direct comparison impossible. The common outcome in such real-world scenarios is a decision based on "choosing from manufacturers we already know."

The difficulty in evaluating power semiconductor suppliers stems from the interplay of three axes: technical assessment, procurement assessment, and business assessment. Leaning too heavily on any one axis can lead to problems surfacing later from the others. This article organizes these three axes and specifies what to look for in each. While SiC MOSFETs are the primary focus, the evaluation framework is also applicable to GaN and next-generation Si devices.

The "Entry Point" to Technical Assessment Lies in Measurement Conditions, Not Specs

When opening a datasheet, the first thing to check is not the spec value itself, but the conditions under which that value was measured. An "on-resistance of 20mΩ" can have vastly different implications for practical application depending on whether the measurement was made at a gate voltage of 15V or 18V. SiC MOSFETs are more sensitive to gate voltage than Si devices, and overlooking these conditions directly translates to a loss of design margin.

Temperature characteristics are similar. The on-resistance (RDSon) of SiC MOSFETs exhibits a positive temperature coefficient, increasing with temperature. While this is advantageous for automatic current balancing in parallel connections, understanding how well each manufacturer's datasheet accounts for increased switching losses at high temperatures requires careful reading.

Comparing spec values is meaningful only after measurement conditions have been standardized. Skipping this step and concluding that "Manufacturer A has lower on-resistance" will lead to time-consuming re-evaluations in later stages.

Short-Circuit Withstand Time: A "Hidden Differentiator"

There's a difference between SiC MOSFET suppliers that is often overlooked: Short-Circuit Withstand Time (SCWT). This is an indicator of how long a device can withstand a load short circuit before failure, and can be understood as the "grace period" before a protection circuit activates.

The shorter this grace period, the faster the response speed required from the protection circuit. The issue is that SiC device dies are smaller and have higher current densities compared to Si, leading to faster temperature rise during short circuits. Engineers with experience designing Si gate drivers may risk device destruction if they set protection timings with the "same mindset."

For instance, Microchip's datasheet states a typical short-circuit withstand time of 3μs under specific conditions for their 700V/1200V SiC MOSFETs. This is significantly shorter than the typical 10μs range for Si IGBTs, implying a fundamental change in the design requirements for protection circuits.

Short-circuit withstand time depends on three conditions: drain-source voltage, gate voltage, and junction temperature. As these conditions are relaxed, the withstand time tends to increase. In other words, the system's operating condition design and short-circuit withstand time cannot be considered in isolation. For supplier evaluation, confirming the "effective withstand time" in conjunction with your company's operating conditions forms the basis for decision-making.

A widely used protection method is the DESAT (desaturation) function. This mechanism monitors the drain-source voltage (VDS) in the on-state and turns off the gate signal when overcurrent is detected. The three key parameters for design are the DESAT threshold voltage (VDESAT), DESAT current (IDESAT), and blanking time. Setting these parameters often relies on application notes provided by the supplier, making the quality of technical support directly relevant to the difficulty of system design.

The On-Resistance vs. Short-Circuit Withstand Time Trade-off: How Manufacturers Address It

Another crucial aspect of technical assessment is the trade-off between on-resistance (Ron) and short-circuit withstand time. Increasing channel density to lower on-resistance tends to reduce withstand time due to increased current flow during a short circuit. The differences in device structures among manufacturers become apparent in how they resolve this trade-off.

Mitsubishi Electric has reported significantly improved short-circuit withstand time by introducing a p-type protective layer in their trench-type SiC-MOSFETs. Rohm aims to achieve both low RonA and high short-circuit withstand time with their proprietary device structure in their 4th generation SiC MOSFETs. While the "correct" approach depends on system requirements, it's evident that choosing solely based on "low on-resistance" has pitfalls.

This graph illustrates the significant "gap" in the design requirements for protection circuits between SiC and Si. Incorporating the cost of redesigning gate drivers and protection circuits into cost estimations when transitioning from Si to SiC can impact the overall project decision.

Procurement Risk Evaluation: Separating Supply Capacity from Technical Dependency

In parallel with technical assessment, it's important to organize the procurement evaluation criteria. Power semiconductor procurement risks can be broadly categorized into two types: "supply risk" and "technical dependency risk." Confusing these will lead to ineffective countermeasures.

Supply risk refers to the risk of not obtaining the desired devices in the required quantities at the necessary time. This is comprised of production capacity, lead times, inventory policies, and the stability of the wafer supply chain. For SiC, wafer (substrate) procurement presents a unique constraint. The limited number of wafer suppliers and the direct impact of wafer quality on device yield mean that examining a device supplier's wafer procurement structure is part of evaluating supply risk.

Technical dependency risk arises from deep reliance on a specific supplier's designs, processes, and tools, making alternative procurement difficult. This risk increases as "ecosystem lock-in" progresses, through proprietary packaging, coupling with proprietary gate drivers, and specialized evaluation boards. Assessing whether the structure leads to a loss of procurement flexibility in exchange for technical advantage serves as material for long-term business decisions.

The priority among these four axes is not fixed. The weight of each axis changes depending on the phase: performance and technical support carry greater weight in the early stages of mass production ramp-up, while supply chain robustness and supply risk become more significant once mass production stabilizes. Being mindful of the current phase and weighting the axes accordingly is key to linking evaluation to practical application.

The Gap Between "It Worked in Prototyping" and "It's Usable in Mass Production"

A common pitfall in supplier evaluation is attempting to complete prototype and mass production evaluations using the same process. At the prototyping stage, the primary evaluation criterion is "whether the device works," but at the mass production stage, the questions become "can the design margin be maintained within the variation range?" and "will the same quality be delivered over the long term?"

Device lot-to-lot variation is defined by the maximum and minimum specifications in the datasheet, but the tightness of the distribution of actual mass-produced items within that range varies by supplier. To verify this, one can request actual measurement data across multiple lots or refer to data from independent reliability testing organizations. The ability to obtain this information before mass production transition significantly impacts the risk of design changes in later stages.

For reliability test data, verifying the test standards and conditions is also essential. The meaning of the numbers changes depending on whether the tests were conducted under JEDEC standards or proprietary conditions. For automotive applications, product compliance with automotive standards (e.g., AEC-Q101) is a prerequisite for selection.

Incorporating the "Roadmap to Next-Generation Devices" into Business Assessment

Alongside technical and procurement assessment, it is increasingly important to analyze suppliers' technology roadmaps within the context of business assessment. SiC device processes are currently undergoing rapid generational updates (trench structures, transition to 8-inch wafers, etc.). When the device adopted today becomes "previous generation" in three to five years, the supplier's ability to support the transition to successor products is a crucial selection criterion from a long-term cost perspective.

onsemi offers a comprehensive portfolio of SiC MOSFETs, SiC diodes, and SiC modules covering voltages from 650V to 1700V. Alongside Infineon's CoolSiC, Rohm's SiC series, and Mitsubishi Electric's trench-type SiC, multiple suppliers are now offering product lineups that cover a range of voltage ratings and applications. This competitive environment signifies not just a "decrease in cost," but a structural change that broadens design options and facilitates alternative sourcing.

The fact that various companies are arriving at different solutions for the technical challenge of balancing on-resistance and short-circuit withstand time also creates an opportunity for evaluators to ask the question: "Which approach best suits our system?"

Establishing a System for Continuous Supplier Evaluation

Finally, it's important to touch upon the timing of supplier evaluation. Evaluation should not be a one-time task performed at the time of adoption decision, but rather a continuous monitoring process throughout the product lifecycle. This ultimately leads to more accurate selection.

Device specification changes (materials, processes, packages) are notified as PCNs (Product Change Notifications). However, in practice, there are cases where mass-produced products change without notice because the PCN reception system is not in place. Whether the operation for managing PCNs is integrated into the supplier evaluation framework is something that should be confirmed from both design and procurement perspectives.

Furthermore, the power semiconductor market, following the surge in automotive demand in 2022-2023 and subsequent inventory adjustments, is seeing shifts in the production investment cycles of suppliers. Information such as in-house wafer production ratios, external procurement ratios, and the operation timing of new fabs serves as material for assessing medium-term supply capacity. Regularly following earnings call presentations and technical conference announcements allows for a three-dimensional understanding of suppliers' "financial health and direction."