Ramping Up Power Electronics For EVs

The rapid acceleration of the power devices used in electric vehicles (EVs) is challenging chipmakers to adequately screen the ICs that power these vehicles.[1]

While progress toward autonomous driving is grabbing the public’s attention, the electrification of transportation systems is progressing quietly. For the automotive industry, this shift involves a mix of electronic components. Among them are complex mixed-signal modules that support battery management systems (BMS), and power ICs that manage the shifts in voltages between charging stations, battery packs, motor drives, auxiliary battery, and braking systems.

Power ICs traditionally were based on silicon technology, either power MOSFET or insulated-gate bipolar transistor (IGBT).[2] More recently, silicon carbide (SiC) and gallium nitride (GaN) based ICs have been making commercial inroads into these systems. These technologies offer higher breakdown voltage, increased mobility, and require less thermal management.

However, the manufacturing processes remain challenging due to the lattice mismatch between the base wafer and the epitaxial layers grown on it. Crystal defects can adversely impact device function. In addition, small wafer size (150mm) and wafer fragility result in a high cost. This motivates engineers to inspect wafers earlier to screen devices and provide feedback on key process steps. In addition, the high voltages and currents pose test handling issues.

The chip industry is working hard to solve these issues because the automotive sector is an important and growing IC market growth.

“The three primary sectors for growth are automotive, computation and data storage, and wireless,” said Jerry Broz, president and director of technology at Advanced Probing Systems. “Although automotive is probably the smallest of these sectors, it is clearly recognized as the fastest growing. This growth is being driven (not fueled) by the fact that all major carmakers are rapidly transitioning into the partially or fully electric vehicle era. I have seen several market predictions that automotive semiconductor devices will account for more than 15% of the global demand within the next 5 to 10 years.”

According to a recent Yole report, the market for power SiC devices is projected to reach $3 billion in 2026, with the automotive market accounting for two-thirds of that amount.

“We’re seeing a big shift in the electrification of vehicles, with an ecosystem emerging around advanced battery management, drive trains and charging stations. Silicon carbide devices and advanced measurements in battery management are some of the most interesting areas, and we’re at the beginning of the wave of innovation,” noted Rick Burns, president of semiconductor test at Teradyne. “Everybody looks at today’s electric vehicles almost like the Stanley steamers of the combustion era. They’re perfectly usable but there are issues around range, weight, recycling and more.”

With a major shift in technology comes a shift in inspection and test needs, as well.

“Industries driving improvements in test are ultimately driven by a step function improvement in product economics and/or product capability. In the case of automotive, the need for both an economic solution and enhanced product capability are required because of the explosion of semiconductor content within the car and the roadmaps to enable software-defined, autonomous driving and electrification,” said Marc Hutner, senior director of product marketing at proteanTecs. “As a result, we are seeing solution providers seek techniques beyond traditional BiST to understand other signatures like device aging, software stress, local variation and other operational parameters.”

High voltage and high current challenges
A primary reason for shifting away from silicon power ICs is the higher voltage tolerance of SiC and GaN. With higher voltages comes high currents, and the testing environment to support them is non-trivial because of range, resolution, and safety.

“New applications come with unique production test challenges,” said Vineet Pancholi, senior director of Test Technology at Amkor Technology. “For instance, EV power switches made with SiC and GaN allow for controlling higher voltages and currents. Test instrumentation to test transient response for voltages up to 2,000V and currents up to 800 to 1,000A, with an economical test time and parallelism, is unique.”

The voltage/current range and resolution of power ICs presents ATE vendors with extraordinary challenges. Voltages are in the range of 400 to 10,000 volts, yet the resolution demanded is remarkably small.

“Today, we’re talking about millivolt resolutions. We already see at least one or two orders of magnitude on the horizon,” said Teradyne’s Burns. “Unless there’s a fundamental shift in the battery technology, this quickly becomes microvolt resolution.”

Current is the other axis. “It’s the extremes that are interesting,” said Burns. “For any given battery cell, current levels are relatively small, and we’re looking at measurements of current at high accuracy, microamp. But at the other end of the spectrum is the charging circuit that’s connected to it, which could be trying to operate at very high voltages. I call it the 1,000-volt and 1,000-amp problem. The dynamic range makes testing challenging. We’ve been able to test large advanced digital data center devices, which consume 1,000 amps, but they don’t try to do it at the same time as measuring at microamp resolution. And with EVs it’s this wide dynamic range that’s creating new test challenges. For a commercially viable product you need to have some ability to swing the tools across this range. The companies that manufacture both BMS devices and drivetrain devices want to have test solutions that have some flexibility to operate in all of these corners.”

Test equipment vendors have been making investments to prepare for the expected uptick in volume of power ICs.

Keith Schaub, vice president of technology and strategy at Advantest America, commented on a recent acquisition: “CREA, now a subsidiary of Advantest America, specializes in power semiconductors testing, and they build specialized equipment to actually measure these devices because their test challenges are unique. The two main unique things are the extremely high voltages and currents. So, you’re talking thousands of volts and hundreds or thousands of amps and the thermal challenges that require specialized hardware, which CREA has invented a lot of the necessary technology. Anytime you’re driving that much current, you’ve got to expertly manage extreme amounts of heat.”

There are electrical safety concerns, as well.

“There are a number of specific test challenges for high-power ICs. Due to the high currents needed for testing you can kill your tester, probe card, or neighboring devices when there is a short circuit in the device,” said Dieter Rathei, CEO of DR Yield. “Therefore, the testing process requires applying the currents carefully.”

Others agree on the dangers that need to be managed. “An engineer recently showed me a picture where they’re testing on wafer, and when something goes wrong, there were welding-like sparks flying from the test cell,” said Schaub. “We have to take that environment and move it into high-volume manufacturing. The silicon carbide (SiC) market was around $1.2B in 2022 and is estimated to reach approximately $14.4B by 2030. Traditionally niche, but with the explosive growth of the EV industry, the challenge will be to scale the technology for high volumes.”

Screening for defects
Compared to CMOS SoCs, power ICs contain a handful of devices to screen for defects. And while testing analog performance can screen out defective parts, the screening starts with inspection at the bare wafer level because the dislocations occur during crystalline growth and epitaxy processes. Intercepting wafers after these processing steps provides process feedback and checks wafer quality.

Compound semiconductors like SiC and GaN are expensive to manufacture, and they are inherently brittle. That results in atypical wafer processing.

“In terms of data analytics, it doesn’t matter much whether it’s compound semiconductor material or plain silicon, we’re just dealing with test data,” said DR Yield’s Rathei. “However, there are some unique challenges. For instance, some of these compound semiconductor materials are so expensive, and also so fragile, that one of our customers processes wafers even if they’re broken. You would never do that in a CMOS factory. It’s a challenge on the tool level, because then the chucks have to be able to hold half or a quarter of a wafer. It also poses some challenges for data analytics, because at functional test you have only a fraction of a wafer that’s been tested.”

Another aspect of defects is that some of the crystal/epitaxy defects could be benign. Determining which ones don’t matter improves yield, but it’s complicated to do so.

“These devices aren’t built on a mono-crystalline structure. Before you start putting circuitry down, there are already a lot of defects within the crystalline structure,” said Dave Huntley, business development director at PDF Solutions. “So, the question then becomes, ‘Which of those defects are going to actually affect the final product?’ These are all over the wafer. You can’t afford to exclude every defect. You have to decide which ones are going to matter. But in order to know what’s going to matter, you need to first inspect them, build your devices, and test them to determine if this device is actually good.”

Such investigations are underway. Huntley described a current project on an assembled module comprised of 36 SiC devices. “We are testing individual pieces of the module rather than the entire module,” he said. “Specifically, we can test individual assemblies within that module at final test. Upstream, we are collecting defect information on the raw silicon carbide wafer, which we characterize with machine learning to identify defect categories. Combining these two data sources we are looking for correlations to determine which categories actually affect the final test result. Nobody really knows which defects are killer defects. We’re trying to find that out using inspection, a defect management system, machine learning, and the final test results.”

Conclusion
Electric vehicles require a diverse set of semiconductor technologies, including automotive grade ICs for battery management and drive train controls, and the ICs and SoCs to manage charging stations along roads and highways. Screening CMOS SoCs at high volumes is nothing new. With the expected ramp of electric vehicles power ICs, those composed of SiC and GaN in particular pose significant inspection and test challenges.

Detecting defects due to lattice dislocations necessitates inspection, along with test systems that can support a wide range of voltages and currents. Today’s test solutions have been manageable due to low product volume. But that’s changing, and to support higher volumes engineers must innovate system solutions to balance the high voltage, current, and thermal effects — while reducing test costs.

References

1.  Wikipedia, electric vehicle definition, accessed on June 1, 2023. https://en.wikipedia.org/wiki/Electric vehicle
2.  Semiconductor Engineering, “Power Semiconductors” ebook,  https://semiengineering.com/product/power-semis-march-2023-ebook/

Related articles

The Silicon Carbide Race Begins
As SiC moves to higher voltages, BEV users get faster charging, extended range, and lower system costs

Gearing Up For Next-Gen Power Semis
Aluminum nitride, diamond semiconductors, gallium oxide and vertical GaN are all being readied, each with its own pros and cons.

Inspecting, Testing, And Measuring SiC
Demand for solutions is booming, but the technology is just beginning to gain its footing.

How Quickly Can SiC Ramp?
New capacity planned for 2024, but production will depend on equipment availability.