High Voltage Testing Races Ahead

Voltage requirements are increasing, especially for the EV market. Even devices that might be considered relatively low voltage, such as display drivers, are now pushing past established baselines.

While working with high voltages is nothing new — many engineers can recall yellow caution tape in their workplaces — the sheer number and variety of new requirements have made testing at high voltages a top priority for both foundries and testing houses. Cars, for example, are now multi-domain systems, with voltage requirements ranging from about 40 volts to 2 kilovolts.

“The adoption of EVs is accelerating, and pressure to have similar price points to combustion engines is increasing,” said Kyle Klatka, director of marketing operations at Teradyne. “One major consumer concern is vehicle range. In order to increase that range, the automobile industry needs to decrease the cable weight and the losses from power distribution. Given that the trend is toward higher voltage throughout the car — and that includes batteries and power conversion throughout all the subsystems, as well as individual components such as displays — expertise in high-voltage and high-power design will be key for the semiconductor suppliers. And test equipment companies will need to achieve that domain capability to really enable the EV market to reach its potential.”

All of that needs to happen quickly, as well. “I see milestones that we thought would be further out,” said Dennis Keogh, business development manager at Eagle Test Systems. “There’s a desire to get to higher power more efficiently, and that drives higher-voltage silicon carbide and GaN components in the drivetrain. There’s also a consumer Holy Grail of wanting to charge a battery in the amount of time it takes to fill a gas tank. The estimate is that needs to be about 1,000 volts in the battery/charging mechanism, so the industry wants to get from today’s 400 volts to that 1,000 volts.”

This accelerated roadmap presents a number of challenges for testing companies, such as the possibility that materials used in higher-voltage parts are brittle and more prone to damage during testing.

“Silicon carbide was considered somewhat new and unproven, but Tesla has gone all in on silicon carbide for the cars and applying that technology for their PMICs. That’s really given the signal to the industry that it’s okay to go ahead with silicon carbide, so it continues to make an impact on inverters and other things for high voltage,” says Marc Swinnen, director of product marketing for Ansys.

The chip industry is racing to iron out all the kinks. “There’s a long learning curve with silicon carbide,” said Roland Stele, marketing manager at Infineon Technologies. “The real challenge is to decrease defect density. Working with silicon carbide is not so different on the design side. It’s more about how it’s manufactured and how it’s packaged. Bonding is a challenge, and so is the testing of the end device. There’s also a transition from 6-inch to 8-inch wafers, and an issue with how to cut the wafers so you don’t have much waste.”

Stele said the learning curve began in the 2000s, but it leveled off because there was no sufficient demand for GaN and SiC. He noted that has changed over the past couple years as these materials are deployed in a variety of applications, particularly automotive.

Both GaN and SiC are more brittle than silicon, so working with these wide-bandgap materials requires some changes. “With materials like gallium nitride and silicon carbide, oftentimes working at max means you rapidly damage the part,” said Tom Tran, field product specialist at Teradyne. “You may not see it immediately, but there’s a high probability it can happen, even below the expected maximum for the device. The other issue is the probe needles. With a lot of these parts, we’re dealing with much higher power, as well. For example, silicon carbide for drive trains is going to be dealing with 10 to 25 amps, or even 30 amps. So now you have to touch down with many more probe needles, because there’s a upper limit to what you can feed down through a single contact.”

This can lead to longer testing times. In addition, chipmakers are wrestling with leakage and dynamic RDS(on), so being able to fine-grade results is essential.

“Dynamic RDS(on) is really a question of, ‘How long do you have to apply your stress before you start seeing real RDS(on) delta?’ We’ve seen a lot of instances where the stress could be as low as 10 microseconds or as high as 400 milliseconds. But we typically would see the higher end, sort of like an R&D type of process versus production,” said Tran. “It’s become really interesting in that different requests from different customers are giving us different numbers as to how close can we hit it.”

While some of these challenges are both novel and well-known, they need to be re-thought, said Dave Armstrong, principal test strategist for Advantest America. “Devices always have been burned in. One way is high voltage stressing them. You run a very demanding pattern, high power, high everything, or you take it up to a higher level of voltage than you normally do, even higher that what it’s specified at. High-voltage stress is a very challenging test, because people have routinely done it at a certain voltage. It’s been a constant number for years. Then they drop the VDD, but they don’t drop the HVS limit. Now, all of a sudden, you’ve got more stress on your part, because your part has so much room for over-voltage. And as you get down to a 0.75 or 0.5 volt part, and you take it up above a volt, are you overstressing your device? People end up throwing away parts because they zap them. “It’s unclear if this is the right thing for us to be doing. What is clear is that this is an existing testing strategy which we need to consider carefully as geometries shrink.”

There are additional challenges, which range from the mundane to the dramatic.

On the one hand, there are obvious problems that go back years. “For example, you’re doing breakdown tests and you’ve got a large site count, like 16 or 32 sites,” said Tran. “Then one of the sites fails and your power rail just drops. What we’ve done to mitigate that is to manage the current on the low side, so even if a part fails, we clamp it. That allows our power supply to stay high, and we can still take measurements on all the other sites. It stays parallel. That’s our way of getting around having to go back and do all the measurements one at a time.”

Other issues
Increasing the voltage creates other challenges, as well. “There are arcing problems as it starts to climb in voltage. It can actually burn its way through to the chuck, and that becomes a very expensive repair,” said Tran. “It becomes an issue not just with our equipment where we’re dealing with controlling everything down to the probes, but for the probe card manufacturer to consider. How do you get a pressure chamber in there? How can you control it, especially as your site counts go up, while your needles get closer and closer together? A lot of that is just raw physics, where it’s a lot of time spacing. Obviously, it’s something we’ll see a lot more often as we go from 900 volts to 1,200 volts and even higher.”

EV demands also are affecting components such as display drivers, which are essential for EV cockpit displays. “The key challenge in this operation is to execute testing that reflects the device’s actual operation,” said Yoshiaki Ueda, marketing manager at Advantest. “Automotive systems require high quality for every component, so automotive semiconductor devices must be tested in actual operation conditions. Handling load current is the key to realizing actual operation testing.”

Advantest just introduced an LCD high-performance per-pin digitizer and comparator module. “DDICs for the automotive market have typically contained separate chips for the source and gate drivers and the timing and touch controllers,” the company said. “Now that these functions will all be integrated into a single chip, demand for a tester capable of testing all device functionality in one insertion is growing rapidly.”

The new devices tests to high voltages of ±40V. It’s unlikely display drivers will exceed those requirements in the near future. So for now, testing is still based on existing equipment and processes. “High-voltage testing of display driver ICs (DDICs) is conducted with an LSI tester, as with any other testing,” said Ueda. “A display basically has a source driver and a gate driver, and the gate driver requires high-voltage operation. As the frame rate increases, the speed of the gate driver will also increase, so high-voltage and high-slew-rate operation should be required.”

The future
A mix of traditional and novel hurdles remain, however. “In a lot of cases, our customers IC design cycles are shorter than equipment companies’ instrument and design cycles, and that’s always put some pressure on us,” Keogh said. “It’s hard to predict the combinations — the infrastructure, the battery infrastructure, where to reduce weight, wireless communication for the battery control systems. Everybody’s going that way, but it might drive some things that are really strange that you wouldn’t think would be impacted by voltage. For example, there’s a battery management chip that monitors the state of charge of all the cells and then balances them to make sure that they stay within the correct range, and it’s really critical to delivering electric vehicle range, which means it’s also critical to safety. The industry is moving to wireless BMS to reduce weight in the car by reducing the amount of physical cabling required. Our hypothesis is there will be a Bluetooth chip and a BMS monitor on a control board, but the two will not be combined. Now we hear people talking about putting Bluetooth in the same package as this battery management chip, which needs 150 to 200 volts.”

The push for higher voltages will have consequences for the drive train, as well. “The efficiency drive is so huge that it’s absolutely pushing the battery charges up, so battery monitors are going to have to deal with 1,000 volt systems. From a safety and manageability standpoint, maybe 1,000 to 1,200 might be the upper limit. On the drivetrain side, that means I have to deal with 1,700 to 2,000, maybe even 3,000 volts, said Keough.

Tran added, “I absolutely see my group of customers also increasing current because they have to hit a particular power curve. As a result, maybe current stays the same but voltage just skyrockets. And so now I have to deal with things like how to isolate my really high voltage stuff from my really high current stuff, and how to deal with the relays and the mechanics of it all.”

There’s another high voltage power question that can’t be overlooked for the future, said Swinnen. Once all the testing issues are resolved, and millions of EVs are on the road, how will power be supplied to the charging stations? China may be ahead of the U.S. on the answer, he said.

“High-voltage transmission lines increasingly have solid state controls at either end, rather than traditional transformers,” he explained. “China has gone big on them, they have at least 10 of these very high voltage transmission lines to bring the power from where it’s generated to the coast where it’s needed. The U.S. hasn’t got that many of these lines yet, but as it tries to upgrade the grid to be at least in the 21st century, that will include long-distance high voltage lines, and that will spur a whole development of HV technology as well.”

Conclusion
“We’re on the bleeding edge,” says Tran. “But as soon as it starts to settle, everyone starts thinking economics again.” In that respect, it’s “Back to the Future” for the testing industry. Big test companies were founded to ease what was initially a bottleneck of testing by hand. Now they are helping to streamline the bespoke awkwardness of the EV industry.

“If you looked at where we were about five years ago, there was a much more of ‘choose your own adventure’ path to high-voltage testing for a lot of discrete parts,” said Teradyne’s Klatka. “Now that our customers are going beyond protypes and ramping to millions of parts a year, they’re starting to realize there’s value in having a provider that has thought about the system and instrument aspects of the design.”

—Ed Sperling contributed to this report.