Which Fuel Will Drive Next-Generation Autos?

With gasoline prices hitting uncomfortable highs, consumers increasingly are looking toward non-gasoline-powered vehicles. But what ultimately will power those vehicles is far from clear.

Inside the cabin and under the hood, these vehicles will be filled with semiconductors. Yet what the energy source is for those semiconductors is the subject of ongoing debate. It could be batteries, hydrogen fuel cells, or solar, or some combination of any of those. Those decisions may vary by region, by regulations, and by total cost of ownership.

“A lot of people assume the world’s going all-electric, which may or may not be true,” said David Fritz, vice president of hybrid-physical and virtual systems, automotive, and mil/aero at Siemens Digital Industries Software. “But most of the energy in the United States currently today is generated by burning coal and fossil fuels, so electric cars don’t drive for free. That energy is generated somewhere to charge up the battery, and that’s expensive pollution-wise. We’re a long way from 100% renewable energy. Something else is likely to happen before we get to all-electric, and that easily could be hydrogen. We’re going to have hydrogen hybrids. Generating hydrogen is pretty cheap. It takes a lot less energy. You put positive and negative poles into water, then you get hydrogen and oxygen out. You save the hydrogen, and find a way to distribute it like propane. Then, as the electric charge goes down, the hydrogen actually runs a small engine with an alternator that charges you back up. So a hydrogen hybrid may happen in the future.”

This is important at a chip level because of all the processing needed inside of vehicles. “Imagine monitoring everything that’s happening,” said Fritz. “All the inertia, all of the cameras, all of those other things, and making sure that you have a safe, comfortable environment takes energy. That means we need something on board that’s sustainable. We’ll have hybrids until we finally get a fusion reactor in the car, because that’s all the energy you will ever need your whole life. Until then, it’s going to be a challenge.”

This marks a significant shift in thinking for the automotive tech world, and questions are being raised at numerous automotive and chip conferences about how these complex electronic system of systems will be powered. While electrification of the vehicle is a given, the fuel source for that electrification is in flux.

“The automotive industry is in a disruption phase, moving from gas to electric,” said Paul Graykowski, senior technical marketing manager at Arteris IP. “However that electricity comes to the car doesn’t matter. What matters is the compute power that’s coming on board with cars. In the past, there were a few microcontrollers here and there, a CAN bus, etc. Now there are between 23 and 26 SoCs in the car. Some of these are focused on cameras or radar or lidar, and there’s just a lot of processing power going on. More than that, there’s also a drive toward infotainment, so there are more modems in the vehicle. You’re getting the entertainment aspects, as well as increasing cockpit control, engine control, chassis control.”

As cars move toward Level 2 autonomy, the market opportunity is tremendous, Graykowski noted. “This market is already to the point where it’s almost the size of the smartphone market. It’s an enormous opportunity for tech in general, and everything that’s going on in automotive is going to drive tech into the future — whether it is the car, or the networks that connect all the cars together, or the data centers that are tied to cars. And while we are a bit further out with Level 4 autonomy, it is conceivable. We cannot have full self-driving, especially in cities, because there’s just too much going on. But on highways, that’s all possible. As full self-driving goes into the cities, the infrastructure is going to have to adapt. Cities are going to have to invest in what they did when they went from horses and carriages to cars. Now they’re going to have to go from the car to the self-driving car, and all that infrastructure is going to have to change.”

Many of the improvements in vehicles are based on data processing, Graykowski added. “Today, they include cool features like adaptive cruise control and the like. As more of that is included, it requires extra compute power to make it possible to provide more features for the consumer. So while we’re seeing a lot of specialized hardware being developed, the reality is it’s the software that’s really making all these things happen. The car will get simpler as you get rid of the transmission and you have a single speed, which is different from the old implementation of the vehicle where it was all about building these complex engines with all the latest gadgets on them to get the most horsepower out of it. An electric motor is so much simpler. Now we have all these fancy processors sitting in there and we have a little extra room to do some new stuff on it, so one of the things we’re going to see moving forward is just what these chips can do together.”

This is where zonal computing comes into the picture. Instead of having everything feed all the information back to the processor, smarter sensors only send the information they really need to send, and they do some local processing. In addition, the chips are being optimized for the software. Nevertheless, the more features, the more energy required.

Which fuel source for the future?
The question now is where that energy comes from, and so far one size does not fit all. While sunny climates can rely on solar to generate electricity, that doesn’t work in much of the world. In Europe, the debate over hydrogen fuel cells has been underway for years, with passionate arguments on both sides. Whether hydrogen fuel cell electric vehicles become a significant part of the on-road vehicle mix has yet to be determined.

For one thing, hydrogen is difficult to work with and to store. “Hydrogen is so small it infiltrates through the crystal structure and the result is iron hydride, which is brittle,” said Marc Swinnen, director of product marketing for the semiconductor business unit at Ansys. “Hydrogen may have some specific uses where it’s good but first of all, hydrogen is not a fuel. There’s no free hydrogen. There’s some in gas wells; you can get some hydrogen out of gas wells containing a proportion, but hydrogen is something we have to make. How do we make hydrogen? We can take electricity and electrolyze water, which is very low efficiency. But why not use electricity directly in the battery and skip the middleman? The way that 95% of the hydrogen is made in the world is through a high-temperature reforming of natural gas. You take natural gas, add a high temperature catalyst, and it splits the methane with steam. You put it with steam, and it turns into carbon dioxide and hydrogen.”

That’s hardly a clean fuel, however. “The carbon dioxide from burning the methane still goes out the smokestack,” Swinnen said. “Now you have hydrogen, which you have to then compress. That takes energy. Then you have to transport it, which takes energy, and then put it in your car, and then transfer it back. Why not burn the natural gas right in your car? You can do that today. Skip all of that nonsense, and just burn the natural gas in your car. The same CO2 is coming out the tailpipe, and you skip all of that complexity in the middle. There’s no benefit to going to hydrogen and people say, ‘Oh, the exhaust is water.’ Well, no. That’s all you have. You created all the CO2 somewhere else. There’s no benefit in hydrogen.”

Another argument in favor of hydrogen fuel cells is the energy density. Compared with other fuels, the energy density is high, but that’s not the complete picture. “At the bottom you have energy density per kilogram, mega joules per kilogram, and the other one you have is mega joules per liter, which is volumetric density. Hydrogen is the best fuel per kilogram, but it’s very low per liter. Liquid hydrogen is very poor from a volumetric density point of view, and to make it liquid, you have to make it cold. It has to be compressed, and a huge amount of energy goes into that. Some people even say that the amount of energy it takes to make hydrogen is more than hydrogen contains. Hydrogen is not a fuel. It’s an energy.”

There are specific use cases for hydrogen fuel cells, however. During the Iraq war, there was a battery emergency because all the electronic equipment — radios, night vision goggles — were churning through batteries quickly. To power them required soldiers to carry many pounds of batteries. By adding jet fuel into a small fuel cell, enough electricity could be generated to power that equipment for several weeks, he said.

That’s one side of the hydrogen debate. “Hydrogen is a different energy carrier or vector with high energy density per kilogram (gravimetric density),” said Patrick Leteinturier, fellow in the Automotive Powertrain Systems at Infineon Technologies. “In 700 bar compressed form, the energy gravimetric density of ‘hydrogen + tank’ is 10 times higher than an electric battery pack solution. Also, the refill time of the hydrogen tank can be achieved in less than 10 minutes, as fast as refilling gasoline or diesel. A hydrogen system can be operational at very low temperature e.g. -30°C. Hydrogen is an abundant element that can be produced everywhere. A hydrogen system requires few rare materials and can be recycled easily, unlike batteries.”

At the same time, Leteinturier recognizes the technology challenges of implementation. These include achieving a high-enough production level to achieve the economies of scale needed for the hydrogen system on the vehicle, and getting the hydrogen infrastructure available and affordable so that it is $1.50 per kilogram of hydrogen, for example. Also required are synergies that span different sectors with industry, energy, heating, and mobility, including cars, trucks, trains, ships, planes, etc.

That leaves a big unanswered question — what is the best alternative to internal combustion engines. The answer is, it depends.

“The internal combustion engine, using hydrogen as a fuel, will peak at 50% efficiency from fuel to mechanic, whereas hydrogen fuel cell can provide an efficiency at about 60%, from fuel to electric so there is no unique answer to the question,” said Leteinturier. “It depends on the application, whether it is passenger cars, commercial vehicles — LDV, MDV, HDV, HDV-long-haul, construction vehicles, agriculture vehicles, trains, ships, or airplanes. There is no doubt that a lightweight passenger car with a range below 400 km will be a (BEV) battery electric vehicle. On the other hand, an HDV-long-haul with 44 tons and 800 km range will be better with a FCEV (fuel cell electric vehicle). For passenger cars, the internal combustion engine using hydrogen as a fuel is more a bridge technology.”

Fig. 1: In electrified vehicles, AI shows great benefits in virtual sensor or system modeling use cases. Source: Infineon

Chips matter
No matter the fuel source, the semiconductor and EDA industries play an important role with any alternative to internal combustion engines.

For the hydrogen fuel cell ecosystem, Infineon’s Leteinturier said the semiconductor industry plays a role in the production of hydrogen. “The power electronics are key — alkaline electrolizers, PEM electrolizers, solid oxide electrolizers. In addition, for mobility we use electronics for control, sensing and power such as PEM monitoring and controls, air compressors, DC-DC. Further, the semiconductor industry plays a role in the transport of fuel. Electronics are used for compression or liquefaction. Finally, the semiconductor industry plays a role in the storage of fuel sources, and electronics are used for metering, monitoring, diagnosis and safety.”

On top of that, there are significant opportunities to improve the efficiency of vehicles. “The semiconductor industry is working with the Tier Ones and automotive OEMs to use AI to do that efficiency improvement,” said Ron DiGiuseppe, senior marketing manager for automotive IP at Synopsys. “We see AI could be applied in engine management, battery management, and the charging infrastructure for onboard chargers. In lots of ways, there are opportunities for better efficiency, which is perfect for battery electric vehicles to extend the range, and to improve the efficiency. AI can be used in multiple ways to do that.”

Chips play a role in engine efficiency and management, as well. “Electric vehicles have fewer sensors in the engine and powertrain compared to the traditional internal combustion engines,” said DiGiuseppe. “That speaks to efficiency since a lot of the traditional sensors have been designed out in the conversion from combustion engine to battery electric. Things like solenoids and oil pressure sensors have been designed out, but there’s still a number of sensors that are in battery electric vehicles. For AI, another way to improve efficiency and reduce the cost is use AI and predictive analytics to have virtual sensors. The remaining sensors can be in the electric motor, for example. There is a traction motor inverter which serves as the rotor positioning for the electric motor. Removing the sensor therefore reduces the cost, and adds to the efficiency. Using a virtual sensor is one of the applications in the electric motor as a use case for AI, for optimizing motor control. It makes it more efficient, and AI uses predictive analytics to understand the position of that rotor within that electric motor.”

AI also can be used for battery electric vehicles to improve efficiency. One way to do this is on the onboard charger, which would use AI to do multi-phase frequency control, reducing system costs and improving efficiency.

Then, for EV batteries, which are the most expensive subsystem in an electric vehicle, AI can be used within the battery management system. “There could be 1,000 to 2,000 cells in the big, complex battery system, all of which are connected serially,” DiGiuseppe said. “One low operating cell in that whole battery, because they are connected serially, could have an impact. It’s always the least efficient part that drives the whole system, so having a battery management system that can predict that battery condition of all these subset cells would also improve the reliability and improve the efficiency.

AI is also used in electric power trains, and other areas. “You would think that power trains are more traditional, that they just have MCUs controlling the air flow or something. But the fact is that AI is already very dominant in ADAS, and is at the heart of all these ADAS solutions for vision, radar, lidar, etc. ADAS is all about AI, but also some infotainment features like heads-up displays and driver monitoring systems. AI is also widely used in infotainment. AI is also being used in powertrain and battery management, engine management, and electric motor management. So suddenly now we’re looking at AI being widely used in ADAS and automated driving autonomous vehicles, infotainment, ADAS, powertrain electric vehicles. That’s a pretty wide application of AI all throughout the car, so it’s really expanding.”

Designing the best way to a destination
As driver assistance features increase, trip planning and fuel management become serious concerns.

“How well does this car actually understand my driving behavior?” asks Frank Schirrmeister, senior group director, Solutions & Ecosystem at Cadence. “How accurate is its planning for my destination? How does the car actually do that? The car understands the driver’s general driving behavior. For example, for a particular trip, my vehicle said I drove 84% economical, 14% normal, and 2% aggressive. So the car understands. It has a digital twin of my driving behavior, if you will, and uses this as an estimate of how far I will get on a charge. Where EDA technology comes in relates to the control and compute aspects that need to be built into the system to understand what’s happening so you are not surprised when you don’t make it to your destination because you are one mile too short. Infrastructure, planning, and control are very important from a system-level modeling perspective in automotive.”

As vehicles get smarter, they should never run out of fuel. “Energy management is a system-level challenge within the car, and it needs adjustment,” Schirrmeister said. “My car learns in from a trip what my reach would be based on my driving habits. But somebody else with different driving habits might have a very different scenario. This predictive power looks at where my car is, exactly my situation for the car, what the roads are, and it gives a plan based on a very complex control algorithm. What’s obvious when cycling through different settings is that my vehicle knows my driving style, so the whole notion of economic driving, normal and aggressive driving impact, which then does a calculation from there as to how much power you will use in each of those modes. The rest is math.”

At the end of the day, EDA is always a little industry enabling 500 billion semiconductors, he said, but a bigger impact is possible with the modeling of the scenarios. “It comes down to having the right scenarios properly modeled, in addition to solving the problem of planning your verification — where with constrained random testing, go into the corner cases and figure out where you want to verify. It’s a related problem for generating synthetic test cases for autonomous driving. That’s the next level of scenario modeling for the verification of the car. In the system environment, you want to really understand how your car will behave, and how you will behave with the car.”