Developments in this part of the market will define low power and energy efficiency for years to come.
Electric vehicles are leading the charge for innovation in automotive electronics. Companies that invested and embraced the challenge of EVs are besting their less-nimble, less-open-minded engineering cohorts.
Semiconductors and embedded computers have been controlling the dashboard, mirrors, seats, heating and cooling for years. But with EVs, engineering teams are starting to tackle tasks where many have only limited experience, at best, including battery pack system management, induction motor design, and a complete and total rethinking of the wiring and cabling connecting for all of the vehicle’s systems, from braking to radar systems for collision detection and prevention.
The engineering challenge around the EV battery subsystem alone is huge. “They are like humans,” says Puneet Sinha, automotive manager in Mechanical Analysis Division (MAD) at Mentor, a Siemens Business. “They can’t be too hot or too cold. They need to be thermally managed very carefully.”
The key here is being able to apply knowledge from multiple domains, from electricial engineering to the physics of cooling to the chemistry of hundreds to thousands of lithium-ion battery cells.
“It’s truly a multi-disciplinary task,” says Sinha. “You are looking at heat generation in the battery, which is dictated by lithium-ion cell chemistry and design as a function of duty cycle and of time. And the heat dissipation is dependent on cell designs and coolant loop configuration. On top of that, power draw to support the coolant flow, especially in pure EVs [as opposed to hybrid electric] comes from the battery itself, so you have to minimize that draw of battery energy.”
Engineers take into account the total volume of the battery pack and its location in the vehicle architecture. They have to work within a very tight envelope and still produce an efficient design. The thermal management task is sufficiently tough enough to require liquid coolant, not air cooling, and that alone adds components, which boosts both complexity and cost. Done wrong, this can cause a fire.
“In an EV battery pack that consist of hundreds to thousands of cells, you have to ensure graceful failure,” Sinha says. “This means in an event where one of a few cells start to overheat, the sharp temperature rise in that one (or few cells) is not allowed to propagate to the rest of the cells in a pack. This aspect of thermal management includes not only coolant flow, but also bus bar designs, where heat from a few localized cells can be dissipated without allowing it to propagate to rest of the cells.” You are also building in the hardware and software for “efficient cooling of electric motors, inverters and other power electronics, which are important to ensure adequate efficiency and reliability of the whole powertrain.”
Tools and expertise needed
This kind of engineering challenge didn’t exist outside of companies such as Tesla five years ago. Franz Maidl, director of Altium’s TASKING business unit, has been working in the semiconductor industry since 1984. He said he hasn’t ever seen the level of activity and investment in new automotive hardware and software, something that is evident on a drive through Livonia, Mich., just outside of Detroit.
“Everyone is there—brand new buildings, every company you can think of, packed with engineers. It reminds me of my first visit to Silicon Valley in the ’80s,” says Maidl. That said, the Tier 1 car makers and auto OEMs are coming to Silicon Valley to a much greater extent than they did in the ’80s and ’90s.
For automotive electronics, this is largely brand new technology. As such, it requires new methodologies, new tooling, more accountability, and a whole new level of expertise.
“There will be a shortage of electrical engineers in the automotive industry by 2022,” says Paolo Giusto, model-based systems thrust area leader for General Motors research and development. “We need to create virtual ECUs and a virtual development and optimization framework to deal with lots of constraints.”
In effect, auto makers are looking for the same kinds of tools that chipmakers have been using for the past few decades. “We need automated configuration of design parameters,” Giusto said. “We need wiring harness design models to deal with miles of wires. And we need co-simulation of a meta bus for optimization of a framework that includes high-fidelity models and a high-fidelity business model.”
The scale of this opportunity hasn’t been lost on tools companies, of course. All of the major EDA vendors have automotive units—Mentor, Cadence, Synopsys and ANSYS. So do IP vendors, such as ARM and Imagination Technologies. And so do chipmakers such as Intel, which just purchased Mobileye, Xilinx, NXP, Qualcomm, Infineon and Nvidia, and embedded FPGA companies such as Achronix. In fact, automotive electronics is one of the top five new market opportunities cited by chipmakers, along with artificial intelligence/machine learning, cloud computing, virtual/augmented reality and IoT/IIoT. All of those have ties into automotive, as well. And all of these offer huge new opportunities because all of these markets either are starting scratch or being reinvented with new technology.
“Right now self-driving is at level 2 of ADAS,” says Lip-Bu Tan, president and CEO of Cadence. “To get to level 5 you have to think ahead. What kind of sensors do you need? What kind of software do you need? It’s a long way to level 5.”
These kinds of changes impact every facet of the semiconductor industry. Cars are just one piece. And each carries its own set of rules, which semiconductor industry never really grappled with outside of defense and mil/aero.
“The auto industry survived for about 60 to 70 years before any regulation came,” says Simon Segars, CEO of ARM. “It wasn’t until the mid-’60s here in the U.S. where there were any laws about what you could drive on the road and safety standards you had to adhere to.”
Segars adds that in some ways, this is business as usual. But in other ways, it is a radical shift. “The complexity keeps going up and up and up,” he says.
That includes cybersecurity, as well, which has never been a concern in the automotive industry prior to the addition of electronics in cars.
“In the future, 50% to 75% of cars will have over-the-air capabilities,” says Chris Clark, principal security engineer for strategic initiatives in Synopsys’ software integrity group. “The dark side is that introduces vulnerabilities, and there is no way to limit all vulnerabilities. You have firmware, operating systems applications and a human interface.”
EV as a starting point
That adds another layer of complexity into these systems, and nowhere is this complexity more evident than in electric vehicles. These are the extreme edge of automotive development, and that becomes apparent in everything from advances in battery management to wiring harnesses.
“EV cable harnesses often handle very high voltages, so with that you get an increased need for safety vs. a similar vehicle with a gas engine,” says Paul Johnston, technology account manager with Mentor’s Integrated Electrical Systems Division. “In your local dealership’s familiar gasoline/diesel vehicles, although engineers have goals to ensure the wiring cost and weight do not exceed targets, typically there is a tradeoff between proliferation of choice as a result of end-user features for the car and the cost of manufacturing many variations. It is not unusual for a car to have more than 120 different installable dash harnesses in the last 10 years. Engineers fine-tune the content and usually give away some circuitry, which is present but not used in many cars. If this is say, four to six pounds, the car carries this unnecessary mass around forever. But in an EV/PHEV, extra weight like this is anathema. It is eliminated wherever humanly possible.”
So the design partners for Tesla are pretty much the same companies that line up to bid for the harness assembly volume business of the Detroit Big Three. These companies have engineering talent and IT tools that spans from gasoline to hybrid PHEV to EV domains, Johnston said.
But even that could change. As with all new markets, all options are table.
“There is a technology explosion in automotive right now,” says TASKING’s Maidl. “The car guys in general are moving away from distributed architectures to centralized. The cabling in the car is way too expensive when it is centralized. For the ECUs, more and more are getting into a complex SoC kind of supercomputer, meaning a multi-core master controller with a slave that is an ARM or MIPS-based SoC. The controller is the gateway into the telematics and then from there into the drivetrain.”
This architecture simplification pays off.
“You consolidate the driver information, the instrument cluster, and the infotainment units into a single ECU,” says Andrew Patterson, automotive business development manager at Mentor. “The right embedded software enables secure separation of the different domains, and shares the hardware resources available between the different hosted applications using a mutli-core framework or hypervisor.”
That, Maidl says, has entailed software re-builds.
“With the move from single core to multi-core, there is a huge stress in automotive when it comes to software. You are under the belief you can reuse the software, good luck if you are moving single core to multi-core,” he says.
The 16-bit controllers, the one-time bread and butter of Freescale or STMicroelectronics, have really given way to 32-bit ARM Cortex-M core based controllers, from Infineon, or from Japan’s Renesas. Maidl notes the POWER controller architecture that came from Motorola SPS and IBM has less influence in automotive, though it remains in aerospace and defense.
“For Airbus and Boeing, that’s how that industry works. A stable process and stable architecture.”
The larger point is the combination of everything — the powertrain, the EV battery pack needs, the driver comforts and options — with the electronics for self-driving has meant an “explosion of software content in the car” and an explosion of embedded computers with plenty of storage. The self-driving alone, in the new Audi A8 says Maidl, entails 4GHz low range and 77GHz long range radar, 80mm laser and cameras.
“To me, radar is THE application right now,” Maidl adds. “Radar is going to get more and more important. Here you are seeing a lot of development.”
Is it safe, and for how long?
Ahmed Eisawy, a product marketing manager at Mentor, talks about how automakers see if these new approaches will still yield safe and reliable electronics in an ultimately safe and reliable car.
“The race to the next process technology, constrained by strict requirements coming from intelligent control, autonomous driving, ISO 26262 and so on, place significant reliance on the verification methodology to ensure the reliability and safety requirements are met,” says Eisawy. Companies like Delphi have to simulate and validate the resulting “system of electronic systems.”
In the case of EVs, all of the semiconductors and electronic systems must perform as good or better than conventional combustion engine vehicles that car makers put on the road with 10-year, 100,000 mile warranties.
“Designers spend a lot of time verifying their designs will function properly, beyond typical conditions, by focusing on the quality aspects of their designs starting with process variation. Extensive simulations are run, along with Monte Carlo analysis, to sweep all PVT (process, voltage, temperature) scenarios,” Eisawy says. “Then, fault coverage and fault tolerance analysis is performed to detect failures in design intent devices and parasitics. Once the quality aspects are addressed, designers typically analyze circuit reliability, starting with the impact temperature effects will have on their designs. ICs placed close to the engine endure high temperature in a rapidly changing environment, which can affect circuit functionality. Next, to account for the heat generated by power devices electro-thermal analysis must be performed. Finally, to ensure proper functionality after 10 years or more of operation, aging simulations would be run to accurately predict the lifetime of the ICs. However, these aging simulations are also impacted by other effects, like process variations and thermal effects which complicates the verification phase.”
Electric vehicles stand at the vanguard of automotive design. They include designs with the most extreme power management and capabilities.
It’s likely that the technologies, methodologies and tools developed in the EV world will find their way into all cars over time, just as some of the advances in data centers made their way into notebook computers and smart phones, and some of the energy-saving techniques in smart phones found their way back into data centers.
But what also is clear is that this is just the beginning of a development cycle. After more than a century of development, the automotive industry has suddenly veered off in a completely different direction.
Designing SoCs For Hybrids
It’s a whole new ballgame when it comes to designing SoCs for hybrid electric vehicles, from regulatory to technical to ecosystem challenges.
Self-Driving Cars Rattle Supply Chain
Shift from hundreds of MCUs to more centralized computing is creating chaos in the auto industry.
What Can Go Wrong In Automotive
Experts at the Table, part 2: Understanding security risks, ECUs vs. SoCs; dealing with an explosion in data.
LiDAR Completes Sensing Triumvirate
Technology will complement cameras and radar in autonomous vehicles.