//php echo do_shortcode(‘[responsivevoice_button voice=”US English Male” buttontext=”Listen to Post”]’) ?>
Likening the modern automobile to a server on wheels is starting to sound cliché, but we’re just hitting the tip of iceberg when it comes to the number of electronic components that are finding their way into modern automotive design, as well as the supporting infrastructure for autonomous driving in large urban centers.
In the meantime, the ground underneath is constantly shifting: Supply chain constraints, software defined architectures, functional safety requirements, and the changing dynamics among original equipment manufacturers (OEMs), tier 1 suppliers, and semiconductor companies are altering the landscape of automotive electronics. This dynamic environment was the subject of discussion in a recent panel hosted by ProteanTecs, and, judging from that talk, “changing” may be an understatement.
The winds of change
Panelist Peter Mertens, an automotive industry veteran and tech investor, said the automotive industry hasn’t faced this much disruption in a century, thanks to a very strong push toward electrification, as well as a huge demand for compute power in the vehicles because of customers need for additional services and features. Electrical architectures in current vehicles are already quite complex with hundreds of electronic control units (ECUs).
“For each and every little functionality, there’s a single ECU,” he said. That’s about to change drastically as OEMs move to a domain-based architecture with high-performance computers.
Mertens said OEMs are moving toward one or two large computers being responsible for all functionality inside the vehicle, which is a “drastic change” that will include software-defined architectures, that have a tremendous impact on the semiconductor supply chain.
Stephen Kosonocky, a senior fellow at Uhnder, said the broader computer industry provides some guidance as to where vehicle architectures are headed. “The computer industry is really moving more toward domain-specific architectures, where you want to maximize the performance and power efficiency for specific compute domains,” he said, noting that the automobile is seen headed in the same direction. “It will employ very complex systems-on-chip (SoCs) or systems-in-packages, much like a modern cell phone, mobile phone, laptop, or server system.”
These architecture changes combined with constraints in the supply chain will drive innovation, said Dean Bushey, VP and head of Hitachi’s environmental business division in the Americas. Chips will be built from the ground up digitally. “That changes the nature of the way we think about manufacturing.”
Data informs sustainability, functional safety, design
Sustainability is also going to be viewed through a new lens because of data, as the car now has so many sources that will inform optimal charging times and where charging stations are placed.
“You can’t just plug all these cars into the grid automatically and everybody gets charged,” Bushey said. “That’s going to have a big impact on an aging grid structure.”
The same data that’s making electric vehicles more sustainable is also making them safer, added Mertens, and it includes the health of batteries and electric systems, which in turn informs predictive maintenance on complex systems including autonomous capabilities and advanced driver-assistance systems (ADAS).
And, as Kosonocky noted, the automotive industry is increasingly embracing high performance digital processing techniques to implement compute-intensive algorithms necessary for advanced safety and automated driving features, as well as infotainment systems. “To maximize the performance of these systems for power efficiency and performance, the chips will continue to scale toward more advanced CMOS technologies, but as we approach the limits of CMOS technology, understanding and designing for manufacturing variations becomes much more critical.”
Lessons learned from the consumer and server computing sectors will help to address variations in next-generation automotive devices, while collection and analysis of real-time device data through the manufacturing processes is key for making improvements, he said.
Testing is crucial
What’s especially important for automotive chip design is the safety envelope, said proteanTecs general manager Gal Carmel: The chips must last the lifespan of the product, and data is crucial for understanding how products behave in the field in general.
He foresees a great deal of artificial intelligence-enabled testing, validation, and qualification to help close the feedback data loop and correlate designs and actual performance in the field, as well as data feedback for functional safety, which is in its early stages.
As vehicles move toward a domain-based architecture rather than hundreds of separate ECUs, understanding how well electronics content is behaving is even more important, Mertens said.
“If the seat heating has a problem with its tiny ECU, it’s not going to be a big problem. It’s still annoying, but you don’t have any safety issues with it.”
But when systems are condensed into big ECUs and compute, there’s bigger risk that a small failure can cause enormous trouble and become a serious safety issue, he added. “That’s not only for electrification; that’s all the features that are coming in the development, starting with ADAS.”
The more autonomous the vehicle, the more important it becomes to predict and monitor degradation to inform predictive maintenance.
The intelligence that enables a vehicle to execute predictive and prescriptive maintenance will touch everything from the tires to the chips—essentially all the components that are going on in the car, Bushey said.
And it’s about not only safety but also business. “From an operational business perspective, if you have predictive and prescriptive maintenance that is spot on, then you can actually optimize your flow and minimize the downtime of your fleet and your vehicles,” he said.
OEMs want a direct line to automotive design
Part of the disruptive phase that automotive design finds itself in today is not just technology changing significantly, but also business models, Mertens said.
“The relationship between suppliers and OEMs are changing drastically.” That includes more OEMs taking responsibility in critical areas—even developing their own silica and working directly with the big semiconductor companies.
Tesla is an example of a company that’s done just that, he said. “They have a completely different vertical integration. They do everything that matters to them themselves. They do it with partners, but they want to understand the details. This is going to be a model which the traditional OEMs will adapt to.”
Kosonocky foresees higher integration of functionality on a single chip or module, and a reduced dependence on many different suppliers for a given system to work: This will not only improve the device reliability, but also the supply predictability by focusing on fewer parts in the system.
At the same time, the industry trend of democratization of chip design, where smaller companies have much higher capabilities with less people to do design, will allow for different IPs to be integrated into a single system and chip efficiently. Kosonocky said this allows the automotive industry to make use of all the technologies and techniques the computer industry really drove over last three or four decades.
Supply chains are going to evolve, too, especially in the wake of recent disruptions cause by the pandemic and war, Bushey said.
The “just in time” model for manufacturing and purchasing minimizes cost, but when the supply chain is disrupted by something like a conflict or a pandemic, there’s no supply on hand locally and no way to produce locally, either. “We’re headed in the right direction as far as increasing the capacity to build chips or have fab plants here in the United States. But it’s going to take a while.”