91³Ô¹ÏÍø

AutoSens Detroit 2024: Automotive EE Architectures - Quo Vadis

Frank Schirrmeister

Jun 28, 2024 / 7 min read

In late May 2024, key automotive ecosystem players representing OEMs and Tier-1 and Tier-2 suppliers came together at the Huntington Place Convention Center in Detroit for the AutoSens conference, dedicated to vehicle perception and sensing. Co-located this year with the InCabin conference, the three-day program covered everything from hardware to software developments, research into AI, computer vision, image processing, and other safety challenges, including validation and simulation strategies for ADAS and autonomous vehicles. My personal interest and focus were on the evolution of automotive electrical/electronic (EE) architectures, for which I had the privilege to moderate a panel with Rivian, Bosch, Semidrive, and Valens, as well as the future market drivers, specifically from a sensing and data transfer perspective. The future impact on automation capabilities for architecture analysis, optimization, verification, and validation, will be profound.

Automotive EE Architecture Evolution

Naturally, the evolution of EE Architecture was front and center at AutoSens as well. In the run-up to the conference, I discussed the race toward centralized computing in ¡°¡± While there seems to be general agreement that we eventually will reach centralized architectures, the path to get there is the topic of intense debate. With AutoSens and InCabin focusing on perception and sensing to enable future autonomy and driver safety, much of the discussion centered around the usage and the balance of cameras, radars, lidars, and thermal sensing.

Sensor content increases with each advance in autonomy level

Sensor Content by Autonomy Leval (Source: S&P Global, Brock Walquist)

, Senior Analyst at S&P Global, presented the chart above. It nicely illustrates the challenges of sensing and perception in EE Architecture. As the industry progresses across the various levels of autonomy, the overall number of sensors per vehicle is expected to grow to 29+, with the resolution of cameras and radars increasing 12-fold to 12MP and 192 channels, respectively.

A presentation by Yole¡¯s Pierrick Boulay detailed the impact of all this on the semiconductor industry as an enabler. Yole estimates that in 2022 advanced cars already had about 900 semiconductor devices (as shown in one of Pierrick's slides):

  • ~ 50 logic devices for MCUs in powertrain, infotainment and ADAS (130nm to 28nm)
  • > 100 opto and sensor devices for MEMS, magnetic sensors, LEDs and image sensors (300nm and 65nm) technologies
  • > 700 power and analog components (450nm to 40nm)
  • ~ 20 memory components for infotainment and ADAS (40nm to 7nm)

Pierrick also outlined the changes in the automotive design chain, likening the ongoing transformation in automotive to the previous transformation the electronics industry experienced in mobile and data centers. He discussed how the co-development strategy used by Apple with TSMC can apply to automotive, and the chart below nicely illustrates the already ongoing changes in behavior.

 

strategies to source devices

New Behaviors in the Automotive Design Chain (Source: Yole)

All of these transformations as previously seen in other industries like mobile, make for some fascinating times in automotive and definitely a golden age for semiconductors as change always offers new oportunities for all design-chain participants!

Debating Changes in the Automotive Design Chain

Later in the day, I presented all this for further debate in a panel I moderated. We were very fortunate as the panel represented the full design chain with an OEM ¨C Rivian ¨C a Tier 1 ¨C Bosch ¨C two semiconductor vendors ¨C Valens and Semidrive ¨C and I brought the tools and semiconductor IP perspective to the panel.

Rivian¡¯s Sr Staff Functional Safety Engineer, , identified the three top challenges facing OEMs in this rapidly evolving landscape. First on his mind are cost and complexity reductions of the electrical architecture, i.e., reducing the number of ECUs and the complexity of the wiring harnesses and connectors, all influencing weight. Given that they are the owner and the direct interface to consumer experiences, enabling new features and upgrades over the vehicle's lifetime are top issues, too. Customers expect their vehicles to receive new features and upgrades regularly, similar to smartphones. Safety and security are the third big area of concern. Lives are at stake, so safety is a must. As vehicles become increasingly software-driven, cybersecurity vulnerabilities pose significant risks, and OEMs must design the electrical architecture to support this, which requires many innovations. On top of that, standardizing and complying with cybersecurity and safety requirements within OEM organizations becomes even more critical, and also extend to the development chain to prevent potential attacks.

, Sr. Manager Product Management at Bosch, emphasized the coordinating role of the ¡°middle tier¡± in the design chain. Tier 1s have to manage diverse OEM requirements as each OEM has different approaches to E/E architecture design, implementation timelines, and quality assurance standards. Flexibility and adaptability are critical to balance the varying requirements while developing products. On the other side, Tier 1s need to balance OEM requirements with supplier capabilities and roadmaps of the Tier 2s and Tier 3s, including semiconductor manufacturers and software providers. This involves understanding the data requirements, bandwidth, and performance needs to propose the right solutions to OEMs. Thirdly, Auston outlined how Bosch leverages industry insights to help guide OEMs. Working with multiple OEMs and suppliers, Bosch has a comprehensive understanding of industry trends, possibilities, and challenges and uses these insights to guide OEMs toward better solutions that balance cost, scalability, and performance while maintaining a competitive edge. Finally, supporting a mix of centralized and distributed architectures poses a unique challenge. While the industry is generally moving towards centralized architectures, OEMs are at different stages of this transition. Tier 1s must support centralized and distributed architectures, ensuring that their products integrate seamlessly with the OEM's chosen approaches.

Representing the Tier 2s on this panel, , Director of Automotive Customer Support at Valens, and , GM, Global Automotive at SemiDrive, added to the challenges from a semiconductor perspective. Oran described how Valens, as a provider of high-speed connectivity solutions, such as MIPI A-PHY, focuses on balancing performance, cost, and power consumption. They also need to ensure safety and reliability, ensuring that their products meet the stringent automotive safety standards and can operate reliably in harsh vehicle environments. SemiDrive¡¯s Eugene echoed the challenges to balance performance, cost, and power for their   CPU, GPU, and AI computing solutions and added safety to the mix. He also pointed to the rapidly evolving automotive industry in China, which poses its own challenges to meet the unique requirements of Chinese OEMs and Tier 1s. Both also emphasized the importance of partnerships and collaborations across the ecosystem. For instance, successful high-speed connectivity solutions involve collaborating with other semiconductor vendors, Tier 1s, and OEMs to ensure widespread adoption and compliance. Generally, with accelerated automotive development timelines and co-optimization of technologies, the industry shifts towards a more collaborative development model and requires all players to effectively manage partnerships and collaborations between OEMs, Tier 1s, and other semiconductor vendors by sharing information, aligning roadmaps, and jointly developing solutions that address the needs of the entire ecosystem.

Outlook and what¡¯s in it for me?

It was great to hear this, as it directly motivated several challenges in my daily life at Synopsys. As tools and IP providers, we must make development decisions for features in development tools and IP years in advance. For instance, my colleague , Sr. Segment Manager for Automotive at Synopsys, discussed during AutoSens how to incorporate ADAS into the IVI domain using the example of our customer SiEngine, an extensive user of Synopsys IP, in this case addressing the industry trend toward combined ADAS and IVI? silicon combining two chips into one, using a Synopsys NPU powerful enough for ADAS and parking on-chip? functionality. And of course, tools play a critical role too. The automotive industry is fully embracing virtual prototyping and digital twin approaches to development. These methodologies allow for early validation and testing of automotive systems, including software, before physical prototypes are available. They are connected to hardware-based development techniques using emulation and FPGA-based prototyping, allowing for scenario testing and at-speed interface validation.

These development flows facilitate the interaction in the design chain between the various players¡ªOEMs can provide software workloads that stimulate the subsystems and devices provided by Tier 1 and Tier 2 providers on pre-silicon engines before actual physical prototypes are available. We are shifting more and more left, bringing hardware, software, and mechanical, aspects together!

And where do EE Architectures go from here? The lightning round at the end of the panel was quite insightful. The top three challenges of the industry remain new centralized and zonal architectures, software-defined vehicles & over-the-air updates, and new behaviors leading to collaboration and co-development across the ecosystem. While the details of future EE Architectures are fast evolving and timelines differ depending on the OEMs, they can only be implemented with closer collaboration and co-development among OEMs, Tier 1s, semiconductor vendors, and other suppliers. This involves a move away from the traditional hierarchical development model towards a more collaborative approach, where stakeholders work together early in the development process to define requirements, share insights, and jointly develop solutions. This collaboration extends to developing standards, tools, and IP to ensure interoperability and faster time-to-market.

It's a great time to be in automotive Semiconductors.

The EE Architectural Panel at AutoSens in Detroit

The EE Architecture Panel at Autosens in Detroit, from right to left: , GM, Global Automotive at SemiDrive, , Sr. Manager Product Management at Bosch, , Director of Automotive Customer Support at Valens, , Sr Staff Functional Safety Engineer at Rivian and Frank Schirrmeister, executive Director, System 91³Ô¹ÏÍø at Synopsys.

Continue Reading