The Automotive Ethernet Congress in Munich is an important and quite technical event for experts in vehicle networking that want to be informed about the latest developments and meet other industry leaders. The congress welcomed not only European visitors but also a significant number of Asian and American representatives. Its rising popularity reflects the imminent inflection point in the electrical architecture strategy after years of stability and in general peaceful coexistence of multiple buses.
Although German OEMs and Tier1s are the main proponents -BMW was the pioneer introducing 100Mbps Ethernet based on Broadcom´s BroadR-Reach more than 5 years ago-, other global manufacturers such as GM, Hyundai, Jaguar Land Rover, Tesla and many Tier1s have joined the cause since then. In the midst of a major ECU consolidation trend, the benefits of having a common protocol stack for the vehicle backbone that can deal with the enormous data throughput and safety requirements expected in connected and highly autonomous vehicles prompted the industry to create standardizing bodies like Avnu, the OPEN Alliance SIG and the NAV Alliance, which collaborate with the IEEE -in charge of Ethernet physical layers-. They complete the ecosystem with automotive interoperability testing specifications and harnessing requirements, pushing for faster adoption and contributing to the shorter development times of the modern automotive industry.
Safety is critical for autonomous cars, but it is not possible without security. As Michael Johnston from NXP stated, there were last year more malign vehicle attacks than white hat activities to detect weak spots and prevent breaches. He proposed a defense-in-depth approach with multiple layers of security involving external interfaces and communications, domain controllers, etc. Fred Rennig from STMicroelectronics also addressed the security problems related to the evolution from a domain topology to a distributed software or service-oriented architecture. OBD and OTA are known threats, but the infotainment domain is still the main attack surface with several wireless and wired entry points. Hardware solutions like HSMs are very powerful but not that flexible. Thus, key distribution and storage, data authentication and software integrity over the whole product lifetime are critical issues as well. Bob Leigh from RTI (Real-Time Innovations) emphasized the importance of having a robust framework and introduced the DDS security specification, which is now part of the AUTOSAR Adaptive software stack.
Aquantia, Valens and Dryv.io were among the silicon startups pushing the speed limits and showing that Gigabit Ethernet can be a reality in vehicles soon. They challenge and sometimes also collaborate with traditional networking IC vendors such as Broadcom, NXP, Microchip, Marvell or Renesas. The supply chain complexity is growing -like in other automotive domains- and, although there is still uncertainty about which solution or combination of technologies will dominate in the long term, one thing is clear: high-speed Ethernet is here to stay.
However, not everyone was focusing on Multigigabit data communications. As Piergiorgio Beruto from Canova Tech stated, since 90% of the ECUs in current cars use less than 10Mbps, a cheaper 10BASE-T1S network could bridge the gap, replacing legacy low speed buses without the need for additional gateways -the leaf nodes would integrate simple PHYs and natively support Ethernet-.
The most heated discussion revolved around optical vs copper transmission lines. Björn Bergqvist from Volvo highlighted with detailed laboratory results the robustness of optical based 1000BASE-RH against UWB (ultra-wideband) electromagnetic pulses. Ryohei Kawabuchi and Daijiro Yumoto from JASPAR (Japan Automotive Software Platform and Architecture) also shared their conducted and radiated emissions tests results, revealing the EMC superiority of the RH version against unshielded twisted pair. As KDPOF explained at their booth, optical fiber helps reduce weight and even if the PHY chip and connectors might be more expensive, the total point to point cost, including harnessing, would be lower. However, going beyond the backbone, a copper cable would be needed anyway to deliver power to the sensors or cameras at the edge. Besides, many still argue that qualifying optical fiber solutions for high temperatures might be an issue.
In short, as Martin Hiller from Volvo brilliantly demonstrated with his opening note and the audience confirmed during the online polls done by Kirsten Matheus from BMW on the second day, a standardized and scalable Ethernet network can reduce complexity and even costs for car manufacturers in the era of software-defined cars. Most of the attendees consider that proprietary solutions should be avoided, think that CAN is still popular mainly because developers are used to it and not due to its lower cost, and believe that Ethernet will dominate the in-vehicle architecture and change the way cars are developed. Time will tell if Ethernet can replace FlexRay, MOST, LIN and even CAN, becoming a de facto standard not only for backbone and sensor networks. In the end, as Nicolas Morand and Antoine Guittet from Groupe PSA pointed out during their interactive and useful closing session, there is a fundamental change of paradigm between CAN and IP networks that the automotive industry will need to confront, probably using a mix of automotive specific and mainstream IT protocols.