DAC Sunday – For Networking in Cars, Ethernet has an Edge

No, not social networking in cars. I’ll leave that for a different time … This is about data and control carrying networks in cars and where they are going. Yesterday I attended here at DAC the Sunday workshop on “Intra and Inter-Vehicle Networking in Automotive: Past, Present, and Future”. It seems like Ethernet has won the battle, albeit not for all areas in the car.

First of all, this workshop was well organized, very productive and interesting – a big thank you to Paolo, Arko and Haibo for putting it together, and of course Alberto Sangiovanni Vincentelli for his championship and guidance.

What networks in a car you may ask? Most of us have heard of CAN, but there is so much more as I learned in the run-up to this event. In this post here a picture from Renesas with some annotations from our team. There are really five busses to look at today:

• CAN – most spread network in the car, some limitations with 1 Mbps bandwidth and non-deterministic behavior under high load >60%.
• LIN – low cost bus for body applications with 19.2 Kbauds and a UART interface
• MOST – designed for multimedia using optical fiber with bandwidth up to 150 Mb/s
• FlexRay – high performance (10 Mbps), deterministic, and secure network, mainly used in X-by-wire, ADAS, and high performance applications
• Ethernet – mainly used for diagnostics today, high potential for more

The morning session of the workshop did the setup from the user side. Alberto Sangiovanni Vincentelli started the day off with an insightful key note on the mega trends. He also pre-counted Ethernet as “winning 5:1” when previewing the upcoming presentations of the day. One key take away of his key note was the trend to design which fully separates function and architecture, which enables OEMs to ‘sandwich” the Tier 1 suppliers more by overtaking a more significant portion of the functional design in software, which then can be mapped into existing ECUs if there is enough performance left. Alberto was adamant that safety critical areas should be separated fully from infotainment like video and audio, safety being the main driving issue.

Harman International continued with an insightful presentation on the AVB extensions for Ethernet – soon to be fully standardized, followed by Prof. Huss from the University of Darnstadt describing simTD, a field test in Germany for Car-to-X applications with very interesting trial results. Raj Rajkumar showed simulations from CMU’s trials on timing guarantees in Vehicle to Vehicle (V2V) networks. What I found most interesting was that CMUs simulation showed that to be effective, only 7% to 8% of the cars need to be equipped with V2V communication. Raj also showed an App, which self-parks Boss, the  autonomous driving car which won the Darpa challenge. This certainly will come in handy during Christmas shopping when it is available.

National Instruments complemented the presentation from Harman with more details on AVB for Ethernet, TTTech gave a great overview of the Ethernet extensions to make Ethernet ready for timing critical and fault tolerant use. Austriamicrosystems did present a future which will see a symbiosis for in-vehicle networks.

After the morning sessions had kicked off with the design issues around Ethernet and left me with the impression that Ethernet is winning for most areas, I kicked off the afternoon session with an overview of challenges and solutions spanning from AutoSAR based architecture analysis, through signal integrity issues when laying out the network, though Ethernet IP and software development for testing. Mentor followed with their AUTOSAR and harness related offerings, Cadence focused on Ethernet IP. Further design solutions for networking related architecture development were presented by aquintos/Vector and Mirabilis, Symtavision closed the day with an overview of migrating from CAN to FlexRay and Ethernet using their formal techniques for timing analysis.

Most valuable was the Q&A. I bluntly asked whether Ethernet is winning or not and the answer was “mostly”. Ethernet definitely has a bright future for video, audio and infotainment in general, but also for more timing critical applications. It’s only downside seem to be cost and EMC issues. FlexRay is definitely not dead it seems, as it used for some timing related applications still. I also asked what the next bus will be which we tool providers need to prepare for. One answer pointed to wireless protocols as next frontier, which makes sense, as weight of cables is a constant target for cost and fuel reduction. Finally I asked who will actually own the other side of the vehicle to infrastructure (V2I) networking as somebody needs to manage traffic and all the data collected. This seems to be an unclear issue and the parties involved in current field trials mentioned that the OEMs need to get together to stimulate that portion. Lots of government and regulatory issues are at play here. It might be easier to create this infrastructure in areas like China’s future megacities, which are just being  planned and built.

Overall a great workshop with interesting insights. Ethernet definitely seems to have a bright future in cars!