The automotive industry does not follow traditional semiconductor design approaches and that makes even approaching the design that much harder.
To a large extent, the automotive industry does not follow traditional design approaches that are seen in the semiconductor industry, mostly because of the way the automotive ecosystem is structured, combined with a level of complexity not seen even in products like smartphones and other highly sophisticated consumer electronic devices.
Thomas Heurung, manager of technical sales teams in Europe and India at Mentor Graphics, said that from the perspective of the whole vehicle and not considering any of the ‘system of system’ connections it has, car OEMs come from a mechanical background, so they know how to specify doors and make the car look good and make engines that burn fuel. “A lot of the design process is structured in a way that is easy to compartmentalize the different design aspects. It is easy to define something like a door based on its response to the rest of the vehicle.”
However, he said, “when you go into any of the other domains that add value to the vehicle, such as the advanced driver assistance or even a power window control, you can introduce problems into the overall function of a system in a totally different location of the system. You can’t compartmentalize this as such. Everything needs to play together in order to achieve the desired function. Now we are talking about an overall design that includes geometrical CAD aspects, so you have mechanical aspects to that. You typically have electrical aspects in terms of having motors and things like that, you have wires that somehow need to be there to distribute energy and information, you have software doing something, and that is all over the place. It is dynamically allocated to different locations in the vehicle because it is not limited to have a certain function actually implemented in a specific location.”
What this means is that carmakers have teams which think very differently about their design jobs, Heurung said. There are software engineers who talk about model-based design that basically have top-down design processes. Then there are the mechanical engineers who can scale their drawings by doing parametric designs. And there are the electrical engineers, for whom the geometry is important, but they’re looking into all kinds of dynamic behaviors which is sometimes hard to predict in the context of an overall system. You can imagine each group of engineers is speaking in their own programming language. And this may be the biggest challenge—just calibrating the numerous languages required.
“When I am talking about architecture, for example, architecture means something to everybody, and typically it means something different to everybody. Now how do you model that? What is the relevant part of the car door on the mechanical side for the behavior of the door for the aspects of functional safety, for example, that you need to consider on the mechanical side? A lot of that stuff is basically handled on the electrical and software side.”
“That’s the challenge in the automotive space. Many different domains need to be connected. And let’s say a change comes in — there must be a linear design process to cope with that. Conversely, semiconductor designs typically start off with a kind of architecture level design and you break it down. You typically don’t iterate between too many different domains. This is where the issues come up. By the time change has been communicated, a lot of things have been implemented that need to be assessed for what impact the change has,” Heurung added.