Automotive System Design Challenges

New standards and complex ecosystem leaves companies scrambling to fit in.

popularity

The automotive semiconductor market did exceptionally well last year. IHS reported strong vehicle production growth and increased semiconductor content in 2014, and that trend is likely to continue with semiconductor revenue for the automotive segment to reach $31 billion this year, up from $29 billion last year.

The market research company affirmed the fastest growing segments for automotive semiconductors are hybrid electric vehicles, telematics and connectivity and advanced driver assistance systems (ADAS).

Interestingly, the cost of electronics and software content in autos was less than 20% of the total cost a decade ago. Today it is as much as 35%, according to studies by Manfred Broy, a professor of informatics at Technical University, Munich. More importantly, electronics systems continue to contribute more than 90% of innovations and new features.

This strong growth is having an impact on the design of those systems, to be sure, but how that looks depends on where in the car you are at, said Frank Schirrmeister, group director for product management and marketing in in the System and Verification Group at Cadence. “What that means is that there are different components in the car — powertrain, safety, body electronics, etc. — that make up a big part of the overall electronics market in those three. Then there is driver assistance, infotainment and aftermarket.”

He noted that the design chain in automotive is slightly more complex compared with that of a phone. The phone maker is going directly to the semiconductor company to get the chips, but in the automotive space, there is a tiered structure of suppliers.

The automaker sits at the top of the tree as the OEM. Below that in Tier 1 are the large automotive systems suppliers such as Bosch, Continental or Delphi. Below that is the tier of semiconductor suppliers that includes Infineon, Freescale, Texas Instruments, ON Semiconductor, Micron and others. IP providers such as Cadence, Synopsys, Sonics, Arteris and others are also selling into that space.

“The requirements and the design aspects go all the way through, so the OEM may define the requirements they need for Ethernet and that may ripple all the way down to the IP provider,” Schirrmeister said. “So if you look at our IP, essentially when we defined our Ethernet IP to make it applicable to cars, that goes all the way up from the OEMs.”

Complex Automotive Design from Semiconductor IP to the Connected Car (Source: Cadence)

Complex Automotive Design from Semiconductor IP to the Connected Car (Source: Cadence)

Complex ecosystem
The automotive sector has some unique business and financial dynamics. “Depending on the area of the car you are in, there are different timelines and different functional safety aspects,” Schirrmeister said. “There is an interesting dynamic between the players involved. Taking the IP providers out, let’s assume the semiconductor owns it all. In automotive there are things like second sourcing that play a big role. They actually don’t want to rely on only one piece of silicon. They want to have a replacement just for redundancy purposes, but also to negotiate price.”

On top of that there are standards such as AUTOSAR, a set of APIs that makes it more difficult for Tier 1 companies to negotiate with OEMs because the OEMs define the API functionality, which is essentially the software. “And then the semiconductor provider provides the abstraction layer for AUTOSAR to the which the OEM writes, so now the Tier 1 is squeezed in between, which makes for interesting dynamics in terms of who owns the power,” he added. “In that area of the car, the OEM definitely writes the applications toward the AUTOSAR APIs. Then, in principle, everybody who supports that version of AUTOSAR in their chips could replace anybody else.”

Looking into other areas such as infotainment and vehicle-to-vehicle communications, or complete driver assistance, this is becoming more like the mobile space, he said. “First of all, you can’t have the old timelines of several years because when I buy a new car, my touchscreen needs to work for the next five years, at least. In that domain, GENEVI comes in, a specific port of Linux for automotive applications.”

Kurt Shuler, vice president of marketing at Arteris, agreed that new requirements are impacting system design. “It’s actually a big issue because there are the traditional automotive semiconductor suppliers that have been in the business for many years. In the past, from the digital logic side, they were providing MCUs, maybe 32-bit on a 65nm process, and they’ve really locked down the security stuff for the engine control units. They really understand the certification processes, and dealing with the other suppliers in the supply chain really well. However, they’re not experienced at doing things that are computationally intensive and have graphics, and the kind of the stuff that was proven in mobile phones showing up in a car.”

Automotive is, to some extent, the place where companies that didn’t succeed in mobile phones have migrated. The products and architectures they are developing are ending up in cars, but that’s well outside their normal sphere of operation. Their familiarity with the automotive ecosystem is limited.

“For companies outside the automotive industry, who are now doing these complex chips, it’s 30% to 50% more R&D cost to target automotive than it is for a mobile phone. It remains to be seen how well those chips are going to hold their average selling prices, and the volumes are lower than a lot of consumer electronics. But the benefit to these companies is that once you’re in, the OEMs don’t change that quickly.”

Bill Chown, product marketing director at Mentor Graphics, observed the challenge of the automaker today is that they are bringing together subsystems, software, IP, semiconductors within the context of the vehicle. “It is a monumental challenge and I’m seeing them struggle with it. They’ve bitten off a big problem space and weren’t structured to deal with it effectively, so they are struggling to keep up with their own ambitions and what their customers demand.”

In terms of making all the pieces work together, he believes they have a way to go. The industry has a long history as a metal-bashing kind of production process. “You wind back from, ‘I’m going to put all these pieces together so I need the manufacturing bill of materials that leads that manufacturing.’ The pieces have to be put together and sourced to do that, and somewhere back at the beginning of this sequence there’s, ‘Oh yeah, we’d better put some design into this.’ But it isn’t a design-driven process. It’s a production-driven process.”

A lot of the implementation, tooling, and the way people are going about doing stuff still builds up from that, Chown said. As a result, there is a more bill of materials-oriented thinking versus the function the customer wants.

“The customer wants navigation. Well, that’s not a bill of materials item on its own because they want the navigation to know things about the car, as well, so it’s got connectivity and integration with parts of the system. ‘I could go out and source a navigation gadget, and the chips that go in that navigation gadget will be supplied by the vendor who provides it. But, oh, this isn’t what the customer wanted.’ The customer wanted an integrated navigation system.”

To make that happen, the OEM or Tier 1 needs to design and specify that rather than specify a navigation unit. And that’s not the way those industries have operated in the past. “Now it’s no longer a sourcing problem,” Chown said. “It’s really a system-design problem, but I find myself still interacting with those people through the sourcing department. And so there’s still underlying history that structures the way things are happening and there’s still a way to go.”

The big picture is that many of the automotive OEMs are realizing that they can’t rely on the Tier 1 companies to the extent that they used to. They have to put in more of their own resources to understand design and integrate several packages that come from Tier 1s.

“The OEM is having to bring in their own designs. They’re having to deal with the fact that they no longer produce lots and lots of the same thing. In the general case, they produce lots and lots of variants of similar things, and that’s changed the way that they buy stuff. That has changed how they are specifying what they’re buying. They’ve got to be able to deal with variants in a way that is much more pervasive than it once was,” Chown added.

And they have to own this because ultimately they are responsible. It comes back to the top of the tree. The automotive manufacturer is the only entity that the public or the government is going to turn to when something goes wrong.

Chown said the automotive OEMs understand the need for system-level design and the need to be able to describe functions, evaluate those functions with each other, and then allocate those functions to implementations. “That requires a different kind of toolset from the tools that implement silicon or printed circuit boards or mechanical enclosures, so we are seeing more interest in those kind of solutions. For instance, they want to do system-level modeling using XML, with their spin on it. But this can come with its own issues as it lends itself to being a little island of automation in itself, and forget about the fact that its going to need to link to the tools down the flow that do the implementation. OEMs may have had their own tools for this that they can no longer afford to maintain, so they are going out to the market for tools and trying to figure out how it fits with in house work they’d been doing. There are interesting challenges in those sorts of spots moving from point tools to a process flow.”

This is currently a work in process.

The software side
Driving the inclusion of the sophisticated features in vehicles today and going forward, are emissions reduction, safety, reliability and connectivity, according to Marc Serughetti, director of business development for system-level solutions at Synopsys.

“Those drive a lot of what is happening at the system level in the automotive sector because with those four trends, the role of electronics is critical. Of course when you talk about things like emissions, there are going to be things related to material, etc., but it’s considered that embedded software, for example, in the powertrain has a very significant role to play when it comes to reducing emissions. What this means in terms of electronics is that it’s not just software—there’s obviously more hardware content that’s coming in all of this—but hardware by itself is not enough. The software brings intelligence on the top of the hardware. With more hardware and more software, the real challenge is that with all of these systems in the car are connected to each other, they are all interdependent,” he explained.

As far as software content in a midrange type of car, there are more than 200 software tasks that have to be done today, Serughetti pointed out. “Then, you’re talking about more than 500 bus signals to be taken care of. And then if you look at the hardware architecture, you’re starting to look more at 60+ ECUs in the car, multiple networks that brings all of those things together to communicate properly.”

Interdependencies between the systems are the responsibility of the OEMs, which also are bringing more complex design tasks back in-house for a variety of reasons. An example are new systems such as advanced driver assistance systems (ADAS), which OEMs can be a market differentiator. As a result they develop the IP in-house for competitive reasons.

“ADAS is essentially a brain inside the vehicle that makes all the decisions for the driver,” said Serughetti. “Therefore it is more than just a simple subsystem. It connects a number, if not the majority, of subsystems in a vehicle. The interesting part is that the solution is not a solution that the OEM has to figure out. It’s a solution that the industry has to figure out. There’s a supply chain in all of this…and how all of those companies are going to be able to work better together. In the past, many issues could be addressed by throwing more engineering resources at the problem, but we’ve reached a point in the complexity where this is not a solution anymore.”

Drew Wingard, CTO of Sonics, said it will be interesting to see how reliability issues get worked out. “The most commonly deployed techniques for enhancing reliability involve redundancy, so as I reduce supply voltages to save power, I introduce more sources of dynamic error because I’ve lost margin. There is an interesting question about whether I can reduce it so low that I can tolerate those failures by having redundancy. I don’t know that the automotive guys are going to want to be the guinea pig on that, but people like ARM already ship versions of their Cortex-R series processors that can be run in a lock step mode where you basically burn a whole second core whose whole job is to check the operation of the first core. They both execute the same program, and there’s some circuit that determines whether they came up with the same answers. If they don’t come up with the same answers, then you basically halt and catch fire because really the kinds of resilience that’s being built into these systems isn’t about recovery from errors, it’s more about detection of errors, so you don’t let errors compound and have an electronic system make a bad decision. Could I run those processors at a much lower voltage, where I’m going to get a higher error rate but I’m going to assume that I can recover and come back to run that piece of code? Maybe. I think we’re a ways away from that, though.”

Conclusion
As the automotive OEM pulls design and subsystem integration in-house, EDA may have an important role to play given its strengths in simulation and modeling technologies and the potential use in fault-testing and reliability in the automotive space—for hardware as well as software. The methodology of automotive development needs to change and evolve, and new tools need to be deployed to address these issues. Where better to apply the decades of EDA expertise and technology?