Designing For Automotive

As cars include an increasing amount of electronics and electronics subsystems, the number of design challenges involving reliability, cost and power are on the rise.

“Reliability tops the list of concerns for the design team because when you put these electronics in, you must know if they are going to operate efficiently by themselves,” said Aveek Sarkar, vice president of product engineering and support at Ansys-Apache. “And when in the system, will they work properly? Imagine you are buying an off-the-shelf mobile processor to design into your car’s dashboard. You go through somebody who has a state-of-the-art mobile processor that’s powering the smartphone or tablet, and you decide to use it. However, those are designed for a smartphone, which is tolerant if you’re trying to take a picture. If it’s slower, it’s not a deal breaker. But imagine you are trying to back up in your car. You go into reverse but because your car was sitting at 105 degrees ambient, the electronic component that was designed to operate between 55 to 200 degrees optimally slows down outside of that temperature, so when you go into reverse and you expect the rear camera and radar to kick on within a second, it takes five seconds. If you’re so used to the radar and it’s not working as usual, it could be a catastrophic situation.”

It gets even more complex when the embedded core is included along with the electronics, as there are many areas where failure can happen, Sarkar explained. Consider what happens with different antennas, for example. In most cases what designers do is encapsulate the chip within a strong container to try to contain the electromagnetic interference. The outside emissions of that electromagnetic signal from the chip are discontinued but come back and hit the chip itself.

“If you are a chip, the amount of energy you are creating doesn’t go out and harm your neighboring chip on a different part of the car, but it gets reflected back and can affect you. This phenomenon is called electromagnetic susceptibility. Chips are not designed to be electromagnetically immune. You design it for performance, you design it for timing — you don’t design it because the same chips are going through all this mission-critical applications, so now you have to start worrying,” he said.

Further, he pointed out, if a 28nm or possibly smaller node chip is being used in a car, problems such as electromigration and ESD that have been solved for mobile now have to be solved at a higher criticality level. Antennas and modems that are all talking to each other, along with their interaction, are also a concern in automotive design. “There will be scenarios where there is conflict. How do you predict these scenarios before the system is manufactured because it’s too costly? This has to be done in a software space,” he added.

Steve Carlson, group director of marketing at Cadence, agrees that reliability is paramount and is tied closely to safety due to heightened concerns in these areas. “There are safety features for the driver so you see a lot of failures and recalls going on. That’s been a heightened concern. Things like ignition switches and airbags have been more mechanical in nature, but everybody’s sensitivity is heightened nonetheless. The cost of recalls elevates the importance of reliability, and there can be safety factors as well. But definitely if there are dollars involved either way in that case, it gets the executives’ attention and pressure gets put down through the engineering organization.”

And given that more than 50% of semiconductor failures can be traced back to elevated temperature, he said, whatever can be done to reduce that is going to improve the reliability of the system, and that’s going to save money in the long term versus recalls.

“Locally you see, because of the amount of electronics, the advantages of going down to 28nm. In some cases there are projects below 28nm now. All of the advanced low-power features that you’ve been hearing about for cell phones are going to be applicable on the automotive targeted applications,” Carlson predicted.

What exactly that means for functional safety is a big question. “How will the system react and be able to model and have a documentation trail showing that you’ve analyzed this and have anticipated the more likely failure scenarios,” Carlson said. “There are these circuit-level things you can do to mitigate and there’s also the architectural. You see a lot more people talking about fault tolerance architecture and redundancy in those kinds of things. That’s all good, but when you start talking about power and cost in those things, it’s a pretty big hit that you take there. There is a middle ground where people are talking about what they can do to not be resilient in the face of faults, but fail gracefully. The car is not going to run off the road but you shut down gracefully with enough time to pull over to the side.”

In the same vein, as far as cost goes, Sarkar questioned, “When you are designing for the next generation smartphone, the cost may not be that much of an impact versus for a medium-end car. How do you optimize? Just optimizing the chip by itself is probably going to be very difficult. You have to optimize across the system.”

But what happens in the automotive supply chain is that much of the work is done in isolation. People who design the chip frequently do it by themselves. So do the people who design the power train control system or infotainment control system. And then there are the auto manufacturers.

There are three tiers of electronics suppliers. At the top are the Nvidias and the Qualcomms, Infineons and Renesas, which are designing the chips. Then there are system integrators who build the systems, such as Continental, Bosch, TRW. The car manufacturers, meanwhile, tend to treat this as off-the-shelf business. They buy the product. And because this is the way it has operated for many years, simulation is not really part of the core design methodology, and most of the testing gets done in the physical prototype.

Designing today’s autos involves mechanical, fluid, electromagnetic interactions in one environment. Going forward, data will have to be shared across platforms in order to optimize the design across those platforms.

Power management/power optimization
Of course, putting so many high power chips together — think of supercomputer in a car — burns a lot of power, so energy efficiency becomes a key aspect of the system, Sarkar said. Here, the design team must make sure the chip is not wasting any power by design for any particular operating mode.

Anand Iyer, director of marketing for the low power platform at Calypto views, views this power as somewhere between the mobile space and the laptop space. “Usually the time of operations for your mobile device is somewhere in seconds to minutes, while laptops can go from hours to days. Cars are on minutes to hours. The power requirements may not be that similar at first glance because there is a bigger power source in the battery or even in the generator. The bigger picture is essentially the fact that there are multiple systems interacting simultaneously, and there is a lot of information transfer that is happening.”

In order to manage these situations, traditional techniques such as power gating really don’t apply to the automotive space, he said, so the power efficiency needs to be built in from the ground up. “This means using high-level synthesis or using power optimization within the design will become more and more important in this scenario.”

Automation expansion
Along with reliability, cost and power, automation in the automotive sector continues to expand, according to Bernard Murphy, CTO of Atrenta. “I heard at CES this year that even low-end cars are going to have 60+ controllers, and that there will be a wire to virtually every piece of sheet metal on the car. Additionally, all of this stuff is networked, so low-power implementations will be very important. For example, Broadcom recently announced a low-power Ethernet controller for car networking. But you have to remember that automotive applications have long qualification times, so none of this is likely to be FinFET. The Broadcom controller was done at 40nm.”

From an overall perspective, automotive electrification is directly impacting the power requirement in cars, noted Marc Serughetti, director of business development for system-level solutions at Synopsys. At the same time, car intelligence is relying more and more on processor-based electronic systems, which require a higher level of performance and as a result consume more power. Increasingly complex MCU and SoCs are being designed for applications such as powertrain, safety, advanced driver assistance systems and infotainment.

“With more complex MCU and SoC architectures, it has become even more important to understand if the performance and power requirements can be achieved very early on in the design cycle. One trend we expect in the coming year is the need for system architects to explore, optimize and validate MCU and SoC architectures from the perspective of both performance and power, to avoid under- or over-design. Architecture design tools and support for standards with a dedicated system-level power modeling specification (UPF 3.0 release in 2015) will help architects discover and remedy system performance problems related to power much earlier, preventing schedule delays and costly device re-spins,” he concluded.