Experts At The Table: Automotive Electronics

First of three parts: Cost versus quality; working together; accommodating requirements.

popularity

By Ann Steffora Mutschler
System-Level Design sat down to discuss the opportunities in automotive electronics with Alexandre Palus, principal SoC architect at Altera; Aveek Sarkar, VP of product engineering & support at Apache; Mladen Nizic, engineering director, mixed signal solution at Cadence; and Stephen Pateras, product marketing director, silicon test solutions at Mentor Graphics. What follows are excerpts of that conversation.

SLD: What are the unique characteristics of working with electronics in the automotive space?
Palus: I’m an architect. I’ve been almost 20 years in automotive, pretty much since I started in TI France. What happened in automotive is that you have to come first with, can you do the function? And then, can you be reliable for enough time? And can you last on the road for 15 years? Then, how cheap are you? It’s not how expensive how are you, it’s how cheap can you do that? For me, that’s the three key points when you go to automotive.
Nizic: I see the shortening of turnaround time significantly. Before, it was assumed that there are three years until the next car. Designing the car took a longer time and then, of course, all the electronics and chips had a longer time to do that. Now we see that car manufacturers are trying to reduce the time significantly, and that’s putting pressure on all their suppliers. That obviously translates into electronic components and how much time they have to design, and then of course the chips, as well. To reduce the turnaround time, it’s harder now to look at your step—whether it’s system-level design or component design or chip design. You have to look at it entirely. There’s much more pressure on the entire ecosystem to work more closely together and to try to squeeze any budgets. When we talk about automotive today, it’s diverse because the reliability is one aspect that spreads through all of that. Even if it’s a chip in a radio, it has to be reliable because the cost of fixing it in millions of cars is expensive.
Palus: It’s not only that. When you get into your car, you turn it on, you expect it to work. When you have your cellular phone, your cellular phone can crash—and that’s ok. But if your radio in the car is damaged or you have to reboot it, you won’t accept it. So customer expectations are slightly different in the automotive business than in the consumer business.

SLD: And what if it is an essential system, such as the airbags or brakes?
Nizic: From the business point of view all of these are essential, but from the liability perspective the ABS is more important than the radio. But there still needs to be similar reliability in every aspect.
Pateras: From a test perspective, we’re finding with automotive that the goal is cost effectiveness. Everything has to be cheap. They need very high quality, but they have to do it in a very cost-effective way. I can get any quality you want if you give me the time. I can spend hours testing that part. The question is can you do it very quickly, and in a cost-effective way. It’s a curve of cost versus quality. And with our automotive customers, we’re finding that curve has to be steeper than with other customers from an ROI perspective. We’re actually seeing needs for new techniques to get that steepness increased from a quality perspective. And also to Mladen’s point, more and more parts are going into automotive. Ten years ago you saw TI, NXP, etc., and now one of my customers is asking me about automotive quality. So we have customers you’d never expect asking about automotive quality. Now everybody has to deal with it, which means that our solutions have to be not only more versatile, they have to be more robust, and they have to have more features and capabilities to adhere to the different requirements of different customers. From a marketing perspective, it’s different. I’m not meeting the needs of one or two customers—I’m meeting the needs of 20 or 30 customers. So from an EDA perspective, it’s a lot more challenging than it was in the past.
Sarkar: Along the theme of cost, I was visiting Conti [Continental] in Regensburg [Germany] a couple of months ago and one of the design managers told me that their goal is to design a rocket system at the price of a washing machine. That’s their mandate. The mindset is to have that level of sophistication at as low a cost as possible. I come from the Apache part of ANSYS, and ANSYS historically has worked with more of the system side of things like designing a gearbox or with all these parts manufacturers like treating the whole car as an antenna for EMI radiation, for example, while Apache has been working on the chip side with Infineon, Renesas, Freescale, TI—all of them. We have been fortunate to look at this whole spectrum. So talking about cost, looking at the optimization, like whether you can optimize not just the chip but the package, the board, the cabling and how do you measure the EMI radiation? Going back to reliability, we see people that five years ago did not worry about sharing information with each other, such as IC teams sharing information with the system teams. Now we see that trend completely reversed. The component manufacturers are on their own giving the system guys models to do the system-level EMI analysis and they are seeing that we run the EM analysis with this tool, under these conditions, and the system manufacturer is coming back with the guidelines saying, ‘We want you to do this, this, this.’ There is more understanding on both sides. Trend-wise, depending on which application, typically Bluetooth and Zigbee are already embedded and broadband RF is probably going to be next. As you have higher speed radios, more energy, the EMI is going to become a worse problem.

SLD: What do EDA tools need to do to accommodate the automotive industry requirements?
Pateras: On the test side, it’s quite simple. We have this ISO standard 26262 that has stringent reliability and quality requirements, which can only be met in certain ways. One of them is in system test. So from a test perspective, that means you need to have the ability to apply your test in system and be able to communicate to those test resources and look at results in system—either through CPU access or what have you. That’s driving things like BiST (built-in self-test) for all parts of the design. It’s driving interfaces to CPUs, so you need to be able to access the BIST through a CPU interface. We’re seeing a huge interest now in CPU access to BiST capabilities, which we really hadn’t seen in the past. That really started very recently—in the past 12 months.
Palus: …for you….
Pateras: I know. TI and Freescale have had this for a while with their own in-house solutions.
Nizic: If you look from the system level, you can’t afford to start building something that’s not correct at the beginning, so you have to verify these complex systems. Just take airbags, which are often used as an example. There are so many things that have to be verified there that depend on the interaction of that system with the rest of the electronics and mechanics—the whole car. There are constant improvements in technology to verify and make sure this thing triggers off at the right time, and that it doesn’t trigger unexpectedly. To do that now I have to have quite accurate models. In verification modeling is the key and you have to model it with a purpose. I see more and more movement into standardizing and sharing among the automotive companies as suppliers to ease the load and create something reusable. Then the whole verification methodology is heavily dependent on software—that’s another aspect that in past was separate. Now it has to be joined from the beginning and often even carried through the chips. If you look today, a microcontroller is not that simple anymore. If you talk about reducing consumption of gas and increasing mileage, today controllers are very sophisticated with this scheme. A number of sensors constantly monitor what’s happening with air, temperature, oxygen levels, emissions and tuning for optimal behavior of the engine to most effectively utilize the gas.