Getting To Automotive Grade

Given the amount of activity surrounding automotive semiconductor design today design teams want to know how to approach designs for a market they may not yet be intimately familiar with. EDA vendors are very quickly ramping tools and services to help them get there.

One of these is Mentor, A Siemens Business, which has actually been working in this market segment for two decades.

Andrew Macleod, director of automotive marketing at Mentor, A Siemens Business explained there isn’t a simple answer as to how the EDA community works with users today because, “EDA for automotive is very much a point solution type of business so nobody really approaches the IC houses with a complete flow saying, ‘you can have everything.’ These tools are optimized for different parts of the process so we approach the customer with our vision of the future: You need to build something for sensor fusion, for autonomous drive, and have this big platform and collaboration flow. Another way could be you’re running millions of units a year of embedded micro controllers and the challenge there is that test times are out of control. Testing of an ic can be up to 30 percent of the overall product cost, which is huge. So we can talk about things like how to get the chip to self test and maybe, if you remove one second of test time on a tester for an ic, that’s a cause for celebration for sure.”

“You save a lot of money, but then of course, how do you do that without compromising quality, and how do you do that to improve quality?” Macleod said.

Within automotive, the test process is interesting, he continued. As chips have gotten more complex, more tests have just been added to every generation of chip and the engineering team hasn’t bothered to remove any tests that weren’t really needed because the reward/risk calculation just didn’t make sense. “Now using EDA, we can actually analyze test programs and find out what tests are redundant and start to remove those tests, and also identify where there are weaknesses in the test program as well. Testers are hugely significant parts of the automotive business, and we have a lot of discussion with customers about that. Once you optimize the test software in terms of what tests you have, you want to address it, we can have discussions with the team about improving quality generation to generation, and by doing things like testing chips at the cell level, and getting into the transistor level testing. These types of things can again highlight weaknesses and improve quality even if it sounds counterintuitive.”

There are challenges on the manufacturing side as well such as electrical overstress (EOS) and ESD. “EOS and ESD are typical challenges with an automotive grade IC where it’s operating in a very temperature environment, especially embedded flash memory that has to be certified for 10,000 program/erase cycles, for example, over maybe 10 to 15 years. These are examples of where we get into the physical chip circuitry once we get to silicon and analyze, through design rule checking, how to optimize line spacing and all of the different rules that make an automotive chip different physically, and structurally from a chip used in the consumer/mobile space,” Macleod observed.

Bringing it all together in the context of one of the large domain controllers that are seen in many automotive designs today for autonomous driving, there may be a sensor fusion box, with an FPGA at the front end for capturing all of the sensor data, along with some kind of big SOC processor in the middle for all of the ADAS and autonomous functionality, and then a microcontroller for doing all the safety checking for ASIL-type checking and security, he noted. “Those are three components that all require different approaches in terms of design for the FPGA and the SOC running the machine learning algorithm.”