Paul McLellan

Community Member

Blog Activity
Options

CDNDrive Automotive Solutions: the Front Wheels

• 4 minute read

cdnlive logo breakfast bytesHarold Wilson, then Britain's Prime-minister, once said, "A week is a long time in politics." Well, a year is a long time in the current automotive market. I've written quite a bit about the market in general (most recently about the Chinese market, see my post Trends, Technologies, and Regulations in China's Auto Market and the European, see my post Automobil Elektronik Kongress 2018) but not covered how the Cadence automotive offering has developed. Our automotive marketing is headed up by Robert Schweiger and he presented at CDNLive Silicon Valley, and I promised I'd write about what he presented. Then he presented in Munich at CDNLive EMEA, and I promised I'd write it up. Finally, when we were in CDNLive Japan a couple of weeks ago, he presented again, and...well, you get the picture. At least this time I did it. In fact, this is the first of two posts that cover our automotive solutions.

The Automotive Electronics Market

The automotive electronics market is $142B and is growing at 7%. But ADAS is growing at 19%, almost 3 times as fast.

There are really two different automotive semiconductor markets. Until the last few years, automotive semiconductors consisted mostly of a mixture of analog and 8-bit microcontrollers, built in processes that already a decade of characterization data. That market still exists, of course. Many, maybe most, of those chips are designed in Virtuoso, but that is not the focus of this post.

The new automotive market, needed to support ADAS and autonomous driving, is in more modern processes such as 16nm, 22nm, or 28nm. They have a very large digital component for functions like processing video, lidar, and radar. This cannot be done on a ten-year old process.

There is a sort of culture clash between these two worlds. The old-school market understood every aspect of reliability in depth, but not how to do a modern digital automotive SoC, which is more like a cell-phone application processor. The groups that do know how to design automotive SoCs typically do not have in-depth experience in the reliability issues. Cell-phones and the like do not have to last over 15 years, and failures are warranty claims, not life-threatening.

This is leading to disruption in the automotive supply chain, with not just old-school semiconductor companies doing automotive design, but also tier-1s (like Bosch, Delphi, Denso, etc), and OEMs (like Ford or BMW) doing SoC design, many from almost a standing start. The semis like Renesas and NXP are pushing up, delivering complete ECUs complete with the software stack.

Automotive System Design Enablement Challenges

Automotive systems have all the complexity as any other system, and they have several of their own. Here are a few, and if the code numbers and acronyms are not familiar, I've written about all of them before:

At the system level, an ADAS or autonomous car has several components:

  • sensors, especially cameras, lidar, radar
  • high-speed networking, probably automotive ethernet
  • a powerful central automotive processing unit, containing a lot of neural network processing
  • actuators (brakes, steering, acceleration)
  • an infotainment system, largely independent of the rest (since it is not safety-critical, but probably implementing display functions)

There may be some local pre-processing at the sensors, but there is a big debate about this. I would say current received wisdom is to feed all the raw sensor data to the central unit where the sensor fusion is done, the scene analyzed, decisions taken, and actuators actuated. The challenge, of course, is that this makes the central unit something akin to a datacenter in your trunk. Connectivity is an important part of automated driving for things like map updates at the very least. The main driving function has to be onboard since connectivity may be lost, and the latency is too high. This may change a bit with 5G, but any car needs to be able to safely handle losing its communication link, otherwise that becomes a huge point of failure (and turning on a warning light to show connectivity was lost, like is done when a faulty airbag sensor is detected, is not enough!).

Tomorrow

Well, I guess I'm guilty of false advertising since I never really got to any details of our solutions. That will be the "back wheels" tomorrow.

 

Sign up for Sunday Brunch, the weekly Breakfast Bytes email