Get email delivery of the Cadence blog featured here
A couple of weeks ago Cadence held its second Automotive Design Summit here on the Cadence campus in San Jose. I will cover some of what was said over the next week or two, but I thought it might be good to do a recap on the basics of the automotive industry, which has its own language, its own structure, and is facing several challenges, most of which are rooted one way or another in technology.
You might think that car companies build most of the stuff that goes into cars. Or at least the important stuff. But actually, they really only make the engines and the bodywork themselves. Everything else they buy in. That complicated infotainment system? Purchased. That ABS system? Purchased. The seats? The lights? The wheels? Yup, purchased. You get the idea.
The industry is in transition as a result of ADAS (advanced driver assistance systems) and autonomous driving, so I'll start by describing how the industry used to be.
In industry jargon, what you think of as a car company is known as an OEM (don't worry about what it stands for). That's the company who puts its name on the trunk, makes the engine, and assembles the overall vehicle.
The OEMs buy all that stuff from the next tier down, which are known as tier-1s. Here in this blog, we are only really concerned with electronic suppliers, not seats and wheels. Some of them also supply a lot of non-electronic pieces.
The tier-1s put together the electronic control units (ECUs) that are all over a modern car. A high-end car might have 100+ of these. Typically these contain a microcontroller that the tier-1 has programmed. Each ECU serves a single function, such as moving a seat, or deploying an airbag. The tier-1s don't build their own silicon or even design it. They buy that in from tier-2s.
The tier-3s are people who supply whatever is needed by the tier-2s. Cadence, for example, or the foundries. So it's a bit like the House that Jack Built:
This is the cat, That kill'd the rat, That ate the malt, That lay in the house that Jack built.
This is the cat, That kill'd the rat, That ate the malt, That lay in the house that Jack built.
One thing of importance in all of this is that the automotive companies don't know what is in the ECUs and they don't have the source code. The tier-1s know what is in the chips in some sense, since they need to know that to program them, but they don't have chip design capabilities themselves. When the ECUs just did things like controlling the windows and airbags, that didn't matter too much. When they control all the driving, and when the company with their name on the hood is responsible for accidents, the whole situation changes. So suddenly everyone is interested in software and chip design.
When each ECU provided very basic functionality, then the OEMs (remember, they are just car companies) could pick and choose among them, and assemble a vehicle. The character of the vehicle was largely determined by the engine, the shape of the bodywork, and the selection of ECUs (is there central locking? how many airbags?). Once the prospect of self-driving cars appeared on the horizon, it became obvious that the car companies would need to control some parts of that themselves, or they were handing over the look-and-feel of the vehicle to the tier-1s, who in turn realized that as more and more of the functionality was integrated into SoCs, it was the semiconductor companies who were going to be making those decisions. To make things worse, electric traction meant that eventually, even building internal combustion engines was going to become a liability not an asset.
So the industry is structured flatter for advanced electronics. Some OEMs build their own chips. Some tier-1s are designing their own chips. New semiconductor companies have entered the automotive market.
To shake things up even more, some tech giants and lots of new companies have entered the market. Some of these have partnered with existing OEMs and some have been acquired.
Three big challenges have come along at almost the same time and are disrupting the automotive market.
The first challenge is ADAS and automated driving. I'll cover that in more detail later in this post, since that was the focus of the Automotive Summit.
The second challenge is electric traction. These are known in the industry as EVs for Electric Vehicles (or NEVs in China, for New Energy Vehicles). This is in contrast to internal combustion engine vehicles, which are known as ICE. There are a number of reasons that this is challenging. One is simply that ICE is not going away anytime soon, so managing the transition while both types of traction are required is tricky. There is a risk of having stranded ICE plants that can produce more engines than the market requires. It is also a marketing challenge. If a car company is positioned as super-reliable, or 0-60 in four seconds, or the sportiest cars around, then that is all going away with the engines that provide those characteristics. Electric engines are super-reliable, can go 0-60 faster than a Formula-One car (search on "Ludicrous Mode" for the most extreme examples), and are generally sportier than ICE due to the high torque without going to the redline.
The third challenge is the potential for people not owning cars themselves anymore. The rise of ride-sharing companies is one early sign of people using their own cars less. American teenagers are no longer chomping at the bit to get a driving license the moment they are old enough. Millennials living in big cities find that exclusively using ride-sharing services is cheaper than owning, insuring, maintaining, and (especially) parking/garaging their own car. This could have a major impact on the size of the market since there will be fewer cars required, although they will be driven more intensely than today's personally-owned cars which sit idle 95% of the time. It also changes the dynamic of the industry since selling a fleet of cars to a ride-sharing company (b2b) is very different from making sales to individuals (b2c). It potentially has knock-on effects in areas like TV revenue, since cars and trucks are one of the biggest classes of advertising. Of course, there are big changes in TV due to technology too, but that's beyond the scope of a blog about the automotive industry.
The Society of Automotive Engineers (SAE) has defined different levels of automated driving. Level 0 is 100% manual control. Level 5 is full automation under all circumstances, the level at which cars might be built without steering wheels.
Although level 0 is 100% manual control, for historical reasons, since it's been around forever, regular cruise-control that just fixes the car at a preset speed is put in level 0. It just requires a simple microcontroller-based ECU to control the accelerator.
Level 1 is automatic traffic sign recognition, and perhaps limited automatic emergency braking (AEB). This level requires just one front-facing camera and an ADAS ECU to control it. I drove a rental car in Europe a year or so ago which had a head-up display that projected the current speed-limit onto the windscreen if I was exceeding it. I assumed it was map-based until I discovered it could do it even in construction zones, and I realized it was recognizing the signs.
Level 2 provides lane following and automatic cruise control. Maybe a bit more. Under good circumstances, the car can drive on the freeway or in stop-and-go traffic on its own. It can maybe drive on city streets in good weather (there's a reason car companies like to start in Phoenix and Las Vegas). It is still the driver's job to monitor what is happening, and typically cars at this level insist that you put your hands on the wheel from time to time. Also, if things get too hard then the vehicle will hand control back. This requires a front-camera and a front-radar, plus an ADAS ECU to power it.
Level 3 provides fully automated driving in most circumstances such as city driving, entering and leaving freeways, driving on country roads, and more. Everything up to the "no steering wheel required" level. This requires a lot more in terms of sensors, such as front and rear cameras, front lidar, half-a-dozen front and rear radars, GPS, and a sensor-fusion ECU to pull all that data together. Not to mention high-performance in-car networking. A car is like a datacenter on wheels...but with a tiny power budget.
There is a level, not an official level, between 2 and 3 known as Level 2+ and for now that seems to be the sweet spot that can be delivered with current technology at a non-insane price point. For one recent post of mine discussing this, see my post from just last month Ludwigsburg: It's All About Return-on-Investment.
Level 4 means you, as the "driver", don't need to do any driving. There might be some circumstances such as snow or fog, or breakdown, where the automated system won't be able to drive and you would have to drive, but not like at the lower levels where the system might pass control back to you unexpectedly at any moment. The car could limp to the side of the road and stop on its safety backup computer. The hardware to do anything close to level 4 is way too expensive today and so is likely to first appear in commercial trucks—it is a lot more feasible to add $20,000 of electronics to a $400,000 truck than a $40,000 car.
Level 5 is not something I expect to see in less than 15 or 20 years, except in very limited circumstances like around hospital and university campuses in good weather. There is an experiment like this going on in Singapore on Sentosa island with a couple of fully automated minibus shuttles.
There is a cute description of these levels as:
And with that, I sign off.
Sign up for Sunday Brunch, the weekly Breakfast Bytes email.