Get email delivery of the Cadence blog featured here
After ARM on the first day, the keynote on the second day of DVCon was by NXP. For semiconductor, automotive is still much smaller than mobile, but it is growing fast. It is growing not so much because of inherent growth in the underlying market for cars, but because the semiconductor content in each car is growing fast. Because of reliability and functional safety requirements, verification is far more important than in mobile. If our smartphone crashes, we reboot it. If it fails totally within a few years, we shrug and buy a new one. We expect our cars to remain safe for 20 years, even if something goes wrong. It sounds like a paradox, but automotive semiconductors need to be more reliable than the semiconductors themselves. The keynote on the second day of DVCon Europe was on automotive, given by Jürgen Weyer, the VP Automotive Sales EMEA for NXP (to be Qualcomm in 2017, probably).
Ten years ago, automotive was the forgotten stepchild of semiconductor, just a few percent of the market, not growing much either. But that was then. The industry has now changed completely. In 1995, Jürgen was in mobile. They used foundries and not their own manufacturing. People in Germany in automotive said back then, “We will not use a microcontroller from a foundry.” Today, no car would run without microcontrollers from foundries. He thought he could bring lots of beautiful technologies over from mobile. He was both right and wrong. One area that was new to him was that they would sometimes see problems a few years into production. In mobile, you are already on to the next chip. In automotive, the problem needs to be fixed, but there needs to be verification methodologies that stop that sort of thing happening so that designs are only verified once, before manufacturing starts, and that take all the aging effects into account.
NXP as it is today was created in December last year when the acquisition of Freescale closed. That was another "not going to happen" thing. It is now the largest automotive supplier at around $4B, 30 sites, and 2400 automotive engineers. In 2009-10, both companies were close to bankruptcy, but they both sorted out their financials and became profitable and sustainable companies before merging.
Anyone in automotive will tell you that electronics is a growing portion of the bill of materials for a car. But the really important number is that 90% or more of innovation in automotive is in electronics. This is a threat and an opportunity. Some predictions are that self-driving cars and an Uber type of model could mean that we only need 20% of the number of cars we have now. I have seen reports that this could lead to a lot of "consolidation" but that seems like an understatement. An automobile manufacturing plant will become less useful than a plant for manufacturing chemical film. Even today, especially in Europe, there is a lot of over-capacity. It would be like "consolidation" in the buggy whip industry following the Model T Ford.
Jürgen said that he was surprised that DVCon did not have a really strong focus on security, given the connectivity of cars to everything: V2V (vehicle to vehicle), V2I (vehicle to infrastructure), and V2x (vehicle to everything). Actually I think he should be happy that DVCon is so focused on automotive and not be greedy. Automotive was a backwater and now every conference and workshop I attend is about automotive. My prediction: Next year's DVCon Europe will have lots about security.
One thing that Jürgen was pleased to see is that automotive is starting to re-use stuff from other industries, instead of being a sort of weird Galapagos Island ecosystem of its own, cut off from all other uses of semiconductors. The most obvious is automotive Ethernet on cheap twisted pair, which will displace FlexRay where the cables cost €2 per meter. Consumers expect copy-exact from their cell-phones into the car. He predicts you will see car OEMs making deals with Apple, Samsung, Huawei, and others.
Wiring is a big deal. Wiring harnesses are up to 80-90Kg. Why not wireless? Why do you need all that weight? Plus each harness has a volume of 1 since each car has different options, so there is lots of opportunity. Plus there is a virtuous cycle where we make the car lighter by removing the harness and automatically it is more fuel efficient. There is a need to make cars lighter still. BMW is going with carbon fiber but Jürgen thinks it is very challenging. You can't repair carbon. What looks like trivial damage can cause structural failure (read bike magazines if you want to see some dramatic examples). It is unclear how the supply chain works when your car is in an accident. We have enough trouble coping with hybrid vehicles with high-voltage cables running all over.
Energy is another big issue. The great white hope was diesel, but its days are probably numbered, especially after the VW scandal where it emerged that diesel is simply not as clean as its image was, despite often being 1/3 of what some SUVs in the US need for the same number of miles. Infrastructure for vehicles will take time to build. He said that the infrastructure was great in Germany..."provided there are only a dozen vehicles". There are some countries where things could happen overnight. It is not hard to imagine China deciding that from a certain date, all cars will be electric. Today we have infrastructure for gasoline/petrol and diesel, but nothing close to what is needed for electrical vehicles in those sorts of volumes.
We will see 48V batteries in cars, and not just electric cars. Obviously the battery can be used for hybrid-style regenerative braking, but an electric turbocharger can give you high performance on a small efficient engine. Just 7KW gives you 70KW of boost. There was a big attempt a decade ago to switch from 12V to 42V that failed, largely because there wasn't enough benefit, especially if you had to keep the 12V battery, too, for compatibility. But now it is complementary in hybrids, etc.
The big challenge with electric is that people don't want to spend more time to recharge a car than they spend in a gas station today. And today, people drive off if they have to wait for a pump. Development of batteries might be revolutionized in the next five years. It is a big opportunity for semiconductor since any foreseeable battery requires a lot of battery management, a lot of analog and digital chips. He expects to see some winners who own the battery IP who are not big names today, or perhaps not yet names at all.
The #1 problem/opportunity in automotive is 1.3M fatalities. 50M seriously injured. $3T in property damage. 90% of this is caused by human mistakes, so we need to get the human factor out of the equation. Jürgen admitted that automatic emergency braking had saved him from at least three accidents. "I wouldn't use a car anymore without automatic cruise control, lane departure warnings, and automatic emergency braking." But it will be a long time until we have fully autonomous vehicles under all conditions. Since I live in California, it would be a minor inconvenience if I have to drive when it is raining, especially in recent years. Jürgen moved on to sensor fusion. Ultrasonic, common today, will probably go away once we have other technologies that we can reuse for things like sonar. For sure, putting a $50K laser apparatus on top of the car like Google does is not the way of the future. There is a challenge to put the sensors on a car without holes in the bumpers that designers (and purchasers) hate. One thing I hadn't realized was that in the recent fatal accident involving a Tesla, there was radar as well as vision. But sometimes there is just too much information with radar.
At the top level, for automotive semiconductors, there are four master dimensions:
The goal being to deliver:
NXP feels it is well positioned for security since they can leverage their expertise in banking and eGov. NXP is #1 in secure identification. For example, over 100 countries use NXP in their e-passports, NXP is #1 in payment cards.
There are starting to be big experiments and government mandates. It is starting slowly, it always does in automotive. But, sometimes, things go fast after a slow start, such as electronic stability control. When Mercedes had cars rolling over during the Elk Test (no elks are harmed in doing this test!), the industry went from 0 to 100 overnight. The next day NXP's forecast from customers was "we need this tomorrow" and it is hard to explain that 28nm takes four months to manufacture. At 16nm and below, the cycles only get longer.
Today, customers don't know what kind of compute power they need. NXP has built a system called BlueBox with a high-powered 64-bit ARM and added safety, security, and more. "I don't believe this box will ever go into volume production," Jürgen said. "But it lets people add software and customize it. It is a big problem to understand the requirements of the customers to pass to the design teams.
Obviously the focus of automotive electronics is on autonomous vehicles. But getting more down to earth, here are a few important short-term features:
Previous: What Is the ARM ARM?