Connecting The Car

K. Charles Janac, chairman and CEO of ArterisIP, sat down with Semiconductor Engineering to discuss changes in automotive and how the connected car will affect chip design and a multitude of other markets. What follows are excerpts of that conversation.

SE: What is the biggest change you’re seeing in semiconductors?

Janac: The really big change is that mobility is flattening out. The market is still growing, but it’s growing less, and the number of participants in that market has shrunk by half. The people left are the system houses and a few huge chipmakers. So the question is, where is the growth? The IoT is a bunch of vertical markets that will be interconnected. The vertical market that seems to be getting a lot of traction, investment and attention is automotive.

SE: How about IIoT?

Janac: I don’t see any other killer apps. Industrial IoT is going okay, but it’s not a huge market. For other markets, there are small chips sold in relatively low volumes. There are smart watches, which still doesn’t look huge. There is medical IoT, which ultimately could be large. The huge market is automotive IoT. People are connecting the car full-time to the Internet.

SE: How quickly do you see that evolving beyond companies like Tesla?

Janac: It’s going to happen much faster than people think. In the past, you selected a car on price, performance, and to a lesser extent, the quality of the interior. Now you’re adding new criteria, which is the level of electronic functionality and driver assistance. If you’re a carmaker and you don’t add those new criteria to your business, your market share is going to shrink very quickly compared to those companies that do have it. There are going to be dead car companies.

SE: So who emerges as the big car companies? Is it the existing carmakers like Tesla or Ford, or is it the electronics companies like Google and Apple?

Janac: Nobody knows. Does it follow the path of the PC market, where it’s the component guys who make all of the margin? Or does it go the way of the camera market, where the end user OEM is the biggest value generator and the component suppliers are fighting to get into that ecosystem?

SE: And it may change from driver-assisted vehicles to fully autonomous vehicles, right?

Janac: Yes. There are different segments here, including self-driving trucks, drones and logistics vehicles. Those have to be level 5. They can justify much higher levels of electronic investment than the mainstream market, which will be automated assisted vehicles. Most people today don’t want to own a car that doesn’t let them drive.

SE: But the next generation may not have the same love of cars, right?

Janac: Some do, some don’t. It used to be that getting a car was the best thing that happened to you as a teenager. It was freedom. In the future, the question is whether cars are owned by individuals or by fleets. But overall, the number of cars doesn’t necessarily go down.

SE: This has broad implications for a lot of adjacent industries, as well.

Janac: Yes. If you look at autonomous taxi companies, they will purchase a huge number of vehicles and apply pressure to lower the cost. That will impact the margins of the car companies, which in turn will decide they can serve the self-driving taxi segment by themselves. And that’s just one piece of this. From a societal perspective, autonomous driving will create bigger changes than the smart phone. The infrastructure becomes electronic. It’s much easier to send an electronic signal to a vehicle than to have it recognize a light or a stop sign. So every traffic light and traffic sign has to have an electronic component. At first there will be geo-fenced regions, but the bulk of people will want the assisted model where they can let the car take over.

SE: Where else do you see the impact?

Janac: If you look aging populations, people can be independent longer and more safely. There are major societal implications. For that to happen, these vehicles have to be secure and safe. That means they need secure and safe electronics.

SE: How do you differentiate those?

Janac: Safety protects you from environmental and internal mishaps. Security protects you from man-made attacks or mishaps. Vehicles will keep improving every month because they can be updated constantly. In the mid-’00s, one car company was upgrading its electronics by putting in a new version of the entire command system in its luxury vehicles. That cost about $5,000. Now, there are no hardware modifications. The improvements are based on software downloads. That’s why the car has become an IoT device.

SE: There’s another angle to that around privacy, which is whether you have control over your car’s data.

Janac: These vehicles will be able to obtain downloads from the cloud, but they also will have full-time connection to the cloud for mapping and entertainment. The car becomes a mapping sensor for automated driving maps. So there are huge amounts of data exchange. That raises the issue of who owns the data and your privacy.

SE: So when an autonomous vehicle fails, how do you ensure that it fails without causing an accident?

Janac: One of the things we provide is resilience. There are different levels of it. Ultimately all of these vehicles will require ASIL-D functionality, which is the highest level of functional safety. For ASIL-D you need unit duplication. When the data comes from a processor, which also has its own safety functional lock-step mechanism, you put out two signals. One is functional, the other is a reference signal. You compare the output and make sure they’re the same. The reference signal runs one or two cycles behind. If those aren’t the same, you get an error report. The next step of this is what the automotive industry calls ‘fail operational,’ which is grading the errors to figure out which ones are serious and which ones can be ignored.

SE: This is a new version of the old Tandem and Stratus computer idea, right?

Janac: Yes, on-chip. You still have one computer in the back of another, but it’s duplicated.

SE: What impact does that have on cost and performance and power?

Janac: It doesn’t make it better. There is a price to pay for resilience. You do pay as much as a 35% penalty in terms of silicon area.

SE: But this does decrease the liability impact for electronics, right?

Janac: Yes. With sufficient documentation and functional safety studies—packet injection, fault simulation, FMEA (failure modes and effects analysis—you can provide documentation that you have taken the best possible care for the system to be functionally safe. And all the components in the system have to have that, from the IP blocks to the chip to the subsystem to the car.

SE: Is there enough forethought in architectures that all of this is in place in cars being designed today?

Janac: No. The car itself is good for 20 years, but now you can have a car that is three or four years old that is missing all of this new technology. So one of the unintended consequences is a faster upgrade cycle. Or maybe you lease a car for three or four years and then you turn it in for a new model.

SE: This is compounded by the fact that we now have more electronics in cars than ever before. So even if the failure rate decreases, the chances of having a failure may actually increase, right?

Janac: Yes, and the reliability has to increase faster than the failure rate. But many systems are not designed for reliability. If your smartphone fails, you may be annoyed but you go out and buy another one. A smartphone won’t kill you, but a car can. These systems have to be thought through for functional safety and resilience all the way through to the lowest denominator, which is the IP, all the way through to the user. If you look at the satellite electronics, that doesn’t fail very often. But it does get bigger and it costs more.

SE: How much more?

Janac: I just went to an automated driving conference, and the question being asked was what is an acceptable price for a self-driving system. If you are dealing with a self-driving taxi, that is operating 24 hours a day, the cost could be somewhere around $40,000 per unit. This is the sensors, the ADAS chip and the software. If you’re in a consumer vehicle, the acceptable cost will probably be $5,000 to $7,000. You have to be very careful about how you engineer things for cost, given that it has to be resilient and safe. In a plane you can afford a different solution than in a $35,000 car. In a car, cost has to be a consideration, along with resilience, security and functional safety. You also need to consider performance, because these systems require HPC-level performance. There will be immense amounts of data that needs to be processed in almost real-time. How do you deliver that at an acceptable cost?

SE: How quickly do you see this becoming ubiquitous?

Janac: There always will be a segment of people that want a V8 engine. There are people who still use fountain pens.

SE: Where else are you seeing market opportunities?

Janac: We’re still seeing growth in mobility. We’re also seeing it in cameras, enterprise computing and networking chips. As all these end points grow, there is some ratio of server blades. In a phone, about 600 phones require a new server blade in a data center. It will take fewer cars per server blade. As the IoT end points grow, the data point volume grows. You’ve also got bandwidth and storage issues. But at the end of the day, who owns the data?

SE: We’re hearing that question being asked a lot these days.

Janac: If you have a car company with 1 million cars on the road, each car has eight cameras and LiDAR and radar, and they’re shipping that data back to the data center. If you have the electronic infrastructure to process that, you’ve just gotten into the data mapping business. The car sees a FedEx truck or a UPS truck. That data can be very valuable if it’s mined. On one hand, the car companies may or may not sell fewer cars. But they could be a very big beneficiary of the data. Or it could be a company like Mobileye, and so Intel becomes the best mapping company in the world.

Related Stories
Autonomous Cars Drive New Software
Security, reliability and complex integration are required, but can it happen quickly enough?
Why Auto Designs Take So Long
A design for safety methodology is essential to trim automotive design costs, but at this point it’s a work in progress.
LiDAR Completes Sensing Triumvirate
Technology will complement cameras and radar in autonomous vehicles.
Automotive’s Unsung Technology
Audio technology is making big strides alongside autonomous vehicles and vehicle-to-infrastructure communication