Drive On: Where are the Flying Automobiles
Tom Beckley, Cadence
In the 1960s, the American cartoon The Jetsons was amazingly accurate in predicting our future. Today, we have at our fingertips almost all their inventions, except one, the flying car. However, we are inching our way towards that breakthrough following one simple and overarching principle: System Design Enablement. In this presentation, you’ll learn how Cadence is solving the challenges of designing and verifying the multitude of electronic disciplines that make up System Design Enablement by harnessing the power of sophisticated EDA solutions, high-performance IP, and strategic partnerships, including MathWorks. Working together, we turn childhood dreams into realities.
Published: 16 Mar 2018
So it's my great pleasure to be here with you today.
So the custom IC—this is RF design, analog design, custom digital design, IC package design, and printed circuit board design teams at R&D for Cadence. It's about 1,100 electrical engineers, software developers, and designers in this group. And so a few years ago, exactly three years ago, I started meeting with the leadership of MathWorks to try to figureout how to better connect the system-level world with the world—are we okay? Nope. We're not okay.
How to connect the system-level and architectural-level world with the world of detailed electronics implementation, analysis, and verification. And I'll comment a little bit more on our great unfolding partnership in just a few minutes.
My presentation today has three parts. The first part will be a little bit like Marianne’s presentation—which was excellent, by the way; I learned a lot—about trends from an electrical engineering perspective, especially from the semiconductor perspective. Then I'll tell you a little bit about Cadence. I'll keep that pretty brief, no big marketing spiel. That's not who I am.
And then we'll talk about the autonomous vehicle and some new design fabrics that we see unfolding across some of our customers, especially driven by advanced node semiconductors.
So you can see this, so hopefully everything's working. You might think this is a bit of an odd title for a keynote. What's this about flying cars? I would posit to you that people have been dreaming about the future and applying science to the future for a long time. If you think back to great authors like Isaac Asimov, or Robert Heinlein, or Ursula K. Le Guin, or Ray Bradbury, they were great scientists who took a look at the future.
And today, with the fusion of complex mechanical advanced node semiconductors and complex software, and the ability to connect that at the edge, at the fog, and at the cloud, we can make that future a reality. And Marianne talked about some of those opportunities that are unfolding.
Some of you may be working on the autonomous vehicle. Anybody–you're working on some facet of the autonomous vehicle? Come on. Come on. Some of you are. I am. Right? So you will make it possible for a person like me to walk up to a living room on four wheels and say, “Alexa, take me to Los Angeles.” I'll then work, or sleep, or play, or maybe daydream, like some of you may be starting to do right now.
I won't have to worry about accidents. The route will be optimized. But you won't be able to walk up to that living room on wheels and say, “Okay, Google, take me to London.” That's still going to be a flight. Now, how many people here enjoy going to the airport, going through security, boarding the plane, getting that little package of pretzels, going to the restroom on the plane and waiting in a big queue for it—there's something wrong with this equation here. There's something wrong with it.
So I would posit to you that our challenge is to get beyond the driverless vehicle and to get to the flying vehicle. More on that in just a few minutes.
So we have this great opportunity today because of engineering visionaries, men and women like William Shockley who, in the late 1940s at Bell Labs, created the first commercial transistor. And Shockley's lab was filled with great engineers. They became known as the Shockley Eight. In the late 1950s, they left the lab of their Nobel Prize-winning mentor, and they formed an iconic company called Fairchild Semiconductor.
Now, in time, this group dispersed yet again. And out of that dispersed group came more iconic companies: Intel, AMD, Xicore, Teledyne, National Semiconductor, and more. About this time, I was just a little tyke. I'll let you figure out which of the Beckleys is me. Now, I watch kids today, I watch my grandkids today. They're playing on iPads, right? I played on an Etch-A-Sketch.
But I would posit to you there's something prescient about this tool back then, this toy back then. It used magnetism. It used sand or silicon. Maybe I was dreaming about my first circuits as early as that.
Now, in the early 1960s, there was a popular American cartoon series called The Jetsons. Anybody recall it? Anybody recall The Jetsons? I remember watching it with my brothers and my sister on this big black-and-white TV, big cathode-ray TV. We'd all gather around it every week to watch the episode. It was about the future.
You know what? They predicted smartwatches, robots in the home, FaceTime, flat-screen TVs, and every episode had—well, I guess they also had remote diagnoses with sensors in the home and the physician far away. But every episode had that personal flying vehicle. While it's true cars don't fly yet, I would argue—and I'm going to talk a little bit about how the automobile industry is transforming. There are going to be big winners and there are going to be a lot of losers, and it is very scary for the incumbents today as to what is transforming on that front.
We had an agricultural revolution. We had an Industrial Revolution. I started in 1981 in semiconductors at a place called Harris Semiconductor in Melbourne, Florida. We were doing Bipolar and BiCMOS, hard designs for the space shuttle and other things with Cape Canaveral. I've seen a lot unfold in electronics and semiconductors in 35 years.
I've never seen an opportunity as big as this. This is incredible, what is unfolding, from the driverless vehicle into the Internet of Things, like Marianne talked about, from smart homes to smart buildings, to smart cities, robotics, and more. There will be explosive growth across incredible amounts of segments. AR, VR, of course, is already unfolding.
From NVIDIA to Uber, to Tesla, to Toyota, the entire automotive industry and its tiers is being reshaped as we speak. If we look at semiconductor content alone, today into the next decade, it will go up 4x. Think about that. Four x more semiconductors going into those vehicles in the next decade.
Let's talk about robotics, which Marianne talked about a little bit. There is incredible investment under way in China and Japan for personal assist robotics because of the demographics, because of the aging populations. This area will grow from the low millions of units today into the billions. Now, think about the number of new semiconductors—boards, systems, all the control systems that have to be built, the machine learning, the deep learning that will unfold because of that.
I didn't talk about smart health like Marianne did. I didn't talk about drones. I didn't talk about all the other applications, including that will unfold. It's incredible what is unfolding today.
At the heart of these are next-generation semiconductors that are the brains of these systems that are processing along with ubiquitous memory that is extremely high performance that is making this a reality. These are very difficult to design. So we are designing today—we have production unfolding at 10 nanometer, we have 7 nanometer in design, we have 5 nanometer test chips that are being worked on as we talk about it.
For those of us that are device physicists, we realize that Moore's law, which I've been chasing for 35 years, that doubling of transistors every other year by shrinking transistors and shrinking interconnect, will reach a fundamental limit for CMOS at probably 2 nanometer. That's it. But that's not the end. Then we will go to more than more. It will be multidie, multi-technology, 3D stacking, and more. And you'll all be part of that equation.
But this is expensive. And yet, this is just one small piece of an incredible stack of technologies and stack of software where analog and digital signals traverse across chips, and packages, and modules, and boards, and electronic control units, and power that complex software stack. There are challenges in signal integrity, power integrity, electromagnetics, and more that unfold as a result of all of this.
So what does this have to do with Cadence? Let's go to part two of my presentation. I'll give you a short video on who Cadence is, what we do. I think you know a little bit about us, but we're going to do it from the perspective of the driverless vehicle that's unfolding.
So I mentioned earlier that I have 1,100 R&D engineers in my team. You'll note that that's about one fourth of the R&D engineers at Cadence. And so other groups develop IP that's used for USBs, and that's HDMI, and SATA, and PCI Express, and more that we provide to semiconductor companies, and they build their IP around that.
Other groups work on digital design, and digital analysis, and digital verification. And then when we weave it together with custom, we have a mixed-signal world. And other groups actually use our software to develop chips, packages, boards, and systems to develop hardware accelerators and hardware emulators. So systems companies like Apple, Google, or Qualcomm, or Huawei can actually co-design their hardware while they're developing their software.
We live in a nano world, and most of you live in a nano world, and that's why we can build these incredible products that are unfolding. If you think about flash memory, 64 gigabyte, the latest coming out of Samsung today, it's at 10 nanometer. How many transistors is it? Anybody know? 512 billion. 512 billion. It uses a 10-nanometer gate size.
It is called a FinFET structure. It's a 3D structure that we do with a lithography. And you can see on either side of that gate are the fins, the source, and the drain. Now, let's put all of this in perspective a little bit. 512 billion transistors. Take the population of the planet, multiply it times 68, and have everyone living on the size of a quarter. Almost seems like an airplane ride to me.
10-nanometer gate size. Take a cat's whisker, take your best kitchen knife, slice it in half 10,000 times. I disclaim any issue on the health side with your fingers that may unfold with that. But that's the world, and that's why our opportunity is so significant today.
Now, we can't do all of this by ourselves, so we have a lot of partners. I'll talk more about that. And then, of course MathWorks is a great partner for us. But electrical engineers take all of these transistors and they wire them up. Starting with the IP and their own IP, they wire it together. A lot of that IP comes from a company called ARM. You're familiar with them, as well.
And then they put packages around it, and then they put that into boards, and then those boards go into electronic control units and more. So let me talk a little bit about where MathWorks fits into this equation. I think you know.
So there are two ways we're working very closely with MathWorks today, and there'll be a presentation on some of this, I think, later in the day. The first one is at that chip level you can see that the design of an integrated circuit and advanced node is all about big data—millions and billions of transistors that are being developed. And therefore, we need the power of MATLAB and Simulink inside of our environment as that development is under way.
The other way we work with MathWorks is at the board level, where we're doing modeling and we're doing co-simulation of boards, using Simulink with our board-level simulator PSpice.
Now, if we look at that board level, here is the high-level power architecture for an electric vehicle. What happens, of course, is the analog and digital circuitry, along with software, actually controls the motors and the actuator. And we saw that in Marianne's presentation, kind of a different view of this. What's really key here is understanding the impact from an analysis perspective, and that's where we can bring the architectural system level with Simulink into the world of transistor-level design and backup more seamlessly.
If you're designing an advanced node semiconductor, or any semiconductor, especially if it has analog or RF circuitry on it, you are in a world of trade-offs. You're going to trade off power, and performance, and area, slew rate, and voltage. You're going to have multiple power architectures on that chip. And the number of corners and issues you have to deal with explodes, and in advanced node, it really explodes.
You generate gigabytes, and gigabytes, and gigabytes—50, 100, 200 gigabytes of data working on a chip. So this is where MATLAB, inside of our Virtuoso custom-designed environment, is so powerful. So how do you parse that data? How do you analyze that data? How do you visualize that data effectively? And here's where that unfolds. And you can see here, we're looking at that SRAM with memory cell, and we're doing, in one case, some trade-offs. These trade-offs happen all the time. How do I get the design right the first time?
Now, one of the things I like about my job is we're always on the bleeding edge, because our customer's always pushing the bleeding edge. And we're not going to enable this world by moving data via electrons in copper wiring. That's just not going to happen. And so we have other partnerships. Of course, MathWorks sits at that system architectural level and helps us with that simulation.
But we have other partners where we do photonics design inside of that same Virtuoso custom-designed environment, because the receiving ends and the transmission ends of that data, which is now moving into photons and into light particles, is controlled by analog and digital circuitry at both ends of that link. And we have partners with light simulation, and with curvilinear shapes, and the like that work in that area. We also work with MEMS companies and sensor companies so we can control that information as well.
Machine learning, for us, is very important, and deep learning. I already noticed and commented that we're in a world of big data. You can see that immediately. But customers want to reuse as much information that they've learned from their prior designs, and then they want to know what's relevant from what's just informative.
They also want to collect data, like we saw, in terms of using sensor information so they can improve the next design. When will my product fail? Will it be the electrical circuitry that fails, or is it the mechanical portion that fails? And how do we feed that back into the next generation design loop? So there, we're working with and developing our own software, and working with leaders, and using MathWorks as part of that process for better simulation, improved place and route, and more.
Let me go to the third part of my talk. Let's talk about the automotive industry and how it's transforming, and what advanced electronics, complex mechanical, and software are doing to this industry. So what I will posit to you, and it's absolutely unfolding, is that new design fabrics and new technologies are going to reshape this industry, and we're watching it happen.
So in 1972, I took driver's training, and with my brother started tinkering with this mechanical structure called a Ford Falcon. Isn't that a beauty? The Ford Falcon used a straight-six Ford engine, Holley carburetor with a distributor. I don't know how many of you have ever rebuilt a motor. I have. I don't know how many of you have ever tried to keep one of these things tuned up. I have.
It was incredibly hard. What was so hard about it—the Holley carburetor and the distributor used a vacuum effect to drive the fuel and the proper air mixture into the spark plugs. You got any moisture in the distributor cap, or you got a crack in the wire, and everything stopped working, or one cylinder wasn't firing.
A revolution has already occurred. There are two revolutions, I think, in the automotive industry that have occurred. The first one is the invisible one. The one we're seeing today, which is the driverless one, is visible. So in the late '70s, early '80s, and into the '90s, and continuing, electromechanical subsystems started unfolding.
So let's think about that distributor and that Holley carburetor that I talked about before. So what happens with electronic fuel injection? Today, who changes spark plugs? Nobody. How often do you see a car with blue smoke coming out of the tailpipe? Rarely. Maybe on a truck, maybe on a diesel. That's it. Why?
Because that electronic fuel injection ensures that that engine never runs too lean, never runs too rough, therefore the valves don't get burned, therefore the piston rings don't lose tension, and that engine lasts, and lasts, and lasts, and is very fuel efficient.
Now let's talk for a minute about dynamic stability control. How does it work? So you go hard into a corner with dynamic stability control, four things happen, maybe five. First off, one wheel gets added height to maximize the amount of tire on the road, the amount of rubber on the road. And the opposing wheel actually gets reduced suspension. A third wheel gets a little braking, and a fourth wheel gets a little more power.
All the time that's happening, it's monitoring for traction and skidding. You sail through that corner and you think, I'm one hell of a driver.
Think again. Let's fast forward to today. Move over to this side. So today, lower left-hand quadrant, 93% of the time, your car is parked. You own your car, it's parked. 93% percent of the time, it's sitting in your house, in a parking lot while you're shopping or at work. And some of us question why the youth of today and the millennials of today say, I don't even want to drive.
Can you really question that? When you're going to spend that kind of money and 93% of the time, it's not used? And then you have to pay insurance, and maintenance, and everything else on top of it? Think again. I think they're the smart ones, in my opinion.
So new models unfold—Uber, Lyft. The OEMS—they wake up. Oh, my God. What's happening to my world? People aren't going to buy as many cars. Let's go to the top right-hand quadrant. Everything is shared. Everything is driverless. 70% fewer cars. Far less bridges and roads. Much greener economy. Far less dependence on oil. Think about what that does to the global world.
There's one thing I know—before we get there, all four quadrants coexist. Massive disruption, big losers, big winners. Legal profession wins, and wins, and wins. That one, I guarantee you.
So I have to admit that I started my career in 1979 at General Motors, lonely electrical engineer who left after nine months. And I can remember telling my boss that GM needs to have an electrical engineer as the CEO of the company. Still hasn't happened. I would argue today it needs to be a software electrical engineer, but that still hasn't happened, either.
Back then, we had a 10- to 12-year development cycle for a new vehicle. I was at AC Spark Plug in Flint, Michigan. Probably drank too much of that water, and that's probably why I'm up here. Today, it's a six- to seven-year development cycle. And yet, and yet, who are the new entrants into this market? NVIDIA, Google, Apple, Baidu, Tesla.
They're software companies. They're electrical engineering companies. They can develop an incredibly complex product in a year or in two-year development cycles. These guys are scared to death, and they should be, and their entire tier should be scared to death. It used to be one ECU—Electronic Control Unit—per major function in a vehicle. Not anymore. It's an integrated ECU. It includes safety systems, infotainment, ADAS, and more.
What does that mean? Multiple operating systems, more memory, more sensors, more radios. And yet, it has to be reliable and secure. The only way this happens is with those FinFET chips, advance packages and modules, and boards, using next-generation technologies like MEMS, and silicon photonics, wireless, and more. So let's push into that for a minute.
So the good news, if I'm designing a next-generation ADAS chip, I don't start from scratch. I can get bleeding-edge IP from ARM or from Cadence. We have our own Tensilica Xtensa and Vision processors. They're used for, obviously, vision processing, audio management, radar/LIDAR management, and more. It's all built with functional safety in mind.
In fact, last year in Germany, the combination of our Tensilica processor, a digital signal processor with a convolution neural network algorithm on a Xilinx FPGA, did better than a human in terms of traffic sign recognition, at 99.8%, and set the industry standard.
So I told you that I have a printed circuit board design group, too, and we talked about Simulink and that integration with PSpice. So we have that great partnership with MathWorks. We have partnerships on the component side, so you can populate your boards and get what you need. We have, of course, integration on the mechanical side, because we can't do it all, with good companies like PTC and Dassault, Autodesk, Siemens, and others.
Think about rigid flex designs. Think about an automobile. And think about the small cavities where you have to insert things in it. And of course, our design teams work on really high speed, big performance, 64-layer boards for networks and servers in the cloud, but we also work on really small boards that have to bend and have flexible connectors, like you see in the top right-hand corner there.
The rearview mirror in a car used to be a rearview mirror. Not anymore. It's chock full of electronics. One of the pieces of that electronics is a forward-facing camera module. It's key to the safety system. It's key to the driverless vehicle, long term. Think about where it is. It's up against the windshield. Think about the summertime and the sun is beating on it. Now think about the wintertime, and it's iced over. And that ice has come through and it's on the inside.
Think about thermal expansion, electromagnetics, power integrity, signal integrity. That module's got to last the life of the vehicle, 10 to 20 years, or you’d better know when it's not going to last the life of the vehicle, just like we saw with the examples from Marianne.
Wireless is another new fabric making its way into the vehicle. We all love having Bluetooth and Wi-Fi in the car. For those of us that are observant of other drivers and are not playing with our phone, we're scared to death when we see the person driving and also playing with their phone all the time. A lot of people still think that the steering wheel is somehow connected to the wheels. It's not. That went away a long time ago.
It's all drive-by-wireless technology. It's amazing. We're getting the weight out. We're getting the performance up. We're getting the cabling minimized. Now, there's a process. But for the driverless car to be successful, we're going to need to get a minimum of 25 gigabits of data in and out of that vehicle every hour. That will only happen with photonics at the cloud level and photonics in the car itself.
So I live in Pittsburgh, Pennsylvania, because my spin-out from Carnegie Mellon University was acquired in 2004 by Cadence. And in Pittsburgh, like other great cities across the globe where there are strong universities on the engineering front, there's a lot of interesting things going on. So Google's there. Apple's there. Uber has a huge development team there. Hired 42 professors out of Carnegie Mellon University about 18 months ago, as an example.
Ford is there. GM is there. Uber has 81 driverless test vehicles on the streets of Pittsburgh. Have been on the streets of Pittsburgh for about 15 months. If you order an Uber in Pittsburgh, you may get a driverless one or you may get a driver-based one. Now, if you get a driver-based one, there's a driver. If you get a driverless-based one, there are two people in the front.
There's a comatose backup driver on the left, and there's a deaf, dumb, and mute software engineer on the right. And the code is streaming by. And if you ask him a question, like I do, then you'll never get another driverless-based one because I don't get them anymore. I'm on the list. And what does that loop do? They don't like that.
On the top of that vehicle is that LIDAR system—in fact, Marianne talked about LIDAR systems earlier—that rotating mechanical beacon that's on the top of that vehicle. Let's think about that. It costs more than the vehicle itself, and here comes a new fabric. Let's think what it might look like, since I have some insight. How about a system and package? How about an SoC control module using advanced node design as the brain of that chip?
How about a second chip that has the MEMS state on it, that has the sensors on it, and that's getting thousands of data points per second coming in. And then where are we going to feed that data? How do we feed it up into the cloud and into the local system? We have a third chip, a silicon photonics chip. All these chips are being developed with different partners inside of Cadence. Sometimes it's all Cadence, or other people in the industry.
And then there's a laser chip all packaged up, all put on a board, all put it into an electronic control unit. And while that's being designed, that system architecture is all done in MathWorks, and I can pull in MathWorks toolkits, like a beamforming offload for that algorithm that I need. That's the world, and that's the Cadence world, and that's why we have this partnership unfolding.
So like you, we envision an incredible future. I've never seen so much opportunity. I'm very envious. I took driver's training in 1972. How many of you people were not born yet in 1972? How many hands, not born yet? Oh, God. This is depressing. You have an incredible opportunity, just incredible opportunity. I'm so glad you're here to enjoy it.
So whether it's at the IP, or the chip, or the package, or the board, the subsystem, or the system itself, not by ourselves, but with great partners, we want to work on helping you design that system. We want to work—increasingly important—on how you analyze, simulate, and verify that system because mistakes can be deadly and expensive.
So I had the privilege of having a mother that was an English teacher, and she could recite poem after poem from memory. Of course, I can't do that. But this was one she recited. It was from Robert Frost, penned more than a hundred years ago. "Two roads diverged in a wood, and I; I took the one less traveled by. And that has made all the difference."
William Shockley came to a fork in the road in the late 1940s. He could have stayed on the left and just done a slightly better vacuum tube, slightly better performance, slightly cheaper to make. But he looked at the sand on the right and said, what if that's an electrical circuit?
Now, that's thinking outside of the box. So I hope I have inspired you to think outside of the box, too, as you develop your next autonomous anything, as you transform this world for us. And I'm hoping that in the not-too-distant future, you can make little Tommy and big Tommy much happier. Thank you so much. You've been a great audience. Thank you.
Featured Product
MATLAB
Select a Web Site
Choose a web site to get translated content where available and see local events and offers. Based on your location, we recommend that you select: .
You can also select a web site from the following list
How to Get Best Site Performance
Select the China site (in Chinese or English) for best site performance. Other MathWorks country sites are not optimized for visits from your location.
Americas
- América Latina (Español)
- Canada (English)
- United States (English)
Europe
- Belgium (English)
- Denmark (English)
- Deutschland (Deutsch)
- España (Español)
- Finland (English)
- France (Français)
- Ireland (English)
- Italia (Italiano)
- Luxembourg (English)
- Netherlands (English)
- Norway (English)
- Österreich (Deutsch)
- Portugal (English)
- Sweden (English)
- Switzerland
- United Kingdom (English)
Asia Pacific
- Australia (English)
- India (English)
- New Zealand (English)
- 中国
- 日本Japanese (日本語)
- 한국Korean (한국어)