To find out what driving’s like when you have a sixth sense, I took a radar-equipped Audi A8 around the highways and byways of Stuttgart, Germany. It was great.
I couldn’t help but smile when I pulled behind a huge truck and, resisting the temptation to hit the brakes, focused on steering. The adaptive cruise-control system, which uses a new radar from Robert Bosch that can see hundreds of meters ahead, did the rest. The system gently nestled the car behind the juggernaut and accelerated at my command, so I was able to pull out into the passing lane, all the while getting the most out of the 4.2-liter diesel, which rapidly propelled me to the speed I’d selected.
The system did have its foibles. Once the radar locked onto the car in front of me, and when the car turned hard to the right and then hard to the left, the radar came unlocked. So I took control, applying the brakes well before the emergency braking would have kicked in. That episode was a little disconcerting. Still, I could easily get used to this gizmo.
Most people who have driven for a while using such a radar are loath to ever give it up. And the number of such devotees will only grow as this technology—which now adds about US $1000 to the price of the car—becomes more affordable. The first commercial system appeared in Japan in 1997, on the Toyota Celsior; others soon followed in some top-of-the-line models from the likes of BMW, Jaguar, Lexus, Nissan, and Mercedes. The market has been expanding at about 40 percent a year, and as prices fall, that rate should rise.
Today’s systems can dramatically reduce your risk of rear-ending someone else’s car, and when most cars have such radars, they will also be much less likely to rear-end you. Once every vehicle on the road is able to sense and avoid others, there’ll be no reason why they won’t be able to negotiate tailing distances among themselves. Eventually, they might even be sending radio messages about their intentions to one another and to monitors on the roadway over ad hoc communication networks. Smart roads may thus emerge organically.
The first step in that evolution, the democratization of radar, is clearly under way. Next year Bosch will release a less expensive version of its radar, with a range of 160 meters, two-thirds that of the one I tested. This won’t be a problem, though, because it’s intended for cars that don’t go nearly as fast.
Falling costs are the key, but of course, costs don’t fall by themselves. Engineers have done their part by ditching the expensive compound semiconductors in their radar sets in favor of the old standby, silicon—but a special form of silicon that’s been speeded up.
In the late 1960s workers at Mullard Research Laboratories, in England, developed a car radar system that operated at 10 gigahertz, and RCA used the same frequency in its 1972 system. To make the next step and cram such a radar into a small space—such as under the hood of a car—manufacturers had to shrink the array of antennas, keeping each antenna far enough from its neighbors to allow for good resolution of detail. They accomplished this task by moving first to 34 GHz, then to 50 GHz, and recently to 77 GHz. The choice of frequency has something to do with the absorption of microwaves in the air and a lot to do with legislation: The law places strict limits on power for the lower frequencies, which is why systems in the lower bands can look forward just a few meters, only enough to avoid fender benders in stop-and-go traffic.
To manage the higher frequencies, long-range auto radars have until recently required seven or more gallium arsenide–based chips to generate, amplify, and detect the microwave signals. That set of chips costs from $20 to $60—not all that much, it might seem. But those chips have to be connected and tested, and if one fails to work, it must be rooted out and replaced. This labor adds substantially to the cost of any radar based on gallium arsenide technology.
In 2009, the German chipmaker Infineon Technologies, based in Neubiberg, produced a system designed around a single silicon-based chip. Then it teamed up with Bosch and started supplying a more flexible, two-chip variant for radar systems in 2010 models of the Audi A8, Porsche Panamera, and Volkswagen Touareg. Not only are these new systems less expensive, they also have significantly better performance, allowing them to cover more than four times the area in front of the car, four times as accurately.
Even specialists in the gallium arsenide industry expect that silicon chips will grab most of the car radar market. Asif Anwar, director of the program for gallium arsenide and compound-semiconductors technologies at the market-research firm Strategy Analytics, in England, predicts that over the next three years, silicon’s share of the chip market for automotive radar will grow from nearly nothing to perhaps 60 percent. Although Infineon will then have captured most of the resulting $120 million market for silicon-based radar chips, it already faces the first signs of serious competition: U.S. chipmaker Freescale Semiconductor, in Austin, Texas, has just started sending samples of its silicon-based chip to automotive radar makers. Other companies are surely following suit.
Infineon has thus overturned the conventional view that silicon chips would never be able to generate, detect, and amplify high frequencies. The problem is that electrons move slowly through those chips—which is why a decade ago Infineon and a handful of other companies were using the faster gallium arsenide to build automotive radar chips. But in mid-2002 Infineon got out of the gallium arsenide business. A year later it was in discussions with Bosch about automotive radar chips based on silicon.
“At that time everybody thought this was not possible to do with silicon-based technologies,” recollects Rudolf Lachner, Infineon’s program manager for radar technologies. “But we did some high-speed circuits, such as voltage-controlled oscillators, which worked at 77 GHz.”
To realize such high speeds in a silicon transistor, Infineon’s engineers inserted into the heart of the device a thin layer that was four parts silicon and one part germanium. The idea was hardly new. Indeed, it can be traced to theoretical work that Nobel Prize–winning physicist Herbert Kroemer, now at the University of California, Santa Barbara, did way back in the 1950s. However, the world had to wait until 1975 for the first real device, made at AEG Research Center (now part of Daimler) in Ulm, Germany. Infineon’s claim to fame comes from boosting this kind of transistor to record speeds, thanks to improvements in internal configuration and material quality.
Adding that layer of silicon germanium alloy introduces electric fields that present the moving electron with the equivalent of a downhill path, speeding it up automatically. Now even transistors with 50-nanometer-thick base layers can reach the speeds demanded by 77-GHz automotive radar.
Switching to the new transistor delivers another benefit—very low noise levels. You can speed up conventional silicon transistors by thinning the base layer, but you’ll just impede the flow of electrons and increase background noise. To muffle it, you could try to reduce the resistance of the base by doping the silicon with traces of boron, whose atoms each have three electrons in the outer shell, rather than silicon’s four. Because there aren’t enough electrons to form all the covalent bonds required, you get a “hole,” or virtual positive particle, which moves freely through the crystal, increasing its conductivity. Unfortunately, increasing the base doping this way reduces the amplification, or gain. Working with a silicon germanium base layer gets around this problem because it makes its own contribution to the gain, offsetting the losses caused by doping. You can make the base doping very high, explains Lachner. “And by making it very high, you get a very low base resistance, which improves the noise behavior of your transistor,” he says.
The fundamental insight stemmed from work Infineon did in the early 1990s while developing chips for next-generation mainframe computers. That project never took off. Nor was the company able to market its chips to mobile-phone vendors: As conventional transistors shrank, their lower cost proved more important than the lower power consumption of Infineon’s chips. But soon after, it became clear to Infineon that this technology was a perfect fit for auto radar.
Perhaps it’s the fiendishly high speeds of the autobahns that have made Germany so keen on technology to avoid collisions. Or it could be government aid. In 2004, Infineon began a three-year automotive radar program with 10 million in subsidies from the German government. That project allowed the company to collaborate with automotive radar system makers Bosch and Continental and carmakers BMW and Daimler.
Infineon’s prototype could operate up to only about 80 GHz, good enough for use in an oscillator but not in the amplifier. That’s because for a transistor to deliver reasonable gain at a given frequency, it needs to top out at about three times that value. In 2007, by improving the quality of the boron-doped silicon germanium in the base, Infineon’s engineers increased the transistor’s maximum operating frequency to the requisite level and soon went on to produce the first commercial silicon germanium automotive radar chips, which ran at 77 GHz. Four years later, Infineon continues to churn out the chips at its huge fab in Regensburg, Germany.
Inserting the silicon germanium layer into the device requires no exotic techniques or extraordinary tools: Infineon simply uses 200-millimeter silicon wafers and grows thin silicon films on top using conventional chemical-vapor deposition. At the appropriate point during the process, a valve opens, germanium-based gases flow into the growth chamber, and a silicon germanium film forms.
One such wafer can yield thousands of chips. “This gives us enough headroom to produce as many automotive radar systems as we would like,” explains Lachner. In fact, most of the fab’s output of 10 000 wafers goes to other purposes. If Infineon somehow captured the entire automotive market overnight, it could easily satisfy the demand.
So why do other companies, such as TriQuint Semiconductor, in Hillsboro, Ore., and United Monolithic Semiconductors, in Orsay, France, still produce automotive radar chips based on pricey gallium arsenide? For one thing, gallium arsenide is still the biggest player in the radar market at the moment, and these firms can sell a lot of chips, at least for a few years. Also, these companies don’t necessarily have silicon production lines to switch to, nor would it make sense to build a full-blown silicon fab for car radar alone.
Cost isn’t the only thing driving change. It’s not only cheaper to use one Infineon chip (or two, in the fancier system); it’s also more effective than the handful of gallium arsenide chips it replaces. When Bosch upgraded Infineon’s product during the development of its third-generation long-range radar (dubbed, unimaginatively, the LRR3), both the minimum and maximum ranges of its system got better: The minimum range dropped from 2 meters to half a meter, and the maximum range shot from 150 to 250 meters. At the same time, the detection angle doubled to 30 degrees, and the accuracy of angle and distance measurements increased fourfold. The superiority stems from the significantly higher radar bandwidth used in the systems containing the silicon-based chips, says Thomas Fuehrer, Bosch’s senior manager for strategic marketing for driver assistance: “It is around 200 megahertz on the LRR2, and we are now using 500 MHz on the LRR3.”
Another selling point is the new system’s compact size—just 7.4 by 7 by 5.8 centimeters. “If you are comparing it with the competitor’s systems, this really is a very small masterpiece,” Fuehrer says. What it means is that automobile designers can stick this thing just about anywhere—even in the headlamp assembly.
The system employs four antennas and a big plastic lens to shoot microwaves forward and also detect the echoes, all the while ramping the emission frequency back and forth over that big fat 500-MHz band. (Because the ramping is so fast, the chance of two or more radars interfering is extraordinarily low.) The system compares the amplitudes and phases of the echoes, pinpointing each car within range to within 10 cm in distance and 0.1 degree in displacement from the axis of motion. Then it works out which cars are getting closer or farther away by using the Doppler effect—the change in frequency associated with motion that also causes us to perceive a train whistle to rise in pitch as it approaches us and fall as it pulls away. In all, the radar can track 33 objects at a time.
On the Audi A8, you receive two separate warnings when you get worryingly close to the car in front. First, a high-pitched alarm sounds, and a light appears on the dashboard. If that sound-and-light show doesn’t work, then comes a short, sharp brake to snap you out of your stupor. “Tests and studies show that most drivers will then immediately look forward at the road and notice if they are too close,” says Bernhard Lucas, head of Bosch’s department for developing car radar hardware.
Even braking may not prevent a collision: Statistics gathered by Bosch show that nearly half of rear-end crashes are caused by drivers pressing the brake pedal too softly. But if that happened in the radar-equipped Audi A8, additional braking would be applied automatically.
If worse comes to worst, the braking system goes into action by itself. “In rare cases where the driver is completely unable to do anything—he is helpless or half dead—full emergency braking is applied when the crash is really unavoidable,” says Lucas. Then the car decelerates abruptly, throwing the driver forward into the safety belt with up to six times the force of gravity but minimizing what would otherwise be a catastrophic impact with the car in front.
Of course, you’ll probably never have to call on such emergency powers to save your life. Few people even consider such features when purchasing a car. That’s why the day-to-day operation of the system is important for winning over the driver. Today, the benefits come mainly in the form of a radar-enhanced cruise control. You can set your radar to lock onto the vehicle in front and keep pace with it, braking and speeding up appropriately. You specify the following distance and the maximum allowable speed, which can be as high as 250 kilometers per hour (155 miles per hour).
It is interesting that when Audi, Porsche, and VW started making radar-ready cars last year, all three companies chose to use the radar as a driving aid rather than a full-blown autopilot. They thus reduced their liability for any accidents that might ensue. Today, it’s clear that the main roadblock for a software-based chauffeur are legal worries and perhaps the fear of the unknown. Should any automaker dare to take the plunge, the technology will not be lacking.
This article originally appeared in print as “A Driver’s Sixth Sense.”
About the Author
Richard Stevenson got a Ph.D. in physics at the University of Cambridge. There he focused on gallium nitride and other compound semiconductors, the beat he usually covers for Spectrum. This time, he writes about an advanced yet inexpensive car-radar system from Infineon Technologies that’s based on good old silicon. “With the shift to silicon, we may soon see luxury features like adaptive cruise control on economy cars,” he says.