Design News for Mechanical and Design Engineers

The electronics industry’s next “killer app” is still a mystery. It might be lying quietly in a nondescript development lab in the Silicon Valley, or it might still be rattling around in the sleep-deprived brain of some brilliant grad student in India.

Whatever it is, though, it’s sure to be disruptive. Experts expect it to be an embedded application, probably software. Their guess is it will be wireless. And they know that when it arrives, manufacturers of processors and transceivers and memories and CMOS imagers and a hundred other components will begin scrambling. They’ll build hardware for it, thus allowing it to be applied to multiple industries, across multiple markets, until it touches virtually every consumer, from young to old, rich to poor, creating scores of multi-million-dollar businesses.

“Software will be the driver,” says Sam Fuller, vice president of research and development for Analog Devices Inc. “It always has been in the past, and it will continue to be in the future.”

Experts point to programming languages of the past — such as COBOL and FORTRAN — or software applications — such as VisiCalc and WordPerfect — and proclaim that the new “killer app” will follow the same patterns. It will take otherwise dormant hardware and turn it into money.

Still, no one knows what the killer app will do or what it will look like. They only know that it will need superior performance and low power to be successful.

“Every time we significantly increase the performance of a technology, new applications pop up,” says Gene Frantz, principal fellow at Texas Instruments. “But the really exciting applications are the ones that completely blindside us.”

While no one can foresee those blindside applications, that doesn’t stop the industry’s smartest engineers from gazing into the crystal ball. And most of those crystal-ball-gazers are guessing automotive or medical.

Many believe the long-term killer app is the driverless vehicle. After all, automotive researchers did the unthinkable less than two years ago, enabling six driverless vehicles to cross a 132-mile, off-road desert course without a shred of human intervention.

Those who believe autonomous driving will be a killer app aren’t merely basing that prediction on the results of DARPA’s 2005 Grand Challenge, however. They’re looking at the pace of innovation in that area. DARPA’s first such race, a mere 19 months before the successful 2005 Challenge, was characterized by comical robotic vehicles spinning in circles, stalling atop rocks and toppling over. The national press chortled over the debacle. Their laughter stopped in 2005, however, when Stanford University’s “Stanley” and five other robotic vehicles crossed the finish line. In graphic fashion, it showed the world how far robotics had come in just 19 months.

“The biggest thing Stanley taught us is that autonomous cars really are possible,” says David Stavens, a Stanford Ph.D. student and co-creator of Stanley, which won DARPA’s $2 million purse in the October 2005 race. “Stanley drove flawlessly in the Grand Challenge, and in the national qualifying event, and for many hundreds of miles before the race.”

To be sure, automakers aren’t laying plans for the arrival of the totally autonomous vehicle. But they’ve got autonomous features — adaptive cruise control, lane-keeping, and collision avoidance — on board vehicles, or at least waiting in the wings.

Depending on its rate of arrival during the next decade, autonomous vehicle technology could propel suppliers to boost their manufacturing of certain sensors, such as radar, vision and long-range LIDAR. Then, it would motivate makers of “downstream” electronics to enhance the performance of their devices. Analog-to-digital converters, for example, would be needed to deliver the sensor data to processors. And the processors themselves would need vastly improved bandwidth to act on the data, so actuators aren’t waiting on data when they should be acting.

“It takes a significant amount of signal processing to put a camera on the front of your car, keep it in your lane and prevent it from hitting the obstructions ahead of you,” says Frantz of Texas Instruments.

Frantz foresees the use of higher dynamic range CMOS imagers as being critical to the development of such vision-based applications. Whereas eight to 12 bits per pixel is good enough for those sensors today, he says, future such imagers may require as much as 24 bits per pixel.

“Driving down a road where the lane markers are light gray and the road is light gray and the sunlight is going directly into the sensors, you’ll have to be able to identify everything in your path,” he says. “You’ll have to ignore the sunlight and see the lane markers. And the way to do that is to have more bits per pixel.”

Moreover, experts say, machine vision is going to put a real strain on makers of processors to boost the performance of their products. Rather than using multiple “execute boxes” (i.e., multiply/accumulators) per DSP, for example, makers of such devices will have to adopt new computing architectures.

“We looked at going wider, but there are very few applications that can keep many execute boxes busy from a single instruction stream,” says Fuller of Analog Devices. “So in the future, we’re going to have to look at multiple processors with multiple instruction streams.”

Frantz believes the industry will also see advances in converters. Vision and surveillance, he says, could see the emergence of “analog-to-information” converters. Instead of converting analog images to a digital format and sending that data to a processor, these converters would simply send the information directly to its destination. The technique would be desirable, not only because sending the images isn’t always desirable, but because conversion and re-conversion takes time.

Such devices could have broad appeal, enabling vision systems to step into applications in home security, manufacturing, robotics and autonomous guidance of all sorts. Still, experts foresee vision’s greatest value in the automotive arena.

“If you think about all the latency and bandwidth issues associated with sending all those pictures to a central processor, you realize what a problem it could be,” Frantz says. “In an automotive vision system, the latency might be the difference between life and death.”

Prognosticators also say the so-called killer app could emerge from the medical market. Wireless access to the Internet already started a revolution in the medical device world, where products such as pacemakers and implantable defibrillators are gaining access to the Internet.

Medical device manufacturer, Medtronic Inc., for example, has leveraged advances in wireless transceivers to produce an implantable defibrillator capable of using the Internet to alert physicians to potentially deadly problems, even while patients sleep. Moreover, medical manufacturers plan to take such technologies farther, adding the ability to monitor fluid build-up inside a patient’s thoracic cavity, or recording pressure inside the heart.

The key to developing such products lies in the enhancement of sensors, batteries, software and processors. Medical manufacturers say they’re looking for a new generation of sensors capable of operating reliably within the harsh environment of the human body for five to 10 years. Similarly, they say they need processors and batteries capable of running for a decade, so surgeons won’t have to go back inside a patient merely to change a battery. Software, too, is key, because with more universal software compatibility, it could give doctors the ability to more easily log device data into patients’ medical records.

“It’s the wave of the future,” says Mike Hess, vice president of product planning for Medtronic Inc. “Physicians will track patients in their homes, gather real heart pressure data and make medical changes over the phone, instead of having them come into the clinic several times per week for multiple physical examinations.”

Whether the revolutionary killer app comes from such areas, however, is at best unknown. The electronics’ industry’s smartest minds admit they can’t foresee the emergence of another VisiCalc, WordPerfect, Mosaic or ARPANET, all of which helped fuel an explosion of PC popularity. But they agree that applications such as those drive the electronics’ market simply by creating a demand that induces component makers to follow.

They also agree the next revolution will take place in the embedded space, not in PCs, and that it will make heavy use of wireless technologies. Significantly, though, the hardware will only serve as an engine; software will provide the fuel, they say.

“Wherever software develops, it will engage customers and drive volume,” says Fuller of Analog Devices. “And wherever the software doesn’t come together, there will be components and machines that don’t quite make it.”