Computer Cubed : Electronics: As industry runs out of ways to reduce the size of circuitry, Irvine Sensors’ three-dimensional chip-stacking technology is starting to get lots of attention.
Imagine trying to stack a pile of 128 computer processor chips so precisely that the thousands of tiny circuits on each of the chips match exactly. Then try to fit the package in a space the size of a sugar cube.
That’s what a small chip research firm here called Irvine Sensors has been trying to do for 10 years. It has been a lonely task--and an expensive one: It swallowed more than $30 million in research funds at a company with sales of just $3.5 million last year.
Until now, the company had been working in obscurity. But as the world’s multibillion-dollar chip industry runs out of ways to reduce the size of electronic circuitry, Irvine Sensors’ three-dimensional chip-stacking technology has become the center of attention.
“We developed the technology a couple of years before it was needed,” said James Alexiou, chairman and president. “Now the market is closing the gap.”
Although some industry experts describe the Irvine Sensors technology as promising, its success will depend on the company’s ability to adapt it for mass production. Some experts are skeptical about whether that can be done.
Chips are the building blocks of computers and other electronic equipment. The earliest chips, made in the 1960s, could fit between 100 and 1,000 circuit components such as transistors on a piece of silicon. Semiconductor companies today pack millions of components onto thumbnail-size chips.
During the past decade, chip makers have improved processing speed by reducing the size of the chips. The more dense the circuits, the shorter the distance an electrical signal must travel between chips and the faster the processing time.
Now there are physical limits to size reduction. The tiniest of dust particles can ruin the sophisticated circuitry, and electrical signals will interfere with one another if circuits are packed too densely.
Chip packaging, then, is becoming a hot research topic at such companies as Texas Instruments Inc., Boeing Co., Apple Computer Inc., International Business Machines Corp. and others. The Microelectronics and Computer Technology Corp., an Austin, Tex., semiconductor research consortium, is working on new packaging technology with the goal of ensuring that the United States retains its lead in the field.
The stakes could be high: market researcher Dataquest Inc. of San Jose estimates that the market for conductor packaging could be $18 billion worldwide by the year 2000.
“What (Irvine Sensors) brings to the party is what a lot of companies are striving for, and that’s a three-dimensional package,” said Alan Hirschberg, associate director of engineering at Teledyne Microelectronics, a chip maker in Los Angeles.
Some technology companies are looking at the creation of multiple chips side by side on a single module or at stacking perhaps 10 chips in a single package. But Irvine Sensors’ high-density approach is perhaps the most ambitious scheme, industry experts say.
“It’s an exciting technology, and the Irvine Sensors approach has the most promise to achieve such a level of density,” said Eric Fossum, a research supervisor at the NASA Jet Propulsion Laboratory in La Canada Flintridge. “It’s a very difficult technology problem, but it only takes one bright person to have a good idea.”
The bright person at Irvine Sensors is John Carson, co-founder and chief scientist. He and Alexiou founded the company in 1980 to develop sensors for a missile-defense system even before President Ronald Reagan proposed the controversial Strategic Defense Initiative program (also called “Star Wars”) in 1983.
At the time, however, the technology was way beyond the computer industry’s ability to apply it commercially. Computer makers were finding other, less complicated ways to produce smaller, ever-faster microchips. And Irvine Sensors was unable to find any company willing to help finance its research efforts.
The company, which went public in 1982, lost $8.8 million between 1983 and 1990. It was able to stay afloat mainly with the help of Pentagon research grants. The military was interested in applying the technology for use in future missiles.
Developing the technology took 10 years and three times the money originally expected because of the intricacy of the electronics involved, the company said. For example, the company looked at more than 1,000 different types of glue before it found the right one.
“I don’t think anybody else was crazy enough to do it,” said John Stuart, chief financial officer. “We’re the only ones that had the arrogance.”
The company also had to solve problems such as how to make uniform paper-thin layers of silicon and how to remove from the cube the heat that was generated by the circuitry for operations such as processing data from sensors. Carson said the company has several patents for the techniques that it developed.
Still, although some experts believe in Irvine Sensors’ technology, many are skeptical that the company can find a ripe market that would justify mass production of the cubes.
“Three-dimensional packaging is certainly important as the distance between chips becomes a significant factor in issues of speed,” said Karl Lofgren, chief scientist at Western Digital Corp., a computer components maker in Irvine. “Will the Irvine Sensors technique dominate? The jury is still out. It will come down to a question of price, and it is an unproven technology in terms of mass production.”
The company acknowledges that it has a long way to go before its chip-stacking technology can be used in mass production. Once it finds customers, Irvine Sensors would perform custom manufacturing in low volumes, but it would license its technology to third parties for mass production.
Stuart said one computer manufacturer, which he declined to name, has said it could be manufacturing millions of the cubes by the end of 1992. Each cube would generate a small license fee for Irvine Sensors.
The company believes that the military has little alternative but to use the technology. Since last October, the company has received development contracts from each branch of the armed forces to develop cubes that could be used in sensor-based missile detection and tracking systems, Stuart said.
Those contracts bring the company one step closer to full-scale development of a chip stack for particular military uses that would be mass-produced during the late 1990s. And mass production itself could lead to further uses as the unit costs come down.
“After the Persian Gulf War, there is more interest in their work now because of their involvement in smart weapons,” said Lorraine Maxfield, an analyst at Paulson Investment Co., an investment bank in Portland, Ore.
With prospects developing on numerous fronts, Stuart and Alexiou believe that the company’s years of hemorrhaging money are over. For its fiscal year ended Sept. 30, 1990, the company lost $788,100 on revenue of $3.5 million. But for the most recent quarter, ended March 31, the company reported net income of $26,900 on revenue of $1.2 million--its first quarterly profit in six years.
“This is still not a stock for orphans,” Alexiou said. “It’s a high-risk company, but the risk has gone a hell of a lot, and the potential rewards are still there. We’ve developed a lot of credibility. We think the electronics world will turn down the road of three dimensions. . . .”
“And there will be a toll gate on that road,” Stuart said jokingly.
Chip Sandwich Name: HYMOSS, an acronym for Hybrid Mosaic on Stacked Silicon
What: A space-efficient method to stack semiconductor chips, the basic building blocks of electronic equipment
Size: Same as a sugar cube.
Capacity: Can stack as many as 128 memory or processing chips.
Development cost: $30 million
Years in development: 10
Manufacturing: 52 steps required
Applications: Military sensors; high-powered computer workstations; neural networks, which mimic the functions of the brain; miniature hard-disk drives for laptop computers.
Source: Irvine Sensors Inc.
