eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
PALO ALTO, Calif.— Conventional methods of fabricating semiconductors will remain the same for about a decade, but the Department of Defenses research agency is already considering alternatives, a top government official said Tuesday afternoon.
Robert Leheny, deputy director of the Defense Advanced Research Projects Agency (DARPA), told an audience at the Hot Chips conference here that the government thinks Moores Law will remain viable for about a decade, as the semiconductor industry pushes to 0.20-micron line widths, or even 0.10-micron.
Beyond that point, however, the semiconductor industry must seek alternatives. And although the industrys finest scientists are involved in hunting down a solution, DARPA is investigating a variety of alternative technologies, Leheny said.
DARPAs role is to develop technologies that can be used to bolster the U.S. military forces, especially to improve their ability to process intelligence. For example, AWAC radar aircraft and Aegis cruisers can be used to gather intelligence on enemy forces, but the agencys new emphasis is on small, unmanned devices that can be used in place of soldiers or more-expensive intelligence devices.
“The underlying idea is to reach out into the analog world with an analog sensor and capture a single characteristic of something,” Leheny said. “Typically, [the subject] being observed is recorded using an analog signal, which then must be converted to digital and then sent to the processor and processed for our purposes.”
One example would be a drone that could autonomously seek out and report what it sees, in human terms: “Is it a tank, is it a missile, is it a school bus?” Leheny said. The goal: the elimination of misidentified targets, such as the Iranian airliner the U.S. military accidentally shot down in 1988.
Next page: DARPAs top technology prospects.
DARPAs Top Tech Projects
Leheny said DARPA is investigating a laundry list of various technologies; he didnt indicate whether the agency favored a particular approach or specify the amount of funding for each initiative. DARPA does not officially support any technology, he said.
Among the technologies under DARPA scrutiny:
- Alternative manufacturing processes: Although the vast majority of all semiconductors are manufactured on a CMOS process, Leheny said the agency favored the fast clock speeds rendered via silicon-germanium fabrication techniques. Silicon-germanium is designed for communications, promising lower power consumption than conventional CMOS chips. DARPA and NASA have also investigated doping silicon with indium phosphide, which the space agency has used to fabricate solar cells. DARPA withdrew from SEMATECH, the semiconductor industrys manufacturing consortium focused on manufacturing research, in the late 1990s.
- Photonics: DARPA has projects under way to use photonics in both silicon lasers and in sensors. Tunable silicon-based lasers could be used to transmit optical signals over optical fiber, replacing copper wire, he said. Eventually, optical connectors will replace silicon-based chip interconnects, Leheny said.Photonic sensors pose a more interesting problem. Images from radar transmitters, even from multiple sources, dont offer the type of resolution needed to hazard more than a basic guess about a targets identity. Placing large numbers of CCDs or other optical sensors in a tight array would allow a soldier to beam back high-resolution images, Leheny said. Unfortunately, a lack of battlefield bandwidth currently allows only still images to be sent, he said.
- Microelectrical-mechanical systems (MEMS) After the dot-com bust, interest in this technology abated, Lenehy said. However, DARPA is looking at using MEMS to develop chip-scale atomic clocks.
At some point, Moores Law will run out of steam. Then, Leheny said, DARPA and the semiconductor industry both will have to turn to new approaches. One possibility might be the use of carbon nanotubes, aligned into microwires. And the industry might have to turn again to the ingenuity that began it. “Intel isnt going to just up and quit,” Leheny said.