Imagine a cell phone that never runs out of juice and delivers crystal-clear reception, or an inexpensive handheld computer that can get loads of information off the Web without running out of memory. Don’t start queuing up yet, but the radically new microchips that could make such devices possible are being pieced together even as you read this magazine. And for one of the most intriguing approaches to making such advanced chips, look no further than Seattle’s Impinj.
Founded last year, Impinj ventures onto the microchip scene with an impressive lineage. Caltech engineer Carver Mead, one of the cofounders, is a near legendary chip-design pioneer who has 21 previous startups under his belt and consulted for Intel in its early days. University of Washington computer scientist Chris Diorio, the company’s chairman and cofounder, did his graduate work with Mead. Diorio’s work on computers that mimic the nervous system won a Presidential Early Career Award, among other prestigious grants. And that work also sparked the technology that is central to Impinj’s innovative chips: a silicon transistor that can adapt to its environment, recalibrating itself on the fly in response to changing signal quality or a device’s new needs.
The Lilliputian transistors on today’s computer chips are perfectly suited to processing digital information, says CEO William Colleran. They turn on and off like a light switch to represent the zeroes and ones of a digital signal. But a device like a cell phone must process analog signals as well-the sound of your voice picked up by the microphone, for example-which consist of continuous waves rather than discrete pieces of information. As a result, such devices must include additional, larger transistors for processing analog signals and converting them to a digital format-taking up space, adding to power demands and making the devices more expensive to produce. Impinj’s solution? A digital-sized transistor that can process analog signals as easily as it handles ones and zeroes. Colleran calls it “the transistor equivalent of a dimmer switch,” able to be on, off or any position in between. What’s more, he notes, each transistor can continuously recalibrate its position, allowing the chip to constantly adapt to signal interference, degraded hardware and new tasks. And, as unusual as they are, Impinj’s transistors can still be fabricated with standard chip-making techniques, so manufacturers won’t have to retool their factories.
Backed by $15 million from Seattle-based Arch Venture Partners and Madrona Venture Group, Impinj is betting that this “self-adaptive silicon” will improve analog/digital signal processing, while also offering such perks as longer battery life and improved performance. But the real home run, says Gartner Dataquest analyst Stan Bruederle, might be that the chips’ flexibility could save manufacturers from having to make a new, highly specialized chip for each new type of device-which requires expensive new templates-as is usually the case today. “Imagine, instead of processing a bunch of different transistors for different devices, you can manufacture one product in high volumes that can be programmed within the various devices later on,” Bruederle says. “This is much more cost effective.” A handful of other companies have also come out with chips that can be programmed to different tasks, but unlike Impinj’s chips, those require a new software upgrade for each shift in function.
For the time being, Mead, Diorio, Colleran and their colleagues are only willing to say they’re working on “adaptive communications”-that is, building a better cell phone. However, with all the possibilities opened up by Impinj’s approach to chips, industry observers are keeping their eyes peeled for a lot more. “I know these guys,” says David Gifford, a professor of electrical engineering and computer science at MIT. “I have no doubt that they have other tricks up their sleeve.” When Impinj does decide to show its hand, it could well change the way we think about chips.