BOSTON It's high time displays started adapting to application architectures, not the other way around, says Gary Starkweather, an architect at the small, two-year-old Hardware Devices Group of Microsoft Research (Redmond, Wash.). Credited with having invented the laser printer during his tenure at Xerox Corp., Starkweather these days is busy prototyping his vision of the display of the future so that Microsoft will be ready with the software and graphics structures needed to support it.
Starkweather will give a glimpse into his vision next month at the Society for Information Display conference, in a paper describing what Microsoft calls the IMODS, or integrated MEMS (microelectromechanical system) optical display system. The IMODS is not to be confused with the iMods (interferometric modulators) being developed by Iridigm Display Corp. (San Francisco), which are also MEMS display devices.
"Wouldn't it be nice to have a display as big as your desk or as big as your wall?" Starkweather posited during an interview. "We spent a lot of time brainstorming and looking at all the available display technologies that could enable that, but there is no one display technology that goes all way from handheld devices to wall-size displays."
Large direct-view displays such as LCDs, he said, "get dicey when you go much larger than 30 or 40 inches, just because of having to pattern such large, thin pieces of glass. And forget about 80 inches for a wall display."
But what about projectors?
"Projectors have problems in getting enough pixels," Starkweather said. "Today, DMDs Texas Instruments' digital micromirror devices are good for 1,024 x 768 and maybe can go as high as 1,280 x 1,024, but it's hard to think about them ever getting to 3,000 x 2,000."
Having surveyed the technologies, "we built a logic table and started looking at combinatorial ways of making this happen," Starkweather said, coming up with "in the neighborhood of several hundred possible ways, using all combinations of acoustics, electronics, optics, emission, etc."
Then "it dawned on us that putting silicon and optics together, much as other people are doing for crossbar switches on networks, might have some real payoffs," Starkweather said. "Our logic was that we could make any display by basically putting small display modules together in arrays."
The hardware group is now prototyping the display it thinks the future might hold: little rear-projection modules, measuring between 1 and 4 cubic inches and delivering a display matrix of 50 x 50 pixels or so. The relatively modest 50-pixel/inch resolution would be "coarse for a desktop monitor but quite acceptable for, say, 4 x 6-foot displays."
The light modulation function of the IMODS is provided by MEMS devices being designed by the group's Mike Sinclair, who Starkweather said will give a future paper on the technology. Various kinds of light sources are appropriate, Starkweather said, depending on brightness requirements. "If you want very high screen brightness for outdoor use, for example, you could certainly put a very bright light source in there. They could be lasers or tungsten lamps or LEDs. As long as you don't feed enough energy into this system to overheat or melt something, you can put a lot of light through it."
The MEMS modulators in the IMODS are tiny light shutters, actuated by tiny MEMS motors, Starkweather explained. "You can etch through-holes in silicon and image spots from a lens array through these holes, which can be occluded or not by a little shutter. It's a very simple way to merge mechanics and molded optics to produce precise but nevertheless low-cost modules."
Providing MEMS shutters for a 50 x 50-pixel display would require 2,500 shutters per module, he said, but "putting 2,500 mechanisms on a piece of silicon costs you 25 cents.
"We're trying to get on the silicon power curve. One of the beauties of optics is that aberrations in optical systems scale with the focal length, so a 1-mm lens with the same shape factors as a 100-mm lens has only 1 percent of the effect of the aberrations. You can get away with things on a small scale that you could never get away with on a large scale."
"We're starting to see papers on small integrated optics now," Starkweather said, which supports the group's feeling that "we're certainly hunting in the right part of the forest."
The economic advantage
Building displays of various sizes and formats out of small integrated modules makes economic as well as technical sense, according to Starkweather. "If you look at current LCDs, there's 14-inch, 15-inch, 17-inch, and so on, and because there's no one standard size, there are tooling issues and economies-of-scale issues that keep the cost from plunging as it could."
With a 50 x 50-pixel module used to make 1,000-module displays, "if I sell 100,000 displays, I'll now be selling 10-million modules," he noted. Therefore I can afford to put the tooling into this little display module that I could never afford with the size distribution in other displays."
The concept also "allows one very powerful feature, at least to my mind but no one's proven it yet, Starkweather said. "Current displays usually have a 4:3 or 16:9 horizontal-to-vertical ratio, and if you want something different, there's no chance. The architecture has to conform to the display. Wouldn't it be nice if you could do it the other way?"
With small modules, "quantized in 1-inch increments, you could have a 17:39 format if you wanted, and what this now starts to give you is a much more flexible structure to embed displays in devices, walls, buildings, whatever," he said. "You start to have something with a lot more flexibility to it."
Such issues as physically integrating multiple modules and eliminating any visible seams between them should present no problem, according to Starkweather. "We're still pretty early in our work, but we've tried to look at all the show-stopper issues. At 50 pixels/inch, you'd need to position modules to 1 or 2 thousandths of an inch so you wouldn't see any variation. That may seem too precise over a large area, but it only has to be between modules, not over the entire diagonal of the screen."
The flexibility of small display modules provides another opportunity to free architectures from having to adapt to displays, Starkweather said. The content on today's displays is constantly updated, whether the content has changed or not, he pointed out. That's a burden on the system and an unnecessary throwback to the days of the CRT.
Freed for the future
"If you look at the CRT, where things all started, the raster goes along whether I put anything new up there or not," said Starkweather. "Therefore, I am updating simply because that's the way the display works. If I had a display system where if nothing changes on the display, there's no sense in sending data to it, I could be much smarter about how I parsed out the data on the data bus, and, hence, I could reduce the traffic that goes through the electronics subsystem to only those devices that require changes."
Such a "selective updating" scheme will be part and parcel of a broader architectural shift, according to Starkweather. "What we really want to do is drive more integration into the module itself. Put the graphics processing right on the module itself, for example, and you'd be able to send a rectangle out and let each module figure out whether it needs to do anything or not. Why should I have the OS sitting there trying to prejudge all that stuff?"
Microsoft's stake in future displays is clear. If the future really holds IMODS-like architectures, "they will need software and graphics structures that deal with them in proper ways," Starkweather said. "That's why we're looking at this."
The company, he said, wants to be ready "with the software and user experience tools that make the use of these devices really a positive experience."