A band of start-ups has worked for the past several years now to come up with a practical way to integrate a liquid-crystal display on a silicon chip. They're getting close enough that the first end products to employ these thumbnail-size displays are expected to hit the market later this year and analysts are coming out now with their first bullish forecasts.

The idea simply is to combine a light source and lenses in a microdisplay subsystem and project images through eyepieces or on to a large screen from the polished surface of a liquid-crystal-on-silicon (LCOS) device. Today's microdisplays are simply small LCDs built on glass for such limited applications as virtual reality goggles.

Developers see a host of potential applications for their LCOS microdisplays. Initially they include viewfinders in camcorders and digital cameras; headsets for virtual reality applications; head-mounted displays for military and industrial applications; and large-screen TV projectors.

Longer range applications are more "blue sky." Promoters see huge volumes going into such products as digital TV sets, flat displays for desktop computers, and portable information appliances and next-generation cellular phones that feature power-stingy microdisplays for viewing the Internet and reading e-mail.

So why bother spending a lot of money to develop this new generation of LCDs? Turns out that it's the same set of advantages that come from putting almost any kind of system on a single chip. Putting it all on one chip can significantly reduce cost, improve performance, and in the case of LCOS it can provide a higher image resolution.

There are two basic ways to employ LCOS: In one, the user looks through an eyepiece or viewfinder directly at the chip in such applications as cellular phones and eyepieces for digital cameras and camcorders. In the other, LCOS projects an image on a large display - a solid-state version, if you will, of Texas Instruments Inc.'s micromachined Digital Micromirror Device.

But there are still major hurdles -- big enough in fact that some skeptics don't see the LCOS going very far. For the most part, the LCOS chip fabrication is no longer considered a problem. But major improvements and lower costs are still needed in the optics and reflective lighting technology to realize fully the market potential of these LCDs on chips.

Backers are eternally optimistic, however. They are convinced that LCOS displays could become a major new application area for CMOS ICs early in the next decade once all of their problems are solved. Supporters also acknowledge that consumers also will have to feel comfortable in using headsets or eyepieces on portable systems to view e-mail or surf the Web.

"This is the last frontier for silicon," declares analyst David Mentley at Stanford Resources Inc., in San Jose "Silicon circuits are used to process images, and most recently, we've seen a lot of activity in CMOS imagers. And now about a half-dozen companies have identified the next step: integrating the displays and silicon functions together."

Mentley is convinced now that LCOS is about to take off. From almost zero last year, LCOS sales should rise to $9 million this year, he predicts, shooting up to $175 million in 2003.

Not surprisingly, the potential for LCOS products has caught the eye of some of the larger chip makers. National Semiconductor Corp. has teamed up with Three-Five Systems Inc. in Tempe, Ariz., to push LCOS-based microdisplays for a range of high-volume applications.

Motorola Inc. and Fujifilm Microdevices Co. Ltd. are collaborating on LCOS with Kopin Corp. of Taunton, Mass. And last November, Hewlett-Packard Co. launched an alliance with Displaytech Inc. of Longmont, Colo., to move a ferroelectric liquid-crystal technology into LCOS-based microdisplays this year.

Now end products using LCOS are coming. South Korea's Samsung Electronics Co. has revealed plans to use Displaytech's silicon microdisplays in a new rear-projection TV it will introduce later this year.

Three-Five Systems says more than a dozen potential customers are now evaluating its LCOS chips for use in everything from digital cameras to head-mounted displays. Colorado MicroDisplay Inc., another LCOS startup, is working with Planar Systems Inc. and In Focus Systems Inc. to develop a range of applications for its dynamic nematic LCOS technology. The Boulder, Colo., startup and Planar figure the products they alone are developing will require 1 million microdisplays over the next three years.

While developers dream of such impressive emerging markets for LCOS, some market observers continue to be skeptical over what is being promoted as the largest potential applications -- next-generation cellular phones, for example.

"Cellular phones are the most often mentioned as a killer application," says Stanford Resources' Mentley. "But when it comes down to it, there is no infrastructure to fully take advantage of this kind of display," he says, referring to current wireless-transmission speeds and computing-like functions in handsets.

To make it, liquid-crystal-on-silicon chips also will have to demonstrate price and performance advantages over existing microdisplays built with conventional LCDs that are fabricated in polysilicon deposited on glass or quartz substrates.

The most widely used technology here is high-temperature polysilicon utilizing tiny, transmissive LCDs containing driver circuits on the same glass or quartz substrate. But many analysts believe that this hybrid approach has nearly reached its limits. Some suppliers, however, mostly Japanese manufacturers, seem to be continuing to squeeze more performance and costs from these small- form-factor LCDs.

But shrinking devices and packing more pixels on the polysilicon surface won't be competitive much longer because the speeds of those LCDs are slower than that of silicon devices, according to Dan Schott, vice president of R&D at Three-Five Systems.

"To get the speed up, they have to use very large feature sizes in polysilicon, which means a large device with small apertures, higher expense, and lower resolution," according to the Three-Five Systems vice president. "When you get to very high resolutions, polysilicon becomes very limited because of the relatively poor mobility compared to single-crystal silicon. The big advantage of single-crystal silicon is that you can put all the circuitry that supports the display on the display," he points out. "Since the substrate and the circuit is silicon, you can integrate everything - row and column drivers, shift registers, controllers - the works."

Single-chip display technology will be able to leverage process advancements in the Semiconductor Industry Association's roadmap, believes Alan Marty, manager of Hewlett-Packard's Components Group in Palo Alto. Calif. "The ability to get the same resolution and the same electronics content in a smaller and smaller space makes a huge difference in whether something is attractive to mass markets," he says. HP is betting that Displaytech's LightCaster design, which uses ferroelectric technology, will produce the best image and color intensity at low power.

Kopin and its partner Motorola -- which now supply a hybrid microdisplay called CyberDisplay -- are expected to pursue a LCOS chip design in the future. For now, the CyberDisplay uses a transmissive LCOS structure that's taken off wafers fabricated by its Taiwan foundry, the UMC Group, and deposited on a glass substrate with integrated drivers. The transparent silicon film allows Kopin to use a backlight to display images from the device, while a magnifying lens creates a virtual image to the user through an eyepiece. Motorola currently supplies an integrated display controller and digital-to-analog converters and drivers on a single chip.

In high-volume applications, LCOS will win because of device integration and the Moore's law learning curve, says analyst Mentley. "What a single silicon backplane brings to the party is a higher resolution," he says. "Theoretically, you should get a few million pixels without a lot of cost because you're closer to the state of the art of IC fabrication as far as pixel size and geometry goes," Mentley adds.

But an LCOS device still needs a fairly intense light source to bounce images off its surface. Most developers today are trying to achieve this by using red, green, and blue light-emitting diodes located next to their microdisplay chips, according to Mentley.

Another critical component in a miscrodisplay system is the optical lens, which must provide a sharp image through eyepiece viewers or on to a large-screen display in projection applications. To be successful in creating major consumer applications, Mentley says that the entire cost of this display subsystem -- LCOS chip, light source, and optics - has to run between $25 to $50.

Glen Kephardt, marketing vice president for display products at Kopin, agrees. "If you're going to put something on the shelf, you just can't go above a certain price," he maintains. "We're looking to get display costs in the $20 range. . . . Motorola's got to get down to $3, $4, or $5 for the interface electronics," Kephardt notes. "And you cannot get there without high-volume manufacturing experience and high-volume markets."

For now, it's a chicken-or-egg situation. Costs won't fall until volume goes up, but volume won't go up until prices fall. "Right now, the manufacturing yields on these things are low," observes Barry Young, senior analyst at Display Search Inc. in Austin, Tex. He and other observers expect to see yields go up as consumer-product volumes begin to take off.

One application where LCOS products will find tough competition is in projection-displays. Texas Instruments' Digital Micromirror Device (DMD) reflects images from 500,000 mirrors that are micromachined on its surface. Pixels of light are turned on or off depending upon how each micromirror is tilted by a synchronous DRAM circuit built into the DMD substrate. The Dallas company has sold 150,000 devices in the past three-and-a-half years and expects to ship 100,000 this year, says Ian McMurray, European marketing director for TI's Digital Light Processing Group.

In late April, TI and Hitachi Ltd. announced a major pact to develop what they call the world's first all-digital, large-screen projection system for high-definition TV. Hitachi plans to use the DMD devices in new HDTV sets that will go on sale in Japan and the U.S. in the second half of 2000.

While TI is keeping a close eye on the LCOS efforts, the company figures it now has a three-and-a-half-year lead in the marketplace, claims McMurray, who is based in Northampton, England.

It's certainly a high-stakes business. Over the past decade or so, TI already has spent a whopping $300 million to develop DMD technology and products.

"We know how much it costs to be in this marketplace," McMurray points out. "Can these typically smaller companies afford what it takes?" -- Additional reporting by J. Robert Lineback