In Brief:
High-resolution projection displays developed by IBM Research promise images far
superior to those
produced by traditional cathode-ray tubes. Applications as diverse as graphics workstations, cinema projection and head-mounted devices that project images directly onto the retina are envisaged.
New projection displays developed at the Thomas J. Watson Research Center are producing eye-popping images that are far clearer and more detailed than those from the best cathode-ray-tube (CRT) displays. Because the projection displays marry silicon wafer and liquid-crystal technologies, they can be quite small and lightweight. Potential applications range from tiny devices that project images directly into the eye, to desktop displays and projectors powerful enough for use in cinemas.
With resolutions of four million picture elements (pixels) or better in large-screen desktop units, the new displays offer benefits to several types of likely users. For example, they will help designers who need to see extremely clear detail in their drawings. The technology also brings new levels of performance to front-projection applications, an already booming growth market. And it is readily adaptable to head-mounted applications for mobile users, for whom it could serve, for example, as an alternative display for an IBM ThinkPad® or as part of an entertainment device.
Better than the best
IBM researchers addressed the need for high-resolution, high-information-content displays more than two years ago. They wanted a solution that would improve on CRTs, which are limited in resolution and brightness, as well as by drawbacks in size, weight and cost as the information content increases.
Silicon technology looked like a promising direction, and it was an area in which IBM had considerable experience. "Over the years we've honed our capabilities in silicon technology and in developing optically flat wafers," explains Steven W. Depp, director, subsystem technologies and application laboratory at Watson. "That was helpful in realizing our idea of building a miniature reflective display on top of a silicon chip."
IBM did not pioneer the approach, but, arguably, it has developed the most successful implementation. "We wanted our design to be better than the best CRT," says Robert L. Melcher, leader of the projection display
effort. "The state-of-the-art CRT is the Sony DDM(TM), a 28-inch display used by air-traffic controllers. It can address four million pixels, but, for technical reasons associated with CRT technology, its resolution is only about half that number."
An interdisciplinary Research team of 20 specialists in silicon technology, packaging, liquid crystals, optics, electronics and mechanical design joined forces to create the projection display. In collaboration with development partners at IBM Japan's Yasu facility, the group has built a prototype 28-inch projection desktop monitor, with true four-million pixel resolution. "We are not just addressing all those pixels, we can actually distinguish them," says Melcher. "The success of the effort is truly a tribute to the teamwork and to the depth of Research's expertise in the various technical areas, as well as to the strength of our partnership with IBM Japan," he adds.
Compared with the target CRT, production versions of the projection display will have twice the resolution, half the weight and depth, consume half the power and could ultimately cost half as much.
Four million mirrors
The core of the display is a light modulator consisting of a 1.5-inch square array on a silicon chip. The active area of the surface of the chip contains four million square-shaped mirrors, each 16 microns in width. Each mirror constitutes an addressable pixel controlled by an individual transistor that lies just beneath the mirror. A thin
layer of liquid crystal is placed on top of each mirror; covering the liquid crystal is a glass sheet whose inner surface is coated with a transparent electrode that provides a common electrical potential for all the mirrors. This "sandwich" structure is key to the operation of the display.
When a voltage is applied to each mirror, it induces a corresponding electric field across the liquid-crystal layer, which changes its electro-optic properties. When polarized light is directed at the mirror array, it is reflected from the mirror through the liquid crystal. In the presence of a voltage, the liquid crystal rotates the polarization 90 degrees, allowing it to pass through the system's optics, thus creating a bright spot on the screen. A dark spot is created by keeping the voltage off, so that the polarization of the reflected beam remains unchanged and is therefore blocked by the optical system. Intermediate voltage levels and corresponding degrees of polarization allow 16 levels of gray to be displayed on the prototype.
Because it is a color display, however, the actual setup is a little more complicated. The incoming light is first separated into red, green and blue by a set of prisms, and each color is directed to a separate modulator, with its own array of four million mirrors. The reflected red, green and blue light - each of which is reflected to produce a 16-level gray scale - is then combined into a full-color image, magnified optically and projected onto a screen.
A variant of an IBM CMOS silicon chip-making process, with 1.5-micron line widths, is used to produce the active-matrix modulator. Critical to perfecting the modulator were an IBM proprietary chemical/mechanical polishing process, which assures a smooth, flat surface for the mirrors, and a light-blocking layer that prevents the intense light from disturbing the active-matrix transistors. The previous lack of such techniques had stymied the attempts of others to develop such a projection display.
Mirror design directly affects the overall optical efficiency of the displays. Reflective technology, compared with current state-of-the-art transmission devices (used in most commercial projectors and goggle-type viewers), allows devices to be made smaller and to produce brighter, higher-resolution images. "Our optical efficiency is quite a bit better," says Melcher. "We can reduce pixel size even further while maintaining optical efficiency; transmission devices will always need to be much larger to provide acceptable results at high resolution."
The team faced a major challenge in creating the liquid-crystal layer on top of the silicon wafer. "We started from a new direction in this area because silicon has different properties than the glass substrate found in transmission devices and flat-panel displays," team member Kei-Hsiung Yang explains. Computer simulations were used to find liquid crystals that would both reflect incoming light and operate at a lower voltage than the more commonly used transmissive liquid-crystal mode.
Elegant optical design
Reflective systems require sophisticated optics to illuminate the light modulators uniformly and to meld the red, green and blue images precisely on the screen. The team used several novel components and a few tricks to make the optics lightweight, compact and efficient. "Illumination is provided by a very small proprietary arc lamp," says Derek Dove, manager of optical systems at Watson. "We use a sophisticated beam splitter to polarize the light directed to the light modulators. It then redirects the light of rotated polarization, reflected from the light modulators, to the projection lens."
The optical system will vary somewhat depending on application, notes Dove. "Clearly, we will need a superminiature version for a retinal-projection device that can be worn on eyeglasses."
Handling large data sets
Addressing four million individual pixels, each with 16 or more levels of gray, creates huge data sets and new challenges. "The electronic interface is a major issue here," says Paul Alt, manager, display systems. "How do you get data to the display? How do you manipulate it? In the prototype, every color pixel gets refreshed 74 times a second from its own internal frame buffer, which requires a bandwidth of about 4 gigabits per second. We were able to achieve that by using a highly parallel approach."
Now, the group wants to go further. "On the prototype, we demo a standard, six-million-pixel photo-CD image, but we have to crop it to four million pixels," says Alt. "It would be nice to show the whole picture. Eventually, it may be possible to physically produce a 10- to 20-million-pixel array on silicon, but we haven't shown yet that we can handle
data sets of that size. There's a lot of work still to be done."
In the future, Alt and other researchers may face fundamental design questions. Can larger silicon chips be manufactured reliably? How small can pixels be made before diffraction degrades optical efficiency? What level of electronic integration can be fabricated in the silicon substrate? In the meantime, Research is working closely with the new IBM display business unit in Japan on critical silicon and liquid-crystal technologies, optical design and electronic architecture issues to move this technology into the marketplace.
Michael Sinclair is a freelance
science writer based in North Guilford, Connecticut.
