Miniaturized displays are here, and are about to change the way we think about computer screens, video projectors, and electronic readouts for good. For the first time, displays will be compact, high-quality, and inexpensive enough to use in toys. However, things are not so clear cut from an engineering point of view. There are many different light modulators on the market (more every month) and every one is different: both in the technology it uses and in its advantages and disadvantages for various applications. Which succeed, and which fail, may have implications far beyond the display market.
At the core of all the new mini-displays, and what makes them so cheap, is silicon; all were designed to use standard VLSI processing. In this sense they all work on the same basic scalea few microns to a few tens of micronsthough some designs are easier to reduce than others. When it comes to actually manipulating light, however, two very different technologies have been developed. One, based on micromechanics, involves physically moving structures on the chip so that each pixel reflects or diffracts light in the right direction. One uses liquid crystals as a valve for polarized light. Within these two categories, however, switching speeds, color generation, efficiency, light source requirements and system bulk are all very different.
Speed and greyscale
Though they may seem unrelated, how fast a display can switch can be very much related to how many colors it can display. For instance, there are currently two micromechanical display contenders. The device in development at Silicon Light Machines (Sunnyvale, CA) is based on switchable diffraction gratings1. Their Grating Light Valve consists of tiny silicon-based ribbons, each just a few microns wide, that are alternately floating and fixed. In the device's off state, all the ribbons line up producing a flat reflective surface. This sends the light back into the light source where it is invisible to the viewer. When a pixel is on, its floating ribbons are electrostatically pulled back into the device, producing a groove that diffracts much of the incoming light, sending it out of the display at equal, but opposite angles. These beams can be recombined optically producing clean (black pixels are very black) monochromatic image.
Figure 1. How the GLV works. Figure Copyright Silicon Light Machines; used with permission.
Crucially, this technology is binary: it either diffracts light or it doesn't. In order to produce grey scale, it must do so over time. For instance, in order to produce 8 grey levels at 30 frames per second, the device would simply switch 240 times over that period. One off to seven ons for a particular frame would look like the lowest on grey and so on. Fortunately for Silicon Light Machines, their display can switch in 20 nanoseconds. This gives the GLV a theoretical frame rate of 50 million per second: enough for all the grey scale that most engineers can imagine. According to Rob Corrigan, Vice President of Marketing for Silicon Light Machines, the company is able to take advantage the GLV's switching speed in other ways, too. "Our technology can support 10-bit/channel grayscale on a workstation-resolution display using simple passive matrix driving logic," he says, "No one else that we know about can do that."
Texas Instruments' minidispl the Digital Mirror Device, works by physically switching mirrors rather than gratings.
Texas Instruments' mini-display, the Digital Mirror Device, works by physically switching mirrors rather than gratings, but is inherently incapable of grey scale in the same way as the GLV2. Again, a time division multiplexing approach is required. In comparison to the Silicon Light Machines Device, the DMD is slow: it switches at just 1000 frames per second. This is, conceivably, a future disadvantage. However, given the amount of data that you can communicate to such a tiny display, this is enough to be going on with for now. This fact, plus the device's extremely high efficiency, make the DMD competitive. [See related article, January 1995 OE Reports]
Of the companies using liquid-crystal-on-silicon reflective displays, Displaytech (Longmont, CO) seems the most committed to using ferroelectric materials3. As liquid crystals go, these are very fast: they compete with the TI device in the 1000-frames-per-second range. However, this speed advantage comes with the disadvantage, once again, of binary-only operation. Central Research Laboratories Ltd (CRL-Hayes, UK), which uses generally similar technology, is currently using both ferroelectric and nematic liquid crystals. The latter are currently much slower, but they are analog and can be switched to any grey level as long as the right voltage is applied. These analog systems have an advantage in terms of simplicity.
Thinking in color
However, with all of these devices, the penalty comes in using them to produce color images. The price is either bulk or power consumption and system complexity, all of which can affect the price. One well-known technique is to use three devices (one each for red, green, and blue) with a white light source. Each device has to be aligned with the other on a beam combiner assembly, which, in real terms, means that the color system fills a volume that is the cube of its longest side. This is no disadvantage for higher-end applications, such as video projectors, but the increase in both cost and bulk takes such systems out of the running in the "toy" market. The other approach is to use either a fast-changing color filter or colored light emitting diodes to select red, green, and blue sequentially. In each case, either the system designer is being given an extra constraint or the light source and display have to be somehow combined, making it bulkier. Also, the color filter is impractical because it consumes a lot of power (two thirds of the light is discarded at any particular time).
Perhaps recognizing this weakness, Texas Instruments has chosen to concentrate on the projector market. According to Dr. Larry Hornbeck, TI Fellow and inventor of the digital mirror device, they are concentrating on both ultralight and ultrabright products. Crucially, he sees a major advantage for TI's technology here. Unlike it's competitors, he says, "[The DMD projector] has the potential to achieve 10,000 lumen brightness levels because of its reflective design and ease of dissipating heat."
Undeterred, two companies working with liquid-crystal-based technologies, MicroDisplay (Richmond, CA), and Kopin Corporation (Taunton, MA), have come up with solutions to the color problem. Like Displaytech and CRL, MicroDisplay's devices are reflective. The difference is that is made up of red, green, and blue triplets. Unlike larger-scale displays, the RGB filters are not dyes or gels: they would be impossible to manufacture. Instead, MicroDisplay uses diffraction gratings that have been etched into the silicon substrate. The grating for each color has a different pitch, thus changing the angle at which light comes off the display. Because diffraction is wavelength dependent, white light hitting the grating spreads into a rainbow. The gratings are arranged so that the red part of the rainbow produced by the first grating lines up with the green of the second and the blue of the third: the rest of the light is hidden from view.
This method has the same power disadvantage as any other color filter system: two thirds of the light is wasted. The gratings themselves also put some constraints on the light used to illuminate the display: it needs to come from a point source. On the other hand, embossed holograms (like those used on credit cards) need the same lighting conditions, and they have succeeded despite being viewed in ambient light. At the very least, this means that the lighting needed for the MicroDisplay system can be low-tech.
However, in terms of power and practicality, Kopin's new CyberDisplay may have the edge for some applications. The company uses a transmissive display, made by using standard VLSI processes on a silicon wafer with a proprietary layer. After the electrode circuitry has been fabricated, it is lifted off the silicon and deposited onto a glass substrate. The liquid crystal they use is of the slower nematic type, so the grey levels are created by modulating the voltage across each pixel.
The fact that the display is transmissive has gives Kopin two advantages. First, the CyberDisplay uses a built-in backlight that is uniquely compatible with it. According to Glen Kephart, Vice President of Marketing, they use a triplet of LEDs, in red, green, and blue. The light is combined in a reflective "mixing bowl," much like the curved reflector used on the back of headlights, and then liquid crystal device. To get color, the red, green, and blue LEDs have only to be turned on sequentially for each frame in order to produce full-color images. From a power point of view, this system has the advantage that no light is being filtered out: the right color is simply produced as needed.
Another advantage of a transmissive system is that the front surface of the display only points one way: light does not have to enter from the front, only leave that way. Kopin has exploited this by putting an array of imaging microlenses on the device, which means that it can be viewed without any further optics.
Who wins market share and for what display applications will have as much to do with money, good engineering, and clever marketing as the technologies' inherent advantages and disadvantages. However, the results are already having a knock-on effect in the fledgling industry of optical information processing. Displaytech, for instance, has done much to promote its products for optical correlators. These are devices that compare the Fourier or other transforms of images by creating them optically. The images are displayed on a minidisplay, picked up by a laser beam, and then transformed using lenses or other optical elements.
Though they are being used, none of the fast-binary devices are ideal for the job. The time-division multiplexing approach that works perfectly well in displays is not quite so ideal in a correlator. Speed is good, but the ideal input device has grey levels, too. The more grey levels displayed, the more information is being processed and the better (depending on the application) the results.
Until now, optical correlators have been developed, almost exclusively, by the military and its contractors. If the display industry can produce lots of very cheap light modulators of the right kind, the revolution in displays could be followed by an explosion in machine-vision systems and image processors that literally work at the speed of light.
1. David Bloom, "The Grating light valve: revolutionizing display technology," SPIE Proc. 3013, May 1997.
2. Larry Hornbeck, "Digital light processing for high-brightness, high resolution applications," SPIE Proc. 3013, May 1997.
3. David M. Walba, "Fast ferroelectric liquid-crystal electro-optics," Science Vol. 270, 13 October 1995.