Even though research in CRT projectors continues, the major effort seems to have shifted to AMLCD light-valve projectors. These projectors now provide images with excellent quality and have a number of cost and performance advantages. Since the panels used are geometrically very accurate and a single projection lens is used, they have no registration problems. The projectors can be moved from place to place and can be simply focused like slide projectors. Their size is relatively small. They require no high-voltage power. Their contrast, color rendition, uniformity, resolution, and motion rendition are all excellent.
AMLCD projectors have a number of problems that affect cost and performance. This section discusses how Japanese companies are addressing these problems.
Most of the designs to date use amorphous silicon thin-film transistor (TFT) elements at each pixel with silicon chip drivers external to the panel. Since a-Si is light- sensitive, it must be well shielded from the high-flux-density light passing through the panel. The technique for connecting the chip and the panel can be a cost and yield problem. Consequently most companies are experimenting with polysilicon row and column drivers on the panel. Polysilicon has a high enough mobility to provide adequate rise times to switch the rows and columns. For high-definition column drivers, the switching time is generally not fast enough. Consequently, the line is divided into roughly six segments, and six columns are driven in parallel at a lower clock rate.
The effort at Seiko-Epson (Thompson) uses polysilicon (p-Si) for both the drivers and the pixel TFT. Since polysilicon TFTs are not light-sensitive, they do not need a light shield. In the Sharp (Katoyama, 1988) effort, two redundant pixel TFTs, row lines, and column lines are used to improve yield. A self-alignment structure is used to improve yield by preventing misregistration of successive masks. This all-polysilicon technique uses quartz as a high-temperature substrate. Research is attempting to produce an all- polysilicon structure on a high-temperature glass substrate.
NEC is working on a hybrid structure (Sakamoto et al., 1991) in which a-Si is used for the pixel TFT and laser recrystallized polysilicon is used for the row and column drivers on the same glass substrate.
Table 6.2 shows the techniques currently in use. Most companies are working on all-polysilicon panels, with drivers on the panels, on a high-temperature glass substrate for projectors. NHK is using an optically addressed reflective light-valve projector (Takizawa, 1991), a scheme similar to the one used by Hughes in the United States (Sterling, 1990). It consists of a sandwich with a photoconductor, an insulating reflecting surface, and a liquid crystal. Hughes uses a TN liquid crystal, and NHK uses a polymer-dispersed liquid crystal (PDLC).
Driver Technology for LCD Panels
Most projectors use TN liquid crystals for projector panels. In a projector the reduction in contrast for TN panels at off-axis viewing is not a problem because the illumination and projection lens are on axis. The response time is adequate for moving objects if progressive scan at 60 fields per second (FPS) is used. The efficiency loss from the requirement to polarize the light is a problem.
NHK and Asahi Glass are working on the use of PDLC in projectors because it does not require polarization of the light. However, PDLC has a trade-off between contrast and displayed brightness: High contrast requires a small projection lens aperture and, consequently, limited light output. Thus, according to a rough calculation of optical efficiency based on quoted screen brightness and screen gain, the NHK projector is less efficient than the TN AMLCD projectors.
Projectors that were demonstrated all used some form of progressive scan at 60 FPS. If an AMLCD is addressed with an interlaced scan, a single pixel is on for the full 1/30-sec frame time. This long display time, compounded by the marginal response time of a TN liquid crystal, gives unacceptable image smear. If the display time is shortened to 1/60 sec to reduce the smear but still scanned interlaced, the brightness and efficiency are cut in half. Consequently, progressive scan at 60 FPS is used.
In low-resolution displays that display only the vertical resolution of one field, the information from each field is simply addressed to the same lines in both fields of a frame. However, this degrades vertical resolution compared to a normal CRT interlaced display. In displays that have a full complement of vertical pixels (483 in a 525-line image), the usual addressing system is to address two lines at a time per field and shift the addressed line pair by one line for the interlaced field. This is the inverse of what is normally done on a CCD camera with interlaced scan. It gives a reasonably adequate vertical resolution with good motion rendition. Some research workers referred to an interlace-to- progressive-scan converter with more sophisticated processing. However, it was not clear that this device was being used in any of the projectors we saw.
In AMLCD drivers used for computer displays, a D/A converter was used on each column to give grey scale (usually 4-bit grey scale). For television displays, analog column switching was used for 8-bit grey scale. Some workers reported 8-bit D/A converters on each column for computer displays; however, they were not demonstrated in any of the projectors we saw. In one projector, screen nonuniformity was corrected by the drive circuit to modify black level, gain, and gamma on each pixel on the basis of stored information.
The TN light valves used transmissive optics. The design of the optical systems is shown in Figure 6.1. The light from a projection lamp is divided into three color beams with two dichroic mirrors. These beams pass through three TN AMLCD panels and are recombined with two more dichroic mirrors into a single beam, which passes through the projection lens onto the screen. Because of the need for two dichroic mirrors to recombine the three beams, the back focus of the projection lens must be quite long. This increases its cost and requires more lens elements if a wide- field projection is desired.
Figure 6.1. Optical Path
The projection lamp used is a metal halide high-intensity arc with a life of about 2000 hours. It has line spectra at the three primary colors but has an undesired yellow mercury line. This tends to desaturate the red and green primaries. Removal of this line optically reduces optical efficiency. The color rendition of the images we saw even with the yellow line was quite acceptable for normal program material. The arc uses a cold mirror reflector to remove infrared.
At the time of the JTEC team's visit, HDTEC was developing a back-projection HDTV display with a black matrix lenticular screen. A Fresnel field lens was used behind the lenticular screen to give better uniformity. The cabinet had a depth in centimeters equal to its screen diagonal in inches. Such a display is thinner and much lighter than a CRT display could be of the same size. The screen was developed by DaiNippon Screen. It used 2 1/2 lenticles per projected pixel to minimize moire patterns. Hitachi is also developing a slim rear projector (Fukuda, 1991). These screens are somewhat similar to those developed for the GE Talaria projector in about 1961 (Glenn, 1969, 1970).
Several projector designers were questioned about the brightness limitations of the projector design. None of the projectors produced more than about 200 lumens, which is a rather low light output for large front-projection screens for either consumer or commercial use. The GE MLV Talaria light-valve HDTV projector, for example, has an output of about 3500 lumens. Asked about the maximum light output limit, the designers generally responded that 250 watts was the highest power high-intensity, long-life, metal halide lamp made and that this fact limited the light output. Asked what limited the optical efficiency, they generally named the need to polarize the light and the inefficient aperture of the pixel display area. For 525-line displays, the panel has about 50% of its area transparent, and HDTV panels have about 30% transparency. The remainder is covered with transistors, drive lines, and a black matrix mask.
When asked what would limit the light output if a higher power lamp were available, the designers had no clear answer. However, from discussions about panel temperature, it seems that heating would limit the light output if significantly more light input was used. The greatest source of panel heating seemed to be the visible light energy absorbed by the black matrix surrounding the pixel clear area.