Over the past five years progress in active matrix liquid crystal displays (AMLCDs) has been spectacular. Five years ago the questions were whether these complex devices could be made and whether they would gain market acceptance. Today those questions have been answered affirmatively; the only remaining questions are how low the cost of AMLCDs can be, how fast they will penetrate the display market, and how good their ultimate performance will be.
Liquid crystal displays are inherently simple and are intrinsically capable of the high performance desired for many display applications. However, expecting adequate nonlinearity sufficient to operate high information content displays places an unreasonable burden on the liquid crystal (LC) material itself. The value of using nonlinear circuitry in series with the LC pixel (Lechner, 1971) was recognized quite early in the technology. The use of thin-film transistors (TFT) as the preferred nonlinear element is based on the pioneering work of Peter Brody (1973), the "father" of the TFT active matrix.
Although many active matrix technologies have been explored, the dominant ones today are hydrogenated amorphous silicon (a-Si) thin-film transistors, metal- insulator-metal (MIM) diodes, and low-temperature polysilicon (p-Si) thin-film transistors. Although a-Si TFTs were suggested quite early (LeComber, 1979), the first commercial product was a pocket TV that used polysilicon TFTs (Morozumi, 1985).
The active matrix is a method of addressing an array of simple LC cells--one cell per monochrome pixel. In its simplest form there is one thin-film transistor for each cell. This arrangement is shown in Figure 5.1.
Figure 5.1. Simple TFT Active Matrix Array
A row of pixels is selected by applying the appropriate select voltage to the select line connecting the TFT gates for that row of pixels. When a row of pixels is selected, we can apply a desired voltage to each pixel via its data line. When a pixel is selected, we want to apply a given voltage to that pixel alone and not to any nonselected pixels. Those nonselected pixels should be completely isolated from the voltages circulating through the array for the selected pixels. Ideally, the TFT active matrix can be considered as an array of ideal switches. The operation of this active matrix would be as follows:
All rows are selected in one scanning period. Thus, if there are 500 lines and the time to load data into each selected line is 50 micro sec, then a single scanning period is 25 msec, for a field-scanning rate of 40 Hz.
The performance required of the TFTs in the active matrix depends on the display performance requirements--number of lines, number of gray levels, operating temperature, pixel density, and so forth. The TFT should behave as an ideal switch--zero on resistance and infinite off resistance. We can plot actual TFT on-current per micron of channel width and off-current per micron of channel width as a way to compare different TFTs and to predict their suitability for differing display applications (Firester, 1987). Figure 5.2 is this type of plot with a number of reported TFT data. An "ideal" TFT would be in the upper left corner of this chart.
Figure 5.2. TFT Leakage and Drive Characteristics
The need for address and data line drive circuitry is another general aspect of active matrix displays that should be considered. A 1000 x 1000 simple monochrome active matrix has 1 million TFTs and requires 2000 connections to external drive circuitry. Currently these external circuits use flex-printed circuit board connections, elastomeric interconnects, tape-automated bonding (Tomita, 1989), and even chip-on-glass technology (Ishihara, 1989). Cost projections for these external-drive circuits range from about 40% of the direct material costs of display manufacturing (Mentley, 1989) to about 50% of the total system costs (Firester, unpublished).
Many developers have been pursuing the integration of this drive circuitry with the active matrix itself. The basic supporting argument is that, given yields adequate to fabricate an active matrix with 1 million perfect TFTs, several tens of thousands additional TFTs forming the drive circuitry will not substantially decrease the overall yields. Indeed with redundancy the yields can be enhanced. Nonetheless there is disagreement whether total system yields are enhanced (Mentley, 1989; Morozumi, 1989) or decreased (Ishihara, 1989) by the addition of integrated drive circuitry.
Perhaps the application for which integrated drivers will be most important is LCD projectors. Here the system cost advantages of small LC light valve size push designs to smaller and smaller pixel periodicities, which also strain available fine- pitch external interconnect technologies.