Novel Power-Performance Optimization Technologies for the Display Systems Interfacing with Mobile Computers.
Technological progress has enabled a liquid crystal display (LCD) to be a universal display with various diagonal sizes ranging from 1 in. to 82 in. Whereas the performance characteristics of LCDs have been steadily improving, the sample-and-hold driving architecture coupled with relatively long response times of the liquid crystals still causes motion blur artifacts. Cathode ray tube (CRT) displays with short persistence phosphor are impulsively driven and do not produce motion blur. However, LCDs sustain the image over the frame time, contributing to motion blur. Furthermore, many researchers have been studying 3-D display implementation using LCDs for near-future applications, which require even higher frame rates and thus faster transition times. To improve display quality and extend into new applications, new driving technologies have been developed. Among them, the overdrive technology that was developed in the early 1990s contributed toward the success of the LCD television (TV) business by significantly reducing the response times. Figure 1 shows a graphical illustration of the overdrive concept. Here, GN is data of gray level and LN is the target luminance of GN. As shown in Fig. 1(a) and 1(b), LC responses of transitions from G1 to G2 and G1 to G3 don't reach the target levels L2 and L3 in a frame time (1/60s), respectively. We can see that the LC response of the transition from G1 to G3 reaches L2, the target luminance of G2, after one frame time as shown in Fig. 1(b). Thus, we can get an ideal luminance response for the transition from G1 to G2 if we change data sequence to be G1→G3→G2 rather than G1→G2→G2. Here, G3 is the overdrive value of the transition from G1 to G2. Unfortunately, there were few studies on the overdrive technology during the first decade after its inception, because the LCD TV market was still immature. However, since 2000, interest and research in the overdrive technology have tremendously increased. Recent efforts have been focusing on how to reduce hardware cost to implement overdrive technology.
A conventional OD technology has bottlenecks to implement it in products. It requires a complex circuit to implement and it demands accurate OD values for good performance. It has several look-up tables (LUTs) and interpolation logics. The size of LUTs is as large as 55,488 bits (17×17×6×32) when the dimension of an LUT is 17×17 and the stored data has 32-bit. 32-bit data is composed of overdrive values and parameters needed for interpolation. The increased power consumption and cost due to an additional frame memory and the huge LUTs have prevented panel makers from applying the OD technology to mobile applications. Intel already developed a graphic chipset to support OD technology. Laptop computer makers, however, have not used the OD technology. Set makers need to find out accurate OD values by themselves. It is a time-consuming job. It needs deep understanding of LCDs and an expensive equipment to measure luminance responses. Thus, more practical implementation method should be developed. We know the digital data is not linear to the analog voltage. Therefore, we propose a new overdrive technology in the digital domain. Our goal of the technology is to extract OD values of all transitions accurately in a short time.
We propose to fit all the OD values by shifting a base line according to DPF. If we find all parameters of the base line for a fixed DPF, the only thing we need is the information on how much we shift according to DPF. If the shift is proved to be linear, then only a few parameters are sufficient to estimate OD values of any transitions. Figure 2(a) shows how to define the base line using the third-order approximation for rising transitions with DPF = 0. The base line can be expressed using
In Eq. (1), x = DCF and y = DOD - DPF. Three parameters such as a, b, and c can be obtained if we know at least three different (x, y) points. Equation (1) represents a third-order line that passes at (0, 0). Thus, we select three points with DCF = 80, 144, and 255 as shown in Fig. 2. The upper-right point represents a rising transition from 0 to 255. Thus, the OD value should be 255. If we extract OD values of the transitions from 0 to 80 and 144, we can calculate the three parameters. We shift the base line to both x and y directions to obtain a new fitting line for transitions with different DPF. Note that additional y-axis shifting is not necessary. The y-axis shift is automatically done when DPF changes because y represents DOD - DPF. Figure 2(b) shows how to shift the base line. We expect that an equation for DPF = 16 will be obtained by shifting the base line with DPF = 0 both to x- and y-directions. As shown in Fig. 2(b), the base line with DPF = 0 is shifted by changing DPF to 16, 32, respectively. Then, the line with a different DPF can be expressed as follows:
The x-shift, Δx, is dependent on DPF. The Δx increases if DPF increases for rising transitions. Using this shifted line, we can extract all the overdrive values for rising transitions. We assume that Δx is proportional to DPF. Thus, Δx can be expressed using
In Eq. (3), s is a shifting parameter. Thus, we can extract shifting values depending on DPF.
We have proposed a new overdrive technology for LCDs with a simple architecture using high order approximation method. We can extract OD values fast and accurately by a few measurements. Due to its simplicity, we can realize low-cost OD technology. From our proposed method, we can implement overdrive technology with much simpler hardware than conventional methods, allowing LCD panel maker or set maker can realize good image performance without any additional cost. We hope LCD panel makers and LCD system makers including laptop computer makers and monitor makers to adopt our technology very soon.
