Using LCD Panels For TVs Takes Technology To The Outer Limits

DESIGN VIEW is the summary of the complete DESIGN SOLUTION contributed article, which begins on Page 2.

From the early 1990s through the early 2000s, the requirements and needs of the thin-film-transistor LCD industry evolved gradually. But with the quickly growing popularity of LCD TVs, the market now demands an entirely new set of performance requirements, which in turn is stressing current silicon architectures and interfaces to the limit.

The television market is calling for disruptive electronics technology to solve the frequency and distance requirements of the interface, provide greater than 1 billion colors, improve the response time for high-quality motion video, and achieve better color uniformity. Mainstream TV panel sizes span 27 to 32 in. today, and at least four companies have demonstrated LCD panels larger than 46 in.

The article discusses the main elements that must be addressed in LCD-panel electronics to accommodate today's demands. For instance, as panel size grows, signal integrity becomes a problem. Options for improving the "data eye" include improving the quality of the pc-board material, altering the data-transmission scheme, and moving to a point-to-point architecture. All are explained in detail.

Another sticky point is color depth, as well as the impending massive increase in decode logic and column-driver size when switching to 10-bit color. Described are dithering techniques and a change in architecture.

Poor motion-video capability hampers LCDs. However, the overdrive method discussed can overcome the otherwise blurred edges of a moving image. One other challenge for panel electronics concerns color quality. The author feels that the problem will only be truly solved if electronics manufacturers provide independent luminance-to-voltage conversions for each color in a cost-competitive manner.

HIGHLIGHTS:
The Panel-Size Issue	Though today's differential bus architectures excel at delivering the bandwidth needs across a 15- to 19-in. panel, signal integrity becomes a problem as panel size increases.
The Color-Depth Issue	The trend for televisions is a color depth of 1 billion colors (10-bit color), creating an eightfold increase in decode logic and making column-driver die size and cost nearly prohibitive. One suggestion is to change from the traditional R-DAC architecture.
Motion-Video Performance	Numerous companies are working on proprietary solutions for response-time compensation. These would ultimately raise LCD motion-video performance to match or exceed today's high-end TVs.
Color Quality	Solving color-quality problems typically involves the timing controller, which manipulates incoming digital data via data expansion and dithering. However, a programmable voltage-to-luminance function for each pixel color will probably be the way to go.
Sidebar: A History Of LVDS And RSDS Interfaces	This sidebar discusses how the low-voltage differential signaling and reduced-swing differential signaling interfaces improved LCD data flow.

Full article begins on Page 2

From the early 1990s through the early 2000s, the requirements and needs of TFT (thin-film transistor) LCDs (liquid crystal displays) evolved gradually. But with the quickly growing popularity of LCD TVs, the market now demands an entirely new set of performance requirements, which in turn is stressing current silicon architectures and interfaces to the limit. That's because the expansion of LCD panels into the television market is more than just an extension of the performance required by notebook computers and desktop monitors.

In fact, the television market is demanding disruptive electronics technology to solve the frequency and distance requirements of the interface, provide greater than 1 billion colors, improve the response time for high-quality motion video, and achieve better color uniformity. The current system architecture provides a patchwork solution that's acceptable in the absence of an ideal solution. But the industry is actively pursuing technologies and options that can solve the problems of LCD TVs now-and into the future.

In the beginning, LCDs were primarily used for notebook-computer applications. Early requirements were low power consumption and high portability. Panel sizes typically measured less than 12 in., resolutions were 640 x 480 pixels for VGA (video graphics array) or 800 x 600 pixels for SVGA (super VGA), and color performance was a luxury, not a requirement. Starting in the late 1990s, LCD panels expanded into the desktop replacement market. The thin, light LCD was a significant space improvement over the traditional CRT (cathode ray tube). Market requirements also migrated to include much higher resolutions, from 1024 x 768 pixels for XGA (extended graphics array) up through 1600 x 1200 pixels for UXGA (ultra extended graphics array). Panel sizes grew from less than 14 in. to between 15 and 19 in. Color became a necessity, with 6-bit color (256k colors) dominating notebooks and low-end monitors, and 8-bit color (16.7 million colors) targeted at high-end monitors.

Demands on panel electronics also expanded to include low electromagnetic interference, further power reductions, higher data rates, and lower cost. These incremental changes, though designed to improve the panel performance for computing applications, did not bring a disruptive new set of requirements for the display industry. As the target use for displays migrates away from computing and toward home theater, the visual performance requirements have taxed traditional system architectures. Instead of competing with the performance of standard computer monitors, LCD panels now compete with high-end televisions and plasma displays that advertise capabilities in excess of 1 billion colors, a true cinema-quality picture, and large panel sizes. Mainstream TV panel sizes span 27 to 32 in. today, and at least four companies have demonstrated larger than 46-in. panels. Before exploring the bottlenecks of, and possible improvements to, the current system architecture, we should review the current state of LCD-panel electronics. An LCD panel can be thought of as an array of transistors that modulate the voltage across the liquid crystal, and thereby control the amount of light passing through the panel. Color is achieved with an additional color filter layer that allows red, green, or blue light to pass through a given pixel. Attached to the transistors' gates are the row drivers. These control which row of pixels is being programmed at any given time by applying either an "on" or "off" voltage. The sources of the transistors are tied to the column drivers, which supply the specific voltage required to achieve proper pixel luminance.

All of this is controlled with a timing controller that takes the display data in from the host and transmits it to the column drivers and row drivers though the intra-panel interface. The row-driver interface, which has relatively low speed, continues to use TTL as the signaling level. The column-driver interface requires significant bandwidth, well over 2 Gbits/s for higher-resolution displays, and uses a differential bus architecture-typically Reduced-Swing Differential Signaling (RSDS) (see the sidebar, "LCD TV Panels: A History Of Their Interface Technology"). A typical LCD display intra-panel interface is shown in Figure 1.

The Panel-Size Issue Today's differential bus architectures excel at delivering the bandwidth needs across the relatively short distances of a 15- to 19-in. panel. But as the panel size grows, signal integrity becomes an issue. Figure 2 shows the eye diagram of a 30-in. LCD TV panel operating under normal conditions. The data eye is acceptable, but will continue to degrade as data rates and panel sizes increase.

Three reasonable options exist for improving the data eye. One is to improve the quality of the pc-board material. A dominating percentage of LCD panels use standard FR4 material for their pc boards. Moving to a pc-board material that's designed for better signal integrity at higher frequencies would certainly upgrade the data eye, but it would also significantly increase the cost of the board. To the best of the author's knowledge, nobody is pursuing this option. Instead, the preference is to investigate other solutions.

The second option involves changing the data-transmission scheme itself. One option is to cascade the data from one column driver to the next (Fig. 3a). With this method, each column driver would have both a receiver circuit as well as a transmitter circuit to regenerate the data and pass it on to the next driver. As such, you'll have virtually unlimited distance because the datapath was reduced from the length of the panel to the distance between two column drivers. It also can use the same timing controller as the current interfaces. However, this method negatively effects the column driver. The number of data interface pins doubles to accommodate the need for both transmitter and receiver blocks. Also, the power requirement of each column driver increases (due to additional transmitters), as does the overall complexity of the column-driver circuitry. The third option is to move away from a bus structure altogether and go with a point-to-point architecture (Fig. 3b). Instead of the timing controller sending data out on a global bus, there will be a dedicated datapath for each column driver. This method carries a number of advantages. It significantly boosts the driving distance, lessens the total number of wires needed for the interface, greatly reduces the data-interface pins on the column driver, and allows the timing controller to customize the control information sent to one column driver independent from all others. The main disadvantage of this method is that it requires a change to the timing-controller architecture, which involves additional memory so that the incoming display data can be reformatted to match the point-to-point architecture.

The Color-Depth Issue The second major hurdle to overcome for panel electronics is the panel's overall color depth. As mentioned earlier, notebooks only require 256k colors (6-bit color), high-end monitors require 16.7 million colors (8-bit color), and the trend for televisions is for more than 1 billion colors (10-bit color). The nonlinear conversion from digital data to the voltage that drives the display traditionally takes place in the column driver using a resistive digital-to-analog converter (R-DAC) architecture. The architecture consists of a resistor with numerous tap points, and decode logic that selects the appropriate tap point based on the incoming data (Fig. 4). The decode logic needs to be repeated for every output (typically about 400 per driver). Going from 6- to 8-bit color created a four-fold increase in the decode logic, and substantial growth in the column-driver die size-up to 40% to 50%.

The impending increase to 10-bit color will create another 4X increase in the decode logic, and make column-driver die size and cost almost prohibitive. As a result, a couple of solutions are currently under investigation. The first solution uses temporal and spatial dithering to achieve 10-bit color using 8-bit column drivers. To generate one of the gray levels not supported by the 8-bit column driver, the timing controller will change the data between the two closest 8-bit levels on a frame-to-frame basis, as well as on a localized spatial basis. This method is commonly used to generate 8-bit color out of 6-bit drivers in lower-end LCD monitors. It works because the human eye does a very good job of averaging luminance over small areas and short time frames.

However, for high-performance applications, dithering sometimes may be seen as a visual artifact. These conditions happen much more often while watching full-motion video than typical computing applications. There's also a marketing drawback in that a true 10-bit system can be billed as having higher quality and performance than a dithered 10-bit system. Nonetheless, several companies are actively investigating dithering as a method to expand the color performance from 8 to 10 bits. A second method for dealing with the large-die-size penalty associated with 10-bit color is to change from the traditional R-DAC architecture. A variety of other DAC architectures are employed throughout the electronics industry, and companies have begun investigating the feasibility of using something other than an R-DAC for 10-bit column drivers.

The most common approach incorporates some capacitive DAC elements into the architecture, either as a part of the overall DAC, or as the entire DAC. This creates a more space-efficient circuit than a traditional 10-bit R-DAC column driver. However, difficulties lie in the nature of the column driver. Each of the approximately 400 outputs must be closely matched to all other outputs on the die, as well as outputs on adjacent die. Building a single high-precision DAC is relatively straightforward. But integrating 400 onto the same IC at low power with high repeatability poses a significantly greater challenge. In addition to the design concerns, a capacitive-based DAC probably demands a new interface. In a traditional differential bus system, each column driver is given 20% of the total line time (total line time is about 20 µs) to receive and convert the data. This is adequate for the R-DAC because minimal settling time is required. For a capacitor-based DAC, the settling time of the capacitor can be significant, and the column driver may require a larger portion of the line time to fully settle and convert the data. The point-to-point interface allows for each column driver to have 100% of the line time to convert and settle. Therefore, a move to an alternative DAC architecture will likely also result in a change to a point-to-point interface and its associated advantages and challenges. If the technical issues can be overcome, the alternative DAC architecture coupled with a point-to-point interface offers the best possibility of solving the 10-bit color issue in LCD TVs.

Motion-Video Performance Traditionally, LCD panels possess excellent resolution and performance with still images, but relatively poor motion-video performance. This is due to the slow response time of the liquid crystal itself. During a motion video, the pixels on the display change their luminance rapidly as the image moves across the screen. The luminance changes are a direct result of the voltage applied to the liquid crystal. Because the liquid crystal acts as a capacitor, and the column line acts as a resistor (typically in the tens of kilohms), there's a nontrivial RC time constant associated with changing the pixel luminance. Under normal conditions, the liquid crystal doesn't fully settle in a single frame and can cause blurring at the edges of moving images.

The common solution these days is to use an overdrive method (Fig. 5). Instead of driving the pixel to the intended voltage, the column driver will drive it to a higher-end voltage so that the pixel will be at the intended voltage at the end of the frame. For this method to work, the timing controller needs to keep track of the previous voltage on the pixel. This requires storage of one entire frame's worth of data (about 16 Mbytes for a mainstream LCD TV). This first-generation response-time compensation (RTC) is good enough to get panel performance to an acceptable level. However, numerous companies are working on proprietary solutions for a second-generation RTC that will bring even faster response times and raise LCD performance to match or exceed the performance of today's high-end televisions.

Color Quality The final challenge for panel electronics in LCD television applications is overall color quality. The red, green, and blue pixels of the mainstream LCD TV liquid-crystal technologies all have slightly different luminance-to-voltage response curves. A traditional architecture uses the same luminance-to-voltage conversion function for all three colors, which is acceptable for computing applications, but not quite good enough for high-quality video. The different functions make the LCD's white output a slightly different color than a CRT. Perhaps the LCD contains a little too much blue. As the image moves from white to black through the various gray levels, the difference between a CRT and an LCD changes so that at the darker gray levels-that is, closer to black-the LCD may now look like it contains a little too much red.

Solving this problem typically involves the timing controller, which compensates for this issue by manipulating the incoming digital data via data expansion and dithering. Several companies have published very successful results in keeping the color consistent across all gray levels. The main drawback to this method is a net loss of gray levels that can only be reclaimed through dithering. As was discussed previously, dithering is only a marginal solution for high-performance panels and may not be acceptable for television performance. However, because this is the best solution right now for production, several LCD panel manufacturers developed their own proprietary algorithms to optimize panel color quality and minimize any visual artifacts associated with dithering.

A better solution is to provide an independently programmable voltage-to-luminance function for each pixel color by incorporating three separate resistor strings in the column driver, one each for red, green, and blue. While this is technically feasible, it would significantly increase the column-driver die size-in the neighborhood of 40% to 50% based on our experience. This method appears to be cost prohibitive, too, and the author is aware of no company pursuing this option. For the color-quality problem to be truly solved, the electronics manufacturers are going to need to provide independent luminance-to-voltage conversions for each color in a cost-competitive manner.