In the world of video displays, many intriguing engineering problems fall outside the mainstream issues of formats and encoding. By way of illustration, here are examples of solutions in three areas that not only represent clever engineering but also may apply to your own future designs.
THROUGH THE HINGE
Serializing the video stream through the hinge between the display and the rest of the system began in laptops, when designers dropped power-hungry and unreliable parallel buses in favor of a variety of serial low-voltage differential signaling (LVDS) schemes.
More recently, as cell-phones shrank and sprouted cameras while their displays grew and offered better resolution with clamshell, cycloidal, and sliding mounts, designers saw the advantages of serial through-the-hinge interconnects, which are less bulky. They also use differential signaling, which helps minimize electromagnetic interference (EMI) with the radio part of the cell phone.
But the serial video link in the cell phone has trod a thorny path. Simply providing LVDS serializer/deserializer (SERDES) chips isn't enough, says Shri Sundaram, manager of business development at Toshiba. As we shall see, adding a buffering function to the link can reduce system power consumption.
Also, the industry is following the usual rocky road toward standardization. Initially, Sundaram explains, major chipmakers developed proprietary interfaces. That inevitably led cell-phone makers to clamor for standardization. But equally inevitably, the process of developing standards leads to give and take among stakeholders, along with delays and conflicts.
The industry is still at the stage where standards are more promise than reality. While several proprietary serial approaches are still in use, there are also competing standards. One, Qualcomm's Mobile Display Digital Interface (MDDI), has become a Video Electronics Standards Association (VESA) open standard. Meanwhile, an industry alliance founded by ARM, Nokia, STMicroelectronics, and Texas Instruments backs the Mobile Industry Processor Interface's (MIPI's) Display Serial Interface (DSI).
Note that MIPI is more than DSI. Some aspects of MIPI are well established. According to Sundaram, most display manufacturers have already adopted the MIPI standard that describes how to convert serial signals to drive display rows and columns.
At the other end of the serial bus, there's more confusion. Baseband processor makers have a more eclectic range of outputs, and most of them would rather not deal with the signal path through the hinge. So, it's up to companies to develop the chips that go in the middle.
The latest versions of these links comprise video buffers and the actual serializers and deserializers. Buffers isolate the baseband processor from the link, of course. But they also enable baseband processors to burst video data and then enter sleep modes, helping extend battery life. Yet this adds another function that the interface chips must support—metering the video between display or camera and the buffer while the baseband processor sleeps.
A sequence of recent Toshiba announcements highlights the evolution toward standards. The company's baseline part, the TC358300, uses a proprietary LVDS signaling technique called TLVDS. It's a simple bridge designed for VGA resolution capable of 250 Mbits/s per each of two channels.
Introduced last year, the follow-on TC358720 MDDI LCD Bridge connects directly to Qualcomm Mobile Station Modem baseband chips and others that use MDDI. It supports data-transfer rates up to 400 Mbits/s. To provide that buffer function, the chip also integrates an 8Mbit embedded DRAM.
The maximum size on the primary display that the TC358720 can support is full VGA (640 by 480 pixels), with a 60-Hz frame rate. Maximum size on the secondary display is QCIF+ (Quarter Common Interface Format Plus: 176 by 222 pixels). The IC also provides programmable gamma correction, RGB format conversion, and backlight dimming. And, it supports 180° rotation and mirroring, all in a 6- by 6mm package.
Earlier this year, Toshiba began sampling its newest chip, the MIPI Display Serial Interface (DSI) TC358730, with expected full production in the second half of the year. For the most part, the new device is based on the core architecture of the TC358720. It replaces the MDDI-host interface with a MIPI-compliant interface and adds a high-speed, serial-output interface.
Specifically, the chip supports the existing MIPI DBI type-B specification for command and video data and the MIPI DPI specification for video data, as well as the MIPI DBI type-C baseband-interface standards for command data. The controller also supports the upcoming high-speed serial MIPI DSI baseband interface for command and video data.
As Sundaram pointed out, allowing the baseband processor to sleep helps reduce power consumption. With that in mind, the TC358730 accepts high-speed, burst-access data transfer from the host, which can then enter low-power sleep mode. While the processor sleeps, the controller updates the connected LCD at the required frame rate.
The new chip also integrates Toshiba's patented "Magic Square" algorithm. It allows an 18-bit LCD panel to produce picture quality equivalent to a 24-bit LCD panel, with up to 16 million colors.
While National Semiconductor is also a MIPI alliance participant, it is biding its time before announcing MIPI products and is winning customers with innovations in packaging (Fig. 1). Recent products include the LM2512 RGB Display Interface Serializer and LM2510 deserializer. The pair use National's proprietary Mobile Picture Link (MPL) LVDS protocol. MPL is a three-wire open standard that currently supports 160-Mbit/s transfers.
BIGGER PIXELS, SAME RESOLUTION
Changing focus from the interface to the display itself, consider that enjoying live video or a downloaded movie on a cell phone or handheld device requires greater luminance dynamic range than simply tolerating Web pages or still pictures taken with a cell-phone camera. The catch is that increasing peak brightness implies higher power consumption, which in turn implies reduced battery life. Not only that, real video requires more resolution, with the trend going from quarter-VGA to half- and then to full-VGA.
To push brightness and resolution in small displays without running through a viewer's battery before the final reel, Clairvoyante's PenTile technology intermixes white LCDs with the reds, blues, and greens. It works by taking advantage of quirks in human vision. Vastly oversimplified, the idea is that the brain processes the blackand-white and color components of an image differently. Black-and-white resolution is what the brain primarily cares about, though.
The PenTile RGBW layout uses each red, green, blue, and white subpixel to present high-resolution luminance information to the eyes' red and green cones. At the same time, it utilizes the combined effect of all the color subpixels to present lower-resolution chroma (color) information to all three cone types. Combined, this optimizes the match of display technology to the biological mechanisms of human vision.
PenTile RGBW displays also exploit other human-vision factors such as the logarithmic representation of luminance values, variable resolution between the center and edge of vision, and the separation and compression of brightness and color differences.
With this concept behind it, the PenTile approach employs fewer and larger subpixels than conventional displays, while adding white subpixels. The resulting RGBW panel increases transmissivity and brightness because more backlight can shine through the larger aperture ratio of the bigger color subpixels. Additionally, there's the potential for extra luminance provided by the white subpixel. Together, this yields nearly twice the white brightness at the same power consumption. A PenTile panel also requires fewer source drivers and backlight LEDs, further reducing power demand along with cost.
How does Clairvoyante reduce the number of subpixels without diminishing resolution? Figure 2 compares writing alternate vertical black and white lines with the arrangement of subpixels in a PenTile array with the same task using the arrangement in a common RGB-stripe LCD. The PenTile RGBW display needs four columns to write a line pair, while the RGB-stripe display needs six. So for the same resolution, the PenTile columns can be wider.
Clairvoyante provides an application note (www.clairvoyante.com/files/pdf/measuring-pentile-displayresolution-vesa-standards.pdf) that applies the measurement techniques of the VESA "Flat Panel Display Measurements" standard. The entire VESA document is available at www.vesa.org/public/Fpdm2/FPDMUPDT.pdf, but the app note includes the four relevant pages on "Resolution from Contrast Modulation" from the standard.
The VESA standard defines contrast modulation as the difference between the white-line luminance and black-line luminance divided by their sum. To obtain those values, it is necessary to compute a luminance profile of the alternating black and white lines in a VESA test pattern and perform a moving window average. The averaging window is one pixel wide. As noted above, it encompasses three subpixels on an RGB-stripe display and two on a PenTile display.
The VESA standard concerns horizontal and vertical resolution. With respect to horizontal stripes, there's a difference between the conventional and PenTile panels, but it doesn't matter. It's true that the horizontal contrast modulation for a 1x1 vertical grille on an RGB-stripe display is theoretically 100%. It's equally true that on a PenTile RGBW display, the same 1x1 vertical grille pattern results in a theoretical horizontal contrast modulation that varies between 85% to 100% because of the moving average.
This isn't enough to affect perceived resolution according to the VESA standard, which defines a contrast modulation threshold of 50% as the minimum necessary for displaying crisp edges on text and graphics. So, both panels can display the same horizontal resolution. When the stripes in the grille are horizontal, both displays are the same.
PenTile involves some extra overhead in digital processing. The additional steps in the algorithms that Clairvoyante licenses consist of converting RGB into RGBW using proprietary gamut-mapping algorithms, which involve compensating for the different gamma of the PenTile display. That's accomplished by a lookup table that takes the white pixels into account. Next, further subpixel rendering algorithms are applied as pipelined matrix manipulations for each 3-by-3 group of pixels.
PAINTING THE BIG PICTURE
Beyond camera-produced video, there's a growing market for individual high-brightness (HB) LEDs for advertising banners and special-effects lighting in architectural and home-theater applications. HB LEDs are a new phenomenon, and they're becoming more affordable. Relatively inexpensive HB LEDs have made many new applications practical, especially the scrolling advertising signage that is redefining the ambience of the modern urban afterdark experience.
A standard LED generates 1 to 3 candelas or, roughly, 13 to 38 lumens (lm). The output of an HB LED is 4 or more candelas and getting stronger with each generation. (There are also high-power (HP) LEDs, which are defined in terms of their power consumption rather than light output. By definition, highpower LEDs consume 1 W or more. Light output tends to correlate with wattage, but not in any standardized way.)
Illuminating signs or throwing colors on a wall involves a lot more than stringing a bunch of LEDs and ballast resistors together and connecting them to a power supply. Color control and consistency is a matter of color mixing as well as dealing with temperature effects and variations among binned LEDs. Cypress Semiconductor's EZ-Color controller is a variation on the company's pSoC configurable microcontroller line aimed expressly at these kinds of applications.
As with other pSoC application targets, Cypress' pSoC Express programming environment facilitates high-level graphical programming. In this case, designers select a color from a gamut presented on-screen. Pre-loaded manufacturers' bin specifications and temperature feedback algorithms are automatically applied to the selected design and programmed into the controller. There are EZ-Color Controllers that support four, eight, or 16 LED channels.
It's instructive to look at the variables that must be considered in designing a HB LED application. It's possible to obtain any color seen by the human eye by varying the intensity of red, green, and blue color sources, though no display ever matches the full color gamut of the human eye.
But we've been a little spoiled by the corner-to-corner consistency of color CRTs and flat panels. The elements in any realworld HB LED triad exhibit much greater part-to-part manufacturing variations, requiring control loops to manage current and pulse-width-modulation characteristics.
Thus, designers must understand (or their development software has to understand) colorimetry. Colorimetry objectifies the subjective perception of colors (known as the gamut of human vision) using three parameters: the tristimulus values, designated X, Y, and Z.
Most engineers recognize Y as simply the luminance, or white level, of a pixel. X and Z are derived from a pixel's luminance and the wavelength of the light it emits. To provide a scalable method of representing chrominance, in 1931, the Commission Internationale de l'éclairage (CIE) devised a twodimensional map of chromaticity: the 1931 Color Space, or CIE XYZ color space (Fig. 3).
Each given color point within the diagram can be represented as an x and y coordinate. Pure white is represented at exactly (1/3, 1/3). This is sometimes called the "white point." An interesting property of the chromaticity diagram is that if one draws a straight line between any two points within the diagram, the color at any point along that line can be created using a mixture of those two points. The midpoint between these two colors would be created using an equal mixture of the two colors.
In two dimensions, the CIE color space doesn't account for luminance. Adding the Y value in the third dimension completes color specification. But while x and z on the color space graph are dimensionless, Y is specified in terms either of lumens or as a percentage to signify a relative flux. Y is controlled by pulse-modulating the current to each LED.
One catch to all of this is that no common RGB source, whether CRT phosphors or LEDs, can produce the entire color gamut seen by humans.
TV and computer makers have to ignore this, but engineers designing HB LED signage can opt to integrate additional colors (yellow/amber, violet/purple, and white) to expand the available gamut—at the cost of complexity and dollars.
Any LED can have an (x, y, Y) vector that specifies its color and flux output at some rated current and junction temperature (e.g., TJ = 25°C and IF = 350 mA). But LEDs of different colors behave differently. For example, a typical red LED may require 1.7 V at 20 mA, while a blue LED needs 3.6 V at 20 mA.
Just mixing colors would be challenging enough by itself, but then there's temperature compensation. The forward voltage drop, color, and luminance of HB LEDs vary with junction temperature (TJ). As with any device, TJ is the sum of the circuit board temperature, TB, and the product of the data-sheet value of thermal resistance, JB, between the junction of the LED and the board times PD, the power being dissipated in the LED—which is peak rated power (forward drop times rated current) times a dimming factor. Some designers will take a conservative data-sheet value for those numbers. Others will measure VF and current on the fly.
Finally, or maybe not so much finally as right in the middle of color mixing and junction-temperature issues, there's binning—the price one pays for using discrete LEDs. Characteristics vary from part to part, leading LED manufacturers to devise systems of codes that delineate the characteristics of a given lot of LEDs and bin accordingly. There are codes for light wavelength (color), forward voltage, and luminous flux, which is the good news. The bad news is that there's little standardization between manufacturers in terms of bin codes.
Cypress set out to simplify the design process with EZ-Color. Using PSoC Express on a PC, designers would select a color from a gamut presented on screen (Fig. 4). Then, preloaded manufacturers' bin specifications and temperature feedback algorithms would be automatically applied to the selected design and programmed into the controller.