Spring-cleaning: Calibration clarifies the picture, Part 2

Imaging science long ago figured out that the human visual system notices luminance errors more readily than it does chrominance inaccuracies. This fact is at the heart of the bandwidth-limited color portion of the NTSC (National Television Standards Committee) broadcast spectrum, and it's also the foundation of the color subsampling that most video-compression algorithms employ (Reference 1). Just because gray-scale accuracy is most important, though, you shouldn't conclude that color accuracy is unimportant. After I cover one remaining luminance issue left over from Part 1, I'll tackle the topic of color (Reference 2).

Ideally, the color temperature that the display outputs should remain constant at 6500°K (also known as D65 and D6500) across the 0- to 100-IRE (Institute of Radio Engineers) range. The IRE level is a means of measuring brightness as a relative percentage of total brightness. A too-high color temperature results in an overly blue picture, whereas a too-low color temperature creates an image with an excessively red tint (Figure 1a). Direct-view CRT monitors can often hold color temperature to within plus or minus a few hundred degrees Kelvin across the IRE range, whereas projectors might exhibit a ±1000°K variance. CRT-based displays will also redden at levels higher than 75 IRE if the current of the blue CRT beam reaches its upper limit.

Displays provide bias, or cutoff, controls, either for red and blue or for red, green, and blue, that adjust color temperature at low IRE values. They also supply gain, or drive, controls to set the video-amplifier gain, which affects color temperature across the entire IRE range. Setting the color temperature is a painstaking process that requires a lot of patience and involves a great deal of interaction among the various colors and between bias/cutoff and gain/drive, along with temperature variability at various IRE levels (figures 1b, c, and d). You also need to ensure that you're measuring a color temperature that represents most of the display's surface area. Both constant and variable IRE gray-scale patterns are useful in identifying and avoiding any atypical cold or hot display regions.

My partner in this project, Robert Busch, founder of Busch Home Theater, used a Philips PM 5639 color analyzer to set the proper color temperature on my Princeton Graphic Systems AF3.0HD (Figure 1e). Our session took place at midday in a room with skylights that allow light to shine on the monitor, so I draped a large blanket over him and the entertainment center. I wish I had taken a picture, because the setup was comical, albeit effective! Every time Busch adjusted color temperature, he'd have to also reconfigure the contrast level—an AF3.0HD quirk that made a monotonous process even more mind-numbing and frustrating.

Once you dial in the color temperature, it's time to set the individual colors. SMPTE (Society of Motion Picture and Television Engineers) and split color-bar patterns are the most common images for this task (Figure 2a). Some displays enable you to individually drive the red, green, or blue beam while shutting off the other two beams. For monitors lacking this feature, some test-pattern suites bundle color filters; you hold these filters up to your eye to approximate the single-beam effect (see sidebar "Do-it-yourself adjustment approximations").

Saturation alters the amplitude of the color difference—that is, Pr or Pb—signals with respect to luminance (Y) (Figure 2b). Adjusting saturation affects viewers' perceptions of an image's vividness or color depth; pastels have minimal saturation, and intense colors have high saturation. Adjusting hue, conversely, alters the colors' shade (Figure 2c). If you properly adjust saturation and hue, the blue components of the blue, cyan, magenta, and white color bars have the same intensity. Similarly, if you look at a multicolor bar-test pattern through a red filter or with the green and blue beams disabled, the magenta, red, white, and yellow bars have a consistent intensity. With only the green beam on or through a green filter, the cyan, green, white, and yellow bars should all look the same.

Interestingly, Busch initially employed no color-test patterns when setting my set's color. Instead, we fired up my copy of the DVD demonstration classic, The Fifth Element. Busch had seen the movie so many times and has such a fine-tuned eye and brain that he knew what various scenes were supposed to look like, and he adjusted my set until it most closely approximated those scenes. After setting the color via the movie, he used a few color-test patterns as a quick check of his results. He noted that due to a past excessively high contrast setting, my display had suffered some phosphor damage. Fortunately, his adjustments significantly corrected this problem.

Next, we adjusted my set's image geometry to maximize its available horizontal and vertical resolution. The AF3.0HD has four possible aspect-ratio settings, and it remembers which of them you've selected for the composite-video, S-Video, dual-component-video, RGB, and VGA inputs. We set up the following aspect-ratio combinations:

• a 4:3 aspect-ratio source image—therefore with vertical black bars on either side of the picture, instead of filling the display with a too-wide and -short picture;
• a wide-screen presentation within a 4:3 aspect-ratio source image, expanding the source both horizontally and vertically to fill the display;
• a wide-screen, SDTV (standard-definition TV)- or DVD-source image; and
• a wide-screen, anamorphic, HDTV-source image (Reference 3).

Over-scan patterns, which Part 1 of this series explains you can employ to detect interframe distortion resulting from a too-high contrast level, are also useful when setting geometry (figures 3a and b). Additionally, you can use crosshatch and other display-sizing patterns to ensure that the display fills the screen without chopping off image edges and that it doesn't horizontally or vertically distort the image (figures 3c, d, e, f, g, and h). For those of you with 4:3 aspect-ratio displays, these test patterns can also assist you in aligning a wide-screen image within it.

The sharpness, or peaking control, boosts and, therefore, restores high-frequency video information that receivers, decoders, or cabling connecting other video equipment to the display has attenuated during broadcast and transmission. Unfortunately, ill-trained and unscrupulous sales personnel and calibration technicians sometimes dial in an excessively high sharpness setting even if the equipment doesn't require it. The result might appear to be a sharper, more detailed picture, especially if you view the display from a distance. In fact, though, the display is only overemphasizing the edges that mark the border between image regions with luminance and chrominance differences, thereby injecting noise into the image. It thus adds information to the displayed image that doesn't exist in the original, and the effect is particularly troublesome when it magnifies the "blockiness" and other artifacts of lossy image-compression algorithms, such as DTV's and DVD's MPEG-2.

The luma-multiburst pattern is useful for detecting an excessive sharpness setting; when correct, the mid- and high-frequency patterns on the right side of the image should appear no brighter than their low-frequency peers on the left side (Figure 4a). Patterns containing abrupt gray-to-black transitions and narrow 1- to 10-pixel black lines are also useful; "ghost" images around the black lines indicate either a too-high sharpness setting or ringing resulting from improper cable-impedance matching and termination (Figure 4b).

Author Information

Technical Editor Brian Dipert is off to restock his inventory of microwave popcorn and spare-battery sets for his Sampo DVD player's remote control. You can reach him at 1-916-454-5242, bdipert@edn.com.