The Cerulean Crux
There are a number of problems associated with the display of video imagery. For one thing, physical objects have shapes. For another, shadows may be more important than the objects that cast them. And then there are Salem cigarettes.
Fifty years ago, color-television sets first went on sale. Two years later, Salem, the first filter-tipped menthol cigarettes, also hit the market.
It was a different time. What we know today as the NBC Nightly News was then called Camel News Caravan, named for the cigarettes “more doctors smoke than any other brand.”
Tobacco advertisers helped pay for what some consider the “Golden Age” of television. So it was natural that Salem commercials would appear on the new color broadcasts.
Unfortunately, there was a problem. Salem’s marketers wanted to convey the cigarette’s menthol taste in its packaging. They selected a saturated blue-green color suggestive of trees, skies, and bodies of water. Call it aquamarine, azure, cerulean, deep turquoise — whatever it was, it could not (and still cannot) be reproduced on a color-television set.
Color picture tubes reproduce pictures by causing chemicals called phosphors to glow. There is a red phosphor, a green phosphor, and a blue phosphor. Combined in appropriate proportions, they generate white light.
Mapped on a chromaticity diagram, each phosphor forms a point, and the three points form a triangle. Any color within the triangle can be reproduced by adjusting the proportions of the different phosphor glows. Unfortunately, Salem’s cerulean logo falls outside the triangle of color-television phosphors — any color-television phosphors.
The colors that humans can see form a shape something like an elongated capital letter D rotated about 150 degrees counter-clockwise. The D is curved; triangles are not.
It could be possible to formulate a cerulean phosphor that matches Salem’s packaging, but the resulting triangle would lose many more colors. Red-green-cerulean would lose blue colors; red-cerulean-blue would lose greens; green-cerulean-blue would lose most of the spectrum. So viewers — including Salem’s marketers — have learned to live with imperfect reproduction of the cigarette’s logo over the last half century.
That’s a long-known flaw in television displays. A more-recent one involves rubber table legs.
From the advent of broadcast television until the introduction of solid-state imaging chips, there was a perfectly synchronized relationship between camera and display. If the camera was scanning the upper left of the image, so was the display.
The chip cameras, however, captured entire pictures at once, while picture tubes still scanned from top to bottom. As a result, the bottom of the picture would appear on the screen later, in relation to when it was captured, than the top. The delay being only about a sixtieth of a second, it was generally imperceptible to viewers.
If a camera is panned rapidly back and forth across a table leg, however, the temporal difference between full-image capture and scanned-image display can become apparent. A straight wooden leg can appear to bend back and forth as though made of rubber.
That display artifact was brought about by a change in cameras. Today, the issue is more one of displays.
Cathode-ray tubes (CRTs) — both direct-view picture tubes and projection tubes — are still found in the vast majority of all television sets worldwide, but the CRT percentage is falling. Today, consumers with enough money can easily purchase plasma display panels (PDPs), direct-view liquid-crystal displays (LCDs), projection displays based on tiny LCDs, and other projection displays based on DLP (digital light processing) micro-mirror displays. There have also been technology demonstrations of video versions of so-called field-emission displays (FEDs), organic light-emitting diode (OLED) screens, and grating light-valve projectors, among other technologies. And then there’s the shape issue.
The visible world has no shape. Traditional video cameras adopted the image shape of Edison’s original movie camera, 4:3 (four units of width to three units of height).
Over the years, movie images have become wider screen, and, today, there’s widescreen video, too. Unfortunately, there’s still traditional video.
Cinema theaters show movies in a range of shapes, but viewers always see what appears to be a screen that matches the image perfectly. That matching is achieved through the use of draperies and other screen-masking materials.
Although JVC once demonstrated a TV set with rolling slats that could mask off unused portions of the screen, most video displays have no equivalent to theatrical draperies. As a result, images not matching the screen shape either don’t fill the screen, are chopped off, or are stretched to fit, making the people and objects in the image the wrong shape.
Then there’s differential phosphor luminance decay (DPLD). Phosphors grow dimmer with age, and blue phosphors tend to age fastest. That means that, with time, CRT-based video displays eventually become both darker and more yellow. For most viewers, the change is slow enough and subtle enough to be imperceptible. After all, a white piece of paper looks white both under an incandescent lamp and at noon on a cloudy day, despite wild variations in the color of those light sources. Humans are not very sensitive to large-area hue shifts.
On the other hand, humans are very sensitive to small-area hue shifts. In a paint store, it’s easy to distinguish between samples of, say, slightly differently colored whites.
When images of one shape are viewed on a screen of a different shape and are shrunk to fit, the phosphors in the resulting unused areas (black bars) don’t age. Those in the used areas do. So the used become yellowish relative to the unused, and the used/unused border becomes apparent when the entire screen is filled.
The problem is particularly significant in plasma displays due to the types of phosphors used in them (sensitive to ultraviolet stimulation rather than electron beam). Early plasma displays could exhibit this “burn-in” problem after only months of viewing the wrong shapes. Today’s are much better, but the DPLD issue hasn’t gone away completely.
Even after all television switches to a widescreen shape, there will still be issues associated with some of the new technologies that haven’t been noticed in CRTs. Consider edge positioning.
The resolution of a black-&-white picture tube or a color CRT-based projector is based on, among other things, how fine a spot the electron beam can make and the bandwidth of the video amplifier. Broadcast analog video in the U.S. is limited to a bandwidth of 4.2 million cycles per second (4.2 MHz).
That video has 525 total scanning lines per frame and roughly 30 frames per second for a line rate of about 15,750 scanning lines per second. Looked at another way, each scanning line lasts about 63.5 millionths of a second (63.5 us). Take away about 11 us for the horizontal blanking interval (that portion of the scanning line that includes a synchronizing pulse and a color reference), and there are 52.5 us left.
A cycle has a high part (peak) and a low part (trough). If all the cycles on each scanning line are aligned, the result will be a white line (peak) and a black line (trough) on the screen for every cycle. There can be about 220 cycles of a 4.2 MHz signal in 52.5 us. Thus, the maximum horizontal resolution of a U.S. broadcast analog video signal is about 440 lines across the width of the picture.
That seems very low compared to the 640 picture elements (pixels) of the computer VGA standard or the more-than-850 of even a non-high-definition plasma display. But consider what happens when those lines move.
In a black-&-white picture tube or color projection tube, there is no restriction as to where the lines fall. Move them a half-line to the right or left, or a quarter-line, or a millionth of a line, and the lines will be just as sharp as they were in the original location.
Now consider any fixed-pixel-position display: DLP, LCD, PDP, OLED, etc. Assume a vertical pattern of alternating black and white lines such that one column of pixels will be white and the adjacent one black. Now shift the pattern one half line horizontally. Each pixel is now getting both a white line and a black line. The entire display will turn gray.
That’s prevented by filtering, but the filtering reduces horizontal resolution. As a result, the 440 hypothetical lines of resolution of analog broadcast television can actually offer more detail than can be displayed on a fixed-pixel display offering over 850 pixels across the screen.
Then there’s gray scale. CRTs are highly non-linear devices. There is much more signal-voltage range devoted to dark areas than to bright ones. That’s compensated for in cameras, and it’s also handy for digital processing.
Consider an eight-bit digital system, offering 256 different levels. Humans can perceive a small-area contrast change of about 1%. So, in a linear digital system, at level 100, the difference between that level and levels 99 or 101 would just barely be perceptible.
At level 250, however, a human wouldn’t be able to perceive the difference between that level and 249 or 251, so the extra levels would be wasted. And, at level 10, the difference between it and levels 9 or 11 would be ten times too great for smooth gray-scale reproduction. Contouring lines would appear.
The non-linear characteristics of CRTs assure that perceptible changes will be relatively uniform across the entire range. They may be technically non-linear, but they are effectively perceptually linear.
Some of the new display technologies, however, are technically linear and, thus, perceptually non-linear. In a DLP system or a plasma display, pixels are either on or off. Frames are subdivided to provide grey scale. If a pixel is on for half the frame duration, it provides half the light. If it’s on for a quarter of the duration, it provides a quarter of the light.
Unfortunately, such linear light processing results in the level-ten problem noted above. Early plasma displays looked good when reproducing bright images but introduced contouring (overly perceptible borders between different shades of gray) in darker images. Later plasma displays introduced an intentional form of graininess, called dither, to eliminate the contouring. That worked, but in some models the contour lines were replaced with almost equally visible dithering dots. Texas Instruments dealt with the problem in DLP by subdividing frames more finely.
There are other issues in the new display technologies. Early LCD screens had such limited viewing angles that, when they were used in control-room walls, the director might not have been able to see anything from the display in front of the associate director in the very next seat. And then there’s the liquid part.
If you’ve ever done exercises in a swimming pool, you know you can’t move as quickly as you can in air. Similarly, it takes time for a liquid-crystal cell to shift from on to off or vice versa. Compared with aquatic exercise, it’s lightning fast; compared with the phosphors of a CRT, it’s snail slow.
Of course, just as plasma displays have improved with time, so, too, have LCDs gotten wider viewing angles and faster transition speeds. As problems are discovered, they’re getting resolved. And some of the upcoming technologies (such as GLV) promise to bring the best characteristics of CRTs into the 21st century.
Viewers of Lord of the Rings on modern video screens, therefore, don’t have to say, “Display’s the thing on which to catch the contours of the king.”