roberts@nbs-vms.ARPA@caip.RUTGERS.EDU (01/30/86)
From: "ROBERTS, JOHN" <roberts@nbs-vms.ARPA> It seems that a high proportion of the postings on this mailing list are concerned with high-resolution mode flicker, so I thought it would be worthwhile to post the following. ISSUES RELATING TO FLICKER IN SCANNED DISPLAYS PHOSPHORS For a number of reasons, I think that phosphors exhibit close to exponential decay in brightness with respect to time at moderately high levels of stimulus. Shown below is a rough sketch of what such a curve would look like. "Brightness" may be defined as the number of photons emitted per unit area per unit time. | | | x | B | R | I | G | x H | T | N | x E | S | x S | x | x | x | x x | +---|-----------|--------------------------------- (A) (B) TIME |_____| one refresh interval (for example) Note that for an exponential decay and a high initial stimulus, a large part of the decay takes place very shortly after removal of the stimulus. For the refresh interval shown, there is a large difference in brightness between the initial level and the level just before the next refresh. If the refresh interval can be lessened (faster refresh rate), then this difference in brightness will be reduced. This is not always possible, however. High-persistence phosphors attempt to address the problem by changing the shape of the curve to reduce its slope. This has the disadvantage that if too slow it may cause moving images to appear to blur. Another approach would be to reduce the initial level of stimulus (by turning the brightness down). For instance, if the phosphors are stimulated to the level found at point (B) instead of point (A), then the absolute decrease in brightness before the next refresh will be greatly diminished. I think that the high-speed response of the eye is sufficiently linear so that a decrease in flicker will be observed. In addition, I suspect that many phosphors exhibit increased persistence at low brightness levels. HUMAN VISION AND PERCEPTION The process of vision begins when incident photons pass through the lens of the eye and strike the retina. There they are absorbed by special dyes in various types of sensor cells and induce photochemical reactions. These reactions in turn stimulate nerve cells, which conduct limited preprocessing then send frequency-modulated signals to the visual centers of the brain. The sensor cells are called rods and cones, and are distributed unevenly over the surface of the retina. The cones are sensitive to color, produce sharp images, and are concentrated mostly in a small pit at the center of focus, which means that only a small part of what you see around you is sharp and in good color at any given time. (The brain can build up a larger detailed image by moving the eyes to scan an incoming image.) Cones have the ability to detect motion (changes in brightness), and need relatively bright light to work. There are different kinds of cones sensitive to different ranges of the color spectrum. These ranges overlap, and have their peak sensitivities at the colors red, green, and blue. This is why these are the primary colors to humans, and any color can be imitated by a combination of these three. Rods, which are mostly scattered over the areas of peripheral vision, are not sensitive to color, and in their configuration do not produce particularly detailed images. They can work in extremely dim light, and are highly sensitive to motion, such as flicker would induce. The chemical properties of the rods and cones and the timing of nerve impulses affect perception of flicker in several ways. First, images continue to appear with decreasing intensity after a stimulus has been removed, which for example makes a flash of light appear to last for much longer than it actually does. Second, a steady stimulus leads to the creation of a negative afterimage, which can eventually interfere with an incoming image. Third, observed flicker will probably be different for direct and peripheral vision. Fourth, if a stimulus flickers at a rate faster than the photochemicals can respond or the nerve impulses can be modulated in a nonuniform manner, the stimulus will be perceived as nonflickering. The visual processing centers of the brain function as an extremely sophisticated pipeline processor or systolic array. (Humans can process many images thousands of times faster than a Cray, in spite of having only a tiny fraction of the circuit switching speed.) The first stages of processing put together 2-dimensional images, detect motion, etc. Higher levels group images into patterns, and determine relative motion of patterns. Much higher levels construct large-scale detailed three-dimensional images, identify objects, recognize the faces of individuals, etc. All of these levels are to varying degrees automatic, and largely prewired. Flicker and other unusual features of scanned images can interfere with the first stages of image processing in an annoying manner. It is incumbent upon designers of electronic displays to produce images that interact well with the visual processing capabilities of the users, or which the users can learn to use effectively. IMAGE DESIGN TO MINIMIZE FLICKER In an NTSC-type display, if vertically adjacent pixels in alternate fields carry the same image, and if the pixels are placed close enough together, then they will both have an influence on any sensor cell that detects them. This will lead to an apparent refresh rate of 60Hz instead of 30Hz, and the apparent flicker will be greatly reduced. An Amiga programmer using 640x400 mode and writing with elements one pixel wide and two pixels high can therefore enjoy many of the benefits of the full hires mode at greatly reduced flicker. Several people have expressed the erroneous impression that this approach would give no greater detail than the 640x200 mode. Observe the structure of a diagonal line: ## # ## ## # ## ## # ## ## # ## ## # ## ## # ## ## # ## ## # ## ## # ## ## # ## The first pattern represents a line drawn on a 640x400 screen using elements two pixels high. The second pattern represents a minimal line drawn on a 640x200 screen using single-pixel elements. The third pattern represents a line drawn on a 600x200 screen that has been "filled out" by using double-width elements, or a line drawn on a 300x200 screen. Note that the first line looks much smoother and more substantial than the others. By inference, one can see that more detailed images or images with better detail can be drawn using the first technique than the other two, for all except patterns of horizontal lines. One could seek to design text fonts to take advantage of this technique. It is conceivable that one allowing a full 50 lines of text could be devised. If not, one might still be able to obtain a 50% increase (for instance) in the number of lines of text. COLOR, BRIGHTNESS, AND CONTRAST OF TEXT AND BACKGROUND Has anyone come up with colors and relative brightness of text and background that seem particularly useful in minimizing flicker? How well does gray scale work? THE 520ST IN MONOCHROME MODE In addition to its 70Hz refresh rate, I suspect that the ST monochrome monitor may run in noninterlaced mode or use long-persistence phosphors. Does anybody know? DESIGNING NONINTERLACED MONITORS From page 22 of the January 23, 1986 issue of EDN magazine: "DSP Chips Extend Digital-TV Capabilities" "Additions to the Digit-2000 digital-TV chip set from ITT-Intermetall (Freiburg, West Germany, TLX 772715) include...a video memory controller. The VMC-2260 video memory controller stores an entire picture frame in standard 64kx4 dynamic RAMs and provides freeze-frame, multiple picture-in-picture, and zoom capabilities plus the ability to eliminate picture flicker by doubling the TV's vertical scan frequency. For the high-bandwidth color monitors required for teletext or computer display, where vertical scan frequency doubling is not suitable, the RGB-2932 double scan processor allows you to eliminate screen flicker by doubling the horizontal scan frequency of the RGB signals." It seems that something like this could be built into a high-speed, high-resolution, non-NTSC display and used with an unmodified Amiga. I would expect such a display to cost anywhere from the price of an Amiga to several times that much. (Standard disclaimers apply. Much of this information was derived from numerous books and magazine articles that were shockingly lacking in differential equations and such, so you might want to do further study if you're really interested. Some of the explanations are a little more simplistic than an actual application would call for. My description of Amiga graphics was based on certain assumptions concerning how the display works. If there are errors in this area or elsewhere, please feel free to bring them to my attention in a non-flaming manner.) John Roberts roberts@nbs-vms.ARPA ------
jan@looking.UUCP (Jan Gray) (02/02/86)
In article <1131@caip.RUTGERS.EDU> roberts@nbs-vms.ARPA@caip.RUTGERS.EDU writes: >THE 520ST IN MONOCHROME MODE > > In addition to its 70Hz refresh rate, I suspect that the ST monochrome >monitor may run in noninterlaced mode or use long-persistence phosphors. >Does anybody know? The ST monochome display is 640 x 400 *non*-interlaced at 70 Hz (short persistence white phosphor). Jan Gray Looking Glass Software, Waterloo Ont. (519) 884-7473