CompuServe Thread

Video resolution is NOT 640×480. True analog video has no pixels, so there is no horizontal pixel resolution. The vertical pixel resolution is about 486 for NTSC and 578 for PAL, but that's because of the discrete scan lines inherent in the raster-scanning method rather than actual pixels per se. (NTSC actually has 525.5 lines per frame, but some of those are used for vertical blanking [VBL] and so aren't available for image data. PAL has a similar situation.) Many people confuse "lines of resolution" with "pixel resolution." For one thing, "vertical lines of resolution" measures HORIZONTAL resolution, while "horizontal lines of resolution" measures VERTICAL resolution. That may seem backwards at first, but the idea is to see how many DISTINCT lines can be resolved (seen as separate and distinct individual lines). If the lines are vertical, it's the horizontal resolution that resolves them, and vice-versa. Imagine a 640×480 computer screen, say, a VGA card. Now imagine that you run this QBASIC program: DEFINT A-Z SCREEN 12 ' 640×480 2-color mode FOR horiz=0 TO 639 LINE (horiz,0)-(horiz,479),1 NEXT horiz This code draws 640 (0-639) vertical lines (presumably white on a black background) from the top (0) to the bottom (479) of the screen. What do you see when you run it? 640 vertical lines? NO!! You see a COMPLETELY WHITE SCREEN (except for a black border)! The lines blend together to form a solid white area. All the pixels are set to white. There are no black ones left to resolve the lines. If you wanted to see as many vertical lines that were still distinct lines as possible, you would need to make the following change to the above program: FOR horiz=0 TO 639 STEP 2 That causes the loop to skip every other value for "horiz," meaning that only the even-numbered horizontal coordinates actually get lines drawn. But this means only HALF as many lines are drawn — 320, not 640! A 640×480 PIXEL SCREEN CAN ONLY RESOLVE 320 VERTICAL LINES OF HORIZONTAL RESOLUTION! The "vertical lines of horizontal resolution" value is ALWAYS — **ALWAYS** — only HALF the value of the horizontal pixel count! Now remember that SVHS is claimed to resolve "about 400" lines, while Betacam-SP component is claimed to resolve "well over 400" lines. The Targa+'s maximum 32-bit PIXEL resolution of 512 means it can only take full advantage of a measly 256 lines of resolution, and that's counting border overscan! The 720 of the Matrox Illuminator Pro can take advantage of up to 360 lines of resolution, which is about equal to the limit of composite video and approaching the limits of SVHS (or Y/C). The 756 of the ATVista resolves up to 378 lines, well into SVHS's domain. I don't actually have an ATVista, but I do have HiRes QFX 3.0 which supports it. On page 170 of its manual, it lists the ATVista resolutions it supports, and in addition to 512×482, 512×486, 604×486, 640×486, 646×486 (with 12.2727MHz crystal), 720×486 (13.5MHz crystal), and 756×486, there are two other NTSC resolutions listed: 1008×486 and 1512×486! (PAL goes up to 986×578 and 1480×586!) There are also two "high-resolution" modes listed, presumably for computer display use: 1024×768 and 1024×856, both interlaced. Page 172 shows some sample "SET DISPLAYCFG=" commands which imply that the ATVista can do 1024×768 at full 32-bit color (assuming enough ATVista RAM), and if it can do that, it should be able to go all the way to 1512×486 and still be 32-bit, unless there's some factor (dot clock rate?) I'm overlooking. Perhaps 1008×486 and 1512×486 only work in 16-bit mode. (Can anyone give me a definitive answer on this?) If those two modes DO work in full 32-bit mode (or even if they don't and you only need 16-bit for a particular application — text screens, for instance), those modes can resolve up to 504 and 756 LINES OF RESOLUTION, respectively! 756 LINES, *NOT* 756 PIXELS!! Note that there is NO inherent limit to horizontal resolution in Y/C or Component signal types where the Luminance (Y) signal is concerned. The chrominance signal (C) of Y/C, and the entire composite video signal, is limited by the 3.58MHz color burst signal under NTSC. Y/C can resolve objects based on luminance alone, though there may be color fringing if an object is smaller than a color burst cycle. Composite cannot resolve any object smaller than a color burst cycle, which is why most home computers designed for TV use (Commodore 64, for instance) can't display more than 40 characters across or 320 pixels across. Color burst interlacing (not to be confused with scan line interlacing) doubles the apparent horizontal res. of the color burst signal, similar to how scan line interlacing doubles the apparent vertical resolution of the entire picture. I'm not sure if the component color signals (R-Y, B-Y) use a color burst signal or not. If not, then they, too, aren't bound by its resolution, and the maximum resolution of Betacam-SP component is hypothetically unlimited for both luminance and color! The Nyquist principle states that in any digital signal processing application, you need TWICE as much resolution as you might THINK you need at first glance, because if you don't a kind of digital noise will be introduced by a form of "aliasing" resulting from a "bouncing" of the frequencies from the midpoint frequency. This is why CD digital audio is sampled and played back at 44kHz when the human ear can hear at most 22kHz sound, for instance. When scanning a document intended to be printed with a 120 lines per inch halftone screen, you need to scan it at 240 dpi for best results. So it is with video: in addition to the doubling of pixel resolution needed to resolve lines, for absolute best results you should double AGAIN because of the Nyquist aliasing phenomenon. Thus, the ATVista 1512×486 mode, which is 756×486 doubled. Particularly with fine details and small text, you will notice the difference.