> On Fri, May 06, 2011 at 03:12:24PM -0700, Keith Lofstrom wrote: > > I fervently agree with the 4:3 minority, but there are many who > > agree with the screen marketing departments instead. The reason > On Fri, May 06, 2011 at 09:12:08PM -0600, Bill Thoen wrote: > Sometimes I wonder how different things would be today if we had taken a > different path in developing electronic display technology. The original > Cathod Ray Tube and its immediate descendants (oscilloscopes and RADAR > screens) are round. A round display might have been a good thing. With > pixels packed hexagonaly it would lend itself to smoother anti-aliasing and
Memory incontinence from the ancient curmudgeon: Back when we were designing the dinosaurs (BEFORE they roamed the earth) at Tektronix we made storage CRTs - the screen was the memory, one color, binary, but infinitely addressable in theory, electron optics and phosphor graininess the actual limit. You used the electron beam to draw lines on the screen. There was probably a way to use a different electron energy beam to erase the lines, but we were lazy and just erased the whole thing, like a big "hollow state" etch-a-sketch (do they still make those?). No pixels, per se. Well, we did read a "ROM" - a big array of diodes wired into circuit boards - storing the pixels for ascii characters. But it was mostly about "calligraphic vector graphics", drawing pictures with line segments. If you wanted to draw an Olde Englishe capital letter, and the RKO5 disk pack on your PDP8 had the vector description, you could make something beautiful, just like the medieval monasteries used to make - assuming the monks were color blind. Not bad for an era when a big memory on a supercomputer was 192K bytes. No aliasing. 4K*3K addressing on a 25 inch diagonal screen (which took two strong people to lift). Circles were CIRCLES, dammit, if you spent enough time drawing little tiny vectors at 115Kbaud. No jaggies, none of this pixel nonsense. Given 192K byte limits on supercomputers, you could not afford RAM pixels anyway. But memory prices were plummeting, and RAM pixels won. I designed an LSI replacement for the electronics on what would have been the last storage CRT graphics terminal - except that by the time it was done, RAM/pixel/raster graphics was already cheaper, and obviously going to get cheaper still. Storage CRTs were history. Regards hexagonal pixel grids - that is effectively what you have in analog color TVs (look up "interlace" and "color subcarrier"). The way we encoded color on the analog signal offset the color information by "half a pixel" every other line, so the artifacts of the low-bandwidth color signal were not visible. It was a clever way to send a lot of visual detail over limited broadcast bandwidth, fooling the eye and the brain into thinking it was seeing something in real world detail. You still have aliasing, with 3 axis symmetry rather than 2. However, most of our graphics math and analytical techniques still focus on cartesian coordinate space. The X/Y grid used in RAM maps onto the X/Y grid used in screen addressing. When we compress images, we are still fooling the eye, but we do so over larger extents and use a lot more information. Still, the best way to transmit an infinitely precise and detailed line drawing is something like SVG, or spline outlines for fonts, providing drawing instructions instead of pixels. So the CRT, and the analog pixel-less screen, still exists as an abstraction deep inside the core of graphics programs and font rendering engines. It also exists in some 3-D printers, numerically controlled milling machines, navigation systems, and other Real World applications. I suspect we will find ways to computationally "de-pixelize" orbit calculation, chemistry and weather modelling, and find find faster and more accurate ways to compute those, drawing on more interesting basis functions than cartesian grids. Learn More Math, and work on better internal representations. The translation to pixels should occur after all the interesting computation is complete. The screen is just another object to pipe abstractions into, and less interesting than many other real world objects. Keith -- Keith Lofstrom kei...@keithl.com Voice (503)-520-1993 KLIC --- Keith Lofstrom Integrated Circuits --- "Your Ideas in Silicon" Design Contracting in Bipolar and CMOS - Analog, Digital, and Scan ICs _______________________________________________ PLUG mailing list PLUG@lists.pdxlinux.org http://lists.pdxlinux.org/mailman/listinfo/plug