Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!cs.utexas.edu!usc!ucsd!ucsdhub!celit!billd From: billd@fps.com (Bill Davidson) Newsgroups: sci.electronics Subject: Re: I don't need HDTV! Message-ID: <7353@celit.fps.com> Date: 17 Mar 90 00:18:15 GMT References: <8Zx8Ip200ioEMMrHEF@andrew.cmu.edu> <132618@sun.Eng.Sun.COM> <2694@sactoh0.UUCP> <1990Mar13.023805.24765@athena.mit.edu> <1990Mar15.090214.9871@spectre.ccsf.caltech.edu> <7322@celit.fps.co <1990Mar15. Organization: FPS Computing Inc., San Diego CA Lines: 44 In article <1990Mar16.172343.10577@phri.nyu.edu> roy@phri.nyu.edu (Roy Smith) writes: >toddpw@tybalt.caltech.edu (Todd P. Whitesel) writes: [Todd didn't write this, I did --billd] >> square pixels would be nice for computer graphics. > I think the thing to realize here is that there is very little in >common between broadcast TV and computer graphics. First off, you have to >worry about bandwidth. With 1024 x 1024 x 24 bit graphics at 72 Hz, you're >talking about 75 million pixels per second. It may be fine to talk about >100 MHz video bandwidth on each of three coax cables in a lab, but how are >you going to transmit that over the air? One of the absolute limits of >broadcasting is that bandwidth is a finite resouce. Unless you can change >the gravitational constant of the universe by "just doing it", you have to >live with that fact. I never expected digital signals. Even a lunatic resolution freak like me knows that that's too much to hope for. Bandwidth is a problem. How we deal with it is unknown. We can compress the signal. We can also have less channels and transmit diffrent parts at diffrent frequencies. Maybe we can't do it. I'd sure like to see it happen though. > Secondly, with interactive computer graphics, the ratio of sources >to displays is about 1:1. With broadcast TV, there are many, many more >recievers than there are sources of material to watch. Using square pixels >may make life easier for the ray tracer folks, but that's just tough on >them. Is there any fundemental reason why you can't do ray tracing with >rectangular pixels? If using rectangular pixels saves broadcast bandwidth, >or makes it easier to build a reciever, it's A Good Thing. No it's not. There is no problem for raytracing with rectangular pixels (hell, elongated hexogonal pixels would work just fine). Raytracing is not the part of computer graphics that has a problem with rectangular pixels. The problem is with geometric transformations. Rectangular pixels cause a lot of extra work to have to be done to scale the image and this has to be done every time you move anything around on the screen. This costs you a lot of cpu cycles which are usually in very short supply when you are doing graphics animation. This cost is very significant and unnecessary. How do rectangular pixels save broadcast bandwidth? I thought that they just increased horizontal resolution. I don't need a resolution increase at this cost. --Bill Davidson