Path: utzoo!attcan!telly!lethe!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!usc!apple!vsi1!zorch!xanthian From: xanthian@zorch.SF-Bay.ORG (Kent Paul Dolan) Newsgroups: comp.sys.amiga Subject: Re: 24 bit color boards Keywords: Video Master 32, PP&S Message-ID: <1990Dec7.073503.21589@zorch.SF-Bay.ORG> Date: 7 Dec 90 07:35:03 GMT References: <93505@aerospace.AERO.ORG> Organization: SF-Bay Public-Access Unix Lines: 186 huebner@aero.aero.org (Robert E. Huebner) writes: > Okay, since it seems we have a lot of graphics-terms experts on the net, Or at least one terminology freak. ;-) > perhaps you can clear up some abiguities regarding a recent press > release for the Progressive 32-bit board. All text is quotes is > directly taken from the literature. All brackets are my own comments: Okay, I'll try to translate, but take into account I never even _heard_ of this system. > "Video Master 32 offers dual frame buffers, each with resolutions of > up to 1024x1024 in 24 bits with 8 bit overlay [apparently this, not > alpha channel, is where the 32-bit figure comes from?]. The frame > buffer can display out in resolutions up to 800x600 and 1024x512 in > over 16 million colors" Right. "Alpha channel" is a software use for some hardware bits, used especially for encoding "opacity" or "coverage" information on a per pixel basis, for accomplishing transparent/translucent objects, or for antialiasing. Another possible software use for the same eight bits would be "z buffer" depth information, to "make hidden surface calculations much cheaper". (That's not exactly what it does, but that's the effect; what it really does is allow a modified painter's algorithm for all surfaces that don't hit the same depth bin, and only require hidden surface calculations when they do.) It sounds like this design's aren't 32 equal weight bits, but instead are two sets, the 8 bits having priority over the 24 bits. What I would expect from this limited information is that when the overlay bits have the "background" value (probably 0), then the color stored in the other 24 bits would show through. This takes hardware intervention, so unless the hardware is switchable to other modes, the alpha channel or z buffer uses for these eight bits wouldn't be available (without a lot of extra work; I can think of half a dozen ways to do it, but they all make my teeth itch.) >So what good is a 1024x1024 frame buffer is its maximum output resolution >is 1024x512? I find this very misleading. Well, unless you have a square screen, or stubby pixels, you won't normally display stuff at 1K x 1K on a home system (commercial systems can do this; they cost a lot). However, the 1K by 1K data storage capability is still very valuable; you can make a roamable structure bigger than the display; there are lots of uses for pictures you have to scroll to get to, but don't have to recompute to get to. If you are using one of the desktop publishing systems, you can store a bigger piece of a page, and look it over in the previewer without having to recompute until you go off the larger frame buffer surface. You can also make a game or other application in which the extra space is used to store 24 bit images to be quickly swapped into the visible area without recomputation. As a graphics programmer, I'd have no trouble putting this extra space to lots of good uses that would make it worth your money as a customer. > "Video Canvas 24, a real-time 16 million color paint program is > included. The Video Canvas updates the display instantaneously as the > user paints. Video artists may use the Video Canvas 24 to create > graphics in a scrollable area 1024X2048 pixels." [I'm sort of hoping they meant 2048x1024 here, and I'll write as if that were so. What you typed would give a four screen high vertical only scroll; a two screen by two screen scroll would be much more expected.] OK, here they have taken the double buffer and used it instead as a double width screen as well as double height. [If I were designing this hardware, I'd have been sure to put in a software switchable mode that would display every other pixel horizontally and vertically, so that an artist could draw at full resolution, but pop back to see the whole picture to make sure stuff matches up and is shaped right. (I'd also do some really clever stuff to allow painting in overview mode, and cleanup in full resolution mode.) No knowing without more information, but it would be worth asking about, and it is cheap enough to implement.] Now you have the ability to paint a really large picture. On an 8 by ten sheet of printer paper, this gives you 128 pixels per inch by 205 pixels per inch in landscape mode, pretty good for a color inkjet printer. It would also be excellent for a color film recorder, etc. > So if this is a frame buffer, how can it provide real-time feedback > when painting and scrolling. Unless the scrolling is handled by the > internal processors (TI 34082 and 34020). Well, the Amiga does this stuff now; all you have to do is give the video drawing chips new starting upper left hand corner addresses within the frame buffer memory. Since all you have to change is one x,y pair, you can draw _any_ 1024 x 512 rectangular subset of the total 2048x1024 frame buffer every screen refresh time. > "The Video Master 32 offers optional real-time 24-bit video digitizing > in all Amiga video resolutions up to 752x525 [I suppose this is PAL > interlaced overscan], including overscan." > A frame buffer that digitizes video. Seems rather toaster-ish. Sure, and the limited resolution is probably camera derived. > "The Video Master 32's design includes a programming and data RAM > storage area of 1 Megabyte, expandable to 8 megabytes. This area may > be used to run blindingly fast custom 34020 applications, such as 3-D > rendering, animation, ray-tracing, image processing, and ADO effects > generation software." Nice; you can dump a program tailored for the 34020 into its address space, and let the 34020 execute it while your 680x0 is doing other good stuff for you. This is like a blitter on power pills. > What isn't explained is- how much RAM is needed for features and > resolutions mentioned elsewhere. Later it states that the 1 Megabyte > model can only handle 1 frame buffer instead of two. Also, what are > these "custom 34020 applications". Do they exist? Will PP&S develop > them, or are they included (doubt it). Well, two frame buffers 1024x1024x32 is eight megabytes, which wouldn't leave any room for software if that were the same memory being used for the video storage, and one megabyte is nowhere close to a frame buffer. This is a bit confusing, especially with the quote about "one frame buffer instead of two"; is there 8 megabytes for the buffers, plus 1 to 8 megabytes user ram, or do you fight for space in a single eight megabyte space? This needs clarification. You _never_ expect spiffy software with new hardware, though there's always something sold to make it at least minimally usable, like that paint program. If the product builds a good base with its startup software, and is as nice as it sounds, software will follow. That's the way the biz works. > "The Video Master's 8-bit overlay allows for operating system windows > and menu displays in up to 256 colors, overlaying the 24-bit video > image. This design eliminates palette selection problems and allows > for easy editing and painting in 24 bits" This is nice, and was one of the first ways 32bit frame buffers were implemented, but it would be nicer if this could be switched off and on by the software, to allow alternate use for z buffer, alpha channel, or other needs, without interfering with the 24 bit display. This may well be part of the design, but I don't see it here. >This is interesting. So this means the mouse pointer and workbench screen >can appear over the 24-bit graphics? Will this allow for directing painting >on the frambuffer screen? Given the software, you could build an eight bit picture that completely covered the 24 bit picture, clear it to zero, and have the original picture intact. That's what makes this an attractive technique: you can overlay any type of trash on top of the "background" picture, and still not have to recompute/redraw the background picture as the stuff in front is modified, moved, or removed. > "All configurations available in NTSC or PAL" > Overall, I begin to think that the claim of 1024x1024x32 bits is a bit > misleading, since the rest of the article stresses its ability to > output to video using a 1024x512 (Which is nice, but is only half what > they claim earlier on) screen. No, this time it's OK; having a writable area bigger than your displayable area is a well known trick; the Amiga superbitmap mode already supports a 1024x1024 writable picture area with a ( first implementation ) 640x400 window into it shown on the display. Not meant to mislead; they really put all those pixel bits in there for all those pixels, your disply just doesn't let you look at all of them at once. > Overall, this sounded like an exciting product, second only to the > Video Toaster in terms of "drool-factor". No price mentioned, and > availability listed as first quarter 91. Wish I were rich! ;-) > My basic question is- is this a true 24-bit framebuffer like the > Firecracker board? How does this differ from the toaster? I don't > believe the toaster uses any on-board RAM, but I could be wrong. No experience with any of them (poverty), but it sure sounds like a true 24 bit buffer from what you quoted. >If anyone can clarify these points, I'd appreciate it I did my best; I hope that helped. Kent, the man from xanth.