Irgendwer Posted April 28, 2008 Share Posted April 28, 2008 Hi, after supplying the ideas for a ST-dither-mode in GEM-View with 150 colours in ST-Low and the Lynx LCD-emulation mode in 'Handy' I thought yesterday about utilizing the same principle for the A8: Mapping the R,G,B values 'through' palette entries. And yes, it works! I think this kind of dithering is new for the A8, but I quite lost the overview regarding all the 'colour-formats'. So please tell me if this is 'old cheese'.... + no interlace + just a DLI every line changing the GR.9 base colour from R to G to B and so on... + results in 16*16*16 = 4096 colour impressions + quite intuitive scheme for image manipulations + light areas are brighter compared to the common gr.9/gr.11 scheme and not broken by 'dark lines' + quite good contrast - since the A8-base colours are poor saturated, the images are not very brilliant Actual palette is 'laloo'. Base Gr.9 colour may should differ on your machine depending your TV system & adjustment. Comments welcome! Format name? I call this format JAD (just another dither)... The zip contains the G2Fs. JAD.zip 1 Quote Link to comment Share on other sites More sharing options...
Urchlay Posted April 28, 2008 Share Posted April 28, 2008 Hm, so your effective resolution is 80x64 (assuming 192 scanlines)... on the plus side, no flicker/interleave, and you only use 8K of video RAM... I'd be interested to see what a photo of a person's face looks like, in this mode. Got anything you can post? Quote Link to comment Share on other sites More sharing options...
Heaven/TQA Posted April 28, 2008 Share Posted April 28, 2008 do not know...honestly... but looks like APAC for me, judging on screens but haven't seen the G2Fs. Quote Link to comment Share on other sites More sharing options...
Irgendwer Posted April 28, 2008 Author Share Posted April 28, 2008 Hm, so your effective resolution is 80x64 (assuming 192 scanlines)... on the plus side, no flicker/interleave, and you only use 8K of video RAM... Yes, but the resolution is virtually higher than 80x64 (which is only the colour resolution!) since the lightness is mapped directly though to the destination image. Results in colour artefacts but is OK compared to a reduced resolution (examine the green artefact in ADT just above the arc). I'd be interested to see what a photo of a person's face looks like, in this mode. Got anything you can post? Here we go... (please notice the resolution of the 'play-ring'...) CU Irgendwer JAD2.zip Quote Link to comment Share on other sites More sharing options...
Irgendwer Posted April 28, 2008 Author Share Posted April 28, 2008 (edited) do not know...honestly... but looks like APAC for me, judging on screens but haven't seen the G2Fs. Definitely not APAC. There is no GR. 11 line in the screens + more than 256 colour impressions. Please also examine the light background building in the 'underground'-image. There is no 'run-through' of dark lines like in APAC interleaving... CU Irgendwer Edited April 28, 2008 by Irgendwer Quote Link to comment Share on other sites More sharing options...
Rybags Posted April 28, 2008 Share Posted April 28, 2008 So, these are just static lines of R,G, B? Hasn't someone already done that? Thought I saw such a pic in the recent compo. Another mode I thought of trying - cycle each scanline through it's R,G,B components each frame. Similar to the existing "4096" mode, except instead of R,R,R,R... followed by GGGG... then BBBB... you have RGBR followed by GBRG then BGRG etc. Quote Link to comment Share on other sites More sharing options...
fox Posted April 28, 2008 Share Posted April 28, 2008 12 years ago I've seen a program which could display this mode with interlace, so that one frame was RGBRGB..., next one GBRGBR... and next one BRGBRG... This program wasn't new at that time (I think it was made in late 80's). It could optionally use GR.15 instead of GR.9 for increased resolution at the cost of lower color depth. It generally looked much worse than APAC because of more flickering and less saturation (even on a TV). Unfortunately I can't remember the name of this program. Quote Link to comment Share on other sites More sharing options...
Rybags Posted April 28, 2008 Share Posted April 28, 2008 (edited) Just wondering how these combinations might go... 1 line Gr.9 - cycling through R, then G, then B values each frame 1 line Gr.11 repeat or Gr. 10 dispay, using RGB "triads". Pixel layout: /R\GB/R\GB GB\R/GB\R only 8+1 colour combos available, so there could only be 3 levels of 2 components, 2 levels of the third. Or, possibly have the background colour as the third level of the third component. Probably be too "chunky" and the eye wouldn't mix the colours. Maybe then, have Gr.10 with 3 frames of triads, with each physical pixel going through R,G,B each frame. Would mean ~ 24K data for a screen though. Edited April 28, 2008 by Rybags Quote Link to comment Share on other sites More sharing options...
puppetmark Posted April 28, 2008 Share Posted April 28, 2008 Comments welcome!Format name? I call this format JAD (just another dither)... The zip contains the G2Fs. I am not up on all the color formats either but this looks amazing. I can't wait to try this on real hardware. Very nice! Quote Link to comment Share on other sites More sharing options...
Rybags Posted April 28, 2008 Share Posted April 28, 2008 Load the pics in G2F, then you can save them as XEX files. I think my G2F version has a bug - it doesn't create an XEX file unless the same file as G2F exists in the target directory. Quote Link to comment Share on other sites More sharing options...
Heaven/TQA Posted April 28, 2008 Share Posted April 28, 2008 Rybags...and what about the Gr.10-pixel shift? Quote Link to comment Share on other sites More sharing options...
Rybags Posted April 28, 2008 Share Posted April 28, 2008 The Gr.10 suggestion I put in would be Gr.10 only. Maybe forget the "triads" idea - probably too chunky. Better idea would be just to have the rotating frames. Another possiblity - Gr. 9 and have a kernal to change background colour register on the fly. Although having the RGB in 3 or 4 bands across the screen would probably be somewhat annoying to watch. Quote Link to comment Share on other sites More sharing options...
Irgendwer Posted April 28, 2008 Author Share Posted April 28, 2008 Rybags...and what about the Gr.10-pixel shift? I thought about that too. Could improve the resolution impression. Didn't check that, because it is not so easy to test... ...or has anybody a hint for a G2F Gr.9/Gr.10 setup ? Another idea, to get rid of the stripe-look is, to interleave two components per line by using only gr.10: RGRGRGRGRGRGRGRGRGRG -> DLI changes 4 red colours to 4 blue ones GBGBGBGBGBGBGBGBGBGB -> green to red BRBRBRBRBRBRBRBRBRBR -> start over... cons: - only 4 shades per channel (+ninth colour would be black) - needs more DLI time while exchanging 4 registers.. results in 5*4*4 80 colour impressions CU Irgendwer Quote Link to comment Share on other sites More sharing options...
Urchlay Posted April 28, 2008 Share Posted April 28, 2008 I had a thought, reading another thread about "extended" graphics modes... There's a mode that alternates GR.11 and GR.15 every other frame, for 16 chromas * 4 lumas = 64 colors (I think this is known as CIN mode). How about extending that by alternating between GR.11 and Antic mode 4, with every 3 lines (120 characters) having a DLI to change character sets? This would be the same resolution as GR.15, but you get the possibility of an extra color (by setting bit 7 on the characters). Setting the color regs to whites/greys, you can have 5 possible luma values, but each 4x8 pixel cell only has 4 of them (the top 4 or the bottom 4, I suppose). This gives 16 chromas for the GR.11 frames and 5 lumas for the Antic 4, which would be 80 colors... but you lose the "any point, any color" nature of the CIN mode, and you eat extra CPU for the DLIs, and you use more RAM: Need 960 bytes of screen RAM and 8 full 1K character sets, except the last 8 characters of each wouldn't be used. Charsets have to fall on a 1K boundary, so you end up with 8 64-byte holes (the mode 4 display list and DLI code could live in a couple of them, the rest could be used for small tables/subroutines by whatever app is using this mode). The on-disk format wouldn't be as wasteful: could store 2 standard 62-sector images, plus 960 bits (120 bytes) to hold bit 7 of each character of the Antic 4 screen RAM. The rest of the bits don't need to be stored because each 3 lines of mode 4 would always have 0-119 in the bottom 7 bits, in order. Does anyone think this sounds worthwhile, or is it way too much effort for not enough gain? Quote Link to comment Share on other sites More sharing options...
Irgendwer Posted April 28, 2008 Author Share Posted April 28, 2008 I am not up on all the color formats either but this looks amazing. I can't wait to try this on real hardware. Very nice! Thank you. Thought so too... CU Irgendwer Quote Link to comment Share on other sites More sharing options...
Rybags Posted April 28, 2008 Share Posted April 28, 2008 Gr. 10 interleave could be an idea. The RAM usage would be kinda high. Only 8 levels of R/G/B available on the gr.10 lines. But, you'd want alternate frames having Gr. 10/9 then Gr.9/10. And, cycling the R,G,B components means another 3 combos. All up, 6 combos: 1: Gr. 9 R, Gr.10 G, Gr. 9 B 2: Gr.10 G, Gr.9 B, Gr.10 R 3: Gr.9 B, Gr.10 R, Gr.9 B 4: Gr.10 R, Gr.9 G, Gr.10 B 5: Gr.9 G, Gr.10 B, Gr.9 R 6: Gr.10 B, Gr.9 R, Gr.10 G So, that would be 48K for a perceived 160x200 picture. Quote Link to comment Share on other sites More sharing options...
Rybags Posted April 28, 2008 Share Posted April 28, 2008 (edited) Re: Antic Mode 4. I did a thread a few months back about GTIA modes in Antic 4. The colour mapping changes due to the way the AN0-AN2 lines are encoded in Mode 4 vs Mode 2. To get all the colour combinations on the GPRIOR=$C0 scanlines, it relies on bit 7 of the character to be set in some circumstances. So, such a mode with alternating colour only and luma via the 4 PF registers would have it's limitations. Edited April 28, 2008 by Rybags Quote Link to comment Share on other sites More sharing options...
Allas Posted April 28, 2008 Share Posted April 28, 2008 (edited) The format is called RGB pictures. It's an old technique, not very used on Atari. Well, in this diskette there is a loader with examples. It's a Basic program called CSVIEW.BAS Colorview_2.61_B_.zip It seems you have found a better process to translate the vivid colors from the original sources. Great! It's time to do a program converser BMP->RGB A hint: you can find better resolution results using this technique Gr.15 with DLIs. Not much color impressions, but works very well. (Grrr... I can't found my examples) Edited April 28, 2008 by Allas Quote Link to comment Share on other sites More sharing options...
MEtalGuy66 Posted April 28, 2008 Share Posted April 28, 2008 (edited) 12 years ago I've seen a program which could display this mode with interlace, so that one frame was RGBRGB..., next one GBRGBR... and next one BRGBRG... This program wasn't new at that time (I think it was made in late 80's). It could optionally use GR.15 instead of GR.9 for increased resolution at the cost of lower color depth. It generally looked much worse than APAC because of more flickering and less saturation (even on a TV). Unfortunately I can't remember the name of this program. It was called COLRVIEW.. ANd there was an assembler version of the program, that came on a disk along with a converter program. This disk pre-dated the one above and used 3 separate files for each picture. If noone else has it, Ill dig it up when I get home and upload it. And it had more of a "swimming" effect than a flickering effect. And it did a full 80x192x4096color or 160x192x64color.. APAC does 80x96x256color by alternating mode lines of gr.9 and gr.11.. Edited April 28, 2008 by MEtalGuy66 Quote Link to comment Share on other sites More sharing options...
MEtalGuy66 Posted April 28, 2008 Share Posted April 28, 2008 (edited) Heres 3 disks worth of sample pics. If theres any porn in there, I apologize, but these were converted in my younger years.. heheh. RGBpics.zip Edited April 28, 2008 by MEtalGuy66 Quote Link to comment Share on other sites More sharing options...
potatohead Posted April 28, 2008 Share Posted April 28, 2008 These look very good. Nice conversion. Some other pictures I've seen done this way didn't have anywhere near as good color in them. Quote Link to comment Share on other sites More sharing options...
Urchlay Posted April 28, 2008 Share Posted April 28, 2008 To get all the colour combinations on the GPRIOR=$C0 scanlines, it relies on bit 7 of the character to be set in some circumstances. Ah, I meant interleaving frames, not scanlines... the GTIA mode bits would be turned off, for the mode 4 frames. Basically it'd show GR.11 on even frames and mode 4 on odd frames (what people like to call "interlaced", but I don't like to call it that because the A8 doesn't do true interlaced video). Quote Link to comment Share on other sites More sharing options...
Rybags Posted April 28, 2008 Share Posted April 28, 2008 To get all the colour combinations on the GPRIOR=$C0 scanlines, it relies on bit 7 of the character to be set in some circumstances. Ah, I meant interleaving frames, not scanlines... the GTIA mode bits would be turned off, for the mode 4 frames. Basically it'd show GR.11 on even frames and mode 4 on odd frames (what people like to call "interlaced", but I don't like to call it that because the A8 doesn't do true interlaced video). I get your meaning now. I don't think such schemes look so good. You have to display the Chroma at luminence 0 or 2. Since the luma would be a seperate frame, the PAL mixing doesn't occur. I think it would be a flicker-fest and not look so good. One idea might be to use the PMGs in CIN mode (Gr.11, 15 interlaced). In combination with Gr. 15 and colour blending (GPRIOR = $00), it would allow some extra luminences. Problem is you have to stretch the PMGs to 4x resolution to cover the entire screen. Quote Link to comment Share on other sites More sharing options...
Irgendwer Posted April 28, 2008 Author Share Posted April 28, 2008 ...Colorview_2.61_B_.zip It seems you have found a better process to translate the vivid colors from the original sources. Great! It's time to do a program converser BMP->RGB ... Have to check the colorview stuff. I have here allready written a PNG -> 'RGB' ( ) converter to build the examples, but it is very basic and needs some postprocessing. Maybe I publish it later. Thanks for the reply Irgendwer Quote Link to comment Share on other sites More sharing options...
Irgendwer Posted April 28, 2008 Author Share Posted April 28, 2008 ...APAC does 80x96x4096color by alternating mode lines of gr.9 and gr.11.. ... 16 (gr.9) x 16 (gr.10) = 4096 ???? CU Irgendwer Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.