The palette converter for the Archimedes

[litwr]

Colors for the 256-color palette are not arbitrary - there are limitations. Is there a converter that can convert an arbitrary palette to one that matches the Arcimedes hardware?

[NickLuvsRetro]

Essentially, from what I understand, you can set 16 of the 256 colors, with the rest being populated by a 'house mix' of tints that will vary depending on what those 16 are.

Section 8 of this article covers it for 16 color modes, but the code is the same if you do this in 256 modes.
https://www.onirom.fr/wiki/blog/30-04-2 ... ogramming/

After doing this, I have some code to export the entire palette to a .hex file (text file with each color hex on a new line) which I load into an application like ASEsprite to see both the palette, and conform images (if needed) to that palette.

See OS_ReadPalette ( https://www.riscosopen.org/wiki/documen ... eadPalette ) where r0 is the palette index (0-255) and r1 is 16. Result goes into r2 and can be converted to hex as follows. You basically loop 256 times and store each r2 as it is read in and convert to your color format of choice.

Code: Select all

    hexColor = ((r2 >> 8) & 0xFF) << 16 | ((r2 >> 16) & 0xFF) << 8 | ((r2 >> 24) & 0xFF);

[litwr]

Essentially, from what I understand, you can set 16 of the 256 colors, with the rest being populated by a 'house mix' of tints that will vary depending on what those 16 are.

Section 8 of this article covers it for 16 color modes, but the code is the same if you do this in 256 modes.
https://www.onirom.fr/wiki/blog/30-04-2 ... ogramming/

After doing this, I have some code to export the entire palette to a .hex file (text file with each color hex on a new line) which I load into an application like ASEsprite to see both the palette, and conform images (if needed) to that palette.

See OS_ReadPalette ( https://www.riscosopen.org/wiki/documen ... eadPalette ) where r0 is the palette index (0-255) and r1 is 16. Result goes into r2 and can be converted to hex as follows. You basically loop 256 times and store each r2 as it is read in and convert to your color format of choice.

Thank you very much. I need to set as many of the 256 colors as possible for a given image. But it seems there is a big gap in information about 256 color modes. On the page Theo Markettos explains the situation very well. He wrote that the Archimedes/VIDC1 has 4096 palette colors but only 16 palette registers. So you can't assign any palette color to any of 256 logical colors. The wiki page claims the same things. It seems that the math problem of finding an optimal Archimedes palette for a given image is still unsolved. Your materials contain nothing about programming colors in the 256 color modes. It seems that the materials are about newer video chips but not about the VIDC1. Or I missed something?
I also tried to use the intergif converter. It seems this program just ignores the problem.

[SarahWalker]

I've usually just thrown together a converter as-and-when required. I could probably put a generic tool together.

[IanJeffray]

Essentially, from what I understand, you can set 16 of the 256 colors, with the rest being populated by a 'house mix' of tints that will vary depending on what those 16 are.

Within reason, if you were stupidly careful, and didn't care about the colour numbers you ended up with, that's possible.

But VIDC1 has just 16 palette registers, which fulfil all needs for 1, 2 and 4 bit-colour modes, but in 8bit colour VIDC uses a horrible mix of the logical colour number (pixel value), with the actual palette entries.

Page 5-9 of the VLSI manual (VIDC datasheet) lays it out:

From this we see that, for example, Red bit 3 is set for all pixels with bit 3 set. As soon as you start using more than 16 colours, you're forced to have some these colour bits set.

IMVHO, it's a hiding to nothing ever trying to change the palette in 256 colour modes on VIDC1 - the influence you have only affects the lower-order bits of each colour channel (particularly green).

Bottom line - VIDC1 does not have a fully programmable-palette 8bit colour mode.

[litwr]

I just finished a utility that converts an arbitrary image to an image with a palette compatible with the VIDC1. It was just a few hours work, so I am sure the result can be better. It can only work with PPM/P6 format now (just use GIMP, ImageMagick convert, or many other tools for this conversion). It uses only the standard C++ library.
I attach the result of the conversion. The original image has 27993 colors, the conversion reduces this number to 51. Is it possible to get a larger number? Of course, the maximum, 256 colors is rather impossible to reach.
The algorithm of my tool is very simple. First, it converts 24-bit colors to 12-bit by dividing each component by 16. Then it separates 8-bit register parts and divides the 8-bit colors into 16 parts. Each part is assigned a color register. Then the 4-bit parts are added. That's it.

archiecconv.zip

(4.07 KiB) Downloaded 13 times

original.png (148.56 KiB) Viewed 1035 times

archie-vdc1.png (33.98 KiB) Viewed 1035 times

[IanJeffray]

I just finished a utility that converts an arbitrary image to an image with a palette compatible with the VIDC1.

You'll get a big improvement in general appearance for this kind of image by adding some dithering, either simple or FSI.

[BigEd]

(I just happened to read a note about using blue noise to help. See perhaps here and here. Amusing note in the first one about the original use of "dither")

[tom_seddon]

Previously in alternative 256 colour Archimedes palettes: viewtopic.php?t=5212

(that thread has long stuck in my mind, as it was my introduction to how cursed the Archimedes 256 colour VIDC1 mode actually is - and presumably why so many Archimedes games had that specific muted look to them)

--Tom

[IanJeffray]

Previously in alternative 256 colour Archimedes palettes: viewtopic.php?t=5212

(that thread has long stuck in my mind, as it was my introduction to how cursed the Archimedes 256 colour VIDC1 mode actually is - and presumably why so many Archimedes games had that specific muted look to them)

If you're just using Arculator and/or a modern monitor, not CRT - you may be interested in this : viewtopic.php?t=28065

[tom_seddon]

Previously in alternative 256 colour Archimedes palettes: viewtopic.php?t=5212

(that thread has long stuck in my mind, as it was my introduction to how cursed the Archimedes 256 colour VIDC1 mode actually is - and presumably why so many Archimedes games had that specific muted look to them)

If you're just using Arculator and/or a modern monitor, not CRT - you may be interested in this : viewtopic.php?t=28065

In my case, this was my recollection from playing games on a friend's Archimedes in the early 1990s as a boy, comparing them to what I'd seen on ST/Amiga/PC. The Krisalis ST/Amiga conversions always looked good - I guess they reused the 16-colour ST palette, that was probably also often the same on the Amiga for CPU performance reasons? - but a lot of Arc-specific stuff just looked kind of visually samey, more so than you'd expect given the relative colour count, even for the 3D stuff where the lack of professional artists was a bit less of an issue. 256-colour VGA PC stuff always looked way better.

--Tom

[litwr]

Thanks for the tips! Happy 2024 year!
I improved my algorithm a bit. The quality of the results became slightly better, but paradoxically it made less the number of colors. There are now only 48 on the converted image. For comparison, I also converted the original image into several other formats: Amiga HAM (320x256, 421 colors), Amiga Halfbrite (320x256, 60 colors) and VGA (320x240, 256 colors). I used ham_convert to get the Amiga images. It seems that the Amiga 6bpp EHB/Halfbrite mode is very similar to the Archimedes 8bpp mode because the latter usually gives us apporoximately 64 colors. The Amiga/EHB has 32 12-bit color registers and 1 extra bit per color, while the Archimedes/8bpp has 16 8-bit color registers and 4 extra bits.
It seems that I have reached my limits for the development of the conversion utility. So this is rather the final version. My goal was just to investigate mathematically how an arbitrary image can be converted to the VIDC1 format. I doubt that my converter works perfectly, but I hope it is close to the optimum.
I am curious is there a way to view a gif/png/...-file under RiscOS 3.1 in fullscreen mode while using the palette from the image? I used ChangeFSI v1.70, it obviously uses the system palette and dithering, the screenshot is attached.

archiecconv2.zip

(2.29 KiB) Downloaded 8 times

VIDC1 320x256 48 colors

archie-320x256-48.png (35.69 KiB) Viewed 866 times

Amiga HAM 320x256 421 colors

amiga-ham-320x256-421.png (53.2 KiB) Viewed 866 times

Amiga EHB 320x256 60 colors

amiga-ehb-320x256-60.png (51.66 KiB) Viewed 866 times

VGA 320x240 256 colors

vga-320x240-256.png (126.14 KiB) Viewed 866 times

RiscOS 3.1 ChangeFSI v1.70

risco31.png (27 KiB) Viewed 866 times

[IanJeffray]

I am curious is there a way to view a gif/png/...-file under RiscOS 3.1 in fullscreen mode while using the palette from the image?

No, for exactly the reasons discussed above - the VIDC1 palette in 256 colour modes is mostly FIXED, where gif/png contain fully programmable palettes.

[litwr]

No, for exactly the reasons discussed above - the VIDC1 palette in 256 colour modes is mostly FIXED, where gif/png contain fully programmable palettes.

It is an interesting challenge to write a fullscreen GIF viewer using an approximation for the palette. Unfortunately, I am too lazy to do this in the near future.
I did a conversion of the picture using InterGiv v5.19 to the standard Archimedes palette. Indeed the result (it only keeps 32 colors) is much worse than for a good approximation of the palette.

archie-stdp-igif-32.gif (19.32 KiB) Viewed 780 times

Surprisingly, ChangeFSI supports the PPM format. So a result of the palette converter may be used directly.

[litwr]

This thread prompted me to write an article about the photos on the first PC. It, of course, contains some information about the Acorn Archimedes.

[IanJeffray]

This thread prompted me to write an article about the photos on the first PC. It, of course, contains some information about the Acorn Archimedes.

"It is generally accepted that the first computer to actually display pictures on the screen was Jay Miner's second and last computer, the legendary Amiga"

Well let's stop reading right there. What nonsense.

[litwr]

"It is generally accepted that the first computer to actually display pictures on the screen was Jay Miner's second and last computer, the legendary Amiga"
Well let's stop reading right there. What nonsense.

Photographic quality color images are expected. In fact, I'm open to finding out what other personal computers might be the first to do this job.

[IanJeffray]

"It is generally accepted that the first computer to actually display pictures on the screen was Jay Miner's second and last computer, the legendary Amiga"
Well let's stop reading right there. What nonsense.

Photographic quality color images are expected. In fact, I'm open to finding out what other personal computers might be the first to do this job.

That's just your matter of opinion though. Why do you think the Amiga's limited video abilities are good enough? Why is 24bit colour good enough?... I do think it's quite poor form to say "It's generally accepted..."

[paulb]

That's just your matter of opinion though. Why do you think the Amiga's limited video abilities are good enough? Why is 24bit colour good enough?... I do think it's quite poor form to say "It's generally accepted..."

I am wary of getting into discussions about "firsts", although they usually involve people who are not merely content that their favourite manufacturer or brand (typically Apple) had considerable success. Such people seem to be compelled to invent a mythology to demonstrate that their favourite company or product had some kind of total superiority. I'm not saying that any such narrative is involved here, however.

As for systems that could do high-resolution colour graphics, say for broadcast purposes, perusing old computer publications reminded me of Cromemco, who were a prominent manufacturer who focused on broadcast television applications as the broader microcomputer business started to get fiercely competitive. Sure enough, you can find quite capable video cards for their systems:

"Cromemco Super Dazzler", Creative Computing, December 1979.
"High-Resolution Computer Graphics Interface", Cromemco, 1979.
"MODEL SDI Color Graphics Interface", Cromemco Board Level Product Catalog, 1983.

That card had a resolution of 756x484 with 16 colours from a palette of 4096, clearly limited by the 48K video RAM.

By 1985, they had quite a range of graphics cards for their 68000-based systems:

"S-Series 68000 Graphics", Cromemco, 1985.

Interestingly, in Guy Kewney's review of the Amiga in Personal Computer World, he refers to the Pluto board when discussing the Amiga's capabilities. IO Research's Pluto boards were quite well known to readers of such publications:

"Pluto Power", Personal Computer World, December 1982.
"Out-Of-This World Pluto", Computing Today, December 1984.

Readers here will be familiar with the Prisma boards for the Acorn machines, too. And we haven't even considered workstations, either.

[litwr]

@paulb Thank you very much. I didn't know about these Cromenco machines and the graphic cards for them. However my article is about more mainstream stuff.

[paulb]

@paulb Thank you very much. I didn't know about these Cromenco machines and the graphic cards for them. However my article is about more mainstream stuff.

I would always like to encourage people to explore the limits of their knowledge and experience, as I try to do myself, and in doing so discover things that question popular narratives. Given your interest in computing history, I am sure you share this perspective.

So, looking into graphical capabilities in the early to mid-1980s revealed mainstream products like the NEC 7220 which was used in the NEC PC-9800 series and could deliver a resolution of 640x400 in 16 colours from a 4096-colour palette. There was also the HD63484 with more sophisticated capabilities that is described in some detail in the following article:

"GPU History: Hitachi ARTC HD63484"

Then, there is the Professional Graphics Controller card for the IBM PC, available from 1984 and supporting 640x480 with 256 colours. Naturally, this starts to enter workstation territory, but as a card for the PC, which was admittedly fairly expensive, one cannot really claim that it isn't mainstream. Certainly, the Amiga offered capabilities in a convenient package and at a lower price point than had been typical at the time of its introduction, but we can recognise that achievement without diminishing the other achievements of the era.

I don't know what Ian was thinking of when he replied. Sometimes it would be good to get at the underlying insight: a few more words including a couple of keywords would reduce the potential for misunderstandings considerably.

[litwr]

I would always like to encourage people to explore the limits of their knowledge and experience, as I try to do myself, and in doing so discover things that question popular narratives. Given your interest in computing history, I am sure you share this perspective.

So, looking into graphical capabilities in the early to mid-1980s revealed mainstream products like the NEC 7220 which was used in the NEC PC-9800 series and could deliver a resolution of 640x400 in 16 colours from a 4096-colour palette. There was also the HD63484 with more sophisticated capabilities that is described in some detail in the following article:

"GPU History: Hitachi ARTC HD63484"

Then, there is the Professional Graphics Controller card for the IBM PC, available from 1984 and supporting 640x480 with 256 colours. Naturally, this starts to enter workstation territory, but as a card for the PC, which was admittedly fairly expensive, one cannot really claim that it isn't mainstream. Certainly, the Amiga offered capabilities in a convenient package and at a lower price point than had been typical at the time of its introduction, but we can recognise that achievement without diminishing the other achievements of the era.

I don't know what Ian was thinking of when he replied. Sometimes it would be good to get at the underlying insight: a few more words including a couple of keywords would reduce the potential for misunderstandings considerably.

Thank you again. This Hitachi HD63484 was something unbelievable in 1984! It seems many of us still don't have monitors capable to show 4096x4096 images! I didn't even imagine that such graphic cards could then be possible.
I knew about the PGC. Perhaps I should have mentioned it. And I am sure there were other super graphic cards for the PC in 1984-85... I agree with you that they were more in the workstation area. However, the DEC Professional was also more of a workstation. And I agree that it is impossible to write a perfect article...
It seems that the English of my text is a factor that makes it somewhat difficult to read. I can correct my blog entry, so I am open to any suggestions for improvements. I came to the conclusion about the mediocre quality of the text because there is actually a paragraph about the PC-98.
I got different responses to my article. Here for example it is quite positive.

[paulb]

It seems that the English of my text is a factor that makes it somewhat difficult to read. I can correct my blog entry, so I am open to any suggestions for improvements. I came to the conclusion about the mediocre quality of the text because there is actually a paragraph about the PC-98.
I got different responses to my article. Here for example it is quite positive.

I am not going to criticise your prose: it is clear and readable. And I don't want to be in a position where I am correcting anyone's writing, and certainly not in terms of style or formulation. Indeed, I think the treatment of the topic is very thorough, and I appreciate all the effort you have made to convert the image to the different representations.

However, I think the objection was in the nature of the claims that were being made, and in certain venues there would be people questioning things like this, too:

For an arbitrary picture to be perceived as a color photograph, you need presumably more than 100 colors in the palette on the displaying equipment.

It is an interesting claim, but now I want to know how someone arrived at a figure of 100!

By the way, I dug up a 1985 article about the Pluto II board that was already available when its predecessor was reviewed in 1984:

"Pluto II", Personal Computer World, November 1985.

The previously mentioned Computing Today article more helpfully notes that while the original Pluto offered 640x576 in 8 colours on a BBC Micro or 768x288 in 16 colours on an IBM PC or Sirius 1, the Pluto II offered 768x576 in 256 colours from a 24-bit colour palette, and three such boards could be combined to offer that resolution in full 24-bit colour.

Clearly, the Pluto boards will have appealed to the broadcast industry, and I have mentioned the Prisma boards already. In fact, those boards have been discussed on Stardot, and I note that the Prisma 3 used the HD63484:

"Millipede Prisma3", Chris's Acorns.
"BBC graphics fit for TV", Acorn User, September 1987.

Its predecessor, the Prisma 2, was noted in various news articles:

"Prisma 2's 4096-colour palette from Millipede", Acorn User, January 1986.
"Graphics package is based on Master", Acorn User, November 1986.

I imagine that they used the same graphics chip or a related one. Meanwhile, here's a moderately amusing story about the applications of these products in the broadcast realm:

"From Russia with Love", Acorn User, January 1992.

[RobC]

Here's a small news piece on the Pluto board being used with the Beeb from Acorn User November 1985.

I remember watching the Juice TV program on BBC Wales at the time and being amazed that the graphics were done on the Beeb

[litwr]

@paulb Thank you. The idea about the disadvantages of my text came up because you wrote about the PC-98 while the article mentioned this computer. So I had to assume that it was boring for you to read the whole text...
My article is about standard systems that had the ability to display colour photos and this feature was available out of the box. It's hard to argue that the Amiga was the first such computer system. I can't imagine that the IBM PC with the PGC was sold as a standard system.
The number 100 is based on my personal experience. First I came across colour photos on the Amiga, then it was the PC with the VGA card. The VGA is capable of displaying 256 colours at once. Much later I was very impressed by the results of image conversions for the Commodore +4, which is capable of displaying 121 colours. The converter actually produces interlaced images that can have up to 240 colours because of the PAL inversion effect. And I don't get pictures of this quality on systems that have less than 100 colours in their palettes. Of course, 100 is only an approximation. It would be great if anybody could provide examples that show that less than 100 colors is enough for all color photos.
The Pluto II and Prisma 2/3 for the BBC Micro were absolutely fantastic! I didn't know that such super graphics were available for 8-bit systems. I can only assume that something similar was available for the Apple II.
paulb and RobC, thanks for the links to such interesting information. I didn't know most of it. The information about the Acorn Archimedes being used on Russian TV was very surprising to me.

[paulb]

@paulb Thank you. The idea about the disadvantages of my text came up because you wrote about the PC-98 while the article mentioned this computer. So I had to assume that it was boring for you to read the whole text...

No, I just hadn't read the text by then. I had previously been evaluating the claim about the Amiga in its own right. Your article is certainly informative and thought-provoking, but to some of us that means that it just raises other questions!

The first question would have to be which computer, beyond the more affordable microcomputing systems, was actually the first to display colour photographic images at a reasonable level of quality? Clearly, systems like the Quantel Paintbox could do so, but they cost $250,000.

Another question would be this: were colour monitor resolutions really were limited to only 256 or so lines of non-interlaced vertical resolution in the early 1980s? Certainly, low-cost monitors were limited to television-level resolutions, which is how they came to be low-cost, but such a limitation seems very unlikely in general given the ascendancy of CAD and the appearance of suitable workstations.

I didn't look at workstations before, but we can combine these two questions and evaluate materials like this brochure from Apollo whose workstations could support colour displays of 1024x1024 with "true color" (ostensibly 24-bit) in "imaging mode". Very impressive for 1983, but also probably rather expensive, if not exactly Quantel pricing. (Maybe Acorn used one of these systems for their ARM design efforts.)

I think the Amiga was a packaging of capabilities not readily available at such price levels previously. But describing its significance is a bit like describing the Archimedes in terms of its significance to RISC technology. Ignoring obviously false claims about ARM, people have previously claimed it to be the first RISC system or the first RISC microcomputer, neither of which are true. In the end, the consensus settles on the Archimedes being described in awkward ways like "the first RISC-based home computer" which is perhaps somewhat less impressive. That should not take anything away from the achievement of delivering it, however.

[litwr]

Thanks for the new interesting information. But IMHO the proper title for the Acorn Archimedes is the first ARM based computer. The ARM technology is perhaps more important than the whole branch of the RISC technologies.
Thanks to our discussion I have just added information about other 1986-87 computers that were also able to show nice colours.

Japan produced other computers capable of beautifully displaying photographic images. The X68000 succeeded the previous Sharp X1 line of computers, which had been in production since 1982. The top-of-the-line X1 turbo Z, which was in production since 1986, has a 320x200 video mode with 4096 free colours, with a palette of the same 4096 colours. Surprisingly, the main processor of this top-end model was, as on the initial model, the Z80@4MHz, which is clearly too small for the video mode, where the screen image requires almost 96 KB of video memory.

Fujitsu, better known as a major player in the mainframe business, had been producing the FM-7 line of computers since 1982. The top-end versions of these computers were labelled FM-77 and were produced from 1984. Starting with the FM77AV, produced from 1985, they could use a video mode similar to the X1 turbo Z. And the latest models from the end of 1987 could even display 18 bpp (262144 free colours!) in a 320x200 image. The processor for all FM-77s was the Motorola 6809 at frequencies ranging from 1.6 (pre-1987) to 2 MHz (from 1987). Although the later FM-77 models were faster than the Sharp X1, they were not much faster and it was not enough to use 12 bpp graphics effectively, let alone 18 bpp. One wonders why they didn't use the pin-compatible Hitachi 6309 instead of the 6809. The 6309 can be up to 50% faster at the same frequency.

NEC, Japan's largest chip and personal computer manufacturer, had been producing various variants of the PC-8800 series of computers since 1981. Some models of this line, produced from 1987, supported graphics up to 640x408 with 256 free colours or 640x204 with 65536 free colours from a palette of 65536 colours. However, the multicolour modes of the PC-8800, FM-77 and X1 were almost never used. And these models themselves, capable of showing colour photos, were rare even in Japan and completely unknown in other countries.

[RobC]

Another question would be this: were colour monitor resolutions really were limited to only 256 or so lines of non-interlaced vertical resolution in the early 1980s? Certainly, low-cost monitors were limited to television-level resolutions, which is how they came to be low-cost, but such a limitation seems very unlikely in general given the ascendancy of CAD and the appearance of suitable workstations.

This was definitely not the case. I have a load of Digisolve video kit from the early eighties and it came with a high resolution monitor that does far more than 256 lines. I believe the kit was used with an Apple II - details of the early models are here but I have some later models that are far more capable - certainly way beyond the Amiga's capabilities and probably moving into broadcast video applications. The only reference I found to the later kit was related to astronomical observations but I can't find it now.

The guy I bought it off said it was involved in the production of the Channel 4 logo animation (so 1982?) but I've not found any other details.

[litwr]

This was definitely not the case. I have a load of Digisolve video kit from the early eighties and it came with a high resolution monitor that does far more than 256 lines. I believe the kit was used with an Apple II - details of the early models are here but I have some later models that are far more capable - certainly way beyond the Amiga's capabilities and probably moving into broadcast video applications. The only reference I found to the later kit was related to astronomical observations but I can't find it now.

The guy I bought it off said it was involved in the production of the Channel 4 logo animation (so 1982?) but I've not found any other details.

Interesting, but it seems that it was a standard monitor, because the typical European (PAL) monitor was able to display 520-576 lines, but by means of interlacing. Computer systems usually use half that number to avoid flickering.

[RobC]

Interesting, but it seems that it was a standard monitor, because the typical European (PAL) monitor was able to display 520-576 lines, but by means of interlacing. Computer systems usually use half that number to avoid flickering.

No - it's definitely much higher resolution and I'm well aware of how interlacing works.

Don't take the specs in the Digisolve manual I linked to as representative of the specs of the monitor as they're just for the earlier, much simpler board (as I stated in my post if you read it carefully). The monitor was used with a board based on the Hitachi HD63484 and so much higher resolution.