Composite CGA
01/04/2025

01/04/2025

A little update to the composite CGA table. I got another screenshot of the CGA (later) card, and the quality is better. I can see that it looks closer to the original (later) palette, so it now replaces the original.

Additionally, a little more about the ACV-1030 (Yamaha v6355) video card is now known. There seems to be an issue with the composite circuit, where the output is at least twice as bright as it should be. This is why the composite palette is mostly a brilliant white.

The palette on the Yamaha v6355 is 9-bit (3 bits for each of R, G, and B). Once the colour of palette entry 15 is reduced by half (RGB 333 instead of 777), the true composite palette is revealed - and it's the same as the IBM PCjr!

In debug.com, this could be done as follows:

-o 3dd 5e
-o 3de 03
-o 3de 33

 

Here is the raw image. These are (mostly) the RGB values extracted straight from an image of the screen, but I have slightly modified some entries where the colours were too close to discern.

Now with added IBM PCjr

I have room for 6 more palettes in this 256 colour PNG

 

I've now included the black level, where available. Hopefully, this will give a better idea of the baseline values.

The IBM PCjr palette is taken from the ACV-1030 (colour 15 set to RGB 333). Yes, it is showing some incredibly high grey levels for what should be black. The incredibly desaturated palette (if I can get my colour terminology correct this time) is with its default palette after a reset (colour 15 set to RGB 777).

It's said that the ACV-1030 needs a working -5V power line in order to display composite images. But it was tested on two different PCs (one was an IBM 5150), so there probably wasn't any issue with the voltage levels.

 

The IBM CGA and Tandy are now the only palettes taken from the original image, but they cannot be compared directly to my RGB values. So, I've made a manually retouched version that tries to remedy that. Again, it's an approximation, as it was all done by eye.

The colours are more saturated, and the brightness has been toned down on the blacks, whites, and greys. The palette indexes within the file are now neat and tidy.

A shining example

Rebrightened, resaturated, and (otherwise) rejiggled RGB values

 

Here is an updated text file containing the RGB values of the added cards.

 

17/03/2025

Davide has tested six CGA cards with a piece of software that draws composite colour bars, in 640x200 mode (default palette).

What I've done, is take an iconic image (the technological equivalent of American Gothic) from this page and improved it (or perhaps I've made it worse). The RGBI, IBM CGA, CGA (later), PCjr, and Tandy palettes remain from the original image, while the rest is new.

The new values should be taken as approximations. They were displayed on a PAL LCD TV, that was capable of handling NTSC decoding. They were then filmed on a mobile phone, so the true values were subject to the variations of the colour output of the TV, the lighting in the room, the angle of filming, reflections on the screen, the camera and software of the mobile phone, and any compression / decompression used until it reached me.

The videos then had a screengrab taken, at an arbitrary frame, and I used the dropper tool of an art package (IrfanView) to extract the RGB values, one-by-one. I then constructed this image, using another old art package (ProMotion).

This explains why they're so saturated. Use the grey values (colours 05 and 10) as a baseline.

If you've ever wondered why old DOS games did not make a lot of effort to accommodate the differing palettes, this table shows why.

 

Too many palettes spoil the broth

The wide, wide world of Composite CGA palettes. So wide that you need quite a high resolution monitor to fit the whole image onto the screen.

 

It's commonly considered that there are only the original CGA (plus later variant), PCjr, and Tandy palettes, but that is not the case. There seem to be as many different palettes as there are video cards and, not only that, but you can't guarantee that they will all be unique colours, or provide the same range of colours. That is also assuming that you don't have one of the cards that can only output monochrome over the composite connector.

 

Notes:

The IBM CGA (other later) palette is from a later IBM CGA card, but it looks like the original palette, except for colour 11, which is closer to the later palette. Neither of us has an earlier IBM CGA card to compare with, so this is uncertain.

The exact model name of some of the cards is not known, so "CGA Clone (long board)" is just as it is described, while the Paradise PVC 2 might be a Paradise Hi-Res Graphics card, or something similar.

The most unusual is the ACV-1030, powered by the Yamaha v6355 chip. Yes, that is the actual palette that it outputs. The v6355 is quite different from most CGA chips, it may have some other settings I do not know of, in order to fix that palette. But it does have a colourful alternative to composite CGA, that I will make a post about (if I can ever get it to work!)

Here is a text file containing the RGB values of the added cards.

 

(C) Jane McKay, 2025