I find this work fascinating. I wonder if it'd be possible to create a test suite for the IBM PC to cycle through various colors and patterns, then distribute it to known IBM 5153 (monitor) owners and have them capture the output in best possible conditions and upload the results.
The results could include the video file and camera used. Then, iterate over the results programmatically and try to determine an average from them. Then the author's work could be normalized across many different revisions of the CRT from different conditions, instead of working from a single sample monitor. Then compare this with the canonical and revised colors.
I'm sure the differences would be barely noticeable, but just kind of wacky retro experiment I'd love to participate in.
There’s an important difference here. The 1K mode is using the CGA adapter’s composite NTSC output to a standard (American) television set. It’s taking advantage of color television’s design to effectively turn high-frequency pixel value changes into new colors.
The original article discusses the sixteen colors supported by the digital monitor that IBM sold for CGA cards. It’s very different from a television set because the monitor itself converts the CGA’s 4-bit digital signal to CRT control. This monitor is completely unable to display the 1K color hack because, even though it’s also CGA, it’s not an NTSC monitor.
The upside of such a digital monitor was much better picture quality. CGA’s 80*25 text mode was practically unreadable via composite on a TV set, you needed a proper monitor. (They were also expensive. An IBM color monitor cost several thousand dollars, probably like $10k in inflation-adjusted 2023 money.)
It also had other 4-color RGB modes other than magenta, cyan, white, and black, but that was the most commonly used one back in the day (even though it looked terrible IMO).
For example, here's a game (Kings Quest II) where they used the red, green, and brown palette with the background set to blue. A mode like that gave better options for color dithering since R G and B were all available, see here:
this is as close as possible to i will ever get to understanding how the older generation feels about cars from the 1950s.
it's not even so much about the tech it's about the memories i have from being young, and the process of discovering that some of the mysteries of the man made world do actually have answers.
I always liked looking at this old tech with a what-might-have-been viewpoint using the technical constraints of the time combined with the knowledge and hindsight available today. Without a time machine it's essentially a pointless endeavour, but nevertheless rather fascinating.
The Gigatron https://gigatron.io/ is a really good example of people making something pretty amazing using late '70s to early '80s tech.
I spent quite some time trying to figure out what palette I could make using just 4 digital lines, resisters, and diodes. This is what I ended up with, https://fingswotidun.com/images/4bitPal_arduino.jpg Maybe I'll make my own 16 colour computer one day with it.
When I see these TTL CPUs made from very few chips they leave me scratching my head. If it was this easy to build a fast 8 bit computer with a small number of the 7400-series logic chips that were available at the time, why was everyone so jazzed about the introduction of the 8080 and 6502?
The thing is it wasn't easy, it was merely possible.
It seams easier now because we have much better access to information and ideas that were much harder to come by. The Gigatron was a team effort, that team wouldn't have come together without the internet. I'm not sure when that particular ALU design came about but the idea of sticking 8 multiplexers and 2 adders together like that was a little bit of genius in itself.
The PC (1981) predates the first color 3270 terminals like the IBM 3179 (1984). A 3179 had 7 colors: pink, red, blue, yellow, green, white, and turquoise. I used one & remember it looking quite good but different from PCs of its day.
The results could include the video file and camera used. Then, iterate over the results programmatically and try to determine an average from them. Then the author's work could be normalized across many different revisions of the CRT from different conditions, instead of working from a single sample monitor. Then compare this with the canonical and revised colors.
I'm sure the differences would be barely noticeable, but just kind of wacky retro experiment I'd love to participate in.