New display technology -- "RasterBleed"

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

JetSetIlly

Quote from: Andrew Davie on 06 Jul 2026, 08:40 AMThere was simply a missing 'convertColour(...)' call there, as it was a first-pass prototype.
The attached version has this added now, and the colours will be correct on PAL and NTSC. AND SECAM!
Just a note that the development build of Gopher2600 has more correct SECAM decoding. I think Stella has it in the pipeline too. I don't know how that will affect your SECAM results.

https://github.com/JetSetIlly/Gopher2600
@JetSetIlly@mastodon.gamedev.place
@jetsetilly.bsky.social

Andrew Davie

Quote from: JetSetIlly on 07 Jul 2026, 07:09 PM
Quote from: Andrew Davie on 06 Jul 2026, 08:40 AMThere was simply a missing 'convertColour(...)' call there, as it was a first-pass prototype.
The attached version has this added now, and the colours will be correct on PAL and NTSC. AND SECAM!
Just a note that the development build of Gopher2600 has more correct SECAM decoding. I think Stella has it in the pipeline too. I don't know how that will affect your SECAM results.


Latest master... secam


JetSetIlly

Quote from: Andrew Davie on 08 Jul 2026, 02:38 AMThe conversion tool is very complex.

Does the conversion tool use the M4's SIMD instructions? It sounds like the sort of program that would really benefit from them.
https://github.com/JetSetIlly/Gopher2600
@JetSetIlly@mastodon.gamedev.place
@jetsetilly.bsky.social

Andrew Davie

Quote from: JetSetIlly on 08 Jul 2026, 04:13 AM
Quote from: Andrew Davie on 08 Jul 2026, 02:38 AMThe conversion tool is very complex.

Does the conversion tool use the M4's SIMD instructions? It sounds like the sort of program that would really benefit from them.

Dunno. It's just a Python program to me.

Kenneth Cochran

Very impressive. I wonder how difficult it would be to adapt to video? The preprocessing makes it sound like it wouldn't be useable for real time rendering but it might make for an amazing cutscene technique.

Andrew Davie

Quote from: Kenneth Cochran on 09 Jul 2026, 04:15 AMVery impressive. I wonder how difficult it would be to adapt to video? The preprocessing makes it sound like it wouldn't be useable for real time rendering but it might make for an amazing cutscene technique.

The technique seems quite limited by the source image, and the 48 pixel width limit. Its utility above 48 pixels might be questionable; the more pixels horizontally, the harder it would be for a single-colour-per-line to cover things with available combinations of colours/pixels over 3 blended frames. It's at the limit right now.

Suffice to say, selected images can look amazingly good.  Some seem to look average at best. Still, it's going to be useful for some things.

Bomberman94

#21
Quote from: Andrew Davie on 06 Jul 2026, 05:56 PMIt's not suited to everything. But some things come out gorgeous.
Particularly, subtle shading and colours can be awesome.  Here's a great one...

I used the 120 KB binary with the skull with real Atari and TV:

Maybe the PAL conversion isn't perfect (or must be manually adjusted as you stated somewhere) but it shimmers a bit (no big thing) but it seems that is not so colorful (as your pictures suggest) and my image seems divided in different zones of colors (hard to describe). Maybe you can see by this picture (the vertical line besides the writing is from my window in the back shining on the TV, sorry):


Bomberman94

Here are screenshots of the other binary (including different pictures):

https://postimg.cc/gallery/0sfggcG

Andrew Davie

Quote from: Bomberman94 on 11 Jul 2026, 12:30 AMI used the 120 KB binary with the skull with real Atari and TV:

Maybe the PAL conversion isn't perfect (or must be manually adjusted as you stated somewhere) but it shimmers a bit (no big thing) but it seems that is not so colorful (as your pictures suggest) and my image seems divided in different zones of colors (hard to describe). Maybe you can see by this picture (the vertical line besides the writing is from my window in the back shining on the TV, sorry):



Thanks for the review.  In short, the algorithm/tool optimises for a given palette. In this case, NTSC palette -- although, as I mentioned earlier - I can optimise for PAL if I want. But the binary contains the NTSC-optimised version.  So, to get the PAL version I run through my standard "convert PAL colour to NTSC colour" table.  For most use-cases this is perfeclty fine.

But because this display method is extremely coupled in terms of the interaction between different colours and time-blending of those, it is functional when those colours used for the optimisation are the actual colours used to show the data. If you use different colours (i.e., converting at runtime from NTSC to PAL), then you introduce differences which effectively highlight these bands that you describe.

It means that two copies of the data must be maintained - one for PAL, and one for NTSC - OR, just treat PAL users as the poor cousin and let them suffer.  As to "not as colourful" -- this is VERY evident when you compare Gopher output which I use for screenshots (VERY vivid colours) or Stella (quite bland in comparison).  It's an emulator difference between the displayed colours, quite eveident in this particular application.  Me, I prefer the Gopher colours here, but I have no idea what the real world colours are. As @JetSetIlly has said, the issue is quite complex.