Donkey Kong Country - Sega Genesis - TECH DEMO

[buckyohare]

Really very impressive tech demo made by Tiagosc and Pyron, love the fake transparent background solution.

https://www.youtube.com/watch?v=NGqG2QcrXq0

They really change my mind.

[turboxray]

Really very impressive tech demo made by Tiagosc and Pyron, love the fake transparent background solution.

https://www.youtube.com/watch?v=NGqG2QcrXq0

They really change my mind.

What was your mind before? That you couldn't do the venetian blinds trick on the Genesis?

[buckyohare]

Really very impressive tech demo made by Tiagosc and Pyron, love the fake transparent background solution.

https://www.youtube.com/watch?v=NGqG2QcrXq0

They really change my mind.

What was your mind before? That you couldn't do the venetian blinds trick on the Genesis?

I've seen a similar approach in Batman & Robin, but not exactly like this. And mainly the colors, i don't perceive any dither, and the background shading retains all original details.

[TmEE]

I don't really like this method as the interlaced BG layer is very flickery and anything in motion is full of combing artifacts... but it is easy to do although it costs costs between 20...30% of CPU time of a frame, depending on how well optimized the IRQ handler is. IRQ every line that does nothing at all but straight up exits uses up about 13% (64 cycles minimum, there's some extra cycle jitter due to alignment with E clock requirement due to Autovector IRQ method of 68K that MD uses) of the CPU time of that line (488.57 cycles).

[Dwedit]

Is that using the actual interlaced mode on the Genesis?

[lidnariq]

No, it's interleaving two nametables in software on a scanline-by-scanline basis, alternating parity per redraw, which sorta looks like 50-50% blending, but to lots of people looks like a horrible flickery mess. (And it gets worse the greater the brightness difference is).

[Dwedit]

I just wonder if there's a way to use the shadow/highlight mode to fake color math?

[aa-dav]

Impressive. Full-scale third layer is something new for me in SMD world.
It's interesting how it looks on the real CRT.
CRT smooths chessboard-transparency-simulaton very much.
It looks like video is from emulator with CRT emulation shaders, but real hardware can be something different.

[93143]

I'm not sold on it being undetectable to the human eye as TiagoSC claims. I tried something very similar on SNES, trying to get a fairly low-contrast Mode 7 layer to scroll in two directions at once, and it looked unacceptably flickery on a real CRT. Maybe it depends on the TV...

Still, it's certainly impressive, and the music is great - not perfect; there are a couple of wrong notes (unless that was done for copyright reasons), and some of the effects are a little off, but it pretty much nails the general vibe, without the looping artifacts of the original (and do I detect an attempt to show this off with an excessively long pitch bend?). The piano isn't as convincing, but that was never going to happen on an FM synth...

CRT smooths chessboard-transparency-simulaton very much.

That's not what this is. This is just alternating tilemap and priority line by line, which isn't nearly as smooth.

[aa-dav]

That's not what this is. This is just alternating tilemap and priority line by line, which isn't nearly as smooth.

I got it when I saw it. Anyway I want to see it in CRT, not in emulator.
Fog is clever "choice" of layers because fog IS something which interferes with sight, so it must be 'interfering'.
Nothing solid could be simulated in such way.

[psycopathicteen]

Really very impressive tech demo made by Tiagosc and Pyron, love the fake transparent background solution.

https://www.youtube.com/watch?v=NGqG2QcrXq0

They really change my mind.

What was your mind before? That you couldn't do the venetian blinds trick on the Genesis?

I've seen a similar approach in Batman & Robin, but not exactly like this. And mainly the colors, i don't perceive any dither, and the background shading retains all original details.

I wish I can see the colors more clearly in this thing. The fog is hurting my eyes.

EDIT: When I switch to 1080p60fps mode, it's not as bad, except when moving. I think they should add dithering so the foreground doesn't disappear entirely on scanlines that are covered by the cloud layer.

[turboxray]

Impressive. Full-scale third layer is something new for me in SMD world.
It's interesting how it looks on the real CRT.
CRT smooths chessboard-transparency-simulaton very much.
It looks like video is from emulator with CRT emulation shaders, but real hardware can be something different.

This doesn't use the chess/checker board method. EDIT: I just saw that the response to this. NVM.




Frame 1:
all odd lines show a tilemap Plane A that has priority over sprites and Plane B
all even lines: show a different tilemap on Plane A that as lower priority than sprite and Plane B
Next frame:
invert the even and odd lines.

It's a flicker effect, but since it's not flickering an entire screen, just interleaved lines - it's not as noticeable. And that's a true statement; it's not as noticeable as flickering two whole screens. The side effect is that the top layer will always be translucent to the eye on top of the sprites and plane B. You don't have choice. The lowest plane A will be translucent (averaged) against the backdrop color - darkening the colors. You also don't have a choice.

The effect is not new. It's been used in demos for A8, C64, etc and has been around since the 80s. Even some existing Genesis games have used this effect (not quite to this scale tho).

Movement can reveal comb like edges. High contrast difference are more flickery than low contrast differences. If you blink or shift your eyes, you'll catch partial edges of it (like a DLP projector). All the usual stuff for this kind of effect. This kind of effect typically looks worse on an emulator because the illusion is kinda broken when a frame gets dropped (which won't happen on a CRT). And ironically, given the use of this effect in primarily PAL demos, this effect looks better on NTSC than PAL because of the higher refresh rate. Sorry PAL owners.

The sprites and plane B themselves don't "flicker". So gaps in the top layer show them as normal. I think the effect works pretty decent for a fog effect on a 60hz display (the CRT itself is not helping you here since there is no blending on the CRT side for this - if anything an LCD with a slow response rate will look better for this kind of effect).

The only real trick here, is that since you can't change the all priorities of all the tiles in a plane without having to manually change them - that having an alt map on a different scroll position (that is normally hidden) with all the tile entries set to high priority is what gets the job done. In the demo, they don't scroll the farthest layer and the fog layer vertically - if they did, they'd have to make a 'sew' line so that each section of the tilemap don't wrap into each other. Not a big deal since they're using h-interrupts already. There are also complications with scrolling this effect vertically - and necessary workarounds. So, the trick for the genesis is getting around the priority thing and it's kinda rare to use h-interrupts on the system.

Shadow can't really be used in place of this - because of how it works (I'm talking about Shadow and priority settings and how it's effectively applied on tile boundaries**, not the shadowing sprites can do). But that's not to say you couldn't mix shadow (based on priority) with this to affect the farthest BG layer. Highlight won't help you here because that comes from sprites (and sprite can't S/H other sprites). Tho I can think of a few effects mixing highlight sprites with this - but you would need to manually blank out even or odd lines in the highlight sprite.

And no, this is not an 480i interlaced effect. You could do that, but ideally you'd want your game to run at 30hz to avoid "combing" on everything.. and other side effects related to 480i. I dunno. On an old consumer TV where the scanlines themselves were thicker than the gaps - it might look better. Those expensive pro and broadcast monitors.. not so much.

** tile boundaries, but scroll offsets still apply.

[tepples]

Would it help to use a 1024x256-pixel (128x32-tile) tilemap, with the front and rear priority tilemaps side by side and bit 9 of each hscroll table entry picking one or the other for each line?

[turboxray]

Would it help to use a 1024x256-pixel (128x32-tile) tilemap, with the front and rear priority tilemaps side by side and bit 9 of each hscroll table entry picking one or the other for each line?

That's a good idea! No more H-INTs.

[TmEE]

The two tilemaps will eat a lot of VRAM space at such a size, but yeah, no more line IRQs. Unfortunately there is no individual tilemap size setting for the BG layers, both share it.