I keep hearing that the NES can't produce a proper yellow

[Alex_the_Brave90]

Then what is this? this looks like a nice bright yellow for the BG.

[Drag]

Yellow is slightly greenish or slightly orangeish depending on which revision PPU you have, and it's almost never that vibrant on a real console, unless you're using an RGB PPU or some kind of custom palette.

[Alex_the_Brave90]

Yellow is slightly greenish or slightly orangeish depending on which revision PPU you have, and it's almost never that vibrant on a real console, unless you're using an RGB PPU or some kind of custom palette.

oh ok, that makes sense. cuz i recently bought an ever drive, and that same stage was not nearly as vibrant as this screenshot. i thought it was the composite cables being unable to display it properly.

[Dwedit]

Here's the NES palette

nintendulator_pal_new2b.png (2.65 KiB) Viewed 3420 times

Here's the YIQ Colorspace for Luma = 0.5

If you take the YIQ color space and divide it into 12 equal slices (look at the intersections at the outside edge), it looks a whole lot like the NES palette. For the particular Luma value used in this illustration, there is no good yellow either.

[Alex_the_Brave90]

That's wild. It's interesting to know how Nintendo arrived at that color palette.

[Yave Yu]

NES's yellow is a bit weird… Two dark yellow are greenish, two light yellow are (a little) orangish.

[Anna_TeamRocket]

NES's yellow is a bit weird… Two dark yellow are greenish, two light yellow are (a little) orangish.

To be fair, all of the NES colors are a bit impure. There's no proper red, and no real blue color. I think it's even better this way. Bright, pure colors also do not appear in real life.

[Ben Boldt]

NES's yellow is a bit weird… Two dark yellow are greenish, two light yellow are (a little) orangish.

To be fair, all of the NES colors are a bit impure. There's no proper red, and no real blue color. I think it's even better this way. Bright, pure colors also do not appear in real life.

Though I find the NES's graphics unique and charming and I don't disagree with you, I have to point out that the NES excels at being abstract; basically an escape from reality. Nothing remotely related to real life or natural-looking color can be shown by the NES. I think most of us like it that way.

[Pokun]

The PPU has a lot of blue, a lot of green and enough violet but is indeed lacking in red, orange and especially yellow. Yellow is a pretty narrow wavelength group in the color-wheel so I guess it's hard to hit it without other colors suffering.

I always thought NES games lacked a bit in color but there are some games that manages to use them to great effects and look very colorful. Hardware limitations forces creativity, and the PPU's colors are characteristic enough to allow you to notice that a game is for the NES almost right away when looking at pictures of it.

[Oziphantom]

Yellow is hard in general for TVs that is why Quattron was invented, you just can't do Gold with RGB for example. See also "Brown-fix" for RGBI.

Since most monitors and TVs are sRGB you get this https://en.wikipedia.org/wiki/RGB_color ... arison.svg yellow on the outside.

[Ben Boldt]

I think we need a new technology that displays HSB instead of RGB. H should extend all the way from IR to UV, and it should be impossible to display magenta without dithering. That sure seems like a more correct way to do it.

[93143]

it should be impossible to display magenta without dithering.

That only applies to monochromatic light sources. If more than one frequency at a time is allowed, magenta works fine.

And of course reflectors (i.e. everything you can see that isn't exclusively a light source) can have arbitrary absorption curves, so you can easily do magenta by having an object that reflects both red and blue.

This is also why LED lights have worse CRI values than you'd expect based on colour temperature - the spectrum is lumpy and peaky, which means a light that looks white to a human eye (proper RGB balance) can interact in funny ways with reflecting surfaces, and things can look lurid and wrong even under a wideband white, never mind a narrow-band RGB white. As an illustrative example, consider a hypothetical paint that reflects in an extremely narrow band (±1 nm) at 534 nm, illuminated by a frequency-doubled Nd:YAG laser at 532 nm. This paint will look black under these conditions, even though in sunlight it appears to be basically the same colour as the laser.

If you're looking straight at the light source, as with a video game, this doesn't matter, because you only have three kinds of cone cells, and if you're stimulating them in the proper proportion you're showing the right colour. Of course, RGB phosphors may not be perfect at hitting the exact peaks and may be too wideband to minimize crosstalk, so there may be regions of the visible range that cannot be shown on an RGB display, but switching to a different display colourspace isn't going to help...

How would you even do that? Free-electron laser display with a 5772 K incandescent backlight?

[Ben Boldt]

You provided a lot of interesting points. Color is always such a cool subject.

My comment was based on the idea that using RGB is kind of built on an illusion, the fact that we sense those 3 colors. But it only works because that is the specific way we see. Some animals, and perhaps aliens from outer space, sense more than 3 separate colors. It would be interesting to be able to produce something more “real” than the RGB illusion that works just for us. I am not sure there is any way that makes it practical or useful though.

[Pokun]

Yeah like the mantis-shrimp which I think has 16 photoreceptors for hue (like our 3 cones) and sees far beyond the human visible light wavelength span. Dogs and cats on the other hand has 2 cones, I think they are missing out on red and see mostly only, what we call, cool colors (but they got more or better rods and can see in the dark better).

Also the cones and rods are not identical for all individuals and probably varies a lot, so there is no one way that matches perfectly in how we see.


I guess mantis-shrimps' TV-sets must have a finer shadowmask/aperture grill to fit 16 primary color channels.

[93143]

256 hues from deep red to deep violet. 256 levels per hue. 64 KB per pixel.

Now do 10-bit HDR with 10-bit UWCG from near infrared to near ultraviolet. 1 MB per pixel...

...but this way the aliens see pretty much the same TV picture we do. Worth it?

Maybe we've just explained this...