Comments by "EebstertheGreat" (@EebstertheGreat) on "These Are Not Pixels: Revisited" video.

raelgoth, I know what you meant, but ".28 dpi" would be some seriously poor resolution . . . 0.28 dots per inch, or 1 dot per 3.6 inches.
9
@rfvtgbzhn No, I don't know how many times this has to be explained to you. No frequency can be absorbed by only one cone cell. All visible frequencies are absorbed by at least two, usually all three, in varying proportions. Sure, if you could stimulate each cone cell individually, using microscopic electrodes or something, that would allow you to reproduce any visible color (including many "impossible colors" that cannot be produced by any combination of light). But that can't happen on a TV screen. You can read about this in the link I provided, so you don't have to "think it's just not done because of technical limitations." You can actually find out, which is what I did before I explained it to you. The reason it is not done is because it is not physically possible. Bit depth does not play a role in the size of the gamut. It only determines spectral resolution. Increasing the bit depth only allows for finer discrimination between similar colors. It doesn't give access to colors outside of the old gamut.
4
@rfvtgbzhn "you would have to just find 3 curves that can when put together can produce the same effect as every natural color" For the third time, no such "curves" exist. I don't know why you keep insisting that they do. I am interested in how you think a spectral color could be reproduced by mixing two or three low-saturation "curves." You have to realize that every color is a point in a color space, not a curve. If it's a spectral color (i.e. monochromatic) it will lie on the spectrum, i.e. that upper rounded side of the horseshoe in the hue-sat plane of the chromaticity diagram. Otherwise, it will lie in the interior. You can't subtract light, so if one of my subpixels is giving me a mixture of red and green, I can't subtract that red out later to get a nice saturated green. No matter what I mix that yellow light with, I can't make green. The only way to get pure green is to shine a pure spectral green light. If you want a more mathematical approach, collapse LMS space into two dimensions by equating colors with different brightnesses but the same hues and luminosities. You can do this by first drawing surfaces of constant luminosity in the LMS space, then equating a point on a given surface with the entire ray passing through it from the origin. This is sort of how you get the color spaces you see (albeit not quite). Now when you mix two colors, you are adding two points in LMS space, getting a third point which is brighter than either of the two, but if you project it back down to the same surface (i.e. reduce its brightness to match the brightness of the original colors, without changing its hue or saturation at all), it will lie on a line segment connecting those two colors. This comes directly from the parallelogram rule for adding vectors. Thus, if you have three primary colors, mixing them can only produce colors inside the triangle those primaries define. This is why they were once specified using trilateral coordinates (aka barycentric coordinates). And if you have n primary colors, mixing them can only produce colors inside the n-gon they define. Since the actual observable color space does not have straight edges (except the line of purples, which have luminosity zero anyway), the whole thing cannot be contained in an n-gon all of whose vertices are themselves contained within the color space. It's like trying to fill a circle with an inscribed polygon--you'll always miss some of the circle. The portion of the space which is inside the n-gon is your gamut. Alternatively, you could use a triangle or other polygon bigger than the color space. This could easily contain the whole thing, but the vertices (or at least one vertex) must lie outside the color space itself and therefore be an imaginary color. For instance, an imaginary hypergreen color that only stimulates the M cone could allow us to reproduce all visible colors (and many invisible colors). If you want a discussion on primary colors, check out this page: handprint.com/HP/WCL/color6.html.
4
@rfvtgbzhn The bit depth is not the issue. You can't reproduce the "color" of a cone cell, because cone cells don't have "colors." You are making the mistake of believing that each cone cell is somehow responsible for detecting a single frequency of light. The problem is that three primary colors cannot reproduce the entire visible spectrum, something that has been known for over a century. Suppose I want to have a fully saturated yellow on screen. How can I do that? Ideally, I would have a yellow subpixel, and that would be that. But I only have red, green, and blue. So I have to light up the red and green subpixels. The red light is mostly received by long cone cells in my eye and also slightly by the medium cones. But the green light is received by all three cones! Whereas monochromatic yellow light will not really be absorbed by the pigments in the short cones at all. So these are different colors and can be distinguished by eye. And the same problem will occur whether I have a bit depth of 3 bpp or 300. The only way to improve the gamut of a screen is to increase the spectral purity of the primaries or to increase the number of primaries. Some LEDs already have pretty spectrally pure light, but a majority (at least for consumers) still have only three primary colors, which means the gamut is limited. Some screens have as many as six primaries, but perfect color reproduction would require infinitely many primaries. Look at the horseshoe shape of a chromaticity diagram and tell me how you are going to fill it all with a triangle. (e.g. en.wikipedia.org/wiki/Gamut)
3
A CRT television cannot display every visible color. Using pure spectral red, green, and blue as primary colors, it is possible to get a large gamut in a triangular space called "Wide-gamut RGB color space." However, that is still not perfect, as it can only display three spectral colors after all. But real CRT televisions are limited by the colors actually produced by the phosphors available, and these result in a much smaller gamut in the "Standard RGB color space". This color space is also used for most LCD and LED displays as well as for most digital image and video formats. In reality, the majority of colors cannot be displayed on a consumer screen or even recorded by a consumer video camera.
2
TheBytegeist is completely right. There are a lot of misconceptions going around here. It is also not true that all perceptible colors can be produced by mixing three pure wavelengths of light in differing proportions. Hypothetically, if it were possible to individually stimulate S, M, or L rod cells, then we could use any combination to produce any color, real or not. But because of the way they overlap, this isn't actually possible. While I guess you can purely stimulate the L cone with very red light, it would be very dim, and the same is not possible for M and S. The total range of colors that can be produced by mixing certain primary colors is called the gamut, and there will be some perceptible colors missing from any gamut. Just as a simple example, your screen cannot display violet, and the closest purple it has is still visibly different from true violet. (Magenta, by contrast, is simply defined as an equal mixture of red and blue, which technically makes it a purple. However, like all colors on a real consumer display, the magenta you see will be pretty far from a true, fully saturated purple.)
1
@katieell4084 LEDs generally only emit light at a single frequency corresponding to their bandgap voltage. Although ideas of variable frequency LEDs have been passed around, I don't know if any have been created. Regardless, reproducing most colors requires mixing more than one frequency. You can't have a magenta LED. But sure, if the frequency of each LED could be individually tuned, that would theoretically allow for a really wide gamut display. Recording and encoding the image would be a separate challenge.
1