There’s a big problem with the above functions. Can you see it? What do you think a negative red light source means?
It’s nonsense! That means with this model, given pure RGB lights, there are certain spectral colors that are impossible to recreate.
Yes, but the negative value still has meaning. When the test subjects ran into such an “impossible to match” spectral color, they were allowed to turn another set of knobs that added some of the RGB primary lights to the “spectral” side of the display, until they could get a match. They found out that 500nm doesn’t match any RGB triple, but you can add 1.0 units of 500nm to 0.07 units of red on one side, and 0.07 units of green to 0.06 units of blue on the other side, and the two do look the same. Then they just did the algebra to turn 500nm + 0.07R = 0.07G + 0.06B into 500nm = -0.07R + 0.07G + 0.06B. You may not be able to physically make “negative red”, but you can still use it, as the article says, to quantify something meaningful. It’s not like the negative values were pulled out of a hat, or only exist because a matrix transform says they should.
There are spots along the spectral locus that require small amounts of negative green or negative blue, but they’re very difficult to see on the graph: they both have minima of about -0.0014. Negative red is in much greater demand because of the very substantial overlap between the L cone (which peaks in the red) and the M cone (which peaks in the green). L’s response to the CIE green primary is almost as strong as M’s is! So when we move into the blue-greens, trying to recreate that color using primaries fails because “pure green” is “too red” — that is, even the greenest hue excites our red detectors proportionally too much.
Thanks for that explanation but it has me wondering. How did they decide which wavelength corresponded to “pure” red green and blue? Put another way, why not use a less red green as pure green?
Largely on the basis of convenient standardization: the blue and green CIE primaries correspond to spectral lines of mercury (that is, you can get them by filtering the output of a mercury arc lamp), and the red one was just chosen at a point where the perceived hue changes hardly at all with wavelength, so that any slight error in reproducing that one wouldn’t really matter.
But it’s not that the CIE green primary is a “bad” one. The 546nm mercury line is very green indeed and pretty close to the center of the colors we’d call greens. It’s just that all pure greens also get a substantial response from those long-wavelength “red” cells. Colors start looking bluish rather than greenish long before you get any really substantial separation between L and M.
Yes, but the negative value still has meaning. When the test subjects ran into such an “impossible to match” spectral color, they were allowed to turn another set of knobs that added some of the RGB primary lights to the “spectral” side of the display, until they could get a match. They found out that 500nm doesn’t match any RGB triple, but you can add 1.0 units of 500nm to 0.07 units of red on one side, and 0.07 units of green to 0.06 units of blue on the other side, and the two do look the same. Then they just did the algebra to turn
500nm + 0.07R = 0.07G + 0.06Binto500nm = -0.07R + 0.07G + 0.06B. You may not be able to physically make “negative red”, but you can still use it, as the article says, to quantify something meaningful. It’s not like the negative values were pulled out of a hat, or only exist because a matrix transform says they should.There are spots along the spectral locus that require small amounts of negative green or negative blue, but they’re very difficult to see on the graph: they both have minima of about -0.0014. Negative red is in much greater demand because of the very substantial overlap between the L cone (which peaks in the red) and the M cone (which peaks in the green). L’s response to the CIE green primary is almost as strong as M’s is! So when we move into the blue-greens, trying to recreate that color using primaries fails because “pure green” is “too red” — that is, even the greenest hue excites our red detectors proportionally too much.
Thanks for that explanation but it has me wondering. How did they decide which wavelength corresponded to “pure” red green and blue? Put another way, why not use a less red green as pure green?
Largely on the basis of convenient standardization: the blue and green CIE primaries correspond to spectral lines of mercury (that is, you can get them by filtering the output of a mercury arc lamp), and the red one was just chosen at a point where the perceived hue changes hardly at all with wavelength, so that any slight error in reproducing that one wouldn’t really matter.
But it’s not that the CIE green primary is a “bad” one. The 546nm mercury line is very green indeed and pretty close to the center of the colors we’d call greens. It’s just that all pure greens also get a substantial response from those long-wavelength “red” cells. Colors start looking bluish rather than greenish long before you get any really substantial separation between L and M.
Thanks.