On Monday, July 23, 2007, 6:08:44 PM, David wrote:
DCB> As I already learnt, the main reason is the conversion. And as we agree, DCB> one has to be able to enter different values. However:
DCB> because otherwise you can increase the amount of K DCB> to get the same result, by decreasing CMY by the the smallest of their DCB> values and increasing K by that amount.
They don't necessarily measure as the same colour when you do that.
DCB> I don't see why.
So I suggest that you print out some patches with varying levels of grey component replacement and measure them. or even just look at them, since the visual result is fairly clear.
DCB> I mean, if I have C50 M50 Y50 that should be the same DCB> as K50 in my understanding.
The problem is that your understanding is using a very simplified model that does not correspond to current practice or to the measured results. I believe I am correct that your model is
C' = (1 - R) M' = (1 - G) Y' = (1 - B)
and then K = min (C', M', Y') C = C' - K M = M' - K Y = Y' - K
which ignores completely different RGB color spaces, different ink sets, dot gain, illuminants, ink density, halftone screening, viewing conditions and so on. In other words, pretty much all of color management.
This is a very common simplification, and I understand you may have seen it listed in a book, but it is very far from the current state of the art and gives very poor results in practice.
DCB> I should be able to replace any Cx Mx Yx DCB> with Kx and hence be able to have one CMY value 0 any time. I do know, DCB> it's not handled that way, but I don't see much reason for it.
DCB> Maybe you can clarify this for me!
I am happy to recommend some books on color theory and color management if that would be helpful.