QTR-Quadtone RIP

--- In QuadtoneRIP@yahoogroups.com, Ernst Dinkla <E.Dinkla@...> wrote:
> 
> Could someone explain to me then how extra curves and extra 
> blending isn't touching color mixing ?  I can understand it 
> for split toning but there one hue is replaced by another one 
> along the tonal scale.
> 
> Joost indicates it more or less in his first message:
> "Since I can only interpolate between two curves I can only 
> create tones at the boundary of the color space".
> 
> Either the new blendings do not cooperate with one another 
> (adding just another straight line) or they together describe 
> a slice of the color space in a crude way (an area) or you go 
> the whole 3D way and use a color engine.
> 

Ernst,

I'm not a color space expert. I'm physicist by training. I have given 
color spaces some thought in the past, but my statements below are 
not definite ones. Just a way to try to proceed the discussion. I 
discussed this with a colleague of mine (same background). I address 
single tone curves only. 

I do not understand your problem exactly. The blending IS of course 
touching color mixing, since that's to me the whole point of blending 
curves. But perhaps I misunderstand what you mean. I regard the three 
curves (warm, cool, lab a/selenium) as the three primaries of the 
color space, which is a subspace (a triangle near the origin) of the 
LAB color space. The exact gamut / subcolor space is determined by 
the combination of inkset and paper. Currently, QTR allows to 
mix/blend two ink curves, which is the equivalent of interpolating 
between two primaries in the LAB space. The color "space" one can get 
with two curves/primaries only is actually a line (is that what you 
call a slice?).  Mixing/blending 3 curves allows actually 
interpolating within the 2D area of the color space. Any color point 
within that triangular space is achievable with three curves.

There are many subtleties to the above. I am aware of at least the 
following issues that might arise, at least in theory: 1) It actually 
might be the case that the subcolor space is not exactly a triangle. 
The actual physical mixing is done with inks on paper. This a 
subtractive process that can give a color point that is not on a 
straight line in the LAB space. It resembles more the behaviour of a 
CMYK space. The boundaries in teh LAB space might be curved. 2) Even 
if the interpolation in the LAB space gives a straight line, the 
interpolation is not necessarily linear. For instance, a mix 50/50 
mix of cool and warm inks, will give a different result than first 
printing a 50% cool layer and overprint that with a 50% warm layer . 
3) In theory, it even might be that mixing two colors (with different 
a and b coordinates) with same density (L value) might give a mix 
with different L value, thus affecting the linearity of the curve. 4) 
In reality, the inks are no pure primaries, but having each a color 
spectrum of themselves, allowing for complicated spectral changes.

I'm not sure how important these issues are in reality. I expect 
issue 3) not happen, since that would already undermine the current 2 
curve blending in QTR. I might test this with two curves…  But I 
think the above issues point out, that one can NOT expect that a 
simple interpolation of 3 curves is sufficient to predict the color 
point of the result. In other words, there's no guarantee that let's 
ay a 33/33/33 mix of three curves with lab a and b values (0,9), (0,-
9) and (9,0) gives a color point of (3,0). So the simple mixing model 
has clearly it's limitations. If one wants to achieve a certain tone, 
one has to find the mix empirically. In order to predict a result or 
to use softproofing one needs a full-fledged color model/engine. 

But, even with the limitations pointed out above, I'd really 
appreciate the simple mixing model with more than two curves.

Am I addressing your questions or am I still way out?

Joost