I was trying to remove green from yellow and create red color insted of yellow.
To do this in outputGreen moved inputRed slider to the left.
As You see yellows goes to be orange.
InputR is -0,516
AFAIK the channel mixer isnāt a āwork on numbersā thing like in other tools , but has code to change colours to how our eyes would react to given changes.
or am I confusing the new channel mixer (color calibration) with the old ?
Anyway , the color-calibration channel mixing is not meant to behave exactly like the simple channel mixers in other tools.
There are multiple threads where people try to understand how the algorithm works by playing with created images like these , and it never works :).
Why things turn blackā¦ Could be the algorithm flips out when giving synthetic colours like in this image. Could be that you turn the colours into something that is out of gamut and they arenāt clipped.
Is your image 16bit tiff?
Anyway, donāt expect the channel mixer of Affinity Photo and the one in the color calibration module of DT to behave the same. Controls seem similar, algorithm or colourspace behind it is quite different.
I raised an issue about this on github a while back. The response was it wonāt fix. As jorismak says darktables color calibration has different math to other channel mixers, I think because it is designed to work in scene referred, unbounded, workflow, whereas Chanel mixers arenāt, they are bounded 0-255. Iām not in front of cpu right now, but I think the way to test this is to increase exposure so your highlights are above clipping, then turn on filmic to bring them back in range. If you place the old channel mixer between exposure and filmic those highlights will clip and not be brought back. But if you place color calibration between exposure and filmic, the highlights wonāt clip, and can still be brought back.
The price to pay for this is the black zones you sometimes see as in your example. This happens when you start with highly saturated colours, then push the sliders s lot. The colours go out of gamut, and eventually just black out. Itās not a problem in most real world photography as you donāt typically want to push highly saturated colours that far, but can be a problem working on highly saturated synthetic images, or if wanting to do extreme effects. Youāre most likely to see it first in magenta, because if you look at the horseshoe gamut diagram, you will notice that colour has the shortest distance to white, thus the least room to move before going out of gamut. (Memory is a little hazy, it might go to black when the y axis is pushed below 0). What other channel mixers do is go out of gamut and stay there without blacking out.
What happens if you set all profiles to rec2020ā¦ I will if I have time but DT has a weird path where the display profile is wedged into the pipeline so color space changes up then down and back up can introduce issuesā¦ setting them all to the same will avoid this. The image wont look right on screen but it will confirm if there is this sort of issueā¦ It is why to fully assess gamut you need to do thisā¦ some display profiles have issues so the only way to assess true gamut clipping is to keep a pass through situation for the dataā¦ Or it is likely the impact of negative rgb valuesā¦
I know in the input profile module is something about clipping to gamut , and in the first tab of the color calibration module around the gamut compression slider is a box to clip out of gamut stuff.
The black sounds like a negative number or NaN coming from a calculation.
I tested that setting and I think its the default so unless it got disabled it would be in effectā¦ All it seemed to do visually at the extreme was show the channel value as 0 in the color picker instead of negative. It still jumps to black if you push far enough. I never saw negative values with my display profile. Maybe its doing some clipping but if I set the display to rec2020 I could see the negative value in the green channel begin to appear ā¦when it gets to around almost -100 and then the image jumps to blackā¦ at least for meā¦ You can push the slider even a bit harder if you use the first option in the CC moduleā¦ as D50 pipelineā¦ this doesnāt do the CAT but does do the channel mixing in the selected CAT colorspace where I think there is more control of the gamut maybe ā¦ but eventually you can still get the blackā¦ THe manual also says these negative values can be poorly handled by OpenCL but for me turning it off or on didntā change it ā¦ So it seems if you push hard enough and drive channels negative you will get that black ā¦
