OK, sorry, I didn’t mean to offend you. I still don’t understand why you wrote: ‘If you slide one of the channels to the right (i.e. value = 1) the image goes to monochrome’, but I’ll just chalk that up to me not being a native speaker.
moving the pointer toward the right will increase the value of the channel, e.g to 0.5 or 0.75
It sets the contribution (weight) of the channel.
What the channel is, actually, rather vague (though you don’t see it on the UI), because the output R = G = B values are not necessarily mixed from the input R, G, B (despite the labels on the gray tab).
The input data, the channels to be mixed to produce grey, depend on the adaptation method chosen on the CAT page:
For none (bypass), the input of the grey mixer is in the working RGB space (e.g. Rec2020); for XYZ, the input will be XYZ value; for the others (CAT16 and the two Bradford variants), the data will be the corresponding LMS space (trying the imitate the cone cell responses). That is why the B&W: luminance based preset (just like the film emulation presets) uses XYZ for adaptation, and input R, G, B = 0, 1, 0 as mixer setup: what is labelled as R is the first channel, X; G means the second channel, Y (luminance), and B the third, Z. With such a setup, R = G = B will be simply set to the Y value.
I did wonder about it affecting the white balance unless you use a 2nd instance so I tested it and it does not appear to affect the WB when using the grey tab.’
Did you mean you were worried the outcome would not be shades of neutral tones (grey)? Nope, the code guarantees R = G = B (out[k] is R, out[k + 1] is G, out[k + 2] is B):