Hi :) Graeme Gill wrote:
There two "blue-turns-purple problem"s that I'm aware of: Perceptual hue non-linearity of L*a*b* space.
Yup, that's the one I was referring to - the one which means an L*ab plot of a constant hue with increasing saturation wouldn't necessarily be a straight line.
The best idea I have off the top of my head, is that when calibration curves are being applied, the in-profile ink limiting is unaware of
> exactly what ink is being put down
so you have to err on the cautious
> side when setting the total ink limit, thereby unreasonably limiting the > gamut.Ah, yes, good point - so the test chart samples, for instance, [0, 50, 50, 0] when searching for a saturated red, and the resulting ink coverage could well be as low as [0, 20, 20, 0] once it's gone through the linearization curve and base gamma adjustment. Yup, that explains my problem, I think.
This is also an issue when the profile channels are actually composites of combinations of light and dark.
Yup, I think this issue is why I've seen unexpected paper-soaking in a couple of tests. I also have no idea quite how linear the relationship is between input-value and ink coverage in pl/mm². In Gutenprint's full raw mode I would imagine pretty good, but I imagine any miscalibration of dropsizes could have a big impact there.
Since the very beginning I've had some hooks in place inside Argyll to allow this to be taken into account, but without having settled on a calibration table format, the hooks aren't currently able to be used.
Interesting! Thanks for the input :) All the best, -- Alastair M. Robinson