Klaus Karcher wrote:
Were you using device links with icclink -G ? (I would guess not).
You caught me red-handed ;-) ATM my aim is to make profiles to be used straightforward by "ordinary users" (usually in Photoshop).
Understandable given the way things are at the moment. It would be nice if device links were more widely usable, so that those after higher quality could go to the trouble of using them.
This doesn't explain anything, but it narrows the phenomenon
down to being to do particularly with the nature and resolution
of the B2A table in the destination profile. Some trials
indicates that the resolution of the "bumps" is related
directly to the resolution of the B2A table, and nothing else.
I haven't got to looking at this in any more detail yet. My only other thought was that perhaps in practice, the device curves are not ideal (even though they are the inverse of the "ideal" A2B curves), and that some experiments with different curve shapes might give a hint as to whether this is the case. (The curve shapes affects both the placement of the grid points in incoming space, and the nature of the interpolation within each grid cell.) If this turns out to be part of the answer, then the next question would be "how does one generate "optimal" B2A curves ?".
I've added two new examples to http://digitalproof.info/argyll/k-test/ -- one with "icclink -g" and one with "icclink -G"
It's hard to interpret the "-g" one. It could be simply that because it uses a separate gamut mapping transform combined with the colorimetric B2A table, that the "steps" in the B2A table are being obscured, rather than there being anything fundamentally different going on (ie. that the "bumps" are purely a properly of the B2A tables.)