I compared the Grapher displays of the original Adobe RGB(1998) and a modified version I got within Photoshop with the white point at D50. I expected the same displays (both within relative colorimetry rendering since I understood that the Grapher then performs a chromatic adaptation towards D50 while this chromatic adaptation is already done by Photoshop in the modified profile.
1 - In fact, the 3D-LAB displays are significantly distinct. Could the chromatic adaptation algorithms be different in Photoshop and CT-Pro ? Would not this be annoying ?
2 - The Yxy or Luv 2D-displays are different, too. Why is not the modified profile a perfect triangle ? The green corner is curiously distorted.
3 - The 2D-profiles become the same if the absolute colorimetry rendering is selected for the original profile, which seems puzzling to me (the 3D-displays are then quite different)
This is not something I have directly compared, I’ll have to look into it… In ColorThink, I used the relative colorimetric intent to calculate the gamut; altering of the colorants is automatic. Photoshop would (should) use chromatic adaptation to alter the chromaticity values of the RGB colorants in the Adobe RGB profile when you alter the white point of the profile. If it doesn’t then that would explain the difference…
One of the confusing points of the ICC version 2 spec is how the white point should be recorded in a profile and how the colorants should be altered and the method of altering (the adaptation matrix) should be recorded. This is much clearer in the v4.x spec and also means that we will never see a working space with any white point other then D50.
The Grapher should produce a perfect triangle and doesn’t in some cases. This is a problem that crept into the software when I altered the overall-gamut boundary code… so it needs to be fixed.
… interesting… I’ll have to look into this one as well. I’m getting the impression that an Adobe RGB profile with a D50 white point is not the same as Adobe RGB (D65) color rendered into Lab using the rel col intent. It’s not appropriate to say one is right or wrong though, just that they’re not apples to apples.