2009/7/23 Klaus Karcher <lists@xxxxxxxxxxxxxxxxx>: > I guess the reason for this bad evaluation result is a white point issue. > Probably Spectraview writes D50 as media white point to the profile, but > Argyll expects the actual white point of the display (otherwise the white > point tag is pretty useless and it is impossible to recover the actual > colors from an ICC v2 display profile!). You are right, I inspected the profiles and I saw Spectraview II writes D50 while Argyll writes the monitor temperature (6900K with hardware calibration only and 6600K with the software calibration too). PCS is XYZ (they are all shaper/matrix profiles), so its white point is 0.964, 1.000, 0.825. Curve is 256-point for Argyll and gamma 2.2 for Spectraview II. Despite that I think that Spectraview curve is sRGB instead of gamma 2.2. In such a case I should use -gs instead of -g2.2 in dispcal. This should explain why I obtained worst results adding the software calibration, am I right? I made a screenshot of the curves: http://darkbasic.homelinux.com/images/sv2/curves.jpg Those are the profiles if you want to do some tests on them: http://darkbasic.homelinux.com/images/sv2/profiles.zip > The reason for this unfortunate > situation is a arguable change (or "clarification") in the ICC profile spec > regarding the media white point of monitor profiles (see below). > > Specification ICC.1:2004-10 (Profile version 4.2.0.0): "As specified in > clause 9.2.25, for monitor profiles the mediaWhitePointTag shall be set to > the PCS white point" > > <http://www.color.org/advantagesv4.pdf>: "An important consequence of this > clarification is that Version 4 profiles for RGB displays and working spaces > should only contain D50 tristimulus values in the media white point tag > indicating the transformation to the PCS white point." But those are ICCv4 specs while Spectraview profile is only v2.1!