[argyllcms] Re: Can anyone help with a weird OSX problem?

  • From: Quartz <quartz@xxxxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Thu, 10 Jul 2014 20:52:12 -0400

(Apologies for the slow reply, I forgot to add the list to my spam exceptions)

One was that fast user switching didn't handle the screen correctly
going from one user to another ... giving blue/magenta tinted screens:
Fix for that one was to disable fast user switching.

Not using fast user switching, but that's good to know for future reference.

Some versions of 10.6 handled V4 profiles incorrectly. Not a problem
with Argyll, which produces V2 profiles, but you might want to check
that your other software is producing V2 profiles, not V4.

IIRC some/most of those issues were supposed to have been resolved in later 10.6 patches, of which this machine has all. Xrite's software defaults to v2 profiles anyway, but I didn't see any difference between that and v4's. Both versions of their software as well as dispcal+argyll produce identical profiles, as far as looks and behavior go.

Another problem was that the MBPs of that generation had screen
backlights with primaries that were some distance from the sRGB
primaries ...
one problem with the MacBook Pro display (as with a lot of
Notebook/Laptop displays) is that it has a fairly small gamut
(considerably smaller than sRGB).

I have a feeling that this is part of the problem. From other threads I've been reading and from my own research, the screen's gamut is generally smaller than sRGB, with the exception of the blue point, which sticks out off to the side somewhat. However it's offset *away* from the violet end of the spectrum (that is to say, it's somewhat more "blue" than sRGB "blue").

For reference, here's a simple example of where I see the problem:
In Photoshop I set my working space to my monitor's profile and create a new unmanaged document, which I fill with blue (0,0,255). If I use the DigitalColor Meter app to inspect the pixels in Photoshop's window, it reports 0,0,255 as expected. However, if I set my working space to any other color space (with the exception of ProPhoto RGB), and likewise fill a new managed document with blue, DigitalColor Meter reports 96,0,255.

I don't really want to use my monitor profile as my working profile, but I feel like using ProPhoto is a bad hack in this situation. What I don't understand is why "blue" is having red added to it, instead of just being shifted linearly to sRGB blue, which is closer.

What's weird is that the color shift issue doesn't seem to affect OSs running on the same hardware. It's like there's some sort of mapping error within the OSX 10.6 colorsync subsystem or something.

Another issue (although I'm not sure how much it contributes to the
problem) is that the old i1 Display2 is probably not very well suited
for a white LED backlit display (the i1 Display 2 originates from an era
when there were only CCFL backlights), at least without a correction matrix.

I'm not terribly familiar at all with how colorimeters work under the hood, but doesn't it just average the spectrum frequency and pick the closest K value? I know "white" LEDs have a sort of weird violet/yellow response curve sometimes, but I wouldn't think that would affect things much. Is there any way I can find or create such a matrix?

Basically, I'm trying to figure out what my next step should be here. I don't want to drop several hundred dollars on a new sensor only to find out that it's a software issue. I don't really want to upgrade the whole OS if I don't have to either (lots of custom software that doesn't work well on newer versions). I don't want to do either and then realize the problem is me doing something dumb because I don't understand as much as I think I do.

If providing copies of the profile or screenshots or argyll log dumps or whatever might help, I can do that.

it has a certain smooth-brained appeal

Other related posts: