[argyllcms] Re: Difficulty profiling new iMac

  • From: Graeme Gill <graeme@xxxxxxxxxxxxx>
  • To: argyllcms@xxxxxxxxxxxxx
  • Date: Fri, 14 Sep 2007 18:07:26 +1000

OK, here's a bit of a clue. If I put up a black test square on
my MacBook with a linear Video LUT at maximum brightness, I get
the following with several different instruments:

                    X           Y           Z           L*          a*          
Spectrolino:        0.366415    0.352546    0.692548    3.184534    1.069715    
Eye-One Pro:        0.405160    0.391266    0.731828    3.534291    1.126687    
Eye-One Display:    0.909546    0.913935    1.476462    8.252869    1.108462    
DTP94:              1.010000    1.010000    1.850000    9.074471    1.321416    
White (Eye-One Pro) 207.192652  213.696473  231.775628  133.413707  1.192318    

I'm, guessing that this machine uses LED backlighting.

So the colorimeters seem to be rather different to the spectrometers, as well
as the difference between white and black color being more extreme. An 
for a "kink" in the blue is that by default for an LCD display, dispcal
doesn't aim to make the whole curve neutral (ie. b* = -24, dE to black in b* of 
since this would lighten up the black too much, but it does so down to a
point where it then crosses over to the native black (ie. it does -k 0.0).
This isn't spread evenly over the whole curve, on the basis that it is
desirable to get a good neutral for more of the whole curve. The differences
between the instruments may not mean much as far as the visual effect goes,
although it's another complicating factor.

Another complicating factor with a test ramp starting at
L* = 0, may be the clipping/gamut mapping behaviour of the
CMM when the L* min of the display is 8. This depends on whether
a true gamut mapped perceptual mapping is being used, or
black point compensation, or not.

I can think of two approaches to improve the situation. One would be
to aim for a white point that is more consistent with the color of the
black. (You'd have to convert the black to an x,y value and feed it into
dispcal -w. Hmm. I should add an xy reading mode to spotread!)

The other thing I could do is to make the transition from neutralizing
to not neutralizing more gradual and spread out. I'm not sure if
another parameter is needed, or whether all LCD's will be better off
with a more gradual transition.

Cutting the MacBook brightness back gives:

                    X           Y           Z           L*          a*          
Eye-One Pro:        0.201529    0.195416    0.362346    1.765181    0.529371    
Eye-One Display:    0.397316    0.394413    0.658303    3.562717    0.687412    
DTP94:              0.450000    0.440000    0.820000    3.974504    1.039887    
White (Eye-One Pro) 97.101597   100.017926  108.225388  100.006931  1.145532    

So the disparity betweem instruments remains, and it is still quite blue (dE 15 
in b*).

For comparison, here is the same setup for my older Titanium PowerBook G4:

                    X           Y           Z           L*          a*          
Spectrolino:        0.399623    0.401044    0.684995    3.622620    0.522360    
Eye-One Pro:        0.402123    0.409760    0.670781    3.701349    0.283949    
Eye-One Display:    0.655930    0.685013    1.005142    6.187695    -0.184091   
DTP94:              0.640000    0.680000    1.080000    6.142415    -0.632202   
White (Spectrolino) 94.719391   102.122330  104.044060  100.814896  -6.469523   

I think this machine uses electroluminescent lighting.

The display is half the brightness, and half as blue. The disparity between
the spectrometers and the colorimeters is smaller, and the difference in b*
is dE 8, much more moderate.

Graeme Gill.

Other related posts: