Hi Greg,
OK, i think i was dead wrong! Thinking about some things makes your brain go crazy.
But i don't want to give up. I went through your paper, several times, and through your discussion with Will on the list about "specifying sources".
Additionally i read "Tools for Lighting Design an Analysis".
To get this right from scratch:
1. I have reflectance values for surfaces in XYZ. Measured with an spectrofotometer with a D65 test source inside. Actually it was the same
Minolta CM-2002 as in your paper. I have data with and without the specular component. (SCI or SCE) Which on would you suggest or does it depend on the surface?
2. These surfaces are to be seen in a scene with sources which have color temperatures ranging from 2400 to 3000 Kelvin.
For these sources i calculated the xy coordinates with blackbody.cal.
3. I think this is the first problem. The reflectance values for the surfaces should have been measured with a tungsten test illuminant in the spectrofotometer.
And the spectral data would have been better than XYZ. Is that right?
4. As i understand you in your conversation with Will, the premultiplying part is only efficient if i would have spectral data. If i have only XYZ data, i convert both reflectance values and color values of the sources to Sharp RGB color space like so:
rcalc -f xyz_rgb.cal -f Sharp.cal -e '$1=R($1,$2,$3); $2=G($1,$2,$3); $3=B($1,$2,$3);'
5. Now i render my picture and add the correct primaries to the picture header with your suggested script. (addSharp.csh)
6. When i call 'ximage' with rtrace -I, i should get the right illuminance values or not?
7. To change the primaries for a special display i use 'pcond' later. I guess 'pcond' does a white point transformation?
8. Does 'normtiff' does a white point transformation as well?
Would this procedure be the best what i can get out if my data for "correct" color display?
Many thanks again and regards,
Christian