M107v1-10: Color calibrator

[technix]

This piece of test equipment is for calibrating color rendering across different displays. I am currently at the drawing board stage of this project, but a few questions already arose regarding this board.

The project is effectively an analog color sensor hooked to an Arduino, and all calibration is done using PC software crunching the numbers.

Questions:
1) How do I profile the display properly in the first place?
2) Do I need to calibrate the sensor itself before use, or can I just compare relative profiles of different displays and generate a ICC profile to make two displays render color in the same way?

[rs20]

This isn't trivial. I'm not sure what you're planning to use as a red filter, whether it be a separate red filter than you buy and purposefully lay over a photosensor, or a color photosensor that has a red filter built in. But either way, that red filter has a certain wavelength/transmission curve. If that wavelength/transmission curve differs from the wavelength/sensitivity curve of the relevant rods and cones in your eyes, then two spectra that look identical to the red filter will look different to the human eye, and vice versa.

Unfortunately, real-life filters have significantly different transmission properties to the sensitivity properties of the human eye, partially because detection and transmission are totally different things.

This is why digital cameras can sometimes see colours in a very different way to each other, and the human eye. Very rich, regal purple colours often produce particular problems, appearing blue. Camera manufacturers put a lot of effort into getting their filters just right; if you just buy a random "red-sensitive" photosensor, or use a random red bit of glass, your results will be significantly worse than a digital camera, which is already too poor for display calibration purposes.

Anyway, these problems can be overcome in various ways -- getting expensive, carefully specified filters; or using more than three channels with carefully selected but individually not-so-good filters and mix the results in a way that simulates better filters (most commercial color calibrators have at least 4 sensors for this reason); or go the full hog and use a spectrophotometer, etc. But if you attempt this project without playing with wavelength/absorption/sensitivity curves a lot, you are almost guaranteed to have very disappointing results!

To answer your questions,
1) Eh? Command the monitor to make lots of different colours, and measure what's actually produced? Compute the inverse transformation and load it into an ICC profile?
2) Interesting question. As I mentioned above, if you're not careful, you may set up the monitors so your sensor thinks they're identical, but they're totally different to the human eye. You can also set up your monitor so that 50% brightness is actually brightness, make sure the gamma of the monitor is correct, etc etc. Obviously if you make your two monitors identical, then your two monitors are identical, which you may consider to be a win. But if you develop photos, you might find that they come out a slightly different colour to what you expected, because your monitors are not identical to the relevant international standards. So whether you consider that "calibrated across different displays" is a question of nomenclature.

[technix]

This isn't trivial. I'm not sure what you're planning to use as a red filter, whether it be a separate red filter than you buy and purposefully lay over a photosensor, or a color photosensor that has a red filter built in. But either way, that red filter has a certain wavelength/transmission curve. If that wavelength/transmission curve differs from the wavelength/sensitivity curve of the relevant rods and cones in your eyes, then two spectra that look identical to the red filter will look different to the human eye, and vice versa.

Unfortunately, real-life filters have significantly different transmission properties to the sensitivity properties of the human eye, partially because detection and transmission are totally different things.

This is why digital cameras can sometimes see colours in a very different way to each other, and the human eye. Very rich, regal purple colours often produce particular problems, appearing blue. Camera manufacturers put a lot of effort into getting their filters just right; if you just buy a random "red-sensitive" photosensor, or use a random red bit of glass, your results will be significantly worse than a digital camera, which is already too poor for display calibration purposes.

Anyway, these problems can be overcome in various ways -- getting expensive, carefully specified filters; or using more than three channels with carefully selected but individually not-so-good filters and mix the results in a way that simulates better filters (most commercial color calibrators have at least 4 sensors for this reason); or go the full hog and use a spectrophotometer, etc. But if you attempt this project without playing with wavelength/absorption/sensitivity curves a lot, you are almost guaranteed to have very disappointing results!

To answer your questions,
1) Eh? Command the monitor to make lots of different colours, and measure what's actually produced? Compute the inverse transformation and load it into an ICC profile?
2) Interesting question. As I mentioned above, if you're not careful, you may set up the monitors so your sensor thinks they're identical, but they're totally different to the human eye. You can also set up your monitor so that 50% brightness is actually brightness, make sure the gamma of the monitor is correct, etc etc. Obviously if you make your two monitors identical, then your two monitors are identical, which you may consider to be a win. But if you develop photos, you might find that they come out a slightly different colour to what you expected, because your monitors are not identical to the relevant international standards. So whether you consider that "calibrated across different displays" is a question of nomenclature.

My primary intention was to make my computer's monitor render the same sRGB color in the same way as my iPad or my iPhone's screens. Or, as an extended goal, if I can get the sensor itself calibrated, make my screens reproduce sRGB as truthfully as possible.

My color sensor is the TCS3200 chip.