János, Tóth F. wrote: > Why wight white at the first place? > White is just one color in the 3D space just like any other color. And > they all mixed from the primaries. Because white is the most critical color by far. People judge color by how non-neutral it is, and notice a lot less if a saturated color has a slight error. So if a display is not additive, it's better that the instrument be more accurate on white than anything else. > As you said yourself, RGB would be perfectly enough if every displays > were perfectly additive. Which they are not. But it comes down to what the dominant source of inaccuracy is. > (I this case I think the measurement errors > should be decreased by collecting an averaging more samples of the > same RGB patches, not guessing from other mixed colors.) Sure, but this goes for everything involve :-) The task in creating the software is to do the best job possible given a certain measurement time budget. > And why weight with an uncalibrated white point? Because even an uncalibrated white is closer to any other white than a primary color. > Setting up the white point of a display is one of the most common > calibration task. And the display is often far from the desired white > point before the adjustments. Right, which is why I'm striving to make the instrument as accurate as possible on white, and typically the way you do that is to use the critical color as a test sample, and weight it heavily. Your results suggest that something counter-intuitive is going on. Graeme Gill