@andrewfee: Thanks for bringing up a very interesting point. As we do not own a spectrophotometer that's accurate across all spectral intensities (last I checked it costs north of US $5000 excluding import duties), we used colour bars/ filters/ movie reference scenes to align the colour decoders. Nothing we did to the "Hue", "Saturation" and "Brightness" in the 3D Colour Management submenu on the Toshiba 32C3030D eliminated the red push.
But I'm intrigued by the method you're using, especially if it means we can migrate towards a more objective way of colour decoding adjustment. Could you describe what define the x and y-axis in your first graph, and how did you make the measurements?
Firstly, I realise I had made a wrong assumption - I thought you had been using
Colorimeter HCFR for taking the measurements, but it seems that may not be the case, as I just noticed your graphs are not watermarked. (what software/hardware do you use?)
Secondly, it looks like my results may not be that accurate from the sounds of things, as I am currently using a colorimeter, rather than a spectrophotometer. (an X-Rite DTP94, which was recently purchased as an upgrade from a Spyder2)
Having said that, visually, the results of this calibration certainly do look like an improvement to me. There is no longer a red push to my eyes.
With the "colorimeter hcfr" software, there is a section to measure RGB primaries, and CMY secondaries to plot out the display's gamut, but you can also measure the individual saturation levels of each in 25% steps. (I had created my own patterns using the colour values required, as posted above, rather than using full-screen patterns, as it was quicker - though I have now gone back to using the full-screen ones)
There are three different diagrams that show the results of these measurements.
If you measure the primaries and secondaries first, to plot out the gamut, everything then becomes relative to that. For these measurements, I did not measure the primareies/secondaries, which means that everything is being compared against Rec.709 (which is why 100% has a high DeltaE for most colours)
The first is the saturation one:
(taken from my latest calibration, as I'm constantly tweaking)
The upper graph is the saturation of each colour. X axis is going from 0-100% (though the results are only marked at 25% intervals) and Y is the saturation from -100% to +100%. Normally I would change the scale of this to ±5% as it makes things a lot easier to see. You can hover over each reading to find out exactly how much over, or under saturated a colour is. On this run, my maximum was +3.08% with 75% yellow, and the minumum -3.54% with 100% red.
To adjust this, obviously, I use the Colour and RGBCMY saturation controls on the display. Colour sets the 100% value, with the RGBCMY controls
mostly just affecting 25% - 75%, which is how I was able to bring things in within about ±3% of ideal. (at least according to the colorimeter)
I wouldn't describe saturation as being the "purity" of the colour, but how "strong" it was. Eg if saturation is too high, an input of 75% red might end up actually being displayed as what 80% red should be.
The lower half of this is DeltaE, which measures how accurate the actual colour is. The lower the better, obviously. As you can see here, greens, blues, cyans and yellows are all fairly good, but reds and magentas are not, which is typical for LCDs. (at least ones that aren't using wide colour gamut technology) This is mostly adjusted with the RGBCMY hue controls.
The second graph is a luminance one, and would presumably be affected by the RGBCMY brightness controls, if this set had them. Unlike the WLT66 or the C3030D, the WLT68 only has hue/saturation controls. Presumably the '68 is using newer processing which could explain why I was able to eliminate a red push, whereas you weren't able to.
The third visualisation is on a CIE diagram:
As I mentioned above, for this measurement, I did not plot the gamut, as it means the readings I get are compared to the ideal, rather than being relative to what the display is capable of. This is why red at 100% has such a high DeltaE, and is undersaturated by 3.5%. If I were to make my adjustments relative to that, it would mean all reds would be undersaturated by 3.5%, whereas I have all red up to at least 75% perfectly saturated this way. That's something I would only do if the display's gamut is close to Rec.709 though - I would hate to see how something with a much wider gamut looked this way.
The reason I did this, rather than making the adjustments relative to what the display was capable of is because the end result is more accurate, and results in a more pleasing image, to my eyes at least.
The same applies to hue. I were to have adjusted the yellow hue so that 100% yellow was exactly where it was supposed to be (as it seems you do with the measurements you take) it would mean everything had a green tint. This way, 0-75% yellow is all pure yellow, with only the brightest yellows being slightly off.
It's the same thing with red - rather than having most reds with a slight orange tint, 0-75% is a lot closer to the ideal, with only the brightest reds being affected, rather than all of it.
As the saturation is within around 3% of ideal, you can see that all the points are evenly spaced from the centre to the edge of the gamut. If a display was over-saturating things, they would be pushed towards the outside edge, which is how this was initially. You can see that with yellows, 50% is slightly undersaturated, and 75% is oversaturated, which is why there is a larger gap between them.
I hadn't even considered the fact that my colorimeter is probably not accurate across all spectral intensities though, so these results won't be 100% accurate. That alone now has me thinking about purchasing a spectrophotometer, though I would have gone for the obvious choice of a Gretag MacBethe Eye One. With you mentioning $5000, that is a lot more than one of those costs, so presumably it isn't up to the task then? An Eye One would be pushing things for me, as I had a hard enough time justifying a colorimeter. (and only bought the DTP94 as I got the money for it from selling my Spyder2 Pro - though the increase in speed/repeatability was certainly worth it)
Interestingly, I recently lent my Spyder2 to a friend before selling it, and on his Sony KDL40W2000 using this method he ended up with a colour setting of 60 as being the best, which matches up with your suggested settings. While it may not be 100% accurate, it certainly does seem to help using this method, in my opinion.
