Quote:
Originally Posted by
Smackrabbit 
Here is your HSL data, in a very quick, rough form. I'm guessing this is too complex to look at and quickly understand for most people, so feedback from anyone that's interested would be great, then I can improve it before it hits a review. This is just green, it's all I have the chart for right now. I really need to find a way to automate the chart code.
After studying your Green RGB Output Chart, I came to the following observations:
1) All the data seems to be valid within the 10% (25 counts) and 100% (255 counts).
2) The data looks interesting. It clearly shows that HUE error is not a problem, therefore GREEN does not shift from Yellow-green to Blue-green at all. That's good.
3) The SATURATION changes only slightly for most of the input range 25% - 100%. Not bad.
4) The combined effect of both of these facts would not even show any problems, if viewed on a 2-D CIE Diagram. Not good for making correct assessments.
5) The major contributor to the dE's is LIGHTNESS (or brightness) of Green over most of the input range 10%-90% stimulus. This becomes increasely bad as input goes from low 10% to high 90%. So, in the important 70-80% area it is about dE=10 for (1994) color error. Not good at all.
So, if you were using this player as a calibration source, it would definitely mess up the other source's output to the HDTV. That's bad.
I wanted to look at my HDTV system knowing these facts. So, I thought the S&M Color Bars should be perfect for seeing this error. Sure enough, the Green bar was brighter with the RGB output mode I had been using. Switching back to Ycc=4:2:2 and viewing the same color bar pattern the green bar became darker. This is only a relative and subjective test but it does agree with your objective data. That's nice! Any time the measurements agree to what we see, it is always a good sign.
I noticed that at about the 220 input (count) value the curves seem to all head to 0(%) error. I am wondering if this is the clipping point that I see in the S&M Clipping Green box? I see only a single large square with no internally nested squares. I wish S&M would have put numbers on the squares or at least documentation as to the levels in the boxes on that test pattern.
BTW the data as you have it displayed is very good for making assessments. The only small notes that I would make is to make the dE's lines BOLD so they are different than the HSL curves. Also, if the x-axis could have both counts and xx% labels it makes it easier to relate to actual testing stimuli. Showing the whole data range allow incites that a single dE number would not provide. The data is what it is, having both dE and HSL combined allows for detailed diagnosis. Others may see or make observations that would not be possible with just a dE number. Anyway, excellent work!