Hi, I am trying to understand how the RGB separation graph is computed and what it is telling me. I have read Ted's explanations here, here, and here, but I still have some questions.



My initial assumption was that at each point, it is plotting the expected luminance (x-axis) of both the greyscale and primary RGB patches against the measured luminance (y-axis). This would make sense except for that the greyscale line is always perfectly linear, and I know that my display does not match the target 2.4 gamma exactly (nevermind measurement error/noise, etc.).

My next thought was that the greyscale measurements are used to normalize the measurements at each point, so the greyscale measurements always lie on the perfect 1:1 line and the RGB measurements are computed as (measured R/G/B Y)/(measured grey Y)*3. This would imply that if (at all luminance levels) the RGB balance was perfect and the display was perfectly additive, all three of the R, G, and B lines would match up with the 1:1 greyscale line.

However, this would also mean that if the display was not perfectly additive such that the greyscale luminance was less than the sum of the R, G, and B luminances, we would see the R, G, or B lines above the greyscale line. Ted's says here that will never happen, so I'm wondering if my understanding is wrong or Ted simply meant that it does not (usually) happen in practice, because most real-displays that are not perfectly additive (like WRGB OLED) have the sum of the color luminances less than the greyscale luminance:


In summary, can someone explain how exactly the RGB separation plot is computed?

Lastly, what is the Error checkbox on the graph supposed to do? I am toggling it back and forth and not seeing any difference.


Thanks,
Erik

Please see the 'Reading ColourSpace Graphs' section of the Advanced Operation page of the website.
That should explain everything to do with the ColourSpace graphs.

And the 'Error' toggle is not yet operational.
(I'm sure that is explained somewhere... but to be honest, I can't find that info at the moment.)

Steve

Steve,

Thanks. I read through "Reading ColourSpace Graphs", but I still don't understand why the R,G, and B lines will never plot above the white line.

Let's say I have an imaginary display that cuts the luminance of any greyscale color in half. So (255,0,0), (0,255,0), and (0,0,255) are displayed as expected, but (255,255,255) is very dim. What would the RGB separation plot look like?

That's because a plot above the line would go out of the graph area...

So, the plot is a relative error, not an absolute one; the direction (above/below the line) is irrelevant.

Steve

Ok, understood.

Although if you measure any patches above 100% white the data already goes out of the graph area, as you can see in my first screenshot.

Since you can pan and zoom the graph it would be nice if there was a graph option to disable the "absolute value" of the error so that it is possible for the R,G,B lines to plot above the greyscale line.

This would give us a little bit more information about what is happening with the display if we choose to enable it without changing the fundamental interpretation of the graph (R, G, B lines should lie as close to the greyscale line as possible).

The first graph means you have Extended Range selected.

That is very different, and is correct as 100% is the edge of the graph.

But there is no such thing as a +/- error with RGB Separation - the error is always in the same direction... It is always a relative error.

But it is not a Grey Scale error!
You can have a perfect grey scale, but have bad RGB Separation.
(The plotted error is the relative error between the grey scale value and the equivalent RGB values.)

Steve