Originally Posted by R Harkness
Have I got that anywhere near right?
Seems about right.
One issue with the ANSI CR checkerboard is that it is a test where our eyes have trouble seeing a lot of CR. It doesn't take all that much to satisfy us. It is watching over periods of time where our eyes can adjust a lot more that our eyes can see across a much bigger range and so it is actually the larger range of the projector (on/off CR) where our eyes can adjust enough to see that there is a deficiency in the performance of the projector.
That isn't to say ANSI CR doesn't matter. It does. It is just an extreme test. If we consider it with gamma it calls for 2.5 times as much light as even a full screen 50% video pattern of gray, which is already much brighter than the average movie image. That isn't even counting that non-gray things tend to have less light. The ANSI CR pattern is even more than 2.5 times brighter compared to a 50% full screen green, red, yellow, or just about any color other than gray.
And going back to that 5%/0% mixed pattern on the 2nd Spears and Munsil disk, the regular ANSI CR pattern is about 700 times brighter than that is supposed to be. How could ANSI CR possibly tell us what we need to know about an image that is 700 times dimmer than it?
I think that what you've seen with your own eyes is that when we have a projector and setup that is good (not great) at ANSI CR and good (not great) at on/off CR the images that really bring attention to themselves are the ones that have poor CR compared to what our eyes are capable of seeing. Since the ability of our eyes to notice CR weakness goes down with more light in the images the lack of great ANSI CR doesn't make it itself obvious all of a sudden, but the lack of great on/off CR shows up as pretty obvious sometimes to those who care about black performance. Here an increase to great ANSI CR brings almost nothing to the table to fix the cases where poor CR is really sticking out, since many of those scenes that show obvious problems have less than 1/500th as much light as the ANSI CR pattern.
Instead of going dimmer by lowering the peak level in an image (like the one with 5% peaks I mentioned) we can go the other direction that our eyes can start to see across a bigger range compared to the regular ANSI CR pattern and that is to decrease the amount of white.
Consider an image that is basically 1% white pixels and 99% black pixels. That may seem like hardly any white pixels, but on an 8' wide screen that is a 16:9 rectangle that is 9.6" wide (10% of the width). Here the amount of light that image calls for is 1/50th as much as the ANSI CR pattern. Again, a ways from it. It is also some distance from the patterns used for on/off CR, but in some ways less distance than from the ANSI CR pattern.
For this pattern what would an ANSI CR of 200: 1 vs 500:1 tell me about the contrast ratio I will get with that image? Not all that much unless the on/off CR is at least 5k:1 I would say, since the on/off CR would be the limiting factor anyway. With only 1% of the pixels at white instead of 50% of the pixels as white like in the ANSI CR pattern the washout it describes is only going to limit our CR in that image to maybe 10k:1 if the ANSI CR is 200:1 and 25k:1 if the ANSI CR is 500:1. Those are approximations as there are a lot of dependencies like how far the black a person is looking at in the image is from the white pixels and the causes of ANSI CR being 200:1 or 500:1 (how proximity based it is).
When some of these people claim that on/off CR is irrelevant I wonder if they even have a clue that an image like that could achieve something like 10k:1 with just 200:1 ANSI CR, but won't even come close if they have poor on/off CR.
I truely wonder how much some of these people have actually checked their assumptions. For instance, one standard is based on just shining lights at the screen then blaming the lower ANSI CR they measured for the problems seen. They basically killed ANSI CR and on/off CR then blamed their lowered ANSI CR for the problems they had seen. If they had properly checked their assumptions for cause and effect they would have used reflections instead of other lighting to reduce their ANSI CR without killing their on/off CR, then I am sure all those problems they had seen would not have been there. Yet they should be there from the lower ANSI CR if that were actually the cause. Some of us have actually done testing like this which is one reason I am confident that those who assumed lower ANSI CR caused their image problems would not have seen all those problems if they had only changed ANSI CR.
Unfortunately, that standard is out there for people to see as gospel without having any idea of how poor their testing methodology was. I wouldn't want my name on a flawed standard like that. Fortunately, I don't think the FDA lets drug companies get away with such mistakes or we would all be in trouble (or more trouble) IMO.