Originally Posted by Dave_O
I'm trying to understand why the gamma results for my Panasonic GT30 plasma vary so much depending on using Windows or APL Windows as a pattern source.
Is this a common finding, or do most sets measure the same with APL windows and windows?
Some TVs measure identical video levels (% shown on pattern) as similar lightness (Y) regardless of the measurement pattern, such as certain fixed-backlighting LCD models and similar TV technologies. Other TVs are more like your example, and they tend to vary luminance (Y) measurements depending on the average picture level of the pattern. I can't say how many TVs technologies or settings fall into each category, and the point is simply that different TVs do not all operate similarly.
Lightness variation from changing average picture level is not very difficult to observe by eye, or at least personally I tend to notice adjusting light output in a dim room. Misc. Patterns A5, Dynamic Brightness, can be used to generally observe what happens to the on-screen image as average picture level changes. The bars in the lower right of the pattern are always the same video levels, and the changing background gives five different average picture levels to observe. Another pattern for looking at how average picture level can change lightness (Y measurement) for the same video levels is the "Black Level Bars + Steps + Varying Gray" pattern on the Avia II DVD, which also has constant bars in the image and part of the background changes from black to white. Either the Dynamic Brightness pattern or the Avia II video provide a quick way to look at how average picture level may affect the on-screen image differently for various TVs models or settings.
The gamma graph is simply a comparison of gray lightness (Y) measurements. The Y value is basically the brightness of the measurement, and higher Y numbers indicate more light output from the TV. The intent of the gamma graph is just to relate how lightness (Y) changes between the video level for black and the video level for white.
There is not a lot of difference between the APL patterns and typical windows. The center video levels are exactly the same in either series of patterns. The only difference is that the APL patterns include the other levels to be measured, so average picture level remains constant for the series. Typical windows have the center video level surrounded by black, so as the video level changes the average picture level of the image also changes. Higher average picture levels (higher % windows), or brighter images, are more difficult for plasma to display, and your measurements simply indicate that your TV tends to vary Y as average picture level changes.
My personal opinion is that for most applications, I currently don't see a reason to use typical windows for measurements. Since different displays do not adjust similarly with changing average picture levels, and because I have not found a clearly-defined reason for including changing average picture levels in measurements, I think it generally makes sense to use constant APL measurements. This perspective is in line with what theory I can remember reading from Charles Poynton. For example the gamma articles from http://www.poynton.com/GammaFAQ.html
generally seem to relate video levels with display gamma, and he never address how measured gamma may be affected by items such as various patterns, displays, or settings. Likewise I haven't ran across any experienced calibrators suggesting that there is a clearly-defined way of calibrating gamma on different display types, even though various display technologies like LCD and plasma can clearly measure gamma differently. So my current personal opinion is that if you only want to measure one series of patterns, go with the Large APL. One series of measurements will never give you complete information about a display, but honestly gamma measurement and calibration is not precisely defined at the level people seem to suppose. At least eliminating average picture level fluctuation, by using constant APL patterns, gets rid of some possibly-detrimental effects that can arise with how certain displays or settings react.
EDIT: One situation where APL patterns may not be appropriate would be if you're using a non-contact meter. When using a non-contact meter with the Large APL patterns it would generally be a good idea to try to check if the levels around the edge of the pattern influence the measurements. For example with the color patterns, contamination from the surrounding levels would likely cause different xy measurements compared to typical window measurements.