ajfotofilmagem said:
Don Haines said:
ajfotofilmagem said:
A true test of DR will compare similar cameras
There are lots of valid reasons for testing dis-similar cameras... One might be asking how the 7D2 compares to the 5D3.... one might want to know what the differences are between old tech and new tech.... between FF and Crop, between 4/3 and crop, between Canon, Nikon, and Sony... Lots of questions to be asked here and even more answers sought. We THINK we know the answers, but until testing is done, we are just guessing.
Don, you forgot the most important part of my message: Tests comparing a parameter (for example DR) must minimize all other possible variables.
I understand that some want to know "how much" of a difference between the different categories and equipment purposes. But the bottom line is always getting stuck at sub-tests at ISO100 purposely under-exposed to lift the shadows in PP. :-[ This seems to me an excuse to prove to yourself something like "my dick is bigger than yours"... :-X
I think that's a flawed assessment, though. It also paints everyone who wants more DR as raging egomainacs, which is absolutely not the case. I hope that Michael shares his full-size images, including the highlight areas, so that people can understand what these tests really do.
If done properly, I expect the highlight steps of the wedge to be white. Only the top one should be pure white, the rest should be slightly less than pure white and barely differentiated. It might require some highlight recovery to properly demonstrate that (highlights are attenuated by RAW editors as well, to a small degree), but in the end, that's what the results should be. IF that is indeed the case, then the test was properly done, and the results have nothing to do with Michael trying to purposely bias the results or intentionally "underexpose".
Cameras, including Canon's, have a lot more dynamic range than can be shown on a computer screen. The "bottom" several stops, the shadow stops (and the lowest "blacks" stop) very often contain data that in real life is not black, however those stops show up as black when the image is imported and displayed on a computer screen.
Why? Because with 8-bit rendering, you have 255 discrete tonal levels. You don't have 24 bits worth of tonal levels, you only have 8 bits, since it's the luminance range of each sub-pixel that ultimately matters here. In terms of stops, those 8 bits get you 20 * log(255/1) / 6, or 8.0218 stops. (There are some arguments that dithering can improve the dynamic range of a digital screen...possibly, but there is a lot of variation there, and it really depends on the screen...and empirically, it's easy to test how much dynamic range a computer screen has by, say, photographing a step wedge and seeing how many stops you have to lift the blacks before the bottom two wedges become a barely differentiated "pure black" and not quite so pure black.)
A Canon camera has around 11 stops or so, which means there are at least three stops of additional tonal detail that, rendered linearly, will just show up black or nearly black (in only 8 stops), when in real life it probably isn't. A camera with 13 stops of DR has at least five stops of additional tonal detail that, rendered linearly, will just show up black or nearly black. (This assumes you render linearly to display the highlights without clipping...most RAW editors actually render for the middle tones via a tone curve, with about a stop of recoverable shoulder in the highlights, and the rest in the shadows.)
Throw in the camera profile curves of a RAW processor (non-linear rendering), which attenuates the rendered signal, and all those near-blacks become fully black, and more tones become "shadow". Then you have the screen gamma, which further attenuates the signal, resulting in more information being rendered as black or nearly black.
Why do we lift shadows? To restore tones that, when rendered to screen, appear black on screen but in real life were not black and in the photo should not be black. That doesn't mean that every single tone in a rendered RAW image should be non-black, it just means that there can be stops...many stops in some cases...that should be non-black that, without shadow pushing, end up black in your photo. This isn't about anyone's dick being bigger than anyone else's. This is a fairly clinical subject...it's simple observations of data and results.
Alright, I'm sure if I continue in this thread any longer, things will become a war (and a war not about a clinical subject...it'll get far more personal and hateful than that)...I don't want that. Michael, thanks for the test images. Interesting results. Looks like the 7D II definitely has less banding. Here's to hoping that Canon can improve their read noise in the next generation, and allow more of those shadow wedges to be usable in the future.
