Dmitry Shijan wrote:GH5s RW2 converted to DNG are very strange, Also i notice that extreme highlights are always hard clipped even on under exposed scenes.
For example here s a sample (not mine) from Pocket Camera shoot in similar light conditions as a building above.
You can't really compare unrelated images like that. But more importantly, the real magic isn't necessarily in the sensor itself, even if that constitutes the foundation. Even if the sensor itself has its hard specs, I'm sure surrounding circuitry and components (that might change between manufactures) will impact final performance.
And even THEN, I'd say the real mojo begins in the translation of the linear information to each manufacturer's 'color science'. Here you have your non-linear encoding and color transformations. In the case of raw data, this will take place in post, but if you go with ProRes it will happen in camera.
I'm much more interested in the image data after it's gone through the OETF and it has been encoded non-linearly. I think RED has found a very good solution with their IPPT 2, where when you choose Gamut and Gamma and your footage comes in with no clipping.
Importing BMD's 'Interior House' footage into Resolve 15 with BMD color and BMD film gamma shows severe clipping. The data is there, of course, but I STRONGLY encourage BMD to make 'project settings' available that import CDNGs with no clipping—without user input, like messing with raw sliders. There should be a film gamma for BMD footage that results in images similar to REDWideGamut and Log3G10 as a starting point. Not because it's difficult to wrangle the data back, but I think there's a benefit to getting everyone to the same starting point when it comes to image evaluation, LUTs and other things that are based off of a constant.