Originally posted by newmikey Grabbed a couple of ISO25600 testshots off Dpreview detailed camera reviews (all raw files) and converted them all with no noise reduction, no curves, no sharpening, basically only with the built-in profile Darktable applies and all other modules turned off.
I wanted to look at noise patterns, color noise versus luminosity noise and detail retention but I did not want to look at the clean files after a lot of PP like most of the reviews and comparisons do, thinking that any NR I can apply would likely obliterate any objective differences between the sensor outputs.
I then combined a crop of the same area (ca. 2000x1300 pixels) of the testshot of all 4 cameras without scaling any of them but flipping some images horizontally, vertically or both in order to be able to properly judge identical areas in close proximity without allowing my eye to get distracted by wandering across too much.
The result is below and I find it quite surprising actually. Cameras are the Nikon D750, K-3 II, K-70 and K-P but I'll leave you to decide which quadrant is which camera. (click through to Flickr for full size)
First problem with that assumption is that it has been years since even raw are "cooked" with some post processing so basically this just mean the raws with the most post processing done in camera win.
Second problem is what count is what people typically get. So that either OOC JPEGs, or RAW post processed automatically with noise reduction from lightrom/camera raw or DxO. And this is interesting the raw processing doesn't apply the same setting to all camera because the goal for them is to get the best possible result, not the one where the camera look as worse a possible as to grow the difference.
The idea is more with our raw processing software all camera look good. And they manage it, mostly. There still difference, but not as much as one might think.