Daan B wrote:
garyvot wrote
Take a look at the D5000 sample closely. Notice what is happening with the luminance noise? It is very clumpy and the edges are quite broken up. Most of the Nikon bodies I've tried seem to exhibit this at some ISO, so it may be that Nikon is "cooking" the RAW data in some way. Nikon IS cooking the RAW data. Not only the D5000, but the D90 too. And probably other crop cams as well.
I see this claim bandied about a lot, but never with a quantitative demonstration to back it up. Do you know of one (or better, several)?
I agree that the 5D2 sRAW1 files look pretty good. It also seems that Canon cooks it with some kind of NR applied, because the sRAW1 data is cleaner than when downsizing a 21MP to sRAW1 size. The negatives of sRAW1 are decreased sharpness (when compared to a downsized full RAW file) and they fall apart quickly in PP (highlights, colors).
IIRC, sRAW/mRAW use half of the storage for luminance info and the other half is split between the two chrominance channels, so there would be less info in these files than in a demosaic'd and downsampled image from a full RAW file. Roughly, chrominance data is split on average between two pixels.