Originally posted by rawr I think you are too confident on this point. Merely repeating a statement a lot doesn't make it true
Technically, no one can tell what is going on unless they have access to the K20D and K7 firmware source code, and/or the sensor microcode, and/or any other information about the way the K7 sensor and associated circuitry works compared to the K20D's. No one here or on dpr has any of that information. And looking at the hex code or metadata of the RAW file itself won't tell you.
I think you'd be surprised at what you can tell just by looking at th RAW data itself. NR algorithms will leave a tell-tale trace on the data that you can see by atching what happens to the actual standard deviation of the pixel values as you increase ISO (I *think* that's the basic idea of process, anyhow - probably oversimplified). People who make it their business to analyze this kind of data are pretty confident on this point. We're talking about some amazingly astute individuals putting a *lot* of effort into this task here, not some random people just sort of guessing.
Quote: Making these deductions on the basis of visual output is difficult though because you are always looking at RAW output that has been rendered by a RAW processor (ACR, dcraw, Capture One, Silkypix, whatever) that usually applies its own 'personality' to the RAW output of any camera, depending on what it might decide is appropriate for what it knows of a particular camera model or not, or other decisions of the RAW processor programmers about interpreting stuff like the DNG specifications (in the case of RAW output in that format) etc etc.
The people I am talking about have written their own software to analyze the data, and at least one of them ("GordonBGood") has created his own utility to demosaic the data himself using *no* additional "personality". Like I said, we're talking about going *FAR* beyond simply looking a some pictures from a geneic RAW converter and guessing.
I think unless you've looked at the data yourself with the same degree of scrutiny and have some specific numeric arguments to counter the very convincing analysis by Gordon and the others who have partaken in this analysis, there is really no reason to doubt their findings.
Bottom line: if K-7 images using default settings look noisier than K20D images using default settings, it is primarily because there is no NR applied to the RAW data on the K-7. Meaning that when you add the NR yourself in PP - or even perhaps with the in-camera JPEG's if you don't care as much about IQ at this level - you should be able to get results that are more or less exactly as good from the K-7 as from the K20D.