OK. Could have something to do with the way on-sensor amplification and the ADC are calibrated together. Recall from the thread where kenw reported his findings concerning the GH2 that it amplifies the analog signal up to ISO 640 but only does digital scaling of the output above 800.
Good.
Neither am I. The main point here is that the loss due to the choice of offset is trivial.
See above. I was busy responding to that while you posted your request.
Hello Iliah Borg ,
I thought the image contains Kodak Q13 (or larger Q14) grey step wedge, http://www.imatest.com/docs/q13/
OK, I see what you were describing.
Any neighboring patch unobscured on all the shots would do. "A", "M", and "B" are the most important points to see the relative exposure.
DxO's "measured ISO" tells us how much exposure a camera can handle at a given set ISO without clipping, and DPR's ISO test tells us how much exposure that's necessary to get the 'correct' output brightness (18% grey), so it should be possible to combine the two measures and find out how close the different cameras in DPR's comparison (with matched output brightness) are to the clipping point.
Well, Steen ... don't just talk about it ... do it, then ! I thought that you may have browsed enough related threads to know that attempting to do so is a bit of a "can of worms", indeed?
Well, knowing my limitations, and being a bit lazy, I'll leave the hard work to those who can do it properly, and don't mind to. Don't think it should be too difficult, but quess it'll be easier and more accurate to take a look at DPR's raw files, to see how close the different cameras/sensors are to full saturation when the output brightness is matched. Would be nice/useful to know how much headroom the different cameras have above 18% grey.Well, Steen ... don't just talk about it ... do it, then ! I thought that you may have browsed enough related threads to know that attempting to do so is a bit of a "can of worms", indeed?
Don't be lazy Detail Man. You've already got all the data.Well, Steen ... don't just talk about it ... do it, then ! I thought that you may have browsed enough related threads to know that attempting to do so is a bit of a "can of worms", indeed?
The seemingly anomalous value of 4097 (for a 12 bit ADC) is caused by the errors introduced by the RW2 lossy compression mechanism. There would have been a lower value prior to compression. Similarly for GH2 where ISO is 800 or less, the peak value of 3989 is actually a compression error also. The real maximum is 3981, but the compression introduces a maximum +8/-7 error. For 3989 you should read 3981, 3988 > > 3980 or 3981 etc.
Yes, I recall your posted graph, showing 3981 as the center of a small distribution of clipped photo-sites in your GH2 image. It is harder to see that using RawDigger (even with a logarithmic Y-axis display), but I am aware of that. It's been a while since I have used RAWnalyze now, but I believe that is where I came up with the 3981 figure myself in the first place. RAWnalyze appears to average into X-axis "bins", and appears to find the center of that distribution (at 3981).
Am not sure what you are saying here. Could you explain further ?
I do not know the answer to that.
I didn't know that RW2 used lossy compression? Is this always the case or just on certain cameras or using certain options (as on some Nikons if I recall correctly)? Judging by Detail Man's response, you have posted some information on this subject at an earlier stage. Do you have a link or links to this/these posts?
Anders ,
Thanks! I'll have a look. Does anyone have information on Oly RAWs in this respect? Lossy compression in that case too or only lossless?
My FZ50 ".RAW" image-files have been described in things that I have read in the past as not being compressed - and they are almost exactly 20 Mbytes in size (for a 10 Mpixel image-sensor).
