Suggestions for the comparison of the DR of cameras
I see another thread started by Jack, which is becoming a comparison, even though Jack probably did not intend it as such.
A comparison does not have to be objective, but if it is intended to be, then I suggest following tests in order to achieve measurable and directly comparable results.
1. Scenery (of course with very constant illumination): it should include
a. a smooth, uniformly colored (the best is some greyish), uniformly illuminated surface. It does not need to be large, even 100x100pix is enough, but if larger, then better, for one has to pick a clean and really uniform patch n it. This spot serves for measuring the noise in terms of standard deviation.
b. something fine textured whatever, the best if if is not uniformly lit, i.e. it has transitional shades. Example: silk flower; some vowen greyish cloth. The distance should be such, that the fine details are very clear in good illumination.
2. the lenses and/or the distance should be chosen so, that the pixel size of the fine structure is the same in the participating cameras. This spot has to be in focus as good as possible.
This spot serves to determine how much details are captured at the dark end.
3. Do not bother with fake ISOs; use only 100 with the Sinar and P25+. The P45+ has real ISO up to 800, if someone is interested on the ISO performance, one should shoot separate for each ISO with the P45+.
4. Make one shot based on metering. This does not serve for comparison between the cameras, but for the measurement.
5. Make one shot far (but FAR) above metering, at least two-three stops. This is to determine the saturation levels. It is mostly 65535 with the 16bit sensors, but with some cameras at some ISOs it can be substantially lower. This shot can be of anything, even a lamp.
g. Now the core: underexposed shots, by 1/3 or 1/2 stops (whatever the camera supports), lower and lower. The evaluation will find the darkest shot, in which the noise is still acceptable and the details are good enough. For the DR it is irrelevant, which exposure this will be, the pixel values count. However, the shutter times should not be very far apart between the cameras, that's not fair. I mean if one camera shoots with f/2.8 and shutter speed in the range of thousandths, while the other camera shoots with f/22 and shutter in the range of tenths, then the noise comparison is not valid.
h. Do not care for the highlights.
If such shots become available, I volunteer to analyze the raw images and to demonstrate it, so that others can do the same. I do the analysis based on the adultered raw pixels, i.e. the result does not depend on which raw processor and how is used.
Of course, this does not exclude any other method, and particularly, subjective aspects are important as well.
Gabor
P.s. please do not think, that I am asking for this. I do not plan to purchase any MFDB; I have not used my tank called Zensa Bronica for years, for it became too heavy for me (or perhaps I became too weak to carry it?).
