I also have a comment on the DR figure. MF sensors have larger pixels, so they can store more electrons. That essentially means that they will have less stochastic noise. Now, DxO-mark measures DR defined as
(maximum signal)/(signal at SNR=1)
noise here is defined as read noise in the sensor. But, for normal photography an SNR of 1 is not really useful. At larger SNR photon noise would play a larger role. Therefore an MFDB may perform better regarding DR/noise characteristics compared to DSLRs with smaller pixels. The DxO definition of DR is the technically correct generally accepted definition, by the way.
Eric,
Your point about the cutoff for shadow noise in determining dynamic range is a good one, but MFDBs do not necessarily have larger pixels than DSLRs. For example, the Phase One P65+ has 6 μ pixels, while the Nikon D3s and D3x pixel sizes are 8.34 μ and 5.95 μ respectively. For a given sensor size, increasing the pixel count will decrease the
per pixel DR as defined above, but the DR for a given print size will often not suffer, since downsizing of the higher pixel count image will decrease noise through pixel averaging. This factor is taken into account in the DXO normalization for print size and is discussed at some length by
Emil Martinec.
For dynamic range in the print at a given size, the number of photons collected over the entire image is the main determinant of total shot noise. The P65+ has a sensor area of 53.9 x 40.4 mm, about 2.5x the area of the full frame Nikon D3x. If fill factor and read noises were comparable between the two sensors, the P65+ would collect 2.5 times the number of photons as the D3x. Since SNR varies as the square root of the number of photons collected, the SNR ratio would be sqrt(2.5) = 1.5 times better for the P65+. This difference in sensor area is significant, but would not lead to the astoundingly greater DR claimed by some for MFDBs. The greater image quality claimed by these proponents is likely related to factors other than DR.
