David Arbogast wrote in post #17420065
That is not what I am referring to regarding DR (and I am a people). Dynamic range should be a black-and-white clinical term used to describe the actual measurable upper and lower limits of the sensor's light gathering abilities: the range cut-offs being where the sensor is no longer capable of capturing detail. It is the range that DXO measures and documents.
If we want to discuss "effective" or usable DR, then that speaks to relative personal opinion, and we would do well to make the distinction clear. It does raise a good discussion point: If the 5D III sensor has a measured 12-Evs DR, then what what is its "effective" DR? 10-Evs? We have been having this discussion based on the actual measured DR differences (12-Evs vs 14-Evs) and some are asking does that somewhat small performance gap really make a difference? Well, let's get clearer. If the Canon sensor has a diminished "effective" or usable DR, then we are no longer discussing a mere 2-Evs performance gap - it's actually much greater than that.
DR is intimately tied to signal to noise. Below is pasted from the DxO mark site.
Dynamic range corresponds to the ratio between the highest brightness a camera can capture (saturation) and the lowest brightness it can capture (typically when noise becomes more important than the signal, i.e., a signal-to-noise ratio below 0 dB).
The part I have bolded is your "black and white" cut-off clinical term for measuring it. They have used the word "typically" because where you put that cut-off is a subjective decision - is it where the signal equals the noise? - or where the signal is one standard deviation above the noise, or two standard deviations etc? Once you have made that subjective decision, you make repeatable, clinical measurements of the sensor properties.
Deciding on that cut-off is a fun game! That is a bit like taking the image below (an area of shadow - shooting for the "Worst Image Ever Posted on POTN") and raising the levels: slowly slide the middle "Levels" slider to the left in Photoshop until you see the noise. At what point can you see the signal? Where did you have to move the slider to before you could see the signal? At what point did you feel the signal rose above the noise 1) on an absolute basis? 2) to an acceptable level? etc. (On my monitor, I need to move it to 3.5 or so before I "see" the signal, but much further before I'd say it was acceptable. That's the subjective "usable" cut-off. Once you've made that choice, you can evaluate any future image with the same setting. This obviously does not address any noise that looks like signal (such as pattern noise, banding etc - i.e. noise that's not truly "random")