FEBS said:
sarangiman said:
dtaylor said:
sarangiman said:
Normalized difference is 11.7 vs 14.8 for the D810...
And that's where any knowledgeable person stops reading. >14 stops...from a linear 14-bit ADC...kind of impossible
Nope, that's where an
unknowledgeable person stops reading, one who doesn't understand the math behind resampling.
Or a person who thinks he understands. How could you get more then 14stops out of a 14bit ADC? No math can solve this. Yes you might interpolate, expect what the value might be, but that is not correct. 14 bits means really only 14 bits max theoretical, and practical it will be lower. There is no logic and no math to find the REAL values outside the sampled values.
It's purely a noise thing. This is in a
normalized context. I prefer to know the literal dynamic range of the hardware itself myself as well. That would be DXO's Screen DR measurement, which tells you the per-pixel dynamic range of a given sensor. For cameras with 14-bit ADC units, none of them have a Screen DR above 14 stops.
A normalized context is used to create a comparable basis. Comparing the noise levels of an image with smaller pixels to the noise levels of an image with larger pixels is ignoring the frame size. This is basic equivalence. Sensor size and quantum efficiency are by far the primary factors that affect noise levels in an image. Since a D810 has more total smaller pixels than say a 5D III with fewer total larger pixels, you have to resample the larger image to the same dimensions as the smaller image. By resampling, noise is averaged out, which reduces the per-pixel noise levels. For COMPARISON purposes, this is the only fair way of determining how one camera compares to another. Otherwise your comparing noise produced at a higher frequency with noise produced at a lower frequency...which is basically comparing apples to apples.
Downsampled comparisons have their limitations. For one, they can be misleading as to how much dynamic range on a sensor is actually usable on-scene. If your standing in front of a landscape, and you meter 14.8 stops of DR from the deepest shadows to the brightest highlights, you won't be able to capture that scene with a D810. Even though it's 8x12 "normalized" DR is 14.8, the hardware DR is 13.8. Your a full stop short with the D810, and you would either need to use a 1-stop GND to balance the scene DR, or use HDR, to capture the entire thing. Additionally, you won't be able to lift the single shot with the GND by seven stops in post (as would be indicated by a 14.8 stop Print DR number). You would be able to lift it by at most 5.8 stops, however trying to lift even a D810 shot that much is going to encounter read noise. Realistically, you probably have 4-5 stops of shadow lifting ability without unsightly read noise (which in the case of Exmor-based cameras, is pretty much just random color noise, still no banding.)
Sarangiman is talking about the normalized DR values from DXO. Those values are only valid if you assume an 8x12" 300ppi downsamplng target. When a 5D III and D810 are downsampled to that image size, the 5D III has 11.7 stops of DR and the D810 has 14.8 stops of DR (engineering DR...the raw measure from the RMS of noise to the saturation point...whether all of that DR is fully usable depends on too many factors, which completely reduces any comparison to mush. The actual usable range is dependent upon the tools you use to process, the precision of those tools algorithms, your capability at maximizing the effectiveness of those tools, your personal tolerances for noise, etc....so there is no real objective measure of "usable dynamic range.")
. Which isn't the case here.
). The thing that set me off not long ago was a bunch of scenes where the skies ended up totally blown out when I exposed to preserve some detail in the shadows...and the skies were patchy...not along a nice horizon where a GND filter could actually be used to fix the problem.
) or why glass shatters on the rocks?
