Hi,
DR is actually defined as full well capacity divided by readout noise. As pointed out by BJanes, the width of the ADC sets a limitation on the DR as long as linear representation is used. Sensor vendors sometimes specify the dynamic range of the chip, say for instance 76 dB. You can convert that 76 dB to EV by dividing by 6. So 76 dB -> 12.6 EV. DxO-mark measures DR for cameras and they came up with a DR 12.34 EV. Pretty close. The camera in question was the Leica M (typ 240). It is a 14-bit device, achieving 12.34 EV in DR.
Now, another device has been measured by DxO-mark, and it came in at 11.89 EV. That camera is a 16-bit device known as the Phase One IQ-180.
Which is the best camera regarding DR according DxO-mark? It is the Nikon D810 at 13.67 EV. It is a 14 bit device.
Now, of all these cameras the Phase One IQ-180 has the lowest DR, actually meaning it utilises 12 bits out of the 16 bits it is claimed to have. But, it has quite a lot of pixels. Now, if we make a small print, those pixels would be packed and yield a higher DR. This is simple math, and I don't go into that for now. DxO-mark does this calculation using a small 8MP print as a base, in this print mode we get:
Leica M (typ 240) 13.13 EV
Phase One IQ-180 13.56 EV
Nikon D810 14.76 EV
It seems that the Nikon can transfer more than 14 EV trough it's 14 bit data path, but this is not the case. It is a result of normalisation.
Let's check out Canon's latest creation, the 5DSR. In screen mode it reaches just 11.05EV, in spite of having a 14 bit data path. In the normalised mode it reaches 12.39 EV, a bit shy of the Leica M (typ 240).
Now, we may go up in ISO, somewhere around 1200 ISO the Canon is just as good in DR as the Nikon. At high ISO the signal of the sensor is amplified and noise in ADC is thus reduced.
Best regards
Erik
The dynamic range of a ADC (or DAC) is the ratio bentween the largest expressible value vs. the smallest expressible value. It's is that simple.
If a sensor/ADC is speced to capture 14bit bit depth, than it should make sure the (internal) noise figure is within the half bit range, unless the engineers or the marketing guys lie. Rarely happen for reputable.
Whether the signal it captures has that much of dynalic range is another issue.
Flare, interference, lens distortion, and optical system resolution etc., would affect the final captured image and will affect the effective usable dynamic range and resolution. Both MP and bit depth can suffer the deteriorating.
If you want to fool yourself, you are welcome.