Actually, the sad fact is that Canon sensors do have 14 stops of DR; the limiting factor is the noise of the ISO amplifier and ADC, which degrade the output DR to a little under 12 stops:
which is why it is best to discuss "effective" dynamic range, which is the ratio of signal at saturation over lowest signal above noise floor.
I am sure it won't be too long before sensor design is improved such that there will be high effective DR and we will need high bit depth to digitise. Big files anyone?
brainiac wrote:
All seems to make sense. The assertion that the noise variance is greater than the lowest bit is made without support. How do you know that?
Measurements of read noise. The 1D3 and 1Ds3 both have about 4.5 14-bit raw levels of read nosie at ISO 100, and it only gets higher (in raw levels) at higher ISO.
And why do I get much better behaviour and less posterisation in the deep shadows when pushing a 1Ds3 file, compared to a 5D file?
I'm not sure; do you have an example you could post? A common noise reduction technique is median filtering; that will have the effect of re-introducing posterization of tonal gradients by removing the dithering effect of the noise, see the discussion around figure 17a and below at
Thus posterization, while not present in the raw data, can be brought out by unfortunate choices in post-processing. An advantage of higher bit depth is to confer a bit of robustness to the post-processing workflow, but with care I suspect that equivalent results could be obtained from lower bit depth data.
Quoting from my detailed study last year: "The new Canon 1D3 has a dynamic range of +1.4 bits over the original 1Ds, while the 1Ds2 dominates the 1Ds by a lesser +0.5 bits.
