Re: More About the EOS 5DS & EOS 5DS R
I agree, stronger CFA won't improve the perceived color quality. Not, at least, unstretched RAW images. It could improve color noise, which would affect the color quality of the shadows.
Improved color discrimination should reduce color crosstalk, which should reduce color noise. Color noise is an issue Canon cameras have had for a while, and it got worse when Canon weakened their CFA with the release of the 1D X. Strengthening the CFA may be doing nothing more than restoring it to what it was before the 1D X, I dunno. Regardless, the fact that Canon is doing anything that could reduce color noise is a good thing. I don't know if reducing color noise will improve tttmnbso*...I kind of doubt it, not if other sources of noise remain high. The primary source of Canon read noise is downstream. I don't think their per-pixel amplifiers add much noise, and their CDS units at each column are very good. I think it's the downstream amp and the ADC units that add most of the read noise.
Based on research I've found on future innovations for CMOS sensors, integration of the ADC into the sensor is a big step towards reducing read noise. Most other manufacturers have at least moved to some kind of highly parallel on-die ADC. Current research involves embedding ADC units into each pixel (well, currently, into each group of four pixels), which reduces read noise even further. I'm hoping Canon moved to an on-die ADC system (which they do have patents for). We won't be seeing any kind of per-pixel ADC any time soon, and I don't think we will until they figure out how to get CDS embedded into the pixels as well (currently CDS is per-column).
* TTTMNBSO: The term that must not be spoken of
chromophore said:jrista said:So this is a camera optimized for low ISO performance. Canon is talking about stronger CFA and all of that....(which, given that for color depth above 24 bits is basically indiscernible to the human eye, is a total waste of time IMO. The only potential benefit MIGHT be lower color noise, if they are more finely restricting the colors through each CFA...but color noise would also be reduced if they would reduce their darn read noise.)
But once again...utter silence about read noise and dynamic range. This is, clearly, a Low ISO part from Canon. If they had resolved their low ISO read noise issues...one would think they would be proclaiming it to the heavens. Why the silence? :\
I don't want to lose the last little bit of hope I have in Canon...
The point you make about the read noise being an overriding factor is a good one and I agree.
Given that the information provided thus far about the camera is very preliminary, and we know nothing about the technology in the sensor, much less seen any actual images from this camera, there are "glass half-empty" and "glass half-full" viewpoints:
Pessimist view: Increased color discrimination in this camera will be largely useless because Canon didn't improve the low ISO read noise. Canon did it because all they can do is implement incremental, tangential improvements in their sensor technology, and cannot compete against Sony.
Optimist view: Increased color discrimination in this camera was implemented precisely because Canon improved the low ISO read noise hence making the existing CFA performance inadequate, especially in post processing losses in color fidelity when bumping shadows by several stops.
Which one will turn out to be the case is unknowable at this time. But the answer will most definitely determine the success of this camera.
I agree, stronger CFA won't improve the perceived color quality. Not, at least, unstretched RAW images. It could improve color noise, which would affect the color quality of the shadows.
Improved color discrimination should reduce color crosstalk, which should reduce color noise. Color noise is an issue Canon cameras have had for a while, and it got worse when Canon weakened their CFA with the release of the 1D X. Strengthening the CFA may be doing nothing more than restoring it to what it was before the 1D X, I dunno. Regardless, the fact that Canon is doing anything that could reduce color noise is a good thing. I don't know if reducing color noise will improve tttmnbso*...I kind of doubt it, not if other sources of noise remain high. The primary source of Canon read noise is downstream. I don't think their per-pixel amplifiers add much noise, and their CDS units at each column are very good. I think it's the downstream amp and the ADC units that add most of the read noise.
Based on research I've found on future innovations for CMOS sensors, integration of the ADC into the sensor is a big step towards reducing read noise. Most other manufacturers have at least moved to some kind of highly parallel on-die ADC. Current research involves embedding ADC units into each pixel (well, currently, into each group of four pixels), which reduces read noise even further. I'm hoping Canon moved to an on-die ADC system (which they do have patents for). We won't be seeing any kind of per-pixel ADC any time soon, and I don't think we will until they figure out how to get CDS embedded into the pixels as well (currently CDS is per-column).
* TTTMNBSO: The term that must not be spoken of
Upvote
0