If CCD rendering can be achieved with CMOS where are the examples?

[jonoslack]

What a wonderful thread, with so many fine and intelligent arguments.
I don't have much to contribute, except for two things
Firstly, it takes time to get to grips with a new sensor and its colour (be it CMOS or CCDS), and as one gets to grips with it one learns to maximise its advantages and minimise its disadvantages.
Secondly, with modern sensors, differences between sensors are never a hundredth as important as differences between style and content.

Angels dancing on the heads of pins.

Still fascinating though

 

[fotografz]

The definitive sign that this thread has "Jumped The Shark":ROTFL:

- Marc

 

[Malina DZ]

Well, I tried and failed to match the scene colors between both sensor types in either Sony IDC or LR5. Very different results between these converters by the way.
I still shoot with Sony a100 (CCD) and a850 (CMOS). I'd prefer a100 colors for portraits when developing RAW in LR since red channel is more accurate in a100 compared to an over saturated one in a850 (Adobe Standard profile). I barely feel any need in color adjustment for skin tones with a100, but almost every single file from a850 requires it.
I have neither CCD nor CMOS medium format experience though. Sony's MF sensor looks promising based on the 645Z RAW files. I'm surprised there're no direct CCD vs CMOS comparisons between Phase or Hasselblad backs with RAW files available for download...

If anyone has time to play with CCD/CMOS color balancing, here are two RAW files, shot with the same lens, on a100 and a850 in a crop mode. Click on the image to download a RAW.


CCD



CMOS

I think we need to bring more image samples to this discussion.

 

[jonoslack]

Well, I tried and failed to match the scene colors between both sensor types in either Sony IDC or LR5. Very different results between these converters by the way.


If anyone has time to play with CCD/CMOS color balancing, here are two RAW files, shot with the same lens, on a100 and a850 in a crop mode.[/url]

Well
I don't think it's very easy to match the colour between any two cameras. Whatever their sensor type; so that your inability to match the colours between these two says nothing about the difference between CCD and CMOS just that those cameras are different.

 

[chrismuc]

... okay we need so see sample: Enclosed four pictures, some CCD, some CMOS. Now, which is what, what's your guess? (only two cameras/sensors)
I tried to match as well as possible the skin tone in acr, I took the pics within half an hour at quite the same light.
I kept the files in ProPhotoRGB for good gamut/skin tone reproduction, be aware that your browser must be able to show this profile correctly.

 

[ErikKaffehr]

What I have seen is that the more controlled the conditions are the less differences I see.

Generate a colour profile for each sensor, include a grey card for white balance and little difference I will see.

Best regards
Erik

Well
I don't think it's very easy to match the colour between any two cameras. Whatever their sensor type; so that your inability to match the colours between these two says nothing about the difference between CCD and CMOS just that those cameras are different.

 

[chrismuc]

Actually I am wondering why the sensor type should have anything to do with the color precision/quality of the picture because both CCD and CMOS sensors only "see" the amount of light, no colors. The colors come from the Bayer pattern RGB filter in front of the sensor, so that filter should be the only color effecting part of the sensor, or am I wrong?

Btw, samples 1+3 = Fuji X-T1 + XF 56f1.2 @ f2, samples 2+3 = Contax 645 + IQ180 + 140f2.8 @*f2.8.

 

[ErikKaffehr]

Thanks for the nice demo and the info.

Best regards
Erik

 

[Jay Emm]

There's a word missing in this thread. The C-word actually.

I've never had a CUSTOMER won or lost (to my knowledge) on my use of CCD over CMOS, or vice versa.

 

[Wayne Fox]

Actually I am wondering why the sensor type should have anything to do with the color precision/quality of the picture because both CCD and CMOS sensors only "see" the amount of light, no colors. The colors come from the Bayer pattern RGB filter in front of the sensor, so that filter should be the only color effecting part of the sensor, or am I wrong?

Btw, samples 1+3 = Fuji X-T1 + XF 56f1.2 @ f2, samples 2+3 = Contax 645 + IQ180 + 140f2.8 @*f2.8.

I think it's a little more complicated, because the dyes used to create the filters in front of each sensel affect rendering. So yes, the basic electronics just measure photon count, but which photon's get through which filter can be quite different between sensors based on the dyes and the density of the dyes.

(prob not a ccd vs cmos thing though)

 

[ErikKaffehr]

Hi,

That is what Chrismuc also said. But, it is (quite) a bit more complex than that.

The colours from the sensor are not directly usable, but will be converted to normal RGB colours using some math, basically a matrix multiplication.

On top of that matrix multiplication there will one or more profiles will be applied and those profiles may have significant tuning for pleasantness or "memory colours". The profiles may be matrix- or LUT-based but often a combination of both.

So, the colour rendition is decided by a series of matrix multiplications, the first one (colour conversion matrix) is actually given by the sensor while the rest is visual adaption and tuning.

By the way, DxO presents the colour conversion matrices for each sensor. They also calculate something called System Metamerism Index, which essentially measures how well the sensor can reproduce the colour fields of ColorChecker card. That figure is not a part of the DxO rating, however.

Just to mention, the ColorChecker card is not just a standard used with 16 arbitrary chosen colour fields, but those fields are chosen with great care. The two skin patches, for instance, are said to show very similar spectral response to human skin. So those fields would be a good representation of human skin under almost any illuminant.

Best regards
Erik

I think it's a little more complicated, because the dyes used to create the filters in front of each sensel affect rendering. So yes, the basic electronics just measure photon count, but which photon's get through which filter can be quite different between sensors based on the dyes and the density of the dyes.

(prob not a ccd vs cmos thing though)

 

[tjv]

While it's true that the CMOS image is far lower when it comes to noise, something about the noise structure on the IQ260 image is quite appealing to my eyes. Maybe this is because I'm used to shooting and scanning film at high resolution, so my tolerance for grain is higher? I guess my feeling is that it seems less "plastic" or something? Having said all this, I'm in the market for a CMOS back so it's a moot point for me.
Anyway, :worthless:

This thread on LuLa discusses a comparison made by Doug Peterson just when the IQ-250 arrived. Doug compared IQ-250 (CMOS) with IQ-260 and IQ-280 (CCD). The best comparison between CMOS and CCD I have seen so far.

This image tells the story on shadow noise (IQ-260 to the left)

Best regards
Erik

 

[ErikKaffehr]

Hi,

It is an interesting point. On the other hand I would say that what we see on noise in the shadows is very different from mid tones. Mid tones are dominated by shot noise (which is natural variation of light) so I would expect midtones be better on larger formats (that capture more light). So comparing two ETTR exposures at base ISO I would be pretty sure the larger format backs would have a smoother tone in midtones compared to an 1.3X crop back.

Best regards
Erik

While it's true that the CMOS image is far lower when it comes to noise, something about the noise structure on the IQ260 image is quite appealing to my eyes. Maybe this is because I'm used to shooting and scanning film at high resolution, so my tolerance for grain is higher? I guess my feeling is that it seems less "plastic" or something? Having said all this, I'm in the market for a CMOS back so it's a moot point for me.
Anyway, :worthless:

 

[archivue]

While it's true that the CMOS image is far lower when it comes to noise, something about the noise structure on the IQ260 image is quite appealing to my eyes. Maybe this is because I'm used to shooting and scanning film at high resolution, so my tolerance for grain is higher? I guess my feeling is that it seems less "plastic" or something? Having said all this, I'm in the market for a CMOS back so it's a moot point for me.
Anyway, :worthless:

i feel the same, but playing in C1 with structure, grain… you can have a better look at the end from CMOS.

 

[tjv]

Yeah, I guess that's true. I've never really thought about adding grain, except for gritty black and white work. Does the C1 grain add a film like coloured grain pattern, or is basically just luminance grain?

 

[fotografz]

There's a word missing in this thread. The C-word actually.

I've never had a CUSTOMER won or lost (to my knowledge) on my use of CCD over CMOS, or vice versa.

This is 100% true, IF the customer is allowed to be the arbitrator of your work … which is a personal decision we each make.

I once evoked that argument to my boss regarding some advertising creative, to which he replied, "You wouldn't hire any client as a Junior Art Director, so why would you surrender your taste to them?"

Never forgot that.

- Marc

 

[emr]

I don't really have much to say regarding directly the question of CCD vs. CMOS, but I find it interesting that someone suggested the difference perceived by some might be due to a different noise pattern. I work in the field of medical imaging myself and am involved in the use of many diagnostic scanners of different brands. The image quality varies quite a lot from one piece of equipment to the next one, often in a way that is hard to explain verbally. One particular CT scanner seems to have much more noise in its images with a comparable radiation dose ("exposure") and equivalent image processing settings than a scanner by another brand. We discussed the IQ aspect with our physicists and according to their physics-magic-super-analysis (which I fail to really understand), the images visually appearing to have more noise actually had less. So the processing from "RAW" data does actually affect the perceived noise pattern quite a bit and can give subjectively quite different results, measurable or not. I guess this applies to visible light imaging as well.

 

[tjv]

Honestly, I think you're on to something. I don't personally subscribe to one technology being better than the other – there are too many variables in how people apply the technology and how people use it – but I do think that, at least at base ISO, the noise pattern and subtle variation of texture that that noise adds makes, to me at least, CCD seem more natural. DR is kind of a red herring, I reckon. More headroom is always welcome, but the processing each individual applies to their files is what makes the real difference. I shot E6 film for years and 12 stops of DR is more than enough for me, but for other types of work I'm sure more is welcome. Problem is, you've got to have a personal vision to make use of it. Without vision and the practical skills to edit, all your photos will look average.

(This is all discounting how close the colours and contrast of the RAW file look "out of the box" to how you want your images to look. Some files take far more work to look "good" than others, for sure. And some people actually like (shock horror) the idea of shooting at ISO's higher than 400... CCD is generally not ideal here.)

I don't really have much to say regarding directly the question of CCD vs. CMOS, but I find it interesting that someone suggested the difference perceived by some might be due to a different noise pattern. I work in the field of medical imaging myself and am involved in the use of many diagnostic scanners of different brands. The image quality varies quite a lot from one piece of equipment to the next one, often in a way that is hard to explain verbally. One particular CT scanner seems to have much more noise in its images with a comparable radiation dose ("exposure") and equivalent image processing settings than a scanner by another brand. We discussed the IQ aspect with our physicists and according to their physics-magic-super-analysis (which I fail to really understand), the images visually appearing to have more noise actually had less. So the processing from "RAW" data does actually affect the perceived noise pattern quite a bit and can give subjectively quite different results, measurable or not. I guess this applies to visible light imaging as well.

 

[synn]

The noise characteristics of CCD and CMOS are certainly quite different. My CCD files have a more "random" noise structure while the sony CMOS files have a very uniform noise pattern. I guess that uniformity does play a part in how plasticky and artificial those files look out of the box.

The other thing I have noticed is how CCD highlights "sparkle" in a way CMOS highlights don't. The latter often gives speculars that are spread out like light hitting on a sheet of cellophane.

You can tighten it up a bit with tone curves, but it's not the same.

You can see this phenomenon in portraits, especially the catchlights and speculars on the cheek and hair.

Lastly, most CMOS images, no matter what color profiles you use, tend to bunch similar colors together. You can see this in foliage and on cheeks, where the blush blends into the foundation. Far more shades to make out in CCD files.

None of this is particularly observable in lab test shots of color passports or charts. Which is why repeatedly trying to break it down to a science fails to give any solid answers.

IMO.

 

[PeterA]

yeah lets post some examples of something from someone showing something on someone's monitor which is different to every other someone's monitor and then lets all pretend that we see or don't see something ...

lulz