CCD vs. CMOS

Ian Stuart Forsyth · Jul 7, 2014

John Cafarella said:
Ian Stuart Forsyth wrote:

>After seeing how to make profiles for a specific camera and how calibration of the cameras raw >files with adobe DNG profile editor you can basically take any camera and fine tune the look of a >an raw file to look like any output you want.

I think you're absolutely correct on this Ian. However.......

I had a K10D, and It seems to me that there is a different tonal response from the later CMOS cameras, particularly in the dark-mid tones that gives a characteristic "look" to the pictures when minimal curves adjustment is done in raw conversion.

Within profile editor you can setup the tone curve to your liking, for instance here I have a linear profile

John Cafarella said:
I fully realise that different raw converters will give different results, and it all can be compensated for with careful manipulation, but with the K10D that "look" is just there with minimal work.

I think it would be educational to take photos of a stepped grey card with both cameras and analyse at the numbers in the RAW file.

You'd need to carefully set exposure on both so that white barely clips, then compare the numbers for each grey level (expressed as % so that there was no argument about 12bit and 14bit files).

It would be also interesting to see if any non linearities found are the same in each of the RGB channels.

Unfortunately I don't have a K10D any more.

--
John Cafarella
Melbourne, Australia

--
The Camera is only a tool, photography is deciding how to use it.
The hardest part about capturing wildlife is not the photographing portion; it’s getting them to sign a model release

DAVID MANZE · Jul 7, 2014

I thought we were talking about the CCD's vs CMOS's performance with RAW, the actual sensor performance.

If you include the SOTC Jpgs this says very little about the sensors performance and is more about Jpg engines.

The reason that MF cameras have hung on to CCDs for so long is their rendering of skin tones which in the studio is so important, none of them are using Jpgs.

To compare sensors you must eliminate Jpg processors and stick to RAW!

Ian Stuart Forsyth · Jul 7, 2014

DAVID MANZE said:
I thought we were talking about the CCD's vs CMOS's performance with RAW, the actual sensor performance.

If you include the SOTC Jpgs this says very little about the sensors performance and is more about Jpg engines.

The reason that MF cameras have hung on to CCDs for so long is their rendering of skin tones which in the studio is so important, none of them are using Jpgs.

No

no one produced a MF cmos till sony did that's why

DAVID MANZE said:
To compare sensors you must eliminate Jpg processors and stick to RAW!

There is nothing in raw (Its just counted photons) until we process them into a an output TIFF JPG then we get color and tone

awaldram · Jul 8, 2014

Ian Stuart Forsyth said:
DAVID MANZE said:

I thought we were talking about the CCD's vs CMOS's performance with RAW, the actual sensor performance.

If you include the SOTC Jpgs this says very little about the sensors performance and is more about Jpg engines.

The reason that MF cameras have hung on to CCDs for so long is their rendering of skin tones which in the studio is so important, none of them are using Jpgs.

Click to expand...

No

no one produced a MF cmos till sony did that's why

Agree 100% the only reason MF stayed CCD was lack of investment in R&D nothing related to any inherent perceived superiority of CCD.

MF sensor maker were busy going bust rather than development.

Ian Stuart Forsyth said:
Ian Stuart Forsyth said:

To compare sensors you must eliminate Jpg processors and stick to RAW!

Click to expand...

There is nothing in raw (Its just counted photons) until we process them into a an output TIFF JPG then we get color and tone

Again I agree RAW is not some magic it just numbers (bit counts) you can't even 'see' it as an image until processed so all this 'it in the RAW' is just impossible when it comes to color reproduction.

Now how much valid colour data , resolution, noise are RAW properties but how that data is represented to the viewer is the processing be that jpg / tiff or any other working image format.

Trevor G · Jul 8, 2014

Roland Karlsson said:
Of course.

There is no reason why CCD should get better colors than CMOS.

Are you 100% sure?

According to Illiah Borg the CFAs are changed from CCD to CMOS.

CCD CFAs are more dense, CMOS ones are more transparent, and so lose some of the colour range capabilities of CCD sensors with denser CFA arrays.

http://www.dpreview.com/forums/post/53762887

DAVID MANZE · Jul 8, 2014

awaldram said:
Ian Stuart Forsyth said:

DAVID MANZE said:

I thought we were talking about the CCD's vs CMOS's performance with RAW, the actual sensor performance.

If you include the SOTC Jpgs this says very little about the sensors performance and is more about Jpg engines.

The reason that MF cameras have hung on to CCDs for so long is their rendering of skin tones which in the studio is so important, none of them are using Jpgs.

Click to expand...

No

no one produced a MF cmos till sony did that's why

Click to expand...

That's true!

awaldram said:

awaldram said:

To compare sensors you must eliminate Jpg processors and stick to RAW!

Click to expand...

There is nothing in raw (Its just counted photons) until we process them into a an output TIFF JPG then we get color and tone

Click to expand...

Again I agree RAW is not some magic it just numbers (bit counts) you can't even 'see' it as an image until processed so all this 'it in the RAW' is just impossible when it comes to color reproduction.

awaldram said:

awaldram said:

--
Dave's clichés

Click to expand...

Click to expand...

--
My PPG
http://www.pentaxphotoga

--You are right about MF sensors, there were none!

To me I found the GX10/K10 on skin tones better in RAW than the K20 and K5 which came after, something about the graduation, but that was the only thing, dynamic range was too seriously lacking; too many burnt out whites here where the sun is ferocious.

Jpg engines are however variable in their treatment of images, the Samson GX10 (version K10) was awful, the K20 was miles better as was the K5, some love the treatment of the K3s Jpgs, others don't so much, other makes also get mixed reviews, most reviews finish by saying "to get the most out of the camera use RAWs". I judge images by their RAWs.

That's my call others may see it differently!

Dave's clichés

gaddigad · Jul 8, 2014

audiobomber said:
The K-3 images are the same, except for WB. The OOC K-3 image is clearly too cool, the K100DS and adjusted K-3 are more natural and quite close to each other, certainly closer than the OOC K-3 image.

audiobomber said:

Or maybe not, it's hard to draw conclusions from those shots given the test conditions.

Click to expand...

The differences are plain as day to me. The only unsatisfactory image is the OOC K-3 with its excessive blue cast.

K-7 to k-3 have been producing a cooler colour gamut than the previous Pentaxes, including k20d and possibly the k-x, as well as the Nikons and Sonys using the same sensors, which are obviously CMOS just as well. So I think it's not a CMOS thing, it's the colour profile built into the camera's image processing pipeline by some ingenious Hoya / Ricoh marketing "guru".

audiobomber · Jul 8, 2014

gaddigad said:
K-7 to k-3 have been producing a cooler colour gamut than the previous Pentaxes, including k20d and possibly the k-x, as well as the Nikons and Sonys using the same sensors, which are obviously CMOS just as well. So I think it's not a CMOS thing, it's the colour profile built into the camera's image processing pipeline by some ingenious Hoya / Ricoh marketing "guru".

The K-x has a definite blue cast, the K20D is considerably warmer, in between the old CCD and CMOS. It has the best daylight WB of any Pentax, IMO.

I don't know why the cool blue colour balance, but my Sony a6000 shows it too, so maybe it's a sensor thing? I have a suspicion they do it to wring better SNR numbers out of DXOMark. One thing that should be mentioned is that the later cameras are much more accurate in yellow light. The older bodies were way too yellow in tungsten and candle light. A warm cast can be appealing though, even if inaccurate.

audiobomber · Jul 8, 2014

DAVID MANZE said:
I judge images by their RAWs.

But you can't see a raw image, it's just ones and zeroes. You have to convert it to view it, and the conversion looks however the raw converter interprets it. The purest conversion of Pentax's intent would be via the supplied Silkypix software, because it reads all the camera settings. No other software does.

I used Pentax software for years. You cannot tell an OOC jpeg from a converted raw if you use Pentax software. They are identical. Other raw converters will look different from DCU, but not necessarily better or worse, it depends mostly on the skill of the person doing the conversion.

gaddigad · Jul 8, 2014

audiobomber said:
The K-x has a definite blue cast, the K20D is considerably warmer, in between the old CCD and CMOS. It has the best daylight WB of any Pentax, IMO.

I don't know why the cool blue colour balance, but my Sony a6000 shows it too, so maybe it's a sensor thing? I have a suspicion they do it to wring better SNR numbers out of DXOMark. One thing that should be mentioned is that the later cameras are much more accurate in yellow light. The older bodies were way too yellow in tungsten and candle light. A warm cast can be appealing though, even if inaccurate.

Dan, "white balance accuracy" usually refers to the jpeg engine used to process the raw file, although I have noticed that setting the white balance correctly (or otherwise) somehow miraculously affects my ability to process the raw file later.. At any rate, the colour gamut (or colour signature) of a particular body is not solely a generic function of the sensor (technology) used, as far as I understand. Have a look at the 645z review for a brief discussion of CCD vs. CMOS: http://blog.mingthein.com/2014/06/27/review-the-pentax-645z-part-i/

DAVID MANZE · Jul 8, 2014

audiobomber said:
DAVID MANZE said:

I judge images by their RAWs.

Click to expand...

But you can't see a raw image, it's just ones and zeroes. You have to convert it to view it, and the conversion looks however the raw converter interprets it. The purest conversion of Pentax's intent would be via the supplied Silkypix software, because it reads all the camera settings. No other software does.

I used Pentax software for years. You cannot tell an OOC jpeg from a converted raw if you use Pentax software. They are identical. Other raw converters will look different from DCU, but not necessarily better or worse, it depends mostly on the skill of the person doing the conversion.

--
Dan

--I said what I found from using Jpgs and RAW, I found big differences between them especially with the Samsung GX10 where the Jpgs were next to unusable, now I use uniquely RAW. One thing I don't want is to mimic Jpgs with RAW, it's the flexibility that I am looking for.

I personally didn't find Pentax's software very pleasing.

My comments were about skin tones so I cannot draw any conclusions from garden scenes with that amount detail, to me it was about graduations and rendition of simpler textures.

However CMOS has replaced CCD now and overall I'm very happy with it's advantages.

As I said that's my call others may think differently!

Dave's clichés

wchutt · Jul 8, 2014

The end result (storing electrical charge in a semiconductor potential well) is identical for CMOS and CCD sensors.

CCD sensors use older color-filter Bayer micro lenses. The Bayer rendering interpolation model assumes only red, blue and green light is present. The micro-lenses in older (CCD) sensors were designed to minimize the non-red, non-blue and non-green light frquencies that reached the semiconductor photodiodes.

Then something changed. The analog signal to noise ratio (a.k.a. high ISO performance) became a critical marketing tool. Photographers spent money based on signal to noise rather than color fidelity.

The result was brands were tempted to change micro-lens design. Micro lenses that pass a wider range of light frequencies deliver more signal to the sensor. The signal-to-noise ratio increases. Unfortunately a wider range of light frequencies means the Bayer model for color rendition is less appropriate. This compromise (more signal for less color fidelity) sold cameras.

The color-filter array micro-lense differences, not the semiconductor and micro-circuity design, are primarily responsible for differences in CCD and CMOS color rendition.

Chris Mak · Jul 8, 2014

wchutt said:
The end result (storing electrical charge in a semiconductor potential well) is identical for CMOS and CCD sensors.

CCD sensors use older color-filter Bayer micro lenses. The Bayer rendering interpolation model assumes only red, blue and green light is present. The micro-lenses in older (CCD) sensors were designed to minimize the non-red, non-blue and non-green light frquencies that reached the semiconductor photodiodes.

Then something changed. The analog signal to noise ratio (a.k.a. high ISO performance) became a critical marketing tool. Photographers spent money based on signal to noise rather than color fidelity.

The result was brands were tempted to change micro-lens design. Micro lenses that pass a wider range of light frequencies deliver more signal to the sensor. The signal-to-noise ratio increases. Unfortunately a wider range of light frequencies means the Bayer model for color rendition is less appropriate. This compromise (more signal for less color fidelity) sold cameras.

The color-filter array micro-lense differences, not the semiconductor and micro-circuity design, are primarily responsible for differences in CCD and CMOS color rendition.

Common sense would dictate that a signal with a wider range of light that reaches the sensor, is in itself a good thing. But as I understand from these discussions, it isn't because the bayer model has shortcomings and benefits from a signal with clear RGB separation.

Perhaps it is time that the new sensors that are rumored, and which do away with the Bayer RGB model, come to market...

Bmoon · Jul 8, 2014

DAVID MANZE said:
awaldram said:

Ian Stuart Forsyth said:

DAVID MANZE said:

I thought we were talking about the CCD's vs CMOS's performance with RAW, the actual sensor performance.

If you include the SOTC Jpgs this says very little about the sensors performance and is more about Jpg engines.

The reason that MF cameras have hung on to CCDs for so long is their rendering of skin tones which in the studio is so important, none of them are using Jpgs.

Click to expand...

No

no one produced a MF cmos till sony did that's why

Click to expand...

That's true!

DAVID MANZE said:

DAVID MANZE said:

To compare sensors you must eliminate Jpg processors and stick to RAW!

Click to expand...

There is nothing in raw (Its just counted photons) until we process them into a an output TIFF JPG then we get color and tone

Click to expand...

Again I agree RAW is not some magic it just numbers (bit counts) you can't even 'see' it as an image until processed so all this 'it in the RAW' is just impossible when it comes to color reproduction.

Click to expand...

Brad99 · Jul 8, 2014

Another difference is the tone curve, CCD sensors gave a more linear response at the highlight end, CMOS a more rounded response. CMOS is therefore better at preserving highlights but if your subject's DR is within the capability of a CCD sensor without clipping, it will look better, punchier and more life-like than CMOS as it's not compressed. Colours can be tweaked (+2 Amber on a K-30 sorts that out), but DR compression is a bit of a 2-edged sword, it's both good (less blowout) and bad (less visible DR).

Chris Mak · Jul 8, 2014

Brad99 said:
Another difference is the tone curve, CCD sensors gave a more linear response at the highlight end, CMOS a more rounded response. CMOS is therefore better at preserving highlights but if your subject's DR is within the capability of a CCD sensor without clipping, it will look better, punchier and more life-like than CMOS as it's not compressed. Colours can be tweaked (+2 Amber on a K-30 sorts that out), but DR compression is a bit of a 2-edged sword, it's both good (less blowout) and bad (less visible DR).

This compressed DR, as I remember, was very visible on the Pentax K5/K5IIs at iso 80, that features a DR of over 14,1 stops per DxO at this setting, but the images always needed quite a bit of work from raw to give them a natural look. At default they could look quite compressed, making me prefer the iso 100 setting.

I recently got a Sony A7r, that has the same 14,1 stops of DR at the lowest native iso setting (100 iso on the A7r), but what struck me immediately, was that the images look noticeably less compressed while having this very wide DR of 14,1 stops. The images retain natural contrast (visible DR) better while giving the benefit of the wide DR, no doubt this is due to the (excellent Sony) FF sensor. As the sensor has the same pixel pitch, this was quite a (pleasant) surprise. It made me wonder if the Apsc sensors are compromised more to get them to perform at this level regarding DR and noise, or that the Sony A7r is simply a newer and better sensor than the Apsc sensor used in the K5 series. Whatever the reason, this is another compelling argument for a Pentax FF camera with a similar sensor.

Chris

John Cafarella · Jul 8, 2014

wchutt wrote:

>The color-filter array micro-lense differences, not the semiconductor and micro-circuity design, >are primarily responsible for differences in CCD and CMOS color rendition.

Well, that's a very interesting bit of info... It provides good explanation for what people are observiing.

wchutt · Jul 11, 2014

There is nothing wrong with the Bayer model.

The problem (color tonality/rendering) is caused by the temptation to modify the data until it is just barely compatible with the Bayer interpolation mathematical model.

Increasing the color-frequency range passed by the R,G and B microlenses means the data and the model no longer directly correspond to each other. But more light reaches the semiconductor potential well so the signal to noise ratio increases. Different brands make different compromises. Before the high ISO performance (signal to noise ratio) became an obsession, color reproduction had a higher priority.

At this point in time there is nothing special about CCD semiconductor characteristics whatsoever for conventional photography.

paulkienitz · Jul 11, 2014

zakaria said:
Nice ..but I can't fool my eyes. I think the last pics is more saturated. The output of k 200 is natural and film like. The ccd sensors AS I THINK rendering the skin colors better than cmos sensors. This is why MF cameras LIKE THE CCD.

Maybe some CCDs have this magic, but not any that I used. I thought my K10D was terrible at skin tones -- to me, that was one of the camera's worst weak points. It sometimes made people in full sun look like they were under an office fluorescent light. The K-3 may turn the sky a bit greenish, but to me the OOC color is a lot more natural. This is noticeable not just on faces but also on green plants, another area where the K10D had issues. Other CCD cameras I've had any time to use (admittedly not many, and not modern) were worse. So as far as I'm concerned, CCD sensors can go build sandcastles on the freeway.

Chris Mak · Jul 11, 2014

wchutt said:
There is nothing wrong with the Bayer model.

The problem (color tonality/rendering) is caused by the temptation to modify the data until it is just barely compatible with the Bayer interpolation mathematical model.

Increasing the color-frequency range passed by the R,G and B microlenses means the data and the model no longer directly correspond to each other. But more light reaches the semiconductor potential well so the signal to noise ratio increases. Different brands make different compromises. Before the high ISO performance (signal to noise ratio) became an obsession, color reproduction had a higher priority.

At this point in time there is nothing special about CCD semiconductor characteristics whatsoever for conventional photography.

I understand that, and I am not at all a high ISO shooter, nor do I use live view at all or shoot video with my K3. I would have greatly preferred the emphasis on color reproduction instead of good high ISO performance. Still, the fact that the Bayer model relies on limiting the light signal that falls through the sensor micro lenses in frequency and consequent signal to noise ratio, comes across as an inherent limitation of the Bayer technology.

Chris

CCD vs. CMOS

Senior Member

Veteran Member

Senior Member

Forum Pro

Veteran Member

Veteran Member

Senior Member

Veteran Member

Veteran Member

Senior Member

Veteran Member

wchutt

Guest

Senior Member

Forum Enthusiast

Senior Member

Senior Member

Leading Member

wchutt

Guest

Veteran Member

Senior Member

Similar threads

Keyboard shortcuts