What happened to "CCD is better than CMOS"?!

[bjanes]

Bill,

 The question here isn't whether the abilities of CCD are different from CMOS, it is whether the actually produced MF cameras are different.

 At the moment there is one manufacturer - SONY - of such sensors. One instance isn't guaranteed perfect just because Eric Fossum says so. And the inventor of CMOS sensors would be expected to thing well of CMOS optical sensors, every inventor loves his child most.

Edmund

Edmund,

Your points are well taken, but Dr. Fossum does not state that his invention is better, but merely that silicon is silicon and CFA filters are CFA filters, which is just what Eric also stated. In judging the professor's credibility one must take into account that he likely knows more about solid state imagers than anyone posting on this forum.

Bill

[eronald]

Some of us have to fight a rearguard battle
I guess Cmos got its bad rep from excessive noise reduction and sloppy CFAs used to boost ISO.

Edmund

Edmund,

Your points are well taken, but Dr. Fossum does not state that his invention is better, but merely that silicon is silicon and CFA filters are CFA filters, which is just what Eric also stated. In judging the professor's credibility one must take into account that he likely knows more about solid state imagers than anyone posting on this forum.

Bill

[Christoph B.]

Perhaps you could supply us with some of your references to back up your assertions. One caveat in comparing CCD vs CMOS designs is the date of the comparison. CMOS designs have been dramatically improved recently, and the older literature comparing the two types of sensors may not be current.

For example, here is an article from 2011 showing how CMOS has gained with respect to CCD. Here is another from 2009. This is now 2016. Note that these articles are from the scientific imaging community where CCDs have long predominated over CMOS.

Bill

The CMOS sensors used in consumer cameras aren't sCMOS - and in the first link you provided it even shows a QE graph of sCMOS saying that it approaches the capability of CCD sensors - but it doesn't surpass it. ("FIGURE 1. The quantum efficiency curve of sCMOS cameras is now much closer to that of CCDs than that of the CMOS cameras of just a few years ago.")

Here it states that most CMOS sensors have their max. sensitivity in the NIR-spectrum while most CCD sensors have their max. sensitivity in the visible spectrum (550 nm): https://de.wikipedia.org/wiki/Active_Pixel_Sensor#Unterschiede_zu_CCD-Sensoren (sorry it's in german)
There are tons of infos on astrophotography and infrared photography forums and pages about CCD sensors and their sensitivities as well as CMOS sensors that support that statement.

Keep in mind that I was only trying to find an explanation for the problems with the weak greens and the over-saturated reds in the images of the CMOS backs and the discrepancy in the comparison of the Canon CMOS and the 'old' PhaseOne back - nothing more, nothing less. The general tendency of CMOS sensors to have a low overall QE and their max. sensitivity in the NIR spectrum may be part of the explanation.

If anyone has a better explanation I'm all ears!

[eronald]

Chris,

 When you see a green differentiation issue, it is IMHO a CFA issue due to, paradoxically, a bad quality -too wide, too noisy - red channel. As a result, there is not enough information to separate shades of green (clorophyll absorbs a lot of red). If I am right, the issues will be exacerbated under bluish or greenish light, ie cloudy days, blue sky with no sun, or forest canopies.

All of this is conjecture, but you can do an experiment: Grab Photoshop, and zero or level adjust the channels one by one, and see where the information is; or else just invert the image, that also often gives you a feel for what is going on.

I think in the end it is simply easier to grab a camera one likes than to try and even understand the tech. The same is true of lens depth of field, "snap" and bokeh: They are in the end subjective, and a "bad" old lens can easily surpass a nice new one in our affections.

Edmund

The CMOS sensors used in consumer cameras aren't sCMOS - and in the first link you provided it even shows a QE graph of sCMOS saying that it approaches the capability of CCD sensors - but it doesn't surpass it. ("FIGURE 1. The quantum efficiency curve of sCMOS cameras is now much closer to that of CCDs than that of the CMOS cameras of just a few years ago.")

Here it states that most CMOS sensors have their max. sensitivity in the NIR-spectrum while most CCD sensors have their max. sensitivity in the visible spectrum (550 nm): https://de.wikipedia.org/wiki/Active_Pixel_Sensor#Unterschiede_zu_CCD-Sensoren (sorry it's in german)
There are tons of infos on astrophotography and infrared photography forums and pages about CCD sensors and their sensitivities as well as CMOS sensors that support that statement.

Keep in mind that I was only trying to find an explanation for the problems with the weak greens and the over-saturated reds in the images of the CMOS backs and the discrepancy in the comparison of the Canon CMOS and the 'old' PhaseOne back - nothing more, nothing less. The general tendency of CMOS sensors to have a low overall QE and their max. sensitivity in the NIR spectrum may be part of the explanation.

If anyone has a better explanation I'm all ears!

[synn]

Christoph:

Welcome to Lula where the numbers are everything and artistry doesn't matter.
You'll get used to it, eventually...

[Bart_van_der_Wolf]

The question here isn't whether the abilities of CCD are different from CMOS, it is whether the actually produced MF cameras are different.

But Edmund, this is exactly what people are still confusing (and I'm amazed they do). Silicon is Silicon, with some relatively minor doping differences. The color sensitivity is more or less carved in, uhm silicon. The real differences come from the added hardware (filter stack and CFA) and software (profiling and demosaicing). So any comparison is flawed if those different components are not normalized.

Manufacturers may make some design choices, and some of those can influence the color response, and perhaps things like the suitability for lens shifts. But things like plasticky looks and green separation are mostly bogus, and just the result of (mostly) profiling, based on the supplied CFA data. You don't have to take my word for it, just listen to the inventor of the CMOS image sensor, Dr. Eric Fossum who's comment about color sensitivity was referenced before by Bill.

At the moment there is one manufacturer - SONY - of such sensors. One instance isn't guaranteed perfect just because Eric Fossum says so. And the inventor of CMOS sensors would be expected to thing well of CMOS optical sensors, every inventor loves his child most.

This has nothing to do with a 'not invented here' syndrome, it's just technological progress. That's why virtually everybody (not just Dr. Fossum) is switching to CMOS image sensors once they have milked their investment in the old technology. Lower power consumption, the possibilities of Live View, on chip PDAF, you name it. Having transistors per photosite has many benefits and only a few drawbacks, and it has little to do with color. Besides, Dr. Fossum is way beyond mainstream CMOS image sensors right now, he is inventing very new technology, based on sub-micron JOTS (but that's something for another discussion).

The only area I can think of that could currently benefit some from CCD is Astronomy, because large CCD photosites are slightly less susceptible to damage by gamma radiation, although their power consumption is concern. Color is not the issue, they just calibrate their sensors for multiple spectral bands (sort of profiling) and use diffraction gratings or filter wheels instead of on chip CFAs.

Cheers,
Bart

[Christoph B.]

Christoph:

Welcome to Lula where the numbers are everything and artistry doesn't matter.
You'll get used to it, eventually...

  well it's a thread about numbers and technical stuff anyway

Besides to me the artistic and technical aspects of photography are both important, you need both to take good photos with a solid quality.

I think in the end it is simply easier to grab a camera one likes than to try and even understand the tech. The same is true of lens depth of field, "snap" and bokeh: They are in the end subjective, and a "bad" old lens can easily surpass a nice new one in our affections.

True! I'm only worries that there might be an issue with the colours and PhaseOne and all the other camera producers will stop using CCD altogether in favour of the new CMOS chips as a point of sale(then again consequentially the 80MP CCDs will be much cheaper and available on the 2nd hand market, so there's an upside to all that as well... ).

If one sensor producer takes over the whole MFDB sector that might not be in the best interest for everyone involved - and Sony is about to do that.

[mjrichardson]

But Edmund, this is exactly what people are still confusing (and I'm amazed they do). Silicon is Silicon, with some relatively minor doping differences. The color sensitivity is more or less carved in, uhm silicon. The real differences come from the added hardware (filter stack and CFA) and software (profiling and demosaicing). So any comparison is flawed if those different components are not normalized.

With all due respect, isn't this exactly the point? We cannot buy sensors, whether they be cmos or ccd in order to build our own cameras with normalized equipment to see which is better/worse/different/the same, we buy cameras that have a look to them, regardless of how that is achieved by the manufacturer. A photographer can pick up a piece of equipment and shoot with it and look at the end result he can get from all of the components, factors, hardware and software and decide for himself if that combination works for him. I don't care so much about the technical aspects of individual items in the chain, just the shots produced and how they differ.

I have 2 cameras from the same manufacturer that are ccd and cmos and the resulting images are different and react in different ways to processing producing a different final image, whether the technical theory backs that up or not is largely irrelevant when looking at the image, surely?

I am as technical as I need to be to get the camera to perform how I want it to and respect fully that the details are just as important to other people but saying that silicon is silicon seems to me to miss the point when we don't use silcon alone, we use cameras.

Just my opinion of course!

Mat

[Bart_van_der_Wolf]

With all due respect, isn't this exactly the point? We cannot buy sensors, whether they be cmos or ccd in order to build our own cameras with normalized equipment to see which is better/worse/different/the same, we buy cameras that have a look to them, regardless of how that is achieved by the manufacturer. A photographer can pick up a piece of equipment and shoot with it and look at the end result he can get from all of the components, factors, hardware and software and decide for himself if that combination works for him. I don't care so much about the technical aspects of individual items in the chain, just the shots produced and how they differ.

I have 2 cameras from the same manufacturer that are ccd and cmos and the resulting images are different and react in different ways to processing producing a different final image, whether the technical theory backs that up or not is largely irrelevant when looking at the image, surely?

Hi Mat,

As long as you do not attribute the differences to CMOS versus CCD, no problem. These are different technologies that allow different possibilities, but color reproduction has little to do with it.

Color reproduction is mostly a function of other components in the image chain, like CFA and profiling.

So basing one's choice of tools on the color reproduction differences by CCD or CMOS is totally misguided (just like the so-called 16-stops dynamic range of CCDs in MFDBs, in general CMOS has higher DR capability than CMOS, not the other way around). Start with good profiling, since the hardware is more or less a given. Most likely there will be almost no observable difference, so you can base the choice of equipment on other requirements that allow to get the shot or not.

Cheers,
Bart

[mjrichardson]

Hi Bart

Thanks for the response, so now I understand you guys a little more, I presume I can say that I prefer the images I get from my camera that happens to use a ccd sensor, due to the variances in the pipeline after the silicon and like the images from my camera with a cmos sensor less for the same reasons, but I cannot say I prefer ccd over cmos?

I honestly believe that for a large proportion of photographers, holding 2 cameras with the different sensors, they would say I prefer the ccd or I prefer the cmos because that is how they differentiate them. I can understand that in reality we are preferring what happens around and after the silicon does its bit but for me, I cannot say that the output is the same and regardless of how, I prefer the camera I own that uses a ccd, I'm pretty basic like that!

Mat

[ErikKaffehr]

Hi Bart,

I guess that I got involved in this because there were a lot of statements about MFD that made little sense. The most obvious one was the promise of great DR in combination with very limited high ISO capability.

So that started me digging into these issues. In the end I even got myself a used P45+, in part of wanting to find out and in part because I happened to have some nice Hasselblad lenses and wanted to utilise a significant part of the image circle.

I have been told that I needed to use Capture One to make use the P45+, and it sort of upset me. I buy a back not a workflow, it is my images after all. Also, at that time Capture One had very limited tone mapping capability, needed to tame high DR. A lot of funny things like "film curve", exposure bias. Also, each time I put a memory card into my computer it popped up. I really hated it. Now I feel it is a very good raw converter, which I still don't like…

Now, getting back to that stuff, I really found that theory was true. The P45+ performed exactly as I would have expected. More pixels than my Sony A900, a lot more Moiré. A bit noisy shadows, but mostly no issue for me. What I found out that I liked to shoot with the Hasselblad V/P45+ combo. For two years it was my most used camera, but just one image I took with it made it to the wall and none to the exhibition floor.

Now, I am shooting an A7rII. It gives me a lot of flexibility, the same amount of pixels as the P45+ and gives me Tilt&Shift capability with all my Hasselblad lenses. Yes, the Flexbody also gave me T&S, but I did not feel it was usable in the field.

A couple of well known British photographers, Tim Parkin (who publishes OnLandscape) and Joe Cornish have suggested that the P45+ had really bad reproduction of greens and I did some digging into that. I wanted to find out how much of that is depending on profiles and how much on sensor. My findings were not very clear. Tim Parkin suggested that I write an article on the issue, but the effort sort of ran out, as I could not make any conclusion. A draft of that article is here: http://echophoto.dnsalias.net/ekr/Articles/OLS_OnColor/OnColor.pdf

In recent time, Anders Torger developed a tool called DCamProf and really found that colour profiles were far more important than sensor characteristics. Anders has also demonstrated that Phase One's file format was actually 14 bit, 16-bitness was created by a Logical Shift Left Twice operation. A perfectly good engineering choice but a very misleading marketing lie.

Best regards
Erik

Hi Mat,

As long as you do not attribute the differences to CMOS versus CCD, no problem. These are different technologies that allow different possibilities, but color reproduction has little to do with it.

Color reproduction is mostly a function of other components in the image chain, like CFA and profiling.

So basing one's choice of tools on the color reproduction differences by CCD or CMOS is totally misguided (just like the so-called 16-stops dynamic range of CCDs in MFDBs, in general CMOS has higher DR capability than CMOS, not the other way around). Start with good profiling, since the hardware is more or less a given. Most likely there will be almost no observable difference, so you can base the choice of equipment on other requirements that allow to get the shot or not.

Cheers,
Bart

[Bart_van_der_Wolf]

Hi Bart

Thanks for the response, so now I understand you guys a little more, I presume I can say that I prefer the images I get from my camera that happens to use a ccd sensor, due to the variances in the pipeline after the silicon and like the images from my camera with a cmos sensor less for the same reasons, but I cannot say I prefer ccd over cmos?

Hi Mat,

You can say what you want, but it is indeed the pipeline after the silicon that makes most of the differences in color reproduction.

I honestly believe that for a large proportion of photographers, holding 2 cameras with the different sensors, they would say I prefer the ccd or I prefer the cmos because that is how they differentiate them.

Yes, I understand that, but it may lead to the wrong purchasing desision. Maybe, with a different profile, the more capable camera happens to be a CMOS device. If e.g. tethering, or high ISO, or low noise and high DR, are important, then CMOS would also be a better choice, especially knowing that the color differences are most likely created by something else.

I can understand that in reality we are preferring what happens around and after the silicon does its bit but for me, I cannot say that the output is the same and regardless of how, I prefer the camera I own that uses a ccd, I'm pretty basic like that!

That's fine, as long as you do not also need some of the other benefits that a different camera has to offer. If you reject a camera only based on production method, based on a myth of better color and DR, then you would be selling yourself short. It are the objective features that mattter, and one can only hope that they are well understood.

Cheers,
Bart

[sgilbert]

"In judging the professor's credibility one must take into account that he likely knows more about solid state imagers than anyone posting on this forum."

I'll bet there are three or four regular posters here who would disagree.

[ErikKaffehr]

Hi,

I don't know. I guess we can all look up up Dr. Fossum's publications and other merits. I obviously don't know who you think would disagree so I cannot have any views on their merits.

Of the frequent posters here, I know that Bart van der Wolff was working for a company called Kodak Eastman in a customer advisory role regarding image acquisition. I am impressed with his knowledge and deeply thankful for him sharing his experience.

I am also in debt to Jim Kasson, Anders Torvalds, Jack Hogan, Emil J. Martinec and many others for sharing their knowledge and experience.

Best regards
Erik

"In judging the professor's credibility one must take into account that he likely knows more about solid state imagers than anyone posting on this forum."

I'll bet there are three or four regular posters here who would disagree.

[bjanes]

Hi,

I don't know. I guess we can all look up up Dr. Fossum's publications and other merits. I obviously don't know who you think would disagree so I cannot have any views on their merits.

Of the frequent posters here, I know that Bart van der Wolff was working for a company called Kodak Eastman in a customer advisory role regarding image acquisition. I am impressed with his knowledge and deeply thankful for him sharing his experience.

I am also in debt to Jim Kasson, Anders Torvalds, Jack Hogan, Emil J. Martinez [sic] and many others for sharing their knowledge and experience.

Erik,

Yes we do have many highly knowledgeable contributors (yourself included) to this forum and those you mention are near the top of the list, and I hope that I have not offended them. However, none of them are in disagreement with Dr. Fossum. Perhaps I should not have said that Fossum knows more about digital sensors than anyone posting here, but that was a polite way of avoiding naming the names of the less knowledgeable contributors.  By the way Emil's last name is Martinec; unfortunately, he has not posted that much recently and probably has returned to his day job of string theory and particle physics.

Regards,

Bill

[ErikKaffehr]

Hi Bill,

My intention was merely to mention some of the great contributors. We have quite a lot of very knowledgeable folks sharing their experience, and I felt it was appropriate to remind the forums of this.

No, I don't think you have offended any one by any means. Just wish that all other posters were a bit less ignorant about knowledge and the will to share that knowledge…

Best regards
Erik

Erik,

Yes we do have many highly knowledgeable contributors (yourself included) to this forum and those you mention are near the top of the list, and I hope that I have not offended them. However, none of them are in disagreement with Dr. Fossum. Perhaps I should not have said that Fossum knows more about digital sensors than anyone posting here, but that was a polite way of avoiding naming the names of the less knowledgeable contributors.  By the way Emil's last name is Martinec; unfortunately, he has not posted that much recently and probably has returned to his day job of string theory and particle physics.

Regards,

Bill

[ErikKaffehr]

Hi,

The thread title says  "Equipment & Technology", so I guess it is more about technology than he Holy Bible…

Best regards
Erik

For the LORD is a great God And a great King above all gods...

[eronald]

Erik,

Yes we do have many highly knowledgeable contributors (yourself included) to this forum and those you mention are near the top of the list, and I hope that I have not offended them. However, none of them are in disagreement with Dr. Fossum. Perhaps I should not have said that Fossum knows more about digital sensors than anyone posting here, but that was a polite way of avoiding naming the names of the less knowledgeable contributors.  By the way Emil's last name is Martinec; unfortunately, he has not posted that much recently and probably has returned to his day job of string theory and particle physics.

Regards,

Bill

Bill,

 Please do not hesitate at naming me. I got my maths degree with difficulty, my degree in telecommunications engineering with a specialty in integrated circuit design, with great difficulty as well, my PhD is doubtless a blot on the good name of the school that awarded it , and I am sure that a diagnosis of Alzheimer's would be justified. But I sat through all those lectures and did all those circuit-level Spice simulations because I was actually interested in how things work. The same reason why I took Dr. Hunt's colorimetry course and Norman's Imatest course. Now, I hate authority arguments. If you have actual data about various chips that would allow us to compare them, just post it. If you have images to show, show them. Put up the whole-system (chip+CFA+cover) spectral graphs of the CCD chips people like and of the Sony sensors, and let us compare them. Let's see some comparison of texture preservation. Color is psychophysics and so is the appreciation of any other aspects of imagery, but one can still expect to use metrics.

Edmund

[bjanes]

Now, I hate authority arguments. If you have actual data about various chips that would allow us to compare them, just post it. If you have images to show, show them. Put up the whole-system (chip+CFA+cover) spectral graphs of the CCD chips people like and of the Sony sensors, and let us compare them. Let's see some comparison of texture preservation. Color is psychophysics and so is the appreciation of any other aspects of imagery, but one can still expect to use metrics.

Edmund,

Data are preferable to expert opinion, but when data are not available expert opinion is superior to statements by non-experts who may be emotionally attached to a specific camera or have conflicts of interest from financial or other considerations. Color reproduction involves not only physics but also human perception and is a very complex topic with a myriad of variables; it is very difficult to normalize these factors and perform a well controlled experiment. Your suggestions are far beyond what I could accomplish, and I must defer to experts, tempering their statements with my own experience and that of others whom I trust. I would defer to Eric Fossum's evaluation over that of a Phase One rep or a photographic artist with no scientific background.

Regards,

Bill

[yashima]

Back to topic with an innocent question, continuing on mjrichardson stream of thoughts:

Is the pipeline behind CCD and CMOS completely interchangeable? If not then we can surely talk about color reproduction of CCD-sensor-stack and CMOS-sensor-stack?