The GetDPI Photography Forum

Great to see you here. Join our insightful photographic forum today and start tapping into a huge wealth of photographic knowledge. Completing our simple registration process will allow you to gain access to exclusive content, add your own topics and posts, share your work and connect with other members through your own private inbox! And don’t forget to say hi!

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

I'm not trying to be provocative here, honestly. It's an honest question that I can't for the life of me find any answers to.

If CMOS sensors are capable of rendering an image just like a CCD and the only difference is some color curve that was pre-programmed into the CCD sensor, then it should at least be theoretically possible to create a luminance and color curve that, when applied to a CMOS image, produces a CCD-like look. I have tried to find examples of how to do this but I can't find any.

Maybe I'm not using the right search terms or reading the right posts or something. Maybe someone here that's gone down this road can point me to a discussion of how to take CMOS images and, applying some magic sauce, make them look like older CCD images?

This also leads me to wonder how one goes about analyzing an image for things like luminance curves? I know you can create a histogram from image samples we find on the internet but I'm wondering how you go about figuring out what to adjust if you have a CCD image that you're comparing a CMOS image to. I know there's going to be differences when the image is different. I'm not into creating a lab where I shoot test targets, etc... I'm interested in the real-world differences. Can I take an image sample I find online at some review site and analyze it for what makes it look the way it does, then apply that knowledge to a "similar" CMOS image to approximate the look?
 

torger

Active member
Noone has succeeded in defining what the CCD look is, and thus it's hard to analyse in an objective manner and even harder to replicate.
 

synn

New member
It doesn't need to be defined. Those who know it will recognize it.

That Pentax 645D/ 645Z comparison posted in lula shows this very well. Plasticky skin on the Z while the one from the D has a certain "Pop".

You may or may not notice this for landscapes, but for portraiture, it's there. As I said on lula again, Frank Doorhof's samples from the credo 50 looks the closest and it could be a combination of the Leaf profiles and Frank's post processing skills.
 

iiiNelson

Well-known member
I'm not trying to be provocative here, honestly. It's an honest question that I can't for the life of me find any answers to.

If CMOS sensors are capable of rendering an image just like a CCD and the only difference is some color curve that was pre-programmed into the CCD sensor, then it should at least be theoretically possible to create a luminance and color curve that, when applied to a CMOS image, produces a CCD-like look. I have tried to find examples of how to do this but I can't find any.

Maybe I'm not using the right search terms or reading the right posts or something. Maybe someone here that's gone down this road can point me to a discussion of how to take CMOS images and, applying some magic sauce, make them look like older CCD images?

This also leads me to wonder how one goes about analyzing an image for things like luminance curves? I know you can create a histogram from image samples we find on the internet but I'm wondering how you go about figuring out what to adjust if you have a CCD image that you're comparing a CMOS image to. I know there's going to be differences when the image is different. I'm not into creating a lab where I shoot test targets, etc... I'm interested in the real-world differences. Can I take an image sample I find online at some review site and analyze it for what makes it look the way it does, then apply that knowledge to a "similar" CMOS image to approximate the look?
Doesn't happen. CCD has it's own look and so does CMOS. I've been hearing that story since the Leica M was released after the M9 and I still haven't seen convincing evidence. The closest I've seen to date was the Sony A7 series to replicating CCD like rendering and color. It's still different though.
 

glenerrolrd

Workshop Member
The problem is applying a standard set of adjustments to achieve the CCD color and luminance from a CMOS file . If you take a photograph of a color chart (Passport color checker for example ) with a Leica M9 and Leica M 240 using the same lens and a custom white balance for each …you can easily see the differences in both color and the tone curve .

The first step to matching color is to create a custom camera profile ….if you wanted a Leica M 9 look you would need a color chart that matched the rendering from a M9 . No different that what they do with the film emulation presets . I chose to match both the M9 and the M 240 to a standard passport color checker . This to some extent neutralizes the color differences .

But the tone curves are significantly different ….because the M9 CCD has a narrow DR …Leica has set the out of the camera tone curve to the right …..few steps in the lights ,highlights . The M 240 CMOS has a flatter tone curve with additional steps in the lights highlights . This gives the appearance of a flatter less brilliant file …because it is that way out of the camera .

The problem I have found is making a consistent set of adjustments ..given a single file and with care in establishing profiles etc …I can make an M9 file and an M 240 file look similar …..but I haven t been able to match the CCD look consistently across a broad range of light .

It is also easier to make the M9 and the M240 look similar than it is to make the M240 look like an M9 file .

Sorry for a long winded explanation but my conclusion for now is that to get a CCD look I need a CCD sensor .
 

ErikKaffehr

Well-known member
Hi,

An engineer at Phase One told Michael Reichmann (founder of Luminious Landscape) that CCD color vs. CMOS colour is just a myth.

When I got my P45+, I was not happy with the colours, but I was affected by Tim Parkins article: The Myth of Universal Colour

At a stage I made a small investigation of the accuracy of colour rendition.

P45+ colour rendition

The findings are not very clear, but I would say they indicate that profiles and profiling play a very important role.

The P45+, with Capture One's own linear profile was very accurate on the IT 8 chart I used, almost as good as the Adobe Standard profile for my Sony SLT 99.

Quite a few real world samples in that article, BTW.

Best regards
Erik


Define the "look." Not much a discussion unless you can do that.
 

ErikKaffehr

Well-known member
Hi,

The one I guess you refer to was based on JPEGs from camera.

But you perhaps have a link to raw files?

More significantly, Michael Reichmann stated that he was told by an MFD back designer that the rendition difference between CCD and CMOS was just a myth. My guess that MFD back designers are educated on he issue. (May be a lot of effort to find the link, though)

Update: here is the link: http://www.luminous-landscape.com/forum/index.php?topic=89040.msg726396#msg726396

It says:
"Thanks for your input Eric. I was told by a senior MF back designer a few years ago that CCDs and CMOS don't differ re colour. As you say, it's the CFA and other factors.

Time to put that myth to bed.

Michael"


Best regards
Erik

It doesn't need to be defined. Those who know it will recognize it.

That Pentax 645D/ 645Z comparison posted in lula shows this very well. Plasticky skin on the Z while the one from the D has a certain "Pop".

You may or may not notice this for landscapes, but for portraiture, it's there. As I said on lula again, Frank Doorhof's samples from the credo 50 looks the closest and it could be a combination of the Leaf profiles and Frank's post processing skills.
 
Last edited:

dougpeterson

Workshop Member
Have you read my essay on CMOS development at P1?

The Phase One IQ250 CMOS Fully Realized

Anyone who tries to boil any camera system down to one factor is underestimating the number of factors involved in producing the final look and feel of a given camera system.

A partial list of the Image Quality Chain is below:
[Lens coating > Lens elements/design > Aperture blade design > internal body coating > microlens size/shape > Anti aliasing filter > IR filter thickness, rolloff and cutoff characteristics > CFA design > sensor photo well size/design > sensor read-out (heat-sinking and/or active cooling very important here) > A/D converter type/quality > A/D converter control parameters > (read-out of black calibration file from sensor recorded as adjunct to the image) > debayering algorithm > color profile > deconvolution / detail finding algorithm, noise reduction based on black calibration file > noise reduction based on image data > sharpening.]
 
Last edited:

dougpeterson

Workshop Member
If you're looking for meaningful answers here, the only real path to them is to get hands on with the relevant gear. There are numerous opprutunities to do this via rental, rental towards purchase, leasing, purchase, borrowing, begging, stealing, workshops, trainings, open houses, etc.

As one example you could do a lot of testing (if pre-arranged with us, the show floor itself is pretty awful for real world testing) while in town for the Photo Expo.
 

ErikKaffehr

Well-known member
Hi Doug,

A fascinating article. I feel readers often don't appreciate the work involved with the development of the tools we have. As it happens, I just have posted a link to your article on LuLa a short while ago.

As a small point, I think that there are some characteristics of sensor CFA's system developers can do little about. One of those areas is red/green/yellow separation. In this area CFA-s may differ a bit, but gradient on reds seems to be very steep.

Best regards
Erik

Have you read my essay on CMOS development at CCD?

The Phase One IQ250 CMOS Fully Realized

Anyone who tries to boil any camera system down to one factor is underestimating the number of factors involved in producing the final look and feel of a given camera system.

A partial list of the Image Quality Chain is below:
[Lens coating > Lens elements/design > Aperture blade design > internal body coating > microlens size/shape > Anti aliasing filter > IR filter thickness, rolloff and cutoff characteristics > CFA design > sensor photo well size/design > sensor read-out (heat-sinking and/or active cooling very important here) > A/D converter type/quality > A/D converter control parameters > (read-out of black calibration file from sensor recorded as adjunct to the image) > debayering algorithm > color profile > deconvolution / detail finding algorithm, noise reduction based on black calibration file > noise reduction based on image data > sharpening.]
 
It doesn't need to be defined. Those who know it will recognize it.
See that's part of the problem. You know what it looks like, but can't describe it. That makes it very hard to duplicate. I do know what you mean though. People talk about DR, but it isn't DR. It's more color depth and purity. I only spent a few minutes with an IQ250, but it is the closest that I have personally used. Not quite there, but there is so much else that is great about the back.


... Frank Doorhof's samples from the credo 50 looks the closest and it could be a combination of the Leaf profiles and Frank's post processing skills.
I think very highly of Frank, but I think his image samples are best overlooked and just go with his words. The man is a magician and can do more with a Kodak Instamatic and cheap flashlight than most can do with all the cool gear in the world.
 

jonoslack

Active member
It doesn't need to be defined. Those who know it will recognize it.
I'm not convinced . . I can tell the difference between M9 pictures and M(240) pictures (and pretty much any other pair of cameras you choose). . . . but I can't tell a CCD from a CMOS image.

I think the Sony A7s is rather the clincher in this - lovely "CCD like" files. The reduced DR of a smaller pixel count combined with the big juicy pixels . . .

The fact that one can tell the difference between the images from a camera with a CCD sensor from a camera with a CMOS sensor does NOT mean one can tell the difference between CCD and CMOS sensors.
 

Steve Hendrix

Well-known member
Define the "look." Not much a discussion unless you can do that.

Some might not like me saying this, but I agree. And honestly, I find it tiresome when CCD or CMOS "looks" are discussed and not defined. Some of you guys are worse than shopkeepers! Why such a mystery? "You know it when you see it" doesn't pass the muster to me. Define it if you feel there's a difference.


Steve Hendrix
Capture Integration
 

fotografz

Well-known member
The problem is applying a standard set of adjustments to achieve the CCD color and luminance from a CMOS file . If you take a photograph of a color chart (Passport color checker for example ) with a Leica M9 and Leica M 240 using the same lens and a custom white balance for each …you can easily see the differences in both color and the tone curve .

The first step to matching color is to create a custom camera profile ….if you wanted a Leica M 9 look you would need a color chart that matched the rendering from a M9 . No different that what they do with the film emulation presets . I chose to match both the M9 and the M 240 to a standard passport color checker . This to some extent neutralizes the color differences .

But the tone curves are significantly different ….because the M9 CCD has a narrow DR …Leica has set the out of the camera tone curve to the right …..few steps in the lights ,highlights . The M 240 CMOS has a flatter tone curve with additional steps in the lights highlights . This gives the appearance of a flatter less brilliant file …because it is that way out of the camera .

The problem I have found is making a consistent set of adjustments ..given a single file and with care in establishing profiles etc …I can make an M9 file and an M 240 file look similar …..but I haven t been able to match the CCD look consistently across a broad range of light .

It is also easier to make the M9 and the M240 look similar than it is to make the M240 look like an M9 file .

Sorry for a long winded explanation but my conclusion for now is that to get a CCD look I need a CCD sensor .
I have a similar take on the subject Roger.

It isn't that something can't be made approximate the look of something else … it is the consistent response in widely variable conditions that seems difficult.

Asking for a quantifiable response as to "what that look and feel may be" seems valid, yet we are dealing with aesthetics … that which we respond to in a right brain manner … thus the "I know it when I see it" … not unlike Justice Stewart's response when asked to quantify what constitutes Pornography ;)

Most people would be hard pressed to explain why they personally like one color over the other, or one flavor over another, or most any other "preference"… let alone something as complex as this subject.

"Look and feel" leads to a preference not a scientific fact. It is an experience based response to visual stimulus and how we emotionally memorize it. In the extreme, we are intuitively drawn to some things and repelled by others.

Science wants everything all tidy and methodical … yet art and aesthetics is a messy business rampant with randomness, subjectivity, intuitive wanderings, feelings, and a predilection toward looking at the "whole" of something and ignoring the parts.

Way to many variables in the hardware, software, and the human response to the results, for me to sort out.

So, my conclusion is if I like something, I don't work my rear off trying to make something else be like it … I just choose the original … and that is currently CCD cameras like my S.

I had the same tug of war from film to digital … and since the handwriting is on the wall regarding CCD vs CMOS, it'll probably be a moot point soon enough. I'll just have to get used to it.

But not quite yet :ROTFL:

- Marc
 

ondebanks

Member
The difference in "look" isn't just down to the CFA. It's not even the most important factor. After all, there is a plethora of different CCD CFA spectral resonses, so even though they don't constitute a uniform family, they are somehow supposed to be very much distinct from the equally non-uniform family of CMOS CFAs? Hardly.

No-one has mentioned this, but the signal to noise curve output by the hardware is also a huge factor in the "look".

BTW, about three quarters of Doug's list of imaging chain factors don't apply to this question. Chop it just before the CFA design, and again after the A/D stuff.

Ray
 

ondebanks

Member
This also leads me to wonder how one goes about analyzing an image for things like luminance curves? I know you can create a histogram from image samples we find on the internet but I'm wondering how you go about figuring out what to adjust if you have a CCD image that you're comparing a CMOS image to. I know there's going to be differences when the image is different. I'm not into creating a lab where I shoot test targets, etc... I'm interested in the real-world differences. C
I understand your reluctance to shoot test targets, but I don't see a more practical way of doing it. And while they may not seem "real-world", the calibration info they yield is what makes real-world photos look good.

Ray
 

Amin

Active member
If CCD rendering can be achieved with CMOS where are the examples?
Before someone answers that, I'd appreciate if someone would answer this:

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

I'd like to see an image achieved with CCD that has this special rendering.
 

Shashin

Well-known member
Science does not want anything in particular, messy, tidy, or otherwise. The CCDS/CMOS problem could simple be confirmation bias, and you can test for that. If there is a "look," that can be tested as well. If the science can actually create the machines and get them to work it really tight tolerances, it can be used to analyse the same machines. Naturally, there are a whole bunch of variables that would be more a factor in this question than just the sensor architecture.

Now, I can imagine that someone could have a preference for a camera and the look from that camera or even cameras, and there may appear to be a correlation to sensor type. But I have yet to find any real analysis to find out why or to even define the qualities associated with it--and they can be quantified.

I actually find this a really interesting question. I am wondering with the seemingly endless question for DR, for example, we as photographers are not considering the significance on the aesthetics. If you had an image with a DR of infinity, it would look horrible. I am not saying a limit to camera DR is needed, it is not so simple, but rather as we process those images, how do we maintain a "natural" look? It seems the conversation leads to that point that one camera look "digital" (unnatural) and one looks "film-like" (natural). Ironically, film was not that "natural." I also have a feeling that the folks see this are not all seeing the same thing, which makes the "look" hard to narrow down.

BTW, my favorite camera sports a CCD. But I think that is more to do with the brilliance of Kodak and Pentax.
 

Shashin

Well-known member
Test targets actually exist in the real world. The light hitting a test target is real world light.
 
Top