CCD vs CMOS: Can you tell which is which?{merged}

[Leica Fanatic]

Advertisement (gone after registration)

Excellent test and thanks for taking the time.

 

To me both sets of images are pleasing, each sensor bringing it's own characteristics to the party and to be honest, there's no difference between them (except on a molecular level ).

 

Going off at a tangent, any suggestion in a CCD vs CMOS debate as to which is better (whatever that means), would be purely a personal decision.

Edited February 25, 2015 by Leica Fanatic

[jaapv]

Me too tend to attribute almost any #1 to M9... but the fact that the general consensus is impressively around 50% (from a quick evaluation) it seems to me that the only possible conclusion is that the differences are not so significant... expecially on screen images...

If I had to say a quick conclusion the only phrase that comes to my mind is "dynamic range is different"

 

My criteria were dynamic range (shadow detail) and more pronounced reds and yellows, based on the difference in the camera profiles in ACR.

But I'm sure David knows this as well, and is quite capable of inserting trick images .

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[AlanJW]

David: thank you for this, which to me shows that the differences, if any, are not significant, which also is what I found when I switched from M9 to M(240). I am not saying that those who can see differences in CCD are crazy. Just that I don't (and I expect -- as the survey metrics indicate -- that most don't).

 

Please don't be put off by those who complain that these are monitor only comparisons or want tungsten or other light. You can file that under "No good deed shall go unpunished".

[Paul J]

If your aim is to show how similar jpgs can look on the internet, after adjustment, then sure, it's a worth while test. However, despite difference between the files it does not address one of the main differences: How the files react during processing. Jpgs do not really help in this regard; the end result doesn't say necessarily say much. Video of the process may, though, in some instances.

 

On top of that, you are adjusting the files, you are drastically down sampling them, reducing them down to a constrained colour space and compressing them with jpg. These have jpg colour, it and would be inadmissible by the majority of my clients pre press departments. You have discarded the large majority of available data and left it up to a computer to decide what goes and what stays, what is changed to what. What is even left of the original file? The structure, composition and general tonal and colour palette. The rest is a computer painting by numbers.

 

Sorry, I don't wish to seem negative, but from my perspective this test does not conclude anything aside from which picture do you like the best. Which will most likely be different on a case by case basis for both cameras.The test is commendable for effort and for it's scale though, at least the pictures are of a good sized spectrum and tonality, enough to see the differences. The Raw files would be another case entirely, though, this defeats the purpose of the guessing game doesn't it?

 

The CCD v CMOS conversation is almost pointless. Leica, or any 35mm camera manufacturer are not going back to CCD in any hurry and CMOS is going to get better with every iteration anyway. The amount of people who would buy the features of CMOS greatly outweigh those who want the CCD for it's characteristics. If you want to see the differences look at the raw files and how they react during processing.

[k-hawinkler]

If your aim is to show how similar jpgs can look on the internet, after adjustment, then sure, it's a worth while test. However, despite difference between the files it does not address one of the main differences: How the files react during processing. Jpgs do not really help in this regard; the end result doesn't say necessarily say much. Video of the process may, though, in some instances.

 

On top of that, you are adjusting the files, you are drastically down sampling them, reducing them down to a constrained colour space and compressing them with jpg. These have jpg colour, it and would be inadmissible by the majority of my clients pre press departments. You have discarded the large majority of available data and left it up to a computer to decide what goes and what stays, what is changed to what. What is even left of the original file? The structure, composition and general tonal and colour palette. The rest is a computer painting by numbers.

 

Sorry, I don't wish to seem negative, but from my perspective this test does not conclude anything aside from which picture do you like the best. Which will most likely be different on a case by case basis for both cameras.The test is commendable for effort and for it's scale though, at least the pictures are of a good sized spectrum and tonality, enough to see the differences. The Raw files would be another case entirely, though, this defeats the purpose of the guessing game doesn't it?

 

The CCD v CMOS conversation is almost pointless. Leica, or any 35mm camera manufacturer are not going back to CCD in any hurry and CMOS is going to get better with every iteration anyway. The amount of people who would buy the features of CMOS greatly outweigh those who want the CCD for it's characteristics. If you want to see the differences look at the raw files and how they react during processing.

 

 

 

Thanks Paul. Excellent points.

I like my M9 and will not get an M240 but am very interested in what specific cameras Leica will come up with next.

 

From my point of view the discussion about CCD and CMOS is pretty pointless.

Basic technology developments in this area are primarily driven by the process requirements of the chip making industry and their foundries IMHO.

Whatever they can produce economically and competitively with other technological alternatives is the foundation for the computer and related industries, such as camera sensors.

End of story.

[Doc Henry]

Hi David thank you for this test.

The ideal is that there is no correction

With a film like Kodak Portra and M7, I almost do not correct or I do not correct

Erwin Puts had mentioned that the M9 gives a "vivid" color and the M240 a "neutral" color

and I agree with him.

Best

Henry

... another test here :

http://www.l-camera-forum.com/leica-forum/leica-m9-forum/326647-m240-vs-m9.html

look at the "red" color of flowers in foreground at left

Best

Henry

[Lenshacker]

Advertisement (gone after registration)

... another test here :

http://www.l-camera-forum.com/leica-forum/leica-m9-forum/326647-m240-vs-m9.html

look at the "red" color of flowers in foreground at left

Best

Henry

 

In the above example, the EXIF is included so you can determine that L1006819.jpg is the M9 image. But- before looking at the EXIF, I downloaded both files and compared file size to "Guess" that L006819.jpg was from the M9. The M9 file uses 251Kbytes and L1007035.jpg uses 239KBytes. JPEG files at maximum resolution are lossless, and coefficients are stored to generate all of the frequency information of the image. The M9 file has more frequency content that does the M240 file in this case. The same is true of David's test where I put an "X" next to the filename. David did not use a tripod, the images shifted- so I applied a little bit of a fudge factor on a couple.

 

David did remove all EXIF information from his images. We'll see if the file-size metric has any bearing on the camera used.

[jmahto]

Processed, web sized pics are useless for comparison...

 

Case in point below. These are from (not in this order)

Leica M240 + 50 summilux

Nikon D810 + 50 1.4G

iPhone6 + (PS for background blur)

 

????

its useless...

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

Edited February 25, 2015 by jmahto

[dfarkas]

Good idea.

Not sure if we are only comparing CCD versus CMOS here. IMO we are comparing the complete chain of elctronics in M9 versus M240.

 

 

Kind of the point, I suppose. A camera and the resulting digital file are much more than the sensors. That's why I've been so curious about testing the CCD Look theory, which led me to do this test.

[elmars]

It is useless for sharpness and resolution, but for colors?

[jmahto]

It is useless for sharpness and resolution, but for colors?

 

Color can be manipulated too *to a large extent* to look the same. See my example above (it was a hastily done job to make a point, I could have done a better job).

[dfarkas]

Thanks, interesting test.

 

However:

 

- sRGB conversion castrates the color gamut. This is a real bummer, as color is what most people cite as the biggest plus of CCD sensors.

 

- The difference in dynamic range makes it easier to spot the M9 where color differences are otherwise subtle.

 

- The web resolution hides the differences in shadow noise.

 

In any case, if not the evanescent "CCD look", I am confident I have spotted the "M9 look" in most photos. Especially those with the lovely blue San Francisco sky

 

sRGB is actually ideal in this case. Most monitors can only view colors in the sRGB gamut. My monitor can display close to 100% AdobeRGB, but I'd wager that most Leica users use either a MacBook Pro or iMac, both with Retina displays that display 99% sRGB. The test is perfectly suited to these judging conditions. Any additional gamut information wouldn't be that useful.

 

Digital photographic printing is closer to sRGB as well. CMYK offset press is much less than sRGB. Some fine art output is larger than sRGB with the latest generation Epson printers, but only with certain papers.

 

As far as hiding shadow noise, I stated in my article that I picked conditions favorable to the M9: good, directional naturally light at low ISO. So, there really isn't anything to hide. The files are available at 1800x1200 pixels if you click on them, a bit larger than web size.

[TonyField]

This is an interesting show-and-tell. There are some individuals on this forum that the eye to see very subtle differences among the cameras (and do, in fact, see the subtle differences in rendition between the M9 and M240). I think I wish I had that sensitivity of vision ... but I don't. The only difference I can see is that the M9 out of camera tends to have strong bold colours at low ISO ... sort of like Kodachrome. Sometimes I do like this, sometimes not.

 

I have looked through my many 16x20 prints from the Leica M9, Nikons (both CCD and CMOS), Canons (all CMOS). In the majority of cases, I do not see any images that identify which camera was used (even when comparing the old cameras with low pixel counts).

 

However, I do have a tendency to post-process in a reasonably consistent way. As far as I can see, the camera itself has only a slight impact on the final outcome (other than detail resolution ... I can certainly spot the difference between the 40mp Hasselblad and the 3 mp Fuji S1 on a 16x20 print). The vast majority of image effectiveness comes from the post processing done - that is where "art" really takes place.

 

Of course, there are many people who shoot "out of camera" ... even jpg's only. This is the only situation (IMHO) in which characteristics of the camera are identifiable.

 

All in all, I cannot see any particular advantage to CMOS or CCD as an image technology ... other than the technology's ability to cope with noise and ISO (I guess I could also include dynamic range).

 

On the other hand, I might have a different opinion if my eyes were as sensitive to colour and tone as some of the other folks here who do have this vision developed to a high degree.

[dfarkas]

David, first of all thank you for doing this. The photographs are very close. I wonder how I fared in my results. I do think I can see the difference in most photographs, but on some I have to guess.

 

I did a similar exercise when I played with both cameras. Now the interesting thing to me though was that after opening the photographs in Lightroom, I always spent more time processing the M240 photographs than the M9 pictures. This is what made me pass on the camera possibly until the next model. Would you agree that you feel more need to tweak the M240 files in post at base ISO?

 

I will talk about this in my wrap-up, for sure.

[dfarkas]

Do I remember climbing Lombard Street..... having just walked all the way from Fisherman's Wharf, so puffed I can't remember which of the CCD/CMOS shots you show represents what i saw...

 

Yeah, I also walked from Fisherman's Wharf up Lombard to take these shots. It was quite a trek!

[jaapv]

I bet that the majority of the Clamoring Five Hundred are on Eizo or Nec and use high-end printers.

[Lenshacker]

On the other hand, I might have a different opinion if my eyes were as sensitive to colour and tone as some of the other folks here who do have this vision developed to a high degree.

 

I just want to know how my Wife's algorithm did.

 

IEEE Xplore Abstract - A comparison of wavelet and Fourier descriptors for a neural network chromosome classifier

[dfarkas]

Hi David thank you for this test.

The ideal is that there is no correction

With a film like Kodak Portra and M7, I almost do not correct or I do not correct

Erwin Puts had mentioned that the M9 gives a "vivid" color and the M240 a "neutral" color

and I agree with him.

Best

Henry

 

My premise was to see if I could match the result from the CMOS (M240) camera to emulate that from a CCD (M9). Correction was required to do this. And, as I've learned over the years, there is very little point in evaluating "uncorrected" images. All digital files are "corrected" at some point, whether by in-camera processing, demosaicing algorithms, camera proflies, LR rendering, etc.

 

Even in film days, there were differences: chemistry type, balance and temp, replenishment rate, emulsion variations, etc. could make the same film look different. When considering a negative film like Portra, there is infinitely more variability in printing a positive image.

 

I am not arguing that the default color response is different from different cameras. It is, but that wasn't my test.

[dfarkas]

It depends too on the workflow. I'm not so sure that the comparison images can tell us much. They are of good, straight subject matter but in my experience processing a file requires an understanding of what your endpoint is intended to be and with such straight images (sorry David, I appreciate your efforts) the endpoint will be to achieve a high degree of similarity. My workflow on some images is very different and my endpoint will be likewise.

 

I absolutely agree with you. As I stated, my workflow was to fist optimize the M9 images, then to match the M240 images to that standard. Without the direct comparison, I might have processed the M240 very differently to better suit my taste. I stuck to the premise of the test, though.

 

IMO it is when lenses and sensors are operating under 'stress' (ie with difficult or awkward lighting, contrast and subject matter) that they show their distinctive 'personalities'. As I said in another thread, it is the shift in workflow from one generation of camera to the next which I find frustrating and if a CMOS sensor camera could achieve a very similar workflow under 'stressed' conditions as a CCD sensor camera then I would be happy. As it is I have to learn (and utilise) different workflows for each camera.....

 

And FWIW I have no idea which shot is from which camera.

 

 

Also, as I wrote, I picked conditions that would be favorable to the M9 so that it wasn't a contest of dynamic range or malleability of the file under less-than-ideal circumstances. I am readying a Part 2 that has a much broader range of subject matter, but which doesn't have direct comparisons.

[Advertisement]

Hallo,

Look here: Leica M

[dfarkas]

It's a pity that these are all shots in full daylight mostly sunlit, few shadows or clouded sky, no shots in available light, no tungsten etc. lit situations.

 

Part 2 coming soon....