CCD or CMOS, interesting experiment by David Farkas at red dot forum

[bjanes]

+1

It has been explained before, photons do not care about the sensor that records them. It's the sensor dimensions/filtration and post-processing that defines the 'look'. CMOS devices just produce lower read noise, and hence a higher S/N ratio which benefits low light exposures. CMOS devices also have a few other benefits like lower power consumption (good for the battery requirements) and thus they operate cooler, which helps to reduce dark current from spoiling long exposures.

Cheers,
Bart

+2

The CCD vs CMOS arguments came to the fore when MFDBs were exclusively CCD and the 135 format cameras were rapidly going over to CMOS. In addition to possessing an extra 6 stops of dynamic range the CCDs also had an inimitable advantage in image quality that only the cognoscenti (MFDB proponents) could perceive. It is almost akin to a religious experience and the Russel's Teapot analogy could apply. Rather than having the majority prove that there is no difference, the burden of proof should be on the CCD proponents.

Bill

[Malina DZ]

I'd like to dilute this conversation with some RAW CCD vs CMOS files. Sorry, no Leica in my possession, only Sony a100 (CCD) & a850 (CMOS). a850 was shot in crop mode to yield a closer FOV. You can download files for evaluation from Google Drive.

[ErikKaffehr]

Hi Malina,

Thanks for sharing.

Best regards
Erik

I'd like to dilute this conversation with some RAW CCD vs CMOS files. Sorry, no Leica in my possession, only Sony a100 (CCD) & a850 (CMOS). a850 was shot in crop mode to yield a closer FOV. You can download files for evaluation from Google Drive.

[Chris Livsey]

Rather than having the majority prove that there is no difference, the burden of proof should be on the CCD proponents.
Bill

I would suggest that those preferring CCD be allowed to do so by the majority. The conversion "force" would seem to me to come from the CMOS "camp". If just counting numbers CMOS has won by a handsome margin but doesn't seem to be happy until the "opposition" is willing to sign an unconditional surrender and hand in all their CCD "weapons", next they will want reparation for the time spent in forums bringing down the last few CCD adherents.


Disclaimer: I shoot both 

[JV]

Personally, I am pretty sure that is just a myth.

Erik,

I am just saying what I am seeing and that is that CMOS files tend to look smoother and more digital to me...

BTW, this is an extract from the by you often quoted diglloyd dated March 2014:

DIGLLOYD: there are some things I miss about the M9 image quality; the M9/M9P CCD sensor has a look to it overall that seems more appealing than the M240 CMOS sensor, better acutance and a different look reminiscent of medium format sensors (the CCD that is). See also the comparisons of the two cameras made last year. In its favor, the M240 behaves better at higher ISO and also with the sun in the frame, and the EVF is a huge plus. There are other considerations too.

Please note that this discussion is not about superiority at all as some people here seem to believe... 

For me film, CCD, CMOS, etc are different looks that all have their place.  I would not classify one as being "superior" over another.

I am just trying to understand what I see and if the differences cannot be attributed to the sensors itself to what can they be attributed?

There was a recent GetDPI discussion on the same topic.  Quoting part of that discussion:

Because the CCD/CMOS / Colour thing does not seem to be founded in fact (again, I'm not denying the difference in the colours). Colour is a function of the Bayer filter and the demosaicing (and partly the DR) - NOT a function of the underlying structure of the sensor. . . . . . . But the reason I referred to it as a 'religion' is that it isn't really possible to investigate on any kind of empirical basis.

Off shooting now with my CCD sensor

Thanks, Joris.

[shadowblade]

Erik,

I am just saying what I am seeing and that is that CMOS files tend to look smoother and more digital to me...

That's because they have less noise at any given ISO. Less noise = smoother.

You can always add in more noise afterwards if you want.

DIGLLOYD: there are some things I miss about the M9 image quality; the M9/M9P CCD sensor has a look to it overall that seems more appealing than the M240 CMOS sensor, better acutance and a different look reminiscent of medium format sensors (the CCD that is). See also the comparisons of the two cameras made last year. In its favor, the M240 behaves better at higher ISO and also with the sun in the frame, and the EVF is a huge plus. There are other considerations too.

Please note that this discussion is not about superiority at all as some people here seem to believe... 

For me film, CCD, CMOS, etc are different looks that all have their place.  I would not classify one as being "superior" over another.

I am just trying to understand what I see and if the differences cannot be attributed to the sensors itself to what can they be attributed?

There was a recent GetDPI discussion on the same topic.  Quoting part of that discussion:

Because the CCD/CMOS / Colour thing does not seem to be founded in fact (again, I'm not denying the difference in the colours). Colour is a function of the Bayer filter and the demosaicing (and partly the DR) - NOT a function of the underlying structure of the sensor. . . . . . . But the reason I referred to it as a 'religion' is that it isn't really possible to investigate on any kind of empirical basis.

Off shooting now with my CCD sensor

Thanks, Joris.

At the heart of it, a digital sensor is really just a counting device. Photons don't care what sort of device they're counted on. CMOS has a lower signal-to-noise ratio, so can count more accurately. The appearance of the image has nothing to do with the counting device at the core of it, but how light is modified before it gets to it - in other words, the Bayer filters, antialiasing and other filters that go in front and modify the light before it hits the photosites. These are different between CCD and CMOS sensors - not because they have to be, but because most CMOS sensors are designed to take greatest advantage of their high-ISO capability and give up colour accuracy and colour detail in order to further increase high-ISO capability. CCD sensors are crap at high-ISO, so no-one designs their filter stack to be able to shoot a little better at the top of their ISO range.

Put an image-quality-optimised filter stack in front of a CMOS sensor and it should give you the same colour detail and other characteristics associated with CCD sensors. By all accounts, that's what the low-ISO-optimised 5Ds is doing.

[JV]

That's because they have less noise at any given ISO. Less noise = smoother.

You can always add in more noise afterwards if you want.

At the heart of it, a digital sensor is really just a counting device. Photons don't care what sort of device they're counted on. CMOS has a lower signal-to-noise ratio, so can count more accurately. The appearance of the image has nothing to do with the counting device at the core of it, but how light is modified before it gets to it - in other words, the Bayer filters, antialiasing and other filters that go in front and modify the light before it hits the photosites. These are different between CCD and CMOS sensors - not because they have to be, but because most CMOS sensors are designed to take greatest advantage of their high-ISO capability and give up colour accuracy and colour detail in order to further increase high-ISO capability. CCD sensors are crap at high-ISO, so no-one designs their filter stack to be able to shoot a little better at the top of their ISO range.

Put an image-quality-optimised filter stack in front of a CMOS sensor and it should give you the same colour detail and other characteristics associated with CCD sensors. By all accounts, that's what the low-ISO-optimised 5Ds is doing.

Thanks for the clear explanation!  Much appreciated!

[ErikKaffehr]

Hi,

Dave Farkas now published the third part of his test: http://www.reddotforum.com/content/2015/03/the-great-debate-ccd-vs-cmos-part-3/

Worth reading, I would say. I have just skimmed trough the article, as I am no potential Leica buyer.

He discusses how to achieve that Leica M9 look and gives a recipe for that. Reducing exposure, varmer white balance and some split toning as far as I recall.

The author seems to feel that most of the perceived difference is coming from the extended DR of the CMOS sensor.

He also writes that the Leica M9 makes very good images under the light conditions where it excels.

Best regards
Erik

[Chris Livsey]

He also writes that the Leica M9 makes very good images under the light conditions where it excels.

Best regards
Erik

Much as: most lenses make very good images, if shot at f8 

[ErikKaffehr]

Hi,

Yes, that is indeed true. I wouldn't argue with that.

Best regards
Erik

Much as: most lenses make very good images, if shot at f8 

[voidshatter]

+2

The CCD vs CMOS arguments came to the fore when MFDBs were exclusively CCD and the 135 format cameras were rapidly going over to CMOS. In addition to possessing an extra 6 stops of dynamic range the CCDs also had an inimitable advantage in image quality that only the cognoscenti (MFDB proponents) could perceive. It is almost akin to a religious experience and the Russel's Teapot analogy could apply. Rather than having the majority prove that there is no difference, the burden of proof should be on the CCD proponents.

Bill

+3

Common hypes in MFDB:

a) CCD offers better dynamic range;
b) 16-bit offers superior image quality;
c) CCD offers better color/look etc.

I found these to be lies after I made my own comparisons. Indeed the Sony CMOS sensors can outrun any CCD sensors in most areas.

[voidshatter]

Hi,

Dave Farkas now published the third part of his test: http://www.reddotforum.com/content/2015/03/the-great-debate-ccd-vs-cmos-part-3/

Worth reading, I would say. I have just skimmed trough the article, as I am no potential Leica buyer.

He discusses how to achieve that Leica M9 look and gives a recipe for that. Reducing exposure, varmer white balance and some split toning as far as I recall.

The author seems to feel that most of the perceived difference is coming from the extended DR of the CMOS sensor.

He also writes that the Leica M9 makes very good images under the light conditions where it excels.

Best regards
Erik

I just quickly checked the first 10 comparisons: by downloading the JPG images and pushing the shadow in ACR I got 80% accuracy - the CCD has a much noisier shadow even for JPG files posted on the web.

[bjanes]

Put an image-quality-optimised filter stack in front of a CMOS sensor and it should give you the same colour detail and other characteristics associated with CCD sensors. By all accounts, that's what the low-ISO-optimised 5Ds is doing.

I'm not an expert on sensor technology, but there are some differences between CCD and CMOS regarding blooming and crosstalk as discussed in these links.

Perhaps some of the more technology savvy forum members could comment.

Bill

[shadowblade]

I'm not an expert on sensor technology, but there are some differences between CCD and CMOS regarding blooming and crosstalk as discussed in these links.

Perhaps some of the more technology savvy forum members could comment.

Bill

Yes, there are - I was simplifying it for the purposes of argument, since neither affect the colour and detail we were discussing (crosstalk reducing local contrast and bloom introducing artifacts). Both are less of an issue on CMOS.

Neither CMOS nor CCD record colour - they're mere photon-counters, with CMOS being a more accurate and less artifact-prone photon counter than CCD. Colour is entirely dependent on the filters in front of each photosite, which affect the amount of light hitting the photosite behind it, and the post-processing (both in-camera and out-of-camera), which takes the photon count from each photosite and, using the knowledge of what colour filter is in front of each photosite, uses these counts to generate an image.

[bjanes]

Yes, there are - I was simplifying it for the purposes of argument, since neither affect the colour and detail we were discussing (crosstalk reducing local contrast and bloom introducing artifacts). Both are less of an issue on CMOS.

Neither CMOS nor CCD record colour - they're mere photon-counters, with CMOS being a more accurate and less artifact-prone photon counter than CCD. Colour is entirely dependent on the filters in front of each photosite, which affect the amount of light hitting the photosite behind it, and the post-processing (both in-camera and out-of-camera), which takes the photon count from each photosite and, using the knowledge of what colour filter is in front of each photosite, uses these counts to generate an image.

From these considerations, it appears as if (contrary to CCD enthusiasts' assertions) that CMOS does have some advantages over CCD in addition to their lower read noise. As David Farkas' experiments show, the differences are subtle. However, in his comparing the two cameras, there are other non-controlled variables than sensor type. Certainly, CCD has been abandoned for 135 type sensors, and the advantages of CMOS are being exploited by MFDB vendors when that technology is available for a given sensor size.

Bill

[ErikKaffehr]

Hi,

Your observation pretty much agrees with Dave Farkas's evaluation.

Best regards
Erik

I use both CCD and CMOS Leica M series cameras. The difference I see between the two is that the M9 delivers a slightly punchier file out of camera which is probably down to the difference in dynamic range rather than sensor type. Colour reproduction can vary slightly between the two cameras. My workhorse camera is the M240.

[ErikKaffehr]

Hi Keith,

At the end of this article: http://www.reddotforum.com/content/2015/03/the-great-debate-ccd-vs-cmos-part-3/

Best regards
Erik

Hi Erik, have you a link to David Farkas's evaluation?

[ErikKaffehr]

Hi,

Just to say, the outcome was pretty much what I would have expected. Some of Dave Farkas's observations were interesting, like the M240 sensor having better highlight roll of and the wider dynamic range causing less contrasty image. Those differences may have a lot to do with camera profiles.

I own both CCD and CMOS, in my case a P45+ having CCD and a few Sony cameras with CMOS. I did some pretty careful comparison between the P45+ and the Sony Alpha 99, and found that both were pretty accurate when using an IT8 test target.

Would someone be interested, here is a draft of an Article I planned on writing for OnLandscape: http://echophoto.dnsalias.net/ekr/Articles/OLS_OnColor/OnColor.pdf

Best regards
Erik

[Jack Hogan]

here is a draft of an Article I planned on writing for OnLandscape: http://echophoto.dnsalias.net/ekr/Articles/OLS_OnColor/OnColor.pdf

Excellent comparison Erik.  It is interesting to note that the simple Sensitivity Metamerism Index (ISO 17321 - where the SLT99 appears to be substantially better than the P45+) is not representative of your more complex testing.

Jack

[ErikKaffehr]

Jack,

Thanks for commenting.

One issue with my test based on the IT-8 is that the dyes used in that target are based on the chromogenic process and I guess that they are intended to minimise metameric error.

To me it seems that colour profiles play a very important role in colour reproduction. The enclosed samples shows six different conversion of the same subject. In this case I actually measured the spectrum on the blades and the leafs, spectrum based representations are also included. The purplish blue (I think it is called) colour was reproduced as blue with Capture One 8 on both P45+ and Sony Alpha 99. It may be hard to know what a natural representation is.

Best regards
Erik

Excellent comparison Erik.  It is interesting to note that the simple Sensitivity Metamerism Index (ISO 17321 - where the SLT99 appears to be substantially better than the P45+) is not representative of your more complex testing.

Jack