CCD / CMOS benefits and drawbacks.

[jaapv]

Advertisement (gone after registration)

Take a look at the camera sensor rankings on dxomark.com for details.

Those are not sensor rankings, but camera raw output rankings, including all software corrections done in-camera.

[noirist]

Yes, DxO provides a single overall rating, which is, as you say, not that useful. Fortunately DxOt also ranks camera sensors by color depth, dynamic range, and low light ISO, and it provides specific measurements of ISO sensititivity, signal-to-noise, dynamic range, tonal range, color sensitivity, and color response. And in every one of those measured attributes the best full frame CCD sensor is trounced by the best full frame and APS-C CMOS sensors.

 

The presence or absence of an AA filter is another attribute of the camera's design that doesn't concern the photographer. The photographer cares about the photographic advantages and disadvantages provided by specific cameras, like whether the camera produces images with moire or how much measured resolution the images have. Let's say there was a camera whose images had horrible moire, a high pixel count, and very low measured resolution. Would you really care whether or not it had an AA filter? Or let's say there was a camera whose images had virtually no moire, very high measured resolution, and the highest color accuracy. Would you really care whether or not it had an AA filter?

 

Dxo's one-number-tells-it-all ranking assumes a weighting of attributes that may not match with how you shoot. It doesn't take other relevant factors into account (the absence of an AA filter, or amazingly noise-free shadows at low iso, or non-sensor attributes like size and weight, or the ability to mount a Summicron).

Edited June 24, 2011 by noirist

[Advertisement]

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

[noirist]

As I understand it, the RAW files contain the actual uncorrected sensor measurements.

 

Even in cases like m43rds, the RAW files contain the actual uncorrected sensor measurements plus some additional lens and body information that allows for easy software corrections by follow-on processing. For example, if you look at the RAW images from the Panasonic 20mm f1.7 at f1.7 in fastpictureviewer or dxo optics, you will immediately see vignetting and barrel distortion. If you view the embedded JPGs or convert RAW to JPG with ACR, then the vignetting and barrel distortion is automatically corrected.

Those are not sensor rankings, but camera raw output rankings, including all software corrections done in-camera.

[noirist]

I just double-checked the DxO sensor rankings for color depth, dynamic range, and low-light ISO, and the CCD performance is even worse than I had initially thought. The best full frame CCD sensor performs significantly worse in all three measurements than any of the full frame CMOS sensors introduced in the past five years.

 

And by those measures, the best full frame CCD sensor in a commercially available camera (in the M9) is soundly beaten in every respect by the commercially available full frame CMOS sensors. It's even beaten in many respects by some smaller APS-C size CMOS sensors in budget DSLRs, like the Nikon D5100. Take a look at the camera sensor rankings on dxomark.com for details. So whatever hypothetical advantages that CCD might have versus CMOS are completely obviated by the very real advantages that actual CMOS sensors have over actual CCD sensors.

[mwilliamsphotography]

I just double-checked the DxO sensor rankings for color depth, dynamic range, and low-light ISO, and the CCD performance is even worse than I had initially thought. The best full frame CCD sensor performs significantly worse in all three measurements than any of the full frame CMOS sensors introduced in the past five years.

 

And yet, users continue to be of the opinion that their CCD equipped little M9 has better IQ than any CMOS camera available ... especially in the area of color depth.

 

I've owned/used a Canon 5D-MKII, 1DMKIII, 1DsMK-III, then a Nikon D3/D700 and D3X ... even using Zeiss lenses and Leica R APO and ASPH lenses ... and now a Sony A900 and all the ZA Ziess optics ...

 

None were/are as good as the M9 in terms of IQ ... and the M9 at ISO 640 and 800 with faster aperture M lenses looks better than the 1DsMKIII and Nikon D3X at the same ISOs. The D3/D700 outperformed the M9 at higher ISOs but are only 12 meg and the color gets squirrelly looking while DR quickly deteriorates.

 

There simply is no free lunch in photography. A mentor once told me there are three basics you can provide ... speed, quality and price ... pick two, because you can't have all three. It seems with Leica we get to pick one ... quality. Hopefully, they'll get that up to two with the next M and I'm fairly sure one of them won't be price ... LOL!

 

-Marc

[colonel]

The presence or absence of an AA filter is another attribute of the camera's design that doesn't concern the photographer.

 

I'm afraid the M8 and M9 produce some of the sharpest images at base ISO (and a few stops higher)

 

Many people, myself included, have compared M8/M9 against both D700 and 5Dii and, although those camers are great and soundly beat up everything else below them in the market in higher ISOs, they are not as sharp or clear.

 

Is it the Leica lenses ?

Is it the lack of AA filter ?

Is it the weak, or non-existant, IR filter ?

Is it the CCD sensor ?

 

Definitely the AA filter make a massive difference. There is a company that will remove them from conventional DSLRs,

Optical Products

 

this is study done with a D200 and if you look at the before and after pictures you will see a big difference:

Nikon D200HR

[mwilliamsphotography]

Advertisement (gone after registration)

It may be worth a call to see if they can convert one of my Sony A900s which is now out of warranty anyway ... that would be an interesting experiment if they can do it.

 

-Marc

[kdriceman]

If you are basing your evaluations of the camera on the technical ratings of one component, you are missing the point completely. The Dodge Viper has more power and much more torque than a Ferrari 458 Italia, yet the Ferrari will trounce it at the drag-strip, on a road course and in top speed. Why? Because there is much more to making a high-performance sports car than putting a big powerful engine into a chassis. In the same way, there is more to making a great image than details of a sensor. IMO the M9 from ISO 160 to 1600 makes significantly better files than any camera I have used. If I am going to have to go over 1600, I have to decide if I want to lug around the DSLR and get better image quality, or take the M9 and spend time in post processing.

[xrogers]

The presence or absence of an AA filter is another attribute of the camera's design that doesn't concern the photographer. The photographer cares about the photographic advantages and disadvantages provided by specific cameras, like whether the camera produces images with moire or how much measured resolution the images have. Let's say there was a camera whose images had horrible moire, a high pixel count, and very low measured resolution. Would you really care whether or not it had an AA filter? Or let's say there was a camera whose images had virtually no moire, very high measured resolution, and the highest color accuracy. Would you really care whether or not it had an AA filter?

 

Both the M8 and M9 photos require much less unsharp masking than my Canons, and I prefer the less sharpened look in prints. If you want to measure that as "amount of USM required for printing" rather than yes or no AA filter, that's fine with me. That's a sensor attribute dxo doesn't measure where I much prefer the M9's sensor.

 

Dxo does not consider the attributes of moire, pixel count or final image resolution. Dxo also doesn't measure color accuracy, it measures color sensitivity (the ability to differentiate very similar colors). Once you have enough bits of color sensitivity, then the number of additional bits matters far less than how those bits are interpreted. Once you have enough bits of dynamic range to cover a scene, low noise in shadows can matter more than additional detail separation in shadows.

 

It is unreasonable to say "CCD is worse than CMOS because their Dxo ratings are lower". Dxo isn't the whole story, it is just one small set of measurements, and measurements don't matter nearly as much as pictures.

 

Until later,

 

Clyde

[M. Valdemar]

Well I hope hat Leica stick with the CCD. No expert on this but like many on this forum have used lots of digital cameras and so can with some qualification say that the original Nikon D1 and then D1H ,D1X gave great image quality and low noise with the CCD chip then Nikon brought out the D2x with the CMOS chip and made what to me was a pile of c**p with rubbish image quality and despite all the hype about it my D2x is and always has been the most expensive paperweight I ever bought. So please Leica stick with the CCD and don't be swayed by all the low noise at high ISO talk and remain true to your original values. Re the Nikon D200 test - personally I have never seen truly sharp picture from a D200 and my colleagues dropped their Canon gear ASAP again due to poor image quality and being peed off and jaded by continually trying to get properly focussed and sharp image straight out of the camera.

 

 

Punctuation, my friend. It is a virtue.

[pop]

Punctuation, my friend. It is a virtue.

 

Sentence often verb.

 

If you don't have to say anything, why don't you do it elsewhere?

[pico]

With the M9-P about to be announced and the talk here that this may be the last opportunity to get a CCD sensor M, can anyone explain the advantages of CCD over CMOS?]

 

Who the hell ever mentioned a replacement of the sensor: Citation, please?

 

.

[jonoslack]

I just double-checked the DxO sensor rankings for color depth, dynamic range, and low-light ISO, and the CCD performance is even worse than I had initially thought. The best full frame CCD sensor performs significantly worse in all three measurements than any of the full frame CMOS sensors introduced in the past five years.

 

Hmmmm . . . and that doesn't make you slightly suspicious?

I don't have an axe to grind here, but I can apply a little logic to a situation, and it seems to me that if a testing system comes up with the result that CCD sensors (just like those in most top range Medium Format cameras) are inferior to CMOS sensors . . . then the most likely solution to the problem is that the testing is biased towards CMOS sensors.

 

. . . . or possibly DxO is god.

[noirist]

There's only one CCD sensor in a production full frame digital camera (the M9). All other production full frame digital cameras use CMOS sensors. So maybe the M9 chose a bad CCD sensor or maybe the Nikon/Canon/Sony engineers figured out how to make a better full frame CMOS sensor than they can make a full frame CCD sensor. Bottom line is that CCD vs CMOS is a meaningless debate for photographers. All that matters to us should be the measured performance of the actual sensor, who cares whether that performance is achieved with CCD or CMOS.

 

The DxO sensor measurements seem well-designed ot me (a priori) and I can't imagine that the DxO engineers could care less if CCD or CMOS was better. Also the tiny CCD sensors do fine in the DxO measurements, so I don't think DxO is participating in a conspiracy against CCD.

 

Hmmmm . . . and that doesn't make you slightly suspicious?

I don't have an axe to grind here, but I can apply a little logic to a situation, and it seems to me that if a testing system comes up with the result that CCD sensors (just like those in most top range Medium Format cameras) are inferior to CMOS sensors . . . then the most likely solution to the problem is that the testing is biased towards CMOS sensors.

 

. . . . or possibly DxO is god.

[jonoslack]

All that matters to us should be the measured performance of the actual sensor, who cares whether that performance is achieved with CCD or CMOS.

 

Actually - all that matters to me is the perceived performance, and the image quality in real world situations - which is often a very different thing.

 

Also - the M9 may be the only Full Frame CCD - but as far as I'm aware ALL MF sensors are CCD and not CMOS - which presumably means that Hassleblad and Phase are making the wrong decisions?

 

 

all the best

[movito]

I was surprised to find that the M9 did so badly in DxO's analysis, earning a meager score of 69. Not sure what to make of it, but the D3x is the best sensor they've tested, outgunning even a Hasselblad H3D II 50.

 

Anyone care to expand on the scores?

[jaapv]

DxO is sailing under a false colours imo. It does not measure sensor performance, but camera RAW performance. So they mainly determine the amount of and, when applicable to similar amounts of, quality of firmware, and not the direct output of the sensor. A camera with a high amount of pre-RAW processing e.g. CMOS cameras will score "higher" by their parameters than a camera with a limited amount of in-camera processing e.g Leica M8 and M9 and MF CCD cameras.

The figures on that site should be taken in a low dosage and with caution. Very pretty graphs though...

Hmmmm . . . and that doesn't make you slightly suspicious?

I don't have an axe to grind here, but I can apply a little logic to a situation, and it seems to me that if a testing system comes up with the result that CCD sensors (just like those in most top range Medium Format cameras) are inferior to CMOS sensors . . . then the most likely solution to the problem is that the testing is biased towards CMOS sensors.

 

. . . . or possibly DxO is god.

Edited June 27, 2011 by jaapv

[mjh]

DxO is sailing under a false colours imo. It does not measure sensor performance, but camera RAW performance. So they mainly determine the amount of and, when applicable to similar amounts of, quality of firmware, and not the direct output of the sensor. A camera with a high amount of pre-RAW processing e.g. CMOS cameras will score "higher" by their parameters than a camera with a limited amount of in-camera processing e.g Leica M8 and M9 and MF CCD cameras.

It has to be said that they try their best to get as close to an assessment of sensor performance as possible. If a manufacturer uses some kind of averaging of pixel data to reduce noise, they can spot that by analyzing correlations between adjacent pixels. Standard CMOS stuff like dealing with pixel non-uniformity will go undetected, but since it doesn’t introduce an unfair bias towards CMOS sensors, it doesn’t pose a problem.

 

If I have an issue with DxOMark results it is with the aggregated values that are next to useless. The individual measurements are generally fine and their methodology is sound. You have to interpret the results yourself, though; the numbers aggregated from the actual measurements can be seriously misleading.

 

Obviously, if you care for features such as resolution and sharpness that DxO doesn’t measure at all, DxOMark results won’t help you. A 36 x 24 mm sensor with 12 MP will outperform a 50 MP CCD sans AA filter since for DxO, resolution doesn’t count at all.

[jaapv]

I don't disagree. To compare sensor output would be a whole different level of testing and I don't doubt their methodology. The problem comes when readers of their site take the figures and graphs as gospel for image quality.

[Advertisement]

Hallo,

Look here: Leica M

[zlatkob]

DxO is sailing under a false colours imo. It does not measure sensor performance, but camera RAW performance. So they mainly determine the amount of and, when applicable to similar amounts of, quality of firmware, and not the direct output of the sensor. A camera with a high amount of pre-RAW processing e.g. CMOS cameras will score "higher" by their parameters than a camera with a limited amount of in-camera processing e.g Leica M8 and M9 and MF CCD cameras.

The figures on that site should be taken in a low dosage and with caution. Very pretty graphs though...

 

It sounds like you're able to enjoy M8 & M9 files straight from the sensor, unprocessed by any camera.