M9 high iso

[erl]

Advertisement (gone after registration)

Adan, that talks to me!

 

I am amazed at what that 135 TE can do. DAG is currently fixing (I hope) BF on mine but when it hits the sweet spot, wow!

 

For my shooting, the high iso up to 1000 or 2500 is very workable. I can't comment on Canon other than your example makes me happy with the M9.

[jl4069]

Jamie,

Sorry to but in here but I was wondering what you thought of the non-kodak sony CMOS chip in the new X1? Does the X1 still come off rather poorly in all those qualities you've espoused over the years about the M series? thanks, Jeremy

[Advertisement]

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

[Rick]

Sean Reid at Reid Reviews has an excellent review of the M9 which includes a detailed ISO noise comparison with the M8 and Canon 5D MkII. If you are not a subscriber it might be worth the subscription fee before deciding.

 

www.reidreviews.com

 

My experience with the M8.2 versus my D3/D700 is that the Leica fast lenses and ability to handhold at slower shutter speeds will allow me to "buy" a couple of stops of ISO. However, that is only effective when I have static subjects and don't need extensive depth of field. I shoot mainly social events so I will be using flash with my M8.2 where I would still be shooting available light with my Nikons. Not a criticism of the M8.2 - that is what I grab for my personal shooting. And I love the images I get from it with my Leica glass. That makes it worth having to bring my own light on occasion.

 

Luke, old review by Sean that he hasn't bothered to update! Sean didn't use the newer LR3 beta noise reduction. The M9 files are looking a lot better than what Sean wrote about. The 5D2 comparisons he did are not valid anymore. Also, M8 doesn't equal M9 results for comparison either. Having said that, my 5D2 files have less noise than my M9 files except they have more in camera processing evident. Pick your poison.

[KevinA]

Luke, old review by Sean that he hasn't bothered to update! Sean didn't use the newer LR3 beta noise reduction. The M9 files are looking a lot better than what Sean wrote about. The 5D2 comparisons he did are not valid anymore. Also, M8 doesn't equal M9 results for comparison either. Having said that, my 5D2 files have less noise than my M9 files except they have more in camera processing evident. Pick your poison.

 

What struck me about the difference between Sean's tests was although more noise is present in the Leica files, areas of the 5D were solid black with little noise, were as the Leica was clinging onto detail in the black but with lots of noise.

I also don't find noise tests where lighting is fairly even and plentiful much use in comparing what you are likely to get in the lighting situation you encounter when you need to turn the iso up.

A situation where you have a few bare lights and lots of fall off to deep shadows is the real test, how the camera handles the transition is equally as important as the noise in the deepest shadow.

 

Kevin.

[adli]

What struck me about the difference between Sean's tests was although more noise is present in the Leica files, areas of the 5D were solid black with little noise, were as the Leica was clinging onto detail in the black but with lots of noise.

Kevin.

 

Isn't this some of the basic characteristic of CCD versus CMOS sensors? Canon and Nikons with CMOS sensors have on chip noise reduction in the sensor which gives much better high ISO capabilities but smoothens out the details?

[Jamie Roberts]

Jamie,

Sorry to but in here but I was wondering what you thought of the non-kodak sony CMOS chip in the new X1? Does the X1 still come off rather poorly in all those qualities you've espoused over the years about the M series? thanks, Jeremy

 

Hi Jeremy,

 

I honestly can't say much about the X1; unfortunately I haven't shot it at night or played with the RAW files. From what I read, though, people are generally very pleased with the noise and detail levels, given where the X1 is positioned as a fixed-lens smaller system.

 

As Rick mentioned, there's always some tradeoffs to be made here. I have no axe to grind with CMOS or CCD technology, but I do like the Kodak does colour and detail as directly opposed to Nikon and Canon (and I'm not the only one).

 

But the X1 is intriguing to me, so maybe I'll get to playing with it later. Sorry I can't give you more of an answer here.

 

I have to say I was out shooting the Noctilux in the city with the M9 last night and was very impressed!

[Jamie Roberts]

Advertisement (gone after registration)

Isn't this some of the basic characteristic of CCD versus CMOS sensors? Canon and Nikons with CMOS sensors have on chip noise reduction in the sensor which gives much better high ISO capabilities but smoothens out the details?

 

Yes, and CCDs generally have off-chip noise reduction, since they are inherently less noisy than CMOS chips. There's a great article on this here:

 

CCD vs. CMOS

[Luke_Miller]

Luke, old review by Sean that he hasn't bothered to update! Sean didn't use the newer LR3 beta noise reduction. The M9 files are looking a lot better than what Sean wrote about. The 5D2 comparisons he did are not valid anymore. Also, M8 doesn't equal M9 results for comparison either. Having said that, my 5D2 files have less noise than my M9 files except they have more in camera processing evident. Pick your poison.

 

I agree about LR3 Beta2. I am really impressed with how it handles my M8.2 images. Even using LR2.6 with custom DNG camera profiles I preferred Capture One 5.1 over LR2.6, but LR3.2 is the best choice now IMO.

[Jamie Roberts]

I agree about LR3 Beta2. I am really impressed with how it handles my M8.2 images. Even using LR2.6 with custom DNG camera profiles I preferred Capture One 5.1 over LR2.6, but LR3.2 is the best choice now IMO.

 

Hmmm. I still prefer C1 v5 for a host of reasons, including colour. But LR is catching up, for sure. NR in both is also better than it's ever been, and it's great!

[Alnitak]

Yes, and CCDs generally have off-chip noise reduction, since they are inherently less noisy than CMOS chips. There's a great article on this here:

 

CCD vs. CMOS

 

Actually, because of design CCD's can't have on-chip noise reduction, and because of design, CMOS can. It's got nothing to do with inherent noise of either platform. In addition, while there is an inherent potential for more noise in a CMOS design, in reality for practical daylight photography, this does not come significantly into play. The integrated nature of the CMOS process actually allows for a better suppression of noise at the pixel level in the CMOS sensors--which is why CMOS is used for all the cameras that have the best current low-light, high-ISO performance. This increases the complexity and cost of the CMOS sensors, however.

 

In practice, a lot of the noise reduction with any sensor is still happening in camera. Much of what is ascribed to a CMOS "look" is really nothing to do with CMOS itself, but rather the noise reduction camera makers have done after image capture and after pixel-level electronic noise suppression. This can be demonstrated with the Nikon D3x and Sony A900, where they share a sensor, but Nikon has done a far superior job in post-sensor-capture noise suppression.

 

This is also part of the reason that the M9 is better than the M8 even with virtually the same sensor technology--Leica is now implementing a much better noise suppression regime as the ISO scales up. While I know some don't like it, I actually quite like the approach Leica has taken with noise suppression--a light hand.

 

Jeff

[Alnitak]

I agree about LR3 Beta2. I am really impressed with how it handles my M8.2 images. Even using LR2.6 with custom DNG camera profiles I preferred Capture One 5.1 over LR2.6, but LR3.2 is the best choice now IMO.

 

I would agree, and this is a very good thing IMO. The workflow in C1 is just very clunky. Clearly it appeals to some people, but it just never worked for me, and was so unworkable, that I lived with ACR's slightly lower-quality output as the other tools--LR and PS--were so superior in every other aspect.

 

I particularly like how well the chroma noise reduction works--I find its often all that's needed, leaving a very pleasing luminance grain that still allows details to show through.

 

Jeff

[squarenegs]

.....A situation where you have a few bare lights and lots of fall off to deep shadows is the real test, how the camera handles the transition is equally as important as the noise in the deepest shadow.

 

Kevin.

 

 

Kevin:

 

Well said. I think that is the heart of any discussion of noise; the quality of that "...transition..." or "...fall off to deep shadows..." in high contrast, low light situations.

 

Regards,

 

Sam

[Jamie Roberts]

Actually, because of design CCD's can't have on-chip noise reduction, and because of design, CMOS can. It's got nothing to do with inherent noise of either platform.{snipped}

 

Interesting. According to Dalsa, that's certainly not historically true: CMOS designs, to be successful at all, had to include rather aggressive on-board noise reduction to be useful at all. CCDs are inherently less noisy and more efficient. But of course you can engineer a lot of things.

 

Certainly the incredibly high ISOs of the newest Nikons aren't due, much, to sensor design but to noise reduction, as you've said. Everything else you've said is correct as far as I can verify it; but historically CCDs have better overall system response, and have been cheaper to make (noise being only one parameter) and that's one of the reasons why most of the medium format platforms are still CCD driven today, isn't it?

 

Of course, they can't match the ISO capabilities of modern CMOS designs. But they can (and do) often provide better colour results.

Edited April 16, 2010 by Jamie Roberts

[Alnitak]

Interesting. According to Dalsa, that's certainly not historically true: CMOS designs, to be successful at all, had to include rather aggressive on-board noise reduction to be useful at all. CCDs are inherently less noisy and more efficient. But of course you can engineer a lot of things.

 

Certainly the incredibly high ISOs of the newest Nikons aren't due, much, to sensor design but to noise reduction, as you've said. Everything else you've said is correct as far as I can verify it; but historically CCDs have better overall system response, and have been cheaper to make (noise being only one parameter) and that's one of the reasons why most of the medium format platforms are still CCD driven today, isn't it?

 

Of course, they can't match the ISO capabilities of modern CMOS designs. But they can (and do) often provide better colour results.

 

I think you missed the point. No one, including Dalsa, says that "aggressive on-board noise reduction" had to be used with CMOS. My point is that a CCD is by design unable to have the per-pixel noise reduction that a CMOS sensor can.

 

The primary reason that a CCD can theoretically have lower noise is that they have a higher fill factor: More of the total surface can be dedicated to the actual sensels, as opposed to supporting circuitry, of which CCD's inherently have less. However, the world has changed a great deal. We have back-lit sensors, where the supporting circuitry is now not in the way at all, and gapless microlenses, which have the same effect--both can provide in essence 100% fill factor.

 

With other changes in the sensor technology, the previous disadvantage of CMOS is now not only essentially eliminated, but now can become an advantage, as the chip can be designed to have better per-pixel noise reduction (which is not image-blurring in any way, unlike after-capture noise reduction), something that can't be done with the CCD architecture. In addition, CMOS can leverage fabrication processes and facilities designed for other chips--CMOS is the primary way nearly every IC in the world is made--meaning they can be constructed more cheaply.

 

Finally, there is no inherent reason that a CCD sensor or CMOS sensor would have any better color production. That's a fallacy has that been propagated for a long time, but there is no scientific basis for it. As for the use in MF backs, there are many reasons for that, not the least of which are historical, rather than technical. However, the main technical reason is that its easier to produce a completely uniform image on a pixel level with a CCD sensor, and it costs a lot of R&D to perfect a large uniform CMOS sensor. This means that in the MF world, even if a CCD was worse, it would still likely be used due to cost and simplicity reasons.

 

I wouldn't be surprised at all if the next digital M uses a CMOS sensor. As the bulk of sensors are now CMOS, the principles of Moore's law have begun to come into play, and they are getting cheaper and cheaper to produce at 35mm sizes. Look at how the price continues to plunge for FF digital SLR's with CMOS sensors. At the same time, the image quality on the CMOS sensors--which is already outstanding--is getting better and better. If the X1 is a Sony-sourced sensor similar to the one made for the Nikon D300, I wouldn't be surprised to see an M10 in a few years based on a version of the sensor used in the D3x and Sony a850 and a900. Now, if Leica could also license some of Nikon's knowledge in the noise reduction area, an M10 based on that technology could truly be a low-light king. But this is just dreaming and speculation.

 

Jeff

[Rick]

...Finally, there is no inherent reason that a CCD sensor or CMOS sensor would have any better color production. That's a fallacy has that been propagated for a long time, but there is no scientific basis for it. As for the use in MF backs, there are many reasons for that, not the least of which are historical, rather than technical. However, the main technical reason is that its easier to produce a completely uniform image on a pixel level with a CCD sensor, and it costs a lot of R&D to perfect a large uniform CMOS sensor. This means that in the MF world, even if a CCD was worse, it would still likely be used due to cost and simplicity reasons...Jeff

 

Yet, the M9 colors have just left me jaw dropped over the last month of using it, especially in lower light situations. I can't compare to Nikon but, I was never struck by the colors with my 5D2 like I am with the M9. I have to think that Kodak engineers have learned something about color over the last century.

 

I know that in the CRT projector world, the formulation of the phosphors used in the CRT are critical to the final color mix on the screen. Regardless, of how fine the electronics were in the projector, you couldn't adjust for great color. Certain manufactures even took stock CRT tubes and added filters over the front of the lens resulting in much better color (like Kodak's new red?).

 

No amount of adjusting an individual CRT tube's (there are three, of course R_G_ electronic parameters in the service menu's could make one projector look as good as another that had good color phosphors to start with. We tried with Spider colorimeters to adjust gamma, drive, brightness, contrast of each color tube and you couldn't match the best projectors that had the color right from the start.

 

I don't understand all of the technical aspects of CRT and CMOS like some of the better minds on the forum but, I can sure appreciate what Kodak has done with the color in the M9 and I have to believe Kodak understands how to make good color happen. Anyone else feel this way?

 

Edit: I was looking at some pictures taken yesterday and even daylight color amazes me as well.

Edited April 17, 2010 by RickLeica

[Alnitak]

Rick,

 

Don't confuse what Leica does with color and what Kodak does. For the most part, all Kodak is doing is creating the color filter array (CFA). After that, it's all up to Leica. As I am sure you know, color is as much psychological in perception as it is purely technical. If you read Puts' review of the M9, you will see a good discussion of this fact. Based on a pure technical measure, Nikon and Canon actually do a better job in nailing color. Leica has shifted the colors in a way that many find very pleasing, but is not as technically accurate. You can see this in color reproduction testing. A pretty graph can be seen here: Head-2-Head Reviews. This actually matches my perception quite well, and explains why the M9 can frequently render skin tones a bit too pink/magenta--note how the reds are shifted pretty far too the magenta.

 

So, while you and I might really like the color rendering, it's not something you can ascribe to the sensor or Kodak--it's just Leica's way of interpreting the color. BTW, this is why some people like the D-LUX series and pay a premium over their identical Panasonic roots--Leica is applying their own color interpretation routines to give the image a look that many prefer. All the technology at the sensor level and IC level is the same.

 

Jeff

[adan]

Jeff, it's ironic that on the H2H page you link to, the banner ad from MacBeth shows pinker skin tones for "better color" than for "good color"

 

Every single step in the image chain affects color, and can't be discounted. The spectral transmission of the Bayer filters, the overall filtering for IR and UV, the programming of the A/D converter, the profile used to convert from RAW to JPEG in-camera, or from RAW in one's own computer, the RAW developer chosen. What one gets form the M9 is Kodak/Jenoptik/Leica color (plus whatever personal profile one chooses to add on top of that) and they can't be separated any more than a cake can be separated back into eggs, flour, and milk.

 

So one has to ask - what profile and/or RAW converter was used to generate the files on which Imatest was then used for those graphs? If it was Leica/Jenoptik's own profile - sure, that one pushes red to the pink side. We figured that out last fall, and many of us then made proper ColorChecker profiles of our own that eliminated the problem. Did H2H?

 

As of today, I have tried out the following (less than a year old) CMOS sensors and found them all to be desaturated and subject to banded or textured noise at ISOs as low as 1000: Panny G line, Oly EP line, Sony A900/850, Canon 5D MkII.

 

The Canon 5DMkI from which I posted samples above, bands at the Hi setting (ISO 3200) so is really limited to 1600. And will show patterned grid-like noise even at ISO 200 with any sharpening. Five year-old technology, to be sure, but also big juicy pixels compared to the M9's CCD - yet not as capable.

 

I'm sure CMOS will eventually improve - but in the recent iterations I have tried it doesn't meet my standards yet. Which cameras - as of today - are actually using the Sony Exmor R backlit CMOS sensors?

[rosuna]

This can be demonstrated with the Nikon D3x and Sony A900, where they share a sensor, but Nikon has done a far superior job in post-sensor-capture noise suppression.

 

That is NOT true. Those two sensors are quite different! The A900 sensor hasn't a A/D converter, and this component is placed outside of the sensor (it is a 12 bit converter). The Nikon sensor (manufactured by Renesas) has integrated A/D 14 bits converters. This is only one of the many differences.

[Alnitak]

Adan,

 

I think we are in agreement on most of your points with the exception of your dissatisfaction with CMOS. Clearly you have your set of standards, and that's fine. However, the Panny G line and EP-1 are in a different league, and face issues of pixel density that introduces different variables, and Sony seems to be lagging in their ability to deal with noise (as are Olympus and Panny for that matter). Of all your mentioned cameras, the only one that I think does a good job is the Canon 5D II, and I agree that pattern noise can rear its ugly head, although I found it most objectionable about one stop higher than you do. Nikon is leading the pack at the moment, and the D3, D3s and D3x are at the top of the heap.

 

I have a feeling that the moment Leica uses a CMOS sensor in an M, everyone here who is dissatisfied with CMOS sensors today will suddenly sing their praises. It happened exactly that way in the Nikon world a few years back.

 

Jeff

[Advertisement]

Hallo,

Look here: Leica M

[Alnitak]

That is NOT true. Those two sensors are quite different! The A900 sensor hasn't a A/D converter, and this component is placed outside of the sensor (it is a 12 bit converter). The Nikon sensor (manufactured by Renesas) has integrated A/D 14 bits converters. This is only one of the many differences.

 

Ehh? And where is your source for this information? Nikon is officially on the record (with published statements) that Sony manufactures the D3x sensor for them. Yes, there are differences in the final implementation around processors etc, but the core sensor is the same. Much of Nikon's secret sauce is in the EXPEED processors.

 

Jeff