D700 vs. M9, high iso.

[thrice]

Ulrik is right in upressing the D700 shots to compare noise against the M9. This evens the playing field if we're considering 100% crops, alternatively you can down-res the M9, thus giving it a distinct advantage.

There is an explanation as to why this is a fair testing methodology elsewhere, but I doubt this particular lynch-mob would care.

Think again. Here's a 5000 ISO M9 shot for you:

http://www.nikonweb.com/m9/M9_5000_ISO_L1000465.jpg (760 KB JPEG)

Shot at ISO 2500, exposure set to + 1 in Adobe Camera Raw. People: It's time to realize that Leica stopped at 2500 ISO for a reason. At lower ISOs, the M9 will produce excellent results - but it's not a high ISO camera. If that's what you really need, get a Nikon D700 (or similar) instead.

Jarle

Jarle, here is that image, run through my noise reduction workflow (takes <2 mins for each image). I'd say it compares favourably to the D700 and I'd rather do this and have the benefits of a CCD sensor and M-sized body than use my 5D-Mark II. Ulrik's ISO5000 shots cleaned up much nicer than this one, probably because it's underexposed even when pushed +1.

I think Leica stopped at 2500 ISO because any more than that is honestly unnecessary, if you need more light, get a flash. I use 1600 and RARELY 3200 max.

 

[Tracnac]

I understand what he is saying. At 18mp the M9 will have more resolution and more noise. That is the take home point.

How can 12mp have more resolution than 18?

?! Oh common you have so MANY examples in dpreview or everywhere else... MP does not mean more resolution !!!! My wife have a P&S that have more MP does not mean more detail than my D700 sorry...

My personal taste is that a D700 and M9 is not in the same league. But you can afford with the noise of the M9 is just how you love noise. For example I have a Ricoh GRD (Very crappy noise even at 64 iso...) but when you shoot at 1600 and process in B&W you just love the grain.

Yvan

 

[thrice]

?! Oh common you have so MANY examples in dpreview or everywhere else... MP does not mean more resolution !!!! My wife have a P&S that have more MP does not mean more detail than my D700 sorry...

For a given EQUIVALENT sensor size, given a relatively low base noise, yes it does mean more resolution, at least at base ISO and given enough light. Also assuming similar performing optics are used.

I see you are citing dpreview... their noise comparison tests are completely out of wack.

 

[ulrikft]

As thrice here points out, you have to upres or downres to compare per-pixel 100% crops, that is necessary to even the playing field.

When it comes to detail, well, a tiny sensored compact camera with 12 megapixels is beyond the threshold. The d3x, a900, 5d MKii and 1ds mkIII all show that high megapixel levels in a fullframe camera, is not a problem, but an asset. With more testing, I think I can put the M9 in the same cathegory. I'm not saying that people shall, must or have to like the M9 or its noise nature and levels. But please just compare it to those that are best at noise at present date, on a fair basis.

 

[gnarayan]

As thrice here points out, you have to upres or downres to compare per-pixel 100% crops, that is necessary to even the playing field.

Your results for noise and resolution now entirely depend on how you interpolated. Did you use a spline? Bicubic? Did you use a lanczos kernel. Want to see what sort of differences that makes?
http://www.all-in-one.ee/~dersch/interpolator/interpolator.html

Do you know if there was sharpening applied after you uprezed. How do you know for sure? Did you actually write the code? Did you use Genuine Fractals maybe.

Simply, you changed the resolution and then you want to compare resolution. What the heck do you mean by 100% per-pixel crops when you upres and change what a pixel is? An 18 MP sensor is not a 12 MP sensor. What the heck is an "even playing field" here actually mean?

Cheers,
-Gautham

 

[ulrikft]

Your results for noise and resolution now entirely depend on how you interpolated. Did you use a spline? Bicubic? Did you use a lanczos kernel. Want to see what sort of differences that makes?
http://www.all-in-one.ee/~dersch/interpolator/interpolator.html

Do you know if there was sharpening applied after you uprezed. How do you know for sure? Did you actually write the code? Did you use Genuine Fractals maybe.

Simply, you changed the resolution and then you want to compare resolution. What the heck do you mean by 100% per-pixel crops when you upres and change what a pixel is? An 18 MP sensor is not a 12 MP sensor. What the heck is an "even playing field" here actually mean?

Cheers,
-Gautham

I did not write either photoshop or genuine fractals, no, these rethoric questions are.. quite useless imo.

If you want to compare two cameras with differing megapixel resolutions, the easiest and best way is to print them both on the same sized paper, as I have done (A3+). comparing on a 100% crop basis without resizing would be like comparing a M9 shot printed on a larger paper than the D700. So the problems with resizing are dwarfed by the problems you encounter by pixel-per-pixel comparison between these sensors. A level playing field is for me comparing them as close as possible to a print-test where i print them both on the same sized paper. (as I have done, as said). If you disagree with this, fine, but argue your point instead of being defensive and angry.

 

[gnarayan]

If you want to compare two cameras with differing megapixel resolutions, the easiest and best way is to print them both on the same sized paper, as I have done (A3+). comparing on a 100% crop basis without resizing would be like comparing a M9 shot printed on a larger paper than the D700.

So the problems with resizing are dwarfed by the problems you encounter by pixel-per-pixel comparison between these sensors. A level playing field is for me comparing them as close as possible to a print-test where i print them both on the same sized paper. (as I have done, as said). If you disagree with this, fine, but argue your point instead of being defensive and angry.

Look, if you are happy with the M9, then great. I'd personally rather have it any day over the D700 but that isn't an option. If you are happy with the high ISO, resolution, color and the prints from the M9, then that is all great.

All I'm saying is that your posted images don't really give you any useful information about the noise and resolution of these two cameras. The lenses are different, the sensors are different architectures, the software processing is different, and then you upres using some unspecified method and got two 18 MP files that you compared and posted here. Based on the images you posted, your claims are of the relative merits of the M9 and D700, are just not supportable IMO. That is just a statement. You should not read in any judgement from me about you because there isn't any. I've absolutely no interest in getting angry about you or your posts. I don't know you from Adam. Very bluntly, I'm saying that your pictures of Epson boxes (while very lovely for Epson boxes) are a pretty bad way to compare cameras.That bears no reflection on you. I'm sure you are a lovely person.

There are plenty of good ways to compare resolution and noise and a whole bunch of camera websites will be doing just that in a bit.

 

[ulrikft]

Look, if you are happy with the M9, then great. I'd personally rather have it any day over the D700 but that isn't an option. If you are happy with the high ISO, resolution, color and the prints from the M9, then that is all great.

All I'm saying is that your posted images don't really give you any useful information about the noise and resolution of these two cameras. The lenses are different, the sensors are different architectures, the software processing is different, and then you upres using some unspecified method and got two 18 MP files that you compared and posted here. Based on the images you posted, your claims are of the relative merits of the M9 and D700, are just not supportable IMO. That is just a statement. You should not read in any judgement from me about you because there isn't any. I've absolutely no interest in getting angry about you or your posts. I don't know you from Adam. Very bluntly, I'm saying that your pictures of Epson boxes (while very lovely for Epson boxes) are a pretty bad way to compare cameras.That bears no reflection on you. I'm sure you are a lovely person.

There are plenty of good ways to compare resolution and noise and a whole bunch of camera websites will be doing just that in a bit.

First i just want to ask, to you understand the mechanics behind me upressing/printing the same size?

Secondly: when comparing different camera systems, you cannot use the same lens, and having different sensor designs is a part of this too. It is impossible to take the different sensor designs out of the test, and it would also be counter productive, as the sensor is what I'm testing. Wether or not one like printer boxes (lots of different colors, small detail and different gradients, quite an ok subject imo). Lenses are a different problem. I can't use a leica 50mm on the nikon and i can't use a nikon 50mm on the leica. The sigma is among the best normal lenses in this mount, the leica lens is also great. When it is impossible to remove variables, you minimize them as much as possible. You put unreasonable demands up, which cannot be met, and I think you know that they cannot be met. You are also welcome to give me a place to upload the nikon and leica raw-files, if you want to push, resize, compare yourself.

The main problem with many sites, is that they don't use the resize step. And only compare per pixel noise levels, and as said, this is not a very accurate or precice way of testing this feature.

 

[markgay]

Despite the occasional bar room brawl, I’ve learned a few things from this thread and enjoyed all the posts.

(The launch of the M8 and even the film-based M7 sparked some feuding among the family of rangefinder users. Just like then, some of the neighbours came round to join in the fight).

My 2 cents: Low ISO performance also counts.

CCD sensors arguably perform better in good light. And when you have high quality, fast glass, you can keep using that better mid-range performance for longer. That's Leica's approach.

CMOS seems better for (but also reliant on) in-camera software processing. This is a good and a less-good thing. System-on-Chip allows more features and image processing. Is it a clean image? Is it the closest you can get to a digital negative? Don't come back to me on this. Read about it here.

Another point: It's easier to see noise at high ISO. It shows up nicely in Jpegs so it gets lots of Internet discussion. Does that make it the biggest quality issue with sensors? Arguably, it's just an issue that seems easy to illustrate.

NB it's not that easy:
I respect those posters who focus on the noise issue and, at the same time, aim for the most neutral methodology.

Some people say sensors should not be compared using brightly-lit, daylight shots because that won’t show up the faults (on a jpeg on a PC screen). But it could show up the strengths (though maybe only in a print or on a very big, high quality monitor).

However, there is an even bigger difference than the sensor which affects the quality of image you get. The lenses. It affects every picture you take, regardless of ISO. It's not so easy to illustrate on the 'net (it's hard to do given different camera systems but that does not make the issue go away). And it gets less attention.

(There is yet another, humongous difference in image quality, related to the skill level of the photographer but that gets almost no mention at all because it's rude. 🙂)

Super high ISOs are called 'turning up the gain' by TV cameramen and they know it's best avoided. And they were using sensors in television cameras for years before still photographers got the chance 🙂 .

Moving images are the torture test for sensors. CMOS sensors, for example, have the rolling shutter issue due to the way they capture the image. So have a read of how sensors compare in moving image cameras and you learn a lot about why different still camera makers use what they use.

Finally, lets call high ISO by its true name. Sensors have a fixed sensitivity. If you want to go above that, you literally turn up the amplifier. Strictly, only film has ISO. Digital borrowed the phrase.

So turn up the amplifier if you want, but know what you are doing. You will get more hiss, and then distortion, as the signal to noise deteriorates.

Record a clean image, and you can play it back “louder”. Let it distort in the first place and it'll ever “sound” natural.

Of course, with images, you can smother the noise (see why it’s called noise?), wipe it all over so it isn't so obvious. Record producers use similar tricks.

That’s why we have remastered versions.

Regards,
Mark

 

[Tracnac]

For a given EQUIVALENT sensor size, given a relatively low base noise, yes it does mean more resolution, at least at base ISO and given enough light. Also assuming similar performing optics are used.

I see you are citing dpreview... their noise comparison tests are completely out of wack.

Sorry the answer is just NOT so easy. (low pass filter, process, density etc...) many thing are composing RAW. Even sometime the RAW is not so RAW...

Yvan.