High ISO on M9

[Nicoleica]

Advertisement (gone after registration)

Jaap. Were those from a D3 or a D3s? The different sensor and circuitry in D3s produces high-iso files which put even the original D3 into the shade. But I also think that the lens used, photographic technique and the skill of the photographer play a much larger part in the result when we are talking about cameras of this quality. Leica lenses give a certain something to almost any photograph, and at this level of technical excellence, I feel that this is much more noticeable.

[jaapv]

It was a D3 staff camera. So beat-up too

[Jaybob]

Hmmm- As it happened I got a number of D3 files from a newspaper reporter yesterday, and I found them far less to my taste than even M8 files - it may be jus me as I'm not used to processing them, but they remained flat and digital-looking, whatever I did. Anyway, this lady told me she shot the camera with the flash permanently attached because she was not impressed with the way her higher ISO files looked - smooth yes, but the color and "look"were not to her taste. FWIIW

 

Matrix balanced Fill flash or Speedlight Output aimed directly at the subject as the main light on a D3, especially if not modified in a semi creative way, almost never looks great. Combine bad strobe output with a film speed rating of 2500 and that "reporter" is just asking for bland photos. It sounds more like she was experiencing a white balance or exposure problem, both of which she solved by using blasting her subject.

 

It's not the hammer, it was the carpenter.

[jaapv]

Quite possibly. I do not set up as a Nikon expert

[rosuna]

Well, I daresay in the files I've seen, the Nikon D3 has demonstrably better noise reduction over the 5d2.

 

And I know you're not going to believe this when I say it, but the M9 is a totally different kettle of fish (and more than a generation ahead) than the M8 when it comes to noise and higher ISO. It has more than a full stop advantage IMO, and the color balance seems to help out as well (and having more pixels might help too). The bottom line is that the M9 and D3 are really comparible, as I've said, up to ISO 2000; after that, the D3 is better (as I'm sure the 5d2 is as well). But you don't lose DR and gain much noise with the M9 compared with the M8 (and you know I shot an M8 extensively for years).

 

I would like to test the M9. I have a few pictures taken with a M9 of a friend but they are too few and random to be significant in my opinion. I have detected an improvement though.

 

I think Leica cannot do much more than they did in improving ISO performance for the M9. Firmware cannot improve the output in RAW files. CCDs improvements depends on Kodak, not Leica. Leica (or Jenoptik) might improve the A/D converter if it is placed on the motherboard (they did for the M9). But there is a ceiling to the possible improvements this basic design can reach (and the S2 is a proof of this).

 

The Nikon D3 and D3s have larger pixels, but there are also circuitry on the surface of the sensor. Anyway, the performance of those sensors seems to be much better even at low ISOs, not matter what variable you consider (dynamic range, tonal variation, color gamut). SNR performance is more similar between the M9 and the 5D Mark II, which has smaller pixels.

 

The character of the images from CCDs and those CMOS is different, and the pixel size seems to have a large impact in the final output (even more than the CMOS/CCD difference).

[mjh]

Another difference is this one: the hardware treatment in CMOSes is done at the pixel level. It doesn't affect detail so strongly as software (on digital output) does.

Yes, but this also implies that most kinds of noise cannot be reduced at all that way. The main kind of noise that can be reduced by taking only single pixels into account is fixed-pattern noise due to non-uniformities of the sensor. But this kind of noise is largely a CMOS issue. So on the one hand hand CMOS sensors are quite noisy due to non-uniformities (much more so than CCDs), but on the other hand CMOS sensors can integrate the circuitry necessary to eliminate that kind of noise.

 

If you want to reduce other kinds of noise such as shot noise, for example, you still need to do take neighbouring pixels into account which would be difficult to do pre-quantization.

 

When the digital file is cocked you cannot change the essential properties of the output.

The same goes for the sensor’s analog output. With regard to noise, the main difference between the signal pre and post the quantization step is that the analog signal is (by definition) still free of quantization noise. But that’s just a tiny part of the overall noise.

[Jamie Roberts]

Advertisement (gone after registration)

{snipped}

The Nikon D3 and D3s have larger pixels, but there are also circuitry on the surface of the sensor. Anyway, the performance of those sensors seems to be much better even at low ISOs, not matter what variable you consider (dynamic range, tonal variation, color gamut). SNR performance is more similar between the M9 and the 5D Mark II, which has smaller pixels.{snipped}

 

Ruben, you really should test an M9.... it's pretty eye-opening IMO. Just make sure you shoot enough--and print enough--to really tell the character of what the camera can produce.

 

As a guide for anyone shooting the M9 at high ISOs, I would say consider with LR3 (or careful tweaking in C1) an effective ISO 6400 as your printable noise floor / limit.... IOW if you're shooting ISO 1600 your shadows have two full stops of detail and acceptable printability. For high-contrast subjects, you can go to ISO 2000 and clip the bottom (because you have a third-stop less or there abouts).

 

That works fine for me; however, as I've said, you can shoot the D3 above ISO 2000 because it probably has a stop more (or maybe 1.5 stops) of printable shadow response. I'm sure the D3s is even better in this respect.

 

But--and this is a big but--I can't say for the D3s, but the D3, even at low ISOs, under most lights, has an extremely weird "perceptual" colour rendering, regardless of the RAW processor. And it's extremely sensitive to different light sources, too.

 

Many people I know have rejected the D3 or do what I do--work very hard in post to get decent colour, especially skin tones (we were discussing just how poor Nikon and Canon colour is compared with a DMR over in the digital forum )

 

Don't get me wrong: I still use a D3 and when you work the colour it's very good. But the M9's is much, much "better" and easier to control (which has a real business benefit to me). Perhaps that's due to a smaller gamut, actually--since printing is very gamut-limited in any case. Or maybe Kodak knows a thing or two about colour that Sony doesn't

 

For DR at low ISOs, there absolutely no printable difference that I've seen between and M9 and a D3. None at all, and again, the Nikons have some very hard to quantify color issues especially with skin (it's almost like the sensor is variably sensitive to IR or something, to tell you the truth).

 

As an aside Canon's are equally tweaky for colour, though they're consistently "off" in my view, which makes them a lot easier to deal with, since a single set of corrections for a given white balance pretty much nails it... not true for the Nikon.

 

But we're splitting hairs here. My point is really that with the M9 I can print--at sizes large enough to compare with the D3--at un-thought-about ISO levels with an M9 compared to an M8.

 

Because of that, to bring this back on track, I wouldn't be surprised if the M9 offers a push ISO 5000 in a future firmware. It will be noisy, for sure, but also useful.

[chris_tribble]

Jamie - thanks for thoughtful post. What you say very much reflects my experience working with 5DmkII and M9. With LR3 1600 is VERY useful on the M9 - and 2500 will deliver usable results for my documentary projects. The 5D2 is absolutely not a bad camera - it's very good! But the M9 is now by far and away my preferred option so long as I don't need to use lenses longer than 90 (I like the 135 apo telyt a lot - but compared to my Canon 70-200 and 200 and 300 primes it's a bit hit and miss!)

[Jamie Roberts]

Jamie - thanks for thoughtful post. What you say very much reflects my experience working with 5DmkII and M9. With LR3 1600 is VERY useful on the M9 - and 2500 will deliver usable results for my documentary projects. The 5D2 is absolutely not a bad camera - it's very good! But the M9 is now by far and away my preferred option so long as I don't need to use lenses longer than 90 (I like the 135 apo telyt a lot - but compared to my Canon 70-200 and 200 and 300 primes it's a bit hit and miss!)

 

Hey Chris,

 

I think if you shoot the M9 for events for any length of time (and you're used to a Canon or Nikon), it's a revelation compared with an M8, and a totally worthwhile low-light companion given the strides made in software. There is certainly something lacking in the DxO analysis of the M9... and the quality of the thing shows up in prints all the time.

 

As for the 5d2 being a good camera, I've been very impressed with what I've seen of it, and I know folks who have been using it for pro video with outstanding results.

 

I've been thinking of a 5d2 just for R glass (well, and the 50 1.2L Canon, which to tell you the truth, I miss a lot). The idea of shooting video--even a little personal video--with an 80R Lux or a 180 APO Elmarit or a 19R just "gets" me and won't go entirely away from my brain

 

But then to use some of my other R favourites I'm into shaving mirrors and that kind of stuff--same on the D3.

 

But that's way off topic

[rosuna]

Printing is an equalizing transformation of the outputs. You can also "simulate" this in DxO graphs (they are computed for "screen" or "print"). In fact, as you says and is expected, the results become more similar in "print" mode. But prints (on paper) are not the only form for final pictures in this digital age. You cannot evaluate the capabilities of a system looking only at prints.

 

Moreover, prints have many different characteristics, depending on paper, ink, printer, print technology, etc. And you can create different versions of a picture for different prints, very different prints. In the end, even for prints, the more possibilities in the original file you have, the better.

[chris_tribble]

I've been thinking of a 5d2 just for R glass (well, and the 50 1.2L Canon, which to tell you the truth, I miss a lot). The idea of shooting video--even a little personal video--with an 80R Lux or a 180 APO Elmarit or a 19R just "gets" me and won't go entirely away from my brain

 

Jamie - tempting to use the R glass, but I've been using a Series 1 Canon 85/1.2 L on the 5D2 and it's remarkably good (+ I had to do a bit of video on a recent assignment and was astounded by how effective the combination was - interview stuff, so nothing arty, but great to have one cam to do both jobs)

 

Printing is an equalizing transformation of the outputs. [...]

Moreover, prints have many different characteristics, depending on paper, ink, printer, print technology, etc. And you can create different versions of a picture for different prints, very different prints. In the end, even for prints, the more possibilities in the original file you have, the better.

 

Rosuna - this has always been the case hasn't it? We make the image in the darkroom (analogue or digital) - we capture the image with the camera. The lens / sensor are critical to the drawing of this image - but it's still only gives us source material that has to be interpreted in post-process...