M9 noise performance

[hoppyman]

Advertisement (gone after registration)

Andy I respectfully disagree. My understanding is that the White Balance value recorded with the DNG file, whatever setting the camera may have, does not affect the data recorded in any way. Nor does it have a colour space set for that matter.

 

Remember I am speaking only about DNG. For JPG capture or for any changes after a conversion to TIFF or PSD or whatever it certainly can matter of course, absolutely no dispute on that. Similarly the low quality (JPG) preview that you are seeing on your LCD is affected too of course since it must have a white balance set. Its histogram is limited by that as well (plus by the Colour space that the LCD can represent anyway).

Watch the preview for a couple of seconds and see it apply JPG values.

 

Sandy or someone by all means step in and tell me that I am wrong and why.

[pico]

I don't freaking care about high ISO performance. The M9 operates perfectly well within my habits, as one coming from film. I guess I'm drawing the line for myself. Everything is good.

[Jamie Roberts]

{snipped}Please let's not involve the Nikon D3 - UNLESS it is the D3X. The D3/D3s has lower noise, but also significantly lower resolution (13 vs. 18 Mpixels). It's like trying to compare Provia 400 to Velvia - sure the Provia is faster, but it captures less detail.

 

Well, respectfully, I "involve" the D3 when I discuss these things because that's what I shoot and it's a low-noise benchmark for me. Comparing the 2 is what I do every week

 

It's also a very good low light, low-noise camera... even by today's improved standards.

 

Comparing them is not like comparing different films. And FWIW, I'm not talking about detail... just 2 totally comparable things: noise and colour rendition (though it's true that detail does "go away" as in camera NR is applied on the D3 as it does with the 5d2 and the D3x as well).

 

As I said previously, the D3 has a better noise profile at high ISOs, and noticeably higher room for playing with shadows beyond ISO 2000, but the tradeoff (for me) is color shift as ISO rises.

 

@ Edmund, the M9 is just fine at ISO 1600, though it probably lacks a stop of DR compared with the most recent NIkons. Not that you'd ever see that difference in print anyway (unless you were doing some very strange compression, which folks do these days, it's true).

[Jamie Roberts]

Jamie -- I absolutely respect your view on this, considering this is your profession and you have been doing this for more than 10 minutes. My question is, in the days of film, what did you do? and can't m9 iso of 1600 suffice today?

 

That's a great question, and it's true--short of asking the lab to push film out to ridiculous places you lived with film between 400 and 1600.

 

I shot a lot more black and white then too due to color incompatibility with different light sources. And truthfully, people / clients expected less of low-light photography too at events. Although there were always folks who went the extra distance to make the photography blend with the ambient.

 

But today I want to do both more simple available light AND a lot more light blending in extremely dark venues. Even when I'm using flash I don't want it to really "look" like flash from the old days--and to a great degree cameras like the Nikon and Canon let you do that with their low-light capability.

 

At a clean ISO 3200 and an f1.4 lens I can record ambient light levels at even extremely dark receptions OR touch up the direction or quality of light with multiple flashes, and use very little flash power.

 

I can do this with the M9 as well, but not to the same degree I can with the D3. Sometimes with the M9 I run out of sensitivity or the ambient is a wee bit too low (+2 stops under)--even with a Summilux--and it's much harder to balance with other high-quality light. Right now, I pull the Noctilux out and gain another stop, which is great, but I can only do that with a 50mm lens

 

Anyway, sorry for the long response. The expectations from clients really have shifted from the days when you'd illuminate an entire hall with steady-state lights or very powerful strobes and shoot at ISO 400. These days you often want to just augment the ambient light, and that requires a relatively high ISO many, many times.

[thighslapper]

Andy I respectfully disagree. My understanding is that the White Balance value recorded with the DNG file, whatever setting the camera may have, does not affect the data recorded in any way. Nor does it have a colour space set for that matter.

 

Remember I am speaking only about DNG. For JPG capture or for any changes after a conversion to TIFF or PSD or whatever it certainly can matter of course, absolutely no dispute on that. Similarly the low quality (JPG) preview that you are seeing on your LCD is affected too of course since it must have a white balance set. Its histogram is limited by that as well (plus by the Colour space that the LCD can represent anyway).

Watch the preview for a couple of seconds and see it apply JPG values.

 

Sandy or someone by all means step in and tell me that I am wrong and why.

 

You are correct.....

 

Things like white balance etc. are appended to the DNG file as an offset that is applied to the image when it is rendered by the RAW converter

[nggalai]

Things like white balance etc. are appended to the DNG file as an offset that is applied to the image when it is rendered by the RAW converter

 

Yes, but that wasn’t the point discussed – as far as I “got” Adan.

 

Point is artificial light + high-ISO performance. With pretty much any digital camera you’ll have a less noisy image when using the same ISO setting in light with a full spectrum (say, daylight, full-spectrum studio lighting) compared to most common artificial light situations (say, nighttime living room, concerts).

 

On the M9, an ISO 1600 shot taken in broad daylight might appear noisy in close-up, but you won’t have too much chroma noise. Anyway, it might appear less noisy than an ISO 800 shot taken in a brightly lit room using “normal” consumer bulbs.

 

Hence, in this discussion, white balance correction indicates light quality. This also means necessary white point correction might have an influence on a picture’s overall noisiness simply because it was necessary from a physical point of view to begin with.

 

Different RAW processors and cameras deal differently with it, and use different tricks/algorithms for minimising softness due to noise reduction.

[t024484]

Advertisement (gone after registration)

Andy I respectfully disagree. My understanding is that the White Balance value recorded with the DNG file, whatever setting the camera may have, does not affect the data recorded in any way. Nor does it have a colour space set for that matter.

 

Geoff,

 

Andy was perfectly right in his explaination.

When generating a picture with LR, Photoshop or whatever, the recorded sensor values for the colors Red,Green and Blue will be multiplied with different numbers and mixed, to form the final RGB value at a certain white balance or color temp value.

Changing the white balance means changing the multiplication factors.

The recorded data is not affected, but the transformation matrix to get your final RGB values is affected by changing the white balance.

 

If a light source does have a far from flat flat spectrum, the sensor channel(s) receiving less light because of this will have to be amplified much more, to achieve that white or grey is correctly displayed in the final image, to compensate for this

Noise will also be extra amplified, leading to a noisier picture than would have been produced by a light source with a more even spectrum.

So in effect, depending on the color temp, changing the white balance can result in more noise

 

Hans

[Ronan]

Has anyone said you get Nikon noise or DR at high ISOs? I find that misleading, actually.

 

You're right about dropping DR as you go up in ISO (DxO aside--they're doing something weird however in their tests). But you'd also be right on that with a Canon or Nikon--they also loose gradation, and tonality, and detail too.

 

The point is that my D3 "loses" it between one and two stops higher than my M9. Not 4 stops, not 3 or more. About 1.5... and that's significant yes.

 

The other point is colour shift: Nikons in particular are all over the place as the ISO increases. Yes, I can shoot at ISO 3200 and produce a nice print. I can push it to 6400 if I have to. But by then the files are mush and colour is terrible.

 

So I'll stand by what I said previously: Up to ISO 1600 on the M9 and I don't worry about it at all--it's better than the M8--no doubt about it. ISO 2500 on the M9 requires careful exposure and lack of DR mitigating choices, but if you've ever shot print film you're doing the same thing

 

Nikon territory ranges above that, so that even at equal ISOs (the M9 is 1/3 stop more light sensitive anyway) I have 1.5 more stops of exposure in the shadows if I want it... but generally I don't for a print if I haven't made a mistake in exposure.

 

No-one is saying they're equivalent technically. But practically, they are not as far apart as some people think and there are tradeoffs for really high ISO response. And those are facts too

 

What? My D3 files were clean up to 3400 ISO, above that, a bit of noise supression was needed. The D3s beats the pants of the D3.

 

I never had color shifts on any of my Nikons... what lenses and profiles do you use?

[hoppyman]

Hans and Sascha, thank you for contributing further on this. I understand the ramifications of correcting white balance after the fact and the potential noise differences in different channels as a result in this scenario. You are describing (as Andy is) developing decisions when you are converting and then developing from the DNG. Both descriptions are good insights into that conversion and developing. That's fine.

 

My only contention was and is that the White Balance value set by the camera has no effect on the recorded data in DNG. That is to say whatever developing decisions that you make in converting, that setting in no way limits nor affects the recorded sensor RGB values. For personal interest I am performing the experiment but I think that we are simply talking about diffferent things.

 

If anyone is able to correct me there with explanation then of course I am very willing to learn.

Edited September 12, 2011 by hoppyman

[lct]

I'm confused here. What's the link between what is recorded in DNG and the fact that adjusting WB in PP may add noise?

[Jamie Roberts]

What? My D3 files were clean up to 3400 ISO, above that, a bit of noise supression was needed. The D3s beats the pants of the D3.

 

I never had color shifts on any of my Nikons... what lenses and profiles do you use?

 

Not sure why you're non-plussed; I said the D3 files are just fine to ISO 3200 and workable (but mainly black and white) to 6400, no?

 

Good to know the D3s is even better in this regard. I'm looking forward to seeing what the "D4" will do.

 

I do hope they "fix" or "tweak" the colour response, however. Maybe "shift" is the wrong term to use but I don't know what else to call the D3s colour weirdness as ISO goes up beyond 3200. I use the 85mm 1.4 extensively (and have used the 35 1.4, a couple of different 50s, the 24 1.4 and both 24-70 and 70-200 zooms). The 85 is a very nice lens and doesn't degrade the colour in the slightest that I can tell.

 

For raw development, I use Capture One V6.2 and Nikon Capture. As the ISO goes up, the D3's colour weirdness does as well, to the point where hands and shoulders can be completely different colours entirely in the same light--and different colours than they are at lower ISOs!--which is frustrating to say the least.

 

It's just *weird,* and not exhibited by other cameras from other manufacturers especially (though not exclusively) bad in tungsten. It does weird things with skin and cyan too (the way Canon does with magenta)... I prefer the colour from the M9 (or DMR or M8 for that matter) over the D3.

 

I do appreciate that many people don't care about stuff like this. And I'm not saying I don't like the D3. I do...it's just more work a lot of the time when I use it at higher ISOs. The fact that it works at all is amazing, of course, and I'm sure the next set of cameras from them will be even better.

 

FWIW, I don't use Lightroom for production right now, though I'm actually in the middle of giving the latest version an in-depth test right now. It certainly has the best noise reduction of any of the raw converters I've seen yet--and maybe that's where our Nikon mis-match is coming from. However, I need to be convinced on Lightroom's colour though (that's for every camera I've tried; S2, M9, M8 and D3)...so far it's much better than previous versions but I still prefer Capture NX and C1. We'll see how that goes.

Edited September 12, 2011 by Jamie Roberts

[adan]

My understanding is that the White Balance value recorded with the DNG file, whatever setting the camera may have, does not affect the data recorded in any way.

 

I never said it did.

 

Similarly the low quality (JPG) preview that you are seeing on your LCD is affected too of course since it must have a white balance set.

 

Yep. The original poster said he was making his noise assessment based on the image on the LCD.

_____________

 

Jamie - So if Nikon decided to improve the D3's or D3s's ISO range even further by knocking the camera down to 7 Mpixels (with even bigger pixels for gathering light) - you'd be OK with that?

 

Why not go for a FF 1 Mpixel sensor? You could probably get ISO 160,000 that looks like the M9's 160. Not a lot of detail, but it would be noiseless.

 

An even better "benchmark" than the D3 - but so what? The images would barely fill the screen of an iPad.

 

The D3 has 8.5-micron pixels. The M9 has 6.8-Micron pixels. The proverbial apples and oranges.

[adan]

Things like white balance etc. are appended to the DNG file as an offset that is applied to the image when it is rendered by the RAW converter

 

Only if one has set the RAW converter WB default to "As Shot" (which I never do) - otherwise it is not applied.

[hoppyman]

Yep Andy. I only wanted to clarify that the camera WB setting in DNG has no effect on the adjustments when developing that you have been describing.

This highlights a difference compared to saving as JPG in camera or editing an exported file in a different format.

 

I think that your information would be a very useful addition to the M9 FAQ.

[Jamie Roberts]

{Snipped}

 

Jamie - So if Nikon decided to improve the D3's or D3s's ISO range even further by knocking the camera down to 7 Mpixels (with even bigger pixels for gathering light) - you'd be OK with that?

{snipped}

 

I don't disagree entirely, Andy--it's all about tradeoffs for sure. For instance, I'm sure the new Lytro cameras won't have much more than 1mp and probably less--and what you get there is something beneficial--"light field" capture.

 

But what you've done with your 1mp low-noise example here is still a bit "reduced to the absurd" so to speak For work where you're going to print, there's obviously a cut off / tradeoff on useful resolution.

 

But it's not just down to pixel size and count, though of course that makes a huge difference. Again, practically speaking, a 12mp and 18mp camera aren't that far apart for 90% of the prints I'm going to make.

 

But in a way you're right: when I want more detail, I use the M9. When I want higher ISO capability, I use the D3.

 

Both have enough resolution to print, and since I deliver the results together, it's still practically useful to compare them. I agree though, it's more like comparing different drill bits for different materials, if not apples and oranges.

[ho_co]

Regarding 1 mp performance--remember that the camera aboard the Cassini spacecraft is a 1 mp camera, 1024 x 1024 pixels.

 

With a bit of stitching, that will give you APOD: 2011 September 4 - In the Shadow of Saturn.

 

And with a bit of time to animate some hundreds of thousands of still images, it will give you Watch a Saturn fly-by video composed from real photographs | Tecca.

[rosuna]

Regarding 1 mp performance--remember that the camera aboard the Cassini spacecraft is a 1 mp camera, 1024 x 1024 pixels.

 

It takes pictures of things that are far, far away...

[mjh]

I would be very surprised not to see new sensor technology like a backlit CMos on the M 10

Not backlit surely but otherwise yes, a major switch in sensor technology seems likely.

[mjh]

My only contention was and is that the White Balance value set by the camera has no effect on the recorded data in DNG. That is to say whatever developing decisions that you make in converting, that setting in no way limits nor affects the recorded sensor RGB values.

Sure enough, only nobody really cares about noise in the raw data. Remember that we never see raw data; all we see are images developed from the raw data, whether in-camera or by some raw converter. Now when the inevitable colour tint created by tungsten lighting is corrected for, this amounts to amplifying the blue channel and all the noise contained therein. Yes, that noise added by amplification is created during conversion, but we need to convert our raw data so there’s an image we can see; there is no way around that. Unless we accept that tungsten lighting produces an orange tint and leave it at that.

[Advertisement]

Hallo,

Look here: Leica M

[sblitz]

we can accept that, i have tons of kodachrome slides with a warm orange hue.