5D Mark 2 Noise Question

[jdavtz]

Originally Posted by blanka

I see mostly banding problems. These can be the result of wrong RAW conversion. The noise is not too bad in most pics.
This one however:
http://forums.macrumors.com/attachme...1&d=1354255607
Looks like a mess. Is this ISO 2000? I don't have a canon, but if someone says it is as good in the dark as a d3/D700: if I see this, it is not even near. It has horrible chroma-shifts in the noisy areas.

That one is ISO 25600

[Crazy Badger]

Don't look noisy to me either, but very few are in focus!

[Redneck1089]

Originally Posted by blanka

I see mostly banding problems. These can be the result of wrong RAW conversion. The noise is not too bad in most pics.
This one however:
http://forums.macrumors.com/attachme...1&d=1354255607
Looks like a mess. Is this ISO 2000? I don't have a canon, but if someone says it is as good in the dark as a d3/D700: if I see this, it is not even near. It has horrible chroma-shifts in the noisy areas.

That's a 25, 000 ISO picture I snapped quickly in the basement. In terms of chroma noise and banding, it's actually not too bad for such a high ISO. In fact, the newer 5D3 does the same. You get the banding and chroma aberrations when the sensor is pushed too far and the exposure isn't set properly.

The 5D2 kills the D700 in noise performance, and the D3 is a professional work horse camera with 12 megapixels; 12 megapixels is a lot less than the 21 MP the 5D2 has, and for the D3's lack of megapixels and massive price tag it damned be better at noise at high ISO.

[shigzeo]

Originally Posted by Prodo123

I honestly don't see any noise.
At least irrecoverable noise. Whatever graininess is in those photos can be repaired through RAW processors.

Yikes. If that is considered noise, I can't imagine what camera would satisfy the OP. For 25.800, that is an amazing profile, one I'd happily shoot with. Those images are either nearly noise-free or typical of those ISO settings.

OP: What miracle camera were you using before that showed even less noise? Canon 22D MK17, transferred to our universe by Spock? Those images are clean.

My guess is you came to photography recently. Any new digital camera will create noise-free images like those you posted. Any camera prior to 2007 will produce noise like your 25800 image at about ISO 400-800. If you ever shot film, you will realise that even some of today's point and shoots can produce less noise than typical ISO 400 film.

[OreoCookie]

Originally Posted by shigzeo

If you ever shot film, you will realise that even some of today's point and shoots can produce less noise than typical ISO 400 film.

Seconded. A while ago, I posted 16 MP scans of b&w ISO 125 film (Ilford FP4 Plus, I believe), and the film grain in the photo was not only visible, but strongly visible. So much so, it added character to the result

Being able to shoot at ISO 6,400 back in the film days was ludicrous, the fastest film I remember I have ever used was ISO 3,200 (I think that's ISO 1,000 film which is pushed, but I could be wrong). But such fast film was rather expensive.

[shigzeo]

Originally Posted by OreoCookie

Seconded. A while ago, I posted 16 MP scans of b&w ISO 125 film (Ilford FP4 Plus, I believe), and the film grain in the photo was not only visible, but strongly visible. So much so, it added character to the result

Being able to shoot at ISO 6,400 back in the film days was ludicrous, the fastest film I remember I have ever used was ISO 3,200 (I think that's ISO 1,000 film which is pushed, but I could be wrong). But such fast film was rather expensive.

Personally, I've never shot faster than 1600. I tend to keep film at 400 or less and push if I really want to, but I'm happy enough with motion blur and other artefacts.

[wolfpuppies3]

I use the 5DIII and 5DII as well as a iDIII along with a dozen Canon lenses. I always always shoot raw and only extremely rarely go as high as ISO 800, 400 is my usual upper limit. That is my choice going back to a self imposed ASA 400 limit in my film days. I don't like grain. None of my bodies show any noise. I've never used "picture style settings" as I do not shoot jpegs.

http://imaginethatimages.zenfolio.com/

[snberk103]

@OP 1) Somehow, there is a way to reset the camera to "factory defaults". With a used camera you should do this, because you don't know what kind of settings the previous owner set. All of the advice you are getting are from people who are assuming the camera is at, or close to, the factory defaults.
2) When you are taking pictures of white walls, the camera is underexposing them. Cameras tend to darken pure white to grey, and lighten pure black to grey. Your camera may have been intelligent enough to recognize the black background in the night photos - so it wasn't mucking up that exposure. But the white wall would definitely have confused it. When it underexposed the white wall (to darken it) you would have seen some more noise.
Figure out how the histogram works, and then use it. Rule of thumb is to push it to the right, without clipping. The RAW files will tend to look overexposed at times, but you can adjust that in iPhoto, Aperture, Lightroom, or another RAW processor. But you will maintain all the detail.

Originally Posted by Prodo123

...
Try to keep the shutter speed low because with higher speeds comes a certain graininess caused by the sensor scanning too fast.

That's news to me, and I've been in the business for while ... Do you have a link for that? I'd like to learn something new if this in fact true.

[righteye]

I urge you to go over to diglloyd ( there is other info on the WWW as well, Michael Reichmann of The Luminous Landscape was the person who brought this into the popular domain but not sure who devised the theory) and read what he has written about ETTR (exposure to the right)
Make the effort and you will be rewarded with much better Photographs.
What looks to bright initially,with the correct adjustments will give much less noise in the final photo, the LCD screen may well look Very bright and washed out but this can be whats needed in the RAW file to get a good exposure. In Lloyd's words "ETTR is a must-master digital skill"
His site is Subscription but it will be some of the best money you spend
Also the high iso settings you are using are emergency use only in my opinion, if you are going to use iso that high think about some specialised noise reduction software it will help a lot but there is no free lunch and some detail will be lost.

http://diglloyd.com/prem/prot/DAP/Wo...-exposure.html

OR a free taster is his blog; http://diglloyd.com/ page down till you find ETTR if you get that far as your eye may well be caught on some other useful titbit of information.

EDIT; iso 50 is not recommend on the 5Dii again more emergency use if you have no other means of lowering the exposure such as a neutral density filter

[Prodo123]

Originally Posted by snberk103

That's news to me, and I've been in the business for while ... Do you have a link for that? I'd like to learn something new if this in fact true.

I've only discovered it accidentally this summer when I was shooting a fountain in broad daylight. I was using 1/4000th at ISO 100 and I could not discern any reason for the noise other than the high shutter speed.

Notice the directional nature of the noise. Aperture comprehended it as a perpendicular crosshatch pattern; Lightroom is a lot better at handling this kind of noise and now in the attached photos you're only able to see the horizontal noise, especially in the darks by the bricks. The bricks are patternless and completely solid, devoid of patterns.

But also notice how the mids and the highlights are free from this directional noise! There it is just average, removable luminance noise.

Having only recently switched to Lightroom, I hadn't noticed how good Lightroom was at countering the directional noise I kept encountering in Aperture, so if my explanation for this noise is wrong, I'd like to hear how exactly I'm getting this noise too

(To most this should be a non-issue as it only seems to occur at high shutter speeds)

[snberk103]

Originally Posted by Prodo123

I've only discovered it accidentally this summer when I was shooting a fountain in broad daylight. I was using 1/4000th at ISO 100 and I could not discern any reason for the noise other than the high shutter speed.

Exposure sounds about right. Though, I will note that the bricks are in deep shadow. The camera will have had to really dig deep to pull detail out of those shadows, which may have introduced noise.

Originally Posted by Prodo123

Notice the directional nature of the noise. Aperture comprehended it as a perpendicular crosshatch pattern; Lightroom is a lot better at handling this kind of noise and now in the attached photos you're only able to see the horizontal noise, especially in the darks by the bricks. The bricks are patternless and completely solid, devoid of patterns.

I'm thinking it is as much a moire pattern as anything else. The mortar lines may are creating a very regular pattern. My photography doesn't tend to fall into moire pattern problems, so I don't have a lot of experience with it. But... Lr 4 allows you 'brush' out moire with the brush tool. Try that on the original RAW file and see if that helps.

Originally Posted by Prodo123

But also notice how the mids and the highlights are free from this directional noise! There it is just average, removable luminance noise.

Also, other non-brick wall shadows. If it was sensor stutter, which I have never heard of and which produces no Google results, I would expect to see it in other places as of the same tonality.

Originally Posted by Prodo123

Having only recently switched to Lightroom, I hadn't noticed how good Lightroom was at countering the directional noise I kept encountering in Aperture, so if my explanation for this noise is wrong, I'd like to hear how exactly I'm getting this noise too ...

I'm going to go with Moire Pattern for $64.

[blanka]

Originally Posted by Redneck1089

The 5D2 kills the D700 in noise performance

No way. I do not agree with that. Nor does DXO (don't use that three letters when there is a Canon guy around).
On a Nikon, noise is always clean noise. Never weird shît. You can pull details from shadows and the image stays nice. It has apparently more noise, but also more detail, and the noise is nice. On Canon it isn't. It is a total dirty mess. With the new generation it is total drama. The D800 eats the shadows of the Canon MK3 for breakfast.

[Prodo123]

Originally Posted by snberk103

Exposure sounds about right. Though, I will note that the bricks are in deep shadow. The camera will have had to really dig deep to pull detail out of those shadows, which may have introduced noise.
I'm thinking it is as much a moire pattern as anything else. The mortar lines may are creating a very regular pattern. My photography doesn't tend to fall into moire pattern problems, so I don't have a lot of experience with it. But... Lr 4 allows you 'brush' out moire with the brush tool. Try that on the original RAW file and see if that helps.
Also, other non-brick wall shadows. If it was sensor stutter, which I have never heard of and which produces no Google results, I would expect to see it in other places as of the same tonality.

I suspected moiré at first too, but then I noticed it also occasionally creeps up on shadows and mids on the tree leaves. Since the sensor is twice as sensitive to green than to red and blue, I thought it would have no trouble extrapolating details from the greens, but apparently not.
Maybe it is the camera sensor's lack of dynamic range, since it IS a Canon...

Originally Posted by blanka

No way. I do not agree with that. Nor does DXO (don't use that three letters when there is a Canon guy around).
On a Nikon, noise is always clean noise. Never weird shît. You can pull details from shadows and the image stays nice. It has apparently more noise, but also more detail, and the noise is nice. On Canon it isn't. It is a total dirty mess. With the new generation it is total drama. The D800 eats the shadows of the Canon MK3 for breakfast.

Photography is a subjective thing. We human beings view things subjectively, to our liking. And no matter what the numbers say, it is all meaningless when the subjective opinion defeats it.
That being said, DxOMark does make objective judgements. And they tend to be accurate; both Canon and Nikon shooters agree that Nikon has a clear edge in dynamic range. But when it comes to noise performance, the matter is very different.
DxOMark measures raw noise. It measures how much noise is present and in which manner. This means before any processing is done. More noise to bring out the "details" (in reality graininess giving the illusion of contrast) is an old trick in the book for making things appear sharp. This is what Nikon tends to do; it leaves the noise as-is, doesn't touch it and it appears generally sharper with a lot more subjective noise.
Canon does things a little differently; it removes more noise, smudging the noise and the details to make things appear noise-free. The result is a cleaner, but less-detailed image. It still has noise, but since there is less detail to confuse with noise, it is easier to remove noise.

Since DxOMark measures the raw, unprocessed image, it would give Nikon the edge because it retains the image's details. Canon appears to get a lower score because of the smudging. But when in post, you do have to remove the noise present in Nikon's images. This ultimately results in the same smudging of the details, sometimes more. With Canon, it's already done for you out of the camera.
So objectively, Nikon has less destructive noise out-of-the-camera, but in reality most photographers agree that Canon has better noise performance straight out of the camera.

Also, I predict that the D600 will get a higher noise performance than the 6D on DxOMark for the aforementioned reasons, despite the fact that Canon can produce usable images up to ISO 12800, where the D600 turns completely green (literally).

[Redneck1089]

Originally Posted by blanka

No way. I do not agree with that. Nor does DXO (don't use that three letters when there is a Canon guy around).
On a Nikon, noise is always clean noise. Never weird shît. You can pull details from shadows and the image stays nice. It has apparently more noise, but also more detail, and the noise is nice. On Canon it isn't. It is a total dirty mess. With the new generation it is total drama. The D800 eats the shadows of the Canon MK3 for breakfast.

That's because DXO is biased in the extreme.

And are we talking about the D800 and 5D3, or the D700 and 5D2?

[Prodo123]

Originally Posted by Redneck1089

That's because DXO is biased in the extreme.

And are we talking about the D800 and 5D3, or the D700 and 5D2?

In both cases Canon has the subjective edge while Nikon has the objective edge.

[OreoCookie]

Originally Posted by Redneck1089

That's because DXO is biased in the extreme.

It's not just DxO mark, every review I have seen puts the D700 ahead of the 5D Mark II when it comes to noise and speed. For one, the D700 has much larger pixels (because it has a FF sensor with 12 megapixels rather than 21 megapixels) which alone gives it a clear advantage. Only if you're willing to complicate the discussion (e. g. sampling down the 5D Mark II's images to 12 megapixels), you can perhaps argue that the Canon has an edge. But I doubt this is particularly relevant in real life, and I think it actually misses the strength of the 5D Mark II: the higher resolution. For some applications (where you don't care about speed or high ISO), this is a distinct advantage.

Originally Posted by Prodo123

Since DxOMark measures the raw, unprocessed image, it would give Nikon the edge because it retains the image's details. Canon appears to get a lower score because of the smudging. But when in post, you do have to remove the noise present in Nikon's images. This ultimately results in the same smudging of the details, sometimes more. With Canon, it's already done for you out of the camera.

If Canon were doing that, shouldn't they achieve better high ISO performance in exchange for lower resolution? They don't, according to DxO mark, the D600 has a much better high ISO performance than the 5D Mark III (there are no scores available for the 6D at this point).

Besides, DxO mark does benchmark the sensor rather than the RAW converter which you happen to use. And since every modern sensor has some pre-processing built into it to filter out some noise, I think this comparison is fair and apt. Your suggestion that Canon somehow makes it »easier« to process your images is also non-sensical: no matter if you spend €€€€ on a Canon or a Nikon, you'll probably shoot in RAW and use one of the common RAW converters (Aperture, Lightroom or the converter supplied by the camera manufacturer), so the computer does all the heavy lifting for you anyway. To measure noise then is much harder since you can optimize for different things when you tweak images manually.

Originally Posted by Prodo123

So objectively, Nikon has less destructive noise out-of-the-camera, but in reality most photographers agree that Canon has better noise performance straight out of the camera.

The claim »most photographers agree« is not something you can test, the way you wrote it sounds like it's mostly the faithful of a particular brand (in this case Canon) trying to convince themselves that »their camera« is not worse than some other camera. Similar to when Nikonians used to claim that it's not necessary for them to have full frame sensors in their bodies.

I'm always amazed why people engage in such discussions why, theoretically, their camera is »better« rather than go out and shoot.

Originally Posted by Prodo123

Also, I predict that the D600 will get a higher noise performance than the 6D on DxOMark for the aforementioned reasons, despite the fact that Canon can produce usable images up to ISO 12800, where the D600 turns completely green (literally).

There is no need to predict, the D600 has the third-highest DxO mark at this time, places 1 and 2 being taken by the D800E and D800, respectively.

Originally Posted by Prodo123

In both cases Canon has the subjective edge while Nikon has the objective edge.

What does »subjective edge« even mean? Wishful thinking? Common lore is that Nikon has a history of suppressing Chroma noise better than Canon which supposedly gives it a more »film grain-like« appearance. Whether you find that type of noise more pleasing is a matter of debate.

[Prodo123]

Originally Posted by OreoCookie

It's not just DxO mark, every review I have seen puts the D700 ahead of the 5D Mark II when it comes to noise and speed. For one, the D700 has much larger pixels (because it has a FF sensor with 12 megapixels rather than 21 megapixels) which alone gives it a clear advantage.

Then the D800, which has 36.3 megapixels, has a clear disadvantage, yet it still manages some 800 points higher on DxOMark...

Originally Posted by OreoCookie

If Canon were doing that, shouldn't they achieve better high ISO performance in exchange for lower resolution? They don't, according to DxO mark, the D600 has a much better high ISO performance than the 5D Mark III (there are no scores available for the 6D at this point).
Besides, DxO mark does benchmark the sensor rather than the RAW converter which you happen to use. And since every modern sensor has some pre-processing built into it to filter out some noise, I think this comparison is fair and apt. Your suggestion that Canon somehow makes it »easier« to process your images is also non-sensical: no matter if you spend €€€€ on a Canon or a Nikon, you'll probably shoot in RAW and use one of the common RAW converters (Aperture, Lightroom or the converter supplied by the camera manufacturer), so the computer does all the heavy lifting for you anyway. To measure noise then is much harder since you can optimize for different things when you tweak images manually.
There is no need to predict, the D600 has the third-highest DxO mark at this time, places 1 and 2 being taken by the D800E and D800, respectively.
What does »subjective edge« even mean? Wishful thinking? Common lore is that Nikon has a history of suppressing Chroma noise better than Canon which supposedly gives it a more »film grain-like« appearance. Whether you find that type of noise more pleasing is a matter of debate.

Here's a more realistic test.
The eye discerns more than DxOMark's numbers can. That is what the subjective edge is; what our minds define as good is not the same as what calculated values say are good.
Scan anywhere and you'll see the D600 pale out and get tons of chroma noise while the 5D Mark III and D800 keep their sharpness and contrast. I don't care how good the D600 score is on DxOMark; the 5D Mark III clearly trumps it here. Between the D800 and 5D Mark III, both sides have their respective advantages and are practically equal in high ISO performance.
There's a reason why people say one should try a camera out before buying; numbers don't reveal everything about a camera.

And about the appearance of noise: Again, as you said, it's a personal preference, a subjective matter. Making numbers represent such things, which is what DxOMark does, is thus impractical.

Wishful thinking? Far from it. I honestly don't care; ISO 3200 or 6400 is all I'll ever need, and I'm quite satisfied with what Canon can do at those sensitivities.
Although I sometimes wish I shot Nikon because of the D600's price and autofocus.

Originally Posted by OreoCookie

I'm always amazed why people engage in such discussions why, theoretically, their camera is »better« rather than go out and shoot.

Sounds a bit like yourself, since neither I nor Redneck said that Canon or my camera is superior to Nikon (it's not, at least in overall stills performance).

-----------------

As for everyone else: I know the 5D2's noise performance is decent, so I'm considering it for sports photography. Do you think the autofocus is up to the task?

[OreoCookie]

Originally Posted by Prodo123

Then the D800, which has 36.3 megapixels, has a clear disadvantage, yet it still manages some 800 points higher on DxOMark...

Yes, the D800 has a clear disadvantage, but scores higher regardless of that. The camera which scores highest in the high ISO test is a D3s, though.

Originally Posted by Prodo123

Here's a more realistic test.

What am I supposed to see? (I've compared the test stills at ISO 3,200-12,800.)
I don't see an edge for Canon's at least, in fact, I'd say the D600 and D700 have less chroma noise while the 5D Mark I and Mark II have more. The D800 shows visibly more noise than its Nikon brethren, worse in terms of detail compared to the Canon, but a little less chroma noise. Overall, it seems to me that the D600 and the D4 have the best output. Just to be clear: I would not trust my eyes to give an objective score, but I don't see the advantage that you speak of.

It's a little bit surprising that you prefer the output (in terms of noise, at least) of those cameras which I find the noisiest. Probably that's one of the reasons you shouldn't put too much faith in subjective noise measurements.

Originally Posted by Prodo123

There's a reason why people say one should try a camera out before buying; numbers don't reveal everything about a camera.

And about the appearance of noise: Again, as you said, it's a personal preference, a subjective matter. Making numbers represent such things, which is what DxOMark does, is thus impractical.

No, I don't think you should go that far: measurements of noise do give you a clear indication on some properties of a camera just like acceleration figures do with cars. If two cars have the same engine, but one of them is 50 % heavier, it's a clear indication which one will accelerate faster, have better handling, etc. But performance goes beyond that and you cannot put everything into numbers. But in these cases, you can measure different types of noise, you can make a clear determination which camera has more and which camera has less noise of a given type.

What you cannot put in numbers is what kind of noise is preferable, that's where the argument gets subjective (for good reason).

Originally Posted by Prodo123

Wishful thinking? Far from it. I honestly don't care; ISO 3200 or 6400 is all I'll ever need, and I'm quite satisfied with what Canon can do at those sensitivities.

Bingo, that's exactly the point I was making earlier: know your camera and its limitations, and you will take good pictures.

Originally Posted by Prodo123

Sounds a bit like yourself, since neither I nor Redneck said that Canon or my camera is superior to Nikon (it's not, at least in overall stills performance).

I'm just arguing because I get sick of people who mix scientific and unscientific arguments. It doesn't rub me the wrong way because I shoot Nikon, it rubs me the wrong way because I'm a scientist. (BTW, just to be clear, I think my next camera will be a Fuji X-series mirrorless, so I'm not going to invest in a full frame dslr.)

Originally Posted by Prodo123

As for everyone else: I know the 5D2's noise performance is decent, so I'm considering it for sports photography. Do you think the autofocus is up to the task?

Seeing as the Mark II uses a slightly tweaked version of the AF module of the Mark I, I don't think the Mark II is a good choice for sports photography. (Meaning it takes more skill to get in-focus images.)