The Canon 5D2 Mark II has strong pattern noise at ISO 100. Most people do not notice it because they only use 5-8 stops of dynamic range. If you would like to use 9, 10, or more stops of dynamic range, the pattern noise will stop you in your tracks.
The following example image was developed to use 11.5 stops of dynamic range (click for full size):
Here is the red channel isolated:
And it's not just an artifact of the raw conversion or post processing either. Here is the raw data itself, red channel, before demosaic, with a +8 linear push and sRGB gamma curve applied in Rawnalyze:
And this is the composite view in Rawnalyze at +6:
And here is the raw file so you can download it and see for yourself:
IMG_4379.CR2
I welcome other conversions for comparison. A typical converter will clip almost everything to black, because they default to only 5-8 stops of dynamic range. That does not mean it is underexposed, of course: the highlights are less than 1/3 stop from clipping in the raw data. What it means is that it was exposed to use a full 11.5 stops and not just 6 or 7 like most photos.
This pattern noise occurs in every single 5D2 ISO 100 raw file with deep shadows. Whether you see it or not depends on one thing: do you ever shoot scenes with high dynamic range? If you just do the typical 5-7 stops, with an occasional 8 or 9, then you'll rarely see pattern noise.
On the other hand, if you would like to be able to shoot things that the A900 and D3X manage easily, or at use 12 bits of data, or try to use just 11.5 out of the 14 stops that you had on film, then you will see the pattern noise.
Even digicams such as the LX3 don't have pattern noise. Look at how it is totally random at 10.6 stops:
The closest competition to the 5D2 is the A900 and A850. It does the same dynamic range with no pattern noise to speak of. Furthermore, the Canon 7D has much less pattern noise as well. Less even than the 1D series.
There are some people who dispute these facts. Here are the responses that can be expected from them:
I don't see it, therefore it does not exist
The reason they don't see the problem is because they only use 5-8 stops of dynamic range. Maybe it has been so long since they shot print film that they forgot what it was like to have 14 stops. Or maybe they don't know how to post process an image for 10 stops.
Maybe they are content to limit themselves to over one hundred times less dynamic range than film. If the scene they are shooting has more than 8 stops, and they can't use split ND or exposure blending to get the dynamic range down, they just pack up and go home.
It's similar to how some people will never shoot low light, so they ridicule people who compare cameras at ISO 6400. Luckily, most people can realize that it's valid for ISO 6400 to be important to some photographers, even if they don't use it themselves.
Unfortunately, those same people will turn around and say that 8 stops DR is enough for anyone, and it's invalid to compare cameras at more than 8 stops.
You are pushing the files too hard
This is just another variation of the above. People think that since they are happy with 7 stops of dynamic range, that everyone else should be too.
You lightened the shadows too much
Again, some people have a very hard time imagining a valid use for 9 or more stops of dynamic range, so any time someone goes above 8 or 9, they get criticized.
The scene is too harsh
Another variation on the idea that one should not shoot for too much dynamic range. Whatever they use (7 stops? 8? 9 maybe?) should be enough for everyone.
You should have used grad ND, multi-exposure HDR, fill flash, etc.
These types of responses side-step the problem by saying that you should have used some sort of technique to fit the harsh, high contrast scene into a small window of dynamic range (e.g. 7 stops). But sometimes I am unwilling or unable to use workarounds such as graduated ND, multi-exposure HDR, or fill flash. I would like the option of using 12 stops of dynamic range without pattern noise, like the A850.
You're expecting too much from digital
This response acknowledges that film could do 13-14 stops of DR no problem, but digital is limited to 7-9. The only problem is that other digital cameras do not have this problem. D3X is getting into 13 stops, and the A900 does 12, and the 7D does 12 with far less pattern noise. The 5D2 can't manage even 10 without pattern noise, but it looks like Canon finally addressed this problem in sub-$4,000 cameras with the 7D.
All cameras have noise
This post is not about random noise. All cameras have that. This is about pattern noise.
The other cameras have this problem too
The $2,000 A850 doesn't, nor does the $1700 7D.
http://www.photographyblog.com/reviews_sony_a900_3.php
http://img.photographyblog.com …e_images/sony_a900_03.arw
The other cameras are only better because of Noise Reduction
That's incorrect. If there is any NR on A900 files, it does not remove pattern noise. I have yet to find any NR software that can remove the 5D2 pattern noise, so if you find any that really works without smothering all the detail, send it my way. Most NR makes it worse through edge preservation by removing the random noise and leaving the patterns.
So what if Canon has this problem - Canon is still best at XYZ
This is when people acknowledge that the 5D2 has worse ISO 100 pattern noise than other cameras (e.g. A900), but they change the subject by saying that Canon has better lens selection. Or better noise at high ISO. Or better movie features. Whatever it is, it does not change the fact that Canon is inferior in this one way.
Your camera is early production. Recent ones don't have this.
I wish this were so. My 5D2 is from December 2008, and I checked one from a colleague that got it in April 2009. Again, no one can find any 5D2 raw file with 11+ dynamic range and no pattern noise.
Your photography stinks anyway
If you post an image demonstrating the noise problem, you will get responses criticizing the subject, composition, etc.
It's a problem with your post processing
This one is easily refuted by posting the raw files and literal interpolations such as those from Rawnalyze, dcraw, libraw, IRIS, etc.
Try bias frame subtraction
That does help to remove a little bit of the fixed pattern noise, but the far majority of it is variable and cannot be removed with bias frame subtraction.
Try printing smaller or re-sizing
Unfortunately, the banding often gets worse at smaller print sizes. It's visible even in a 4x6.
You didn't expose correctly
This is just another way of saying that you used too much dynamic range. What they mean by "correct exposure" is exposure for 7-8 stops of dynamic range (maybe 9). The more dynamic range you need, the more they think you are "underexposing".
DPReview says the problem doesn't exist.
Some will quote the 5D2 review at DPReview.com to prove the problem doesn't exist. But their measurement of dynamic range is highly flawed in many ways. Even if they did measure it correctly (they don't) and found 11 stops of dynamic range, it would show the pattern noise.
Your camera is broken
All 5D2 cameras have this problem. I have seen hundreds of files from different 5D2 cameras, but I have never seen a single one with 11+ stops of dynamic range and no pattern noise. Either it has little dynamic range and no pattern noise, or it has lots of dynamic range and lots of pattern noise.
Your lens, flash, or something else is wrong
I'm told that RF interference from some lens AF motors, flashes, and other environmental sources can cause pattern noise. This can make the pattern noise even worse than most cameras, but given the fact that all 5D2 cameras have strong pattern noise, it cannot account for every case.
Canon rules, the other camera drools
There are invariably a wide variety of fanboi responses.
I like to know and discuss the problems and limitations of the gear that I own.
(nikoncafe has some examples somewhere with a pool table).
