New 5DII, Noise, Moire, Noise, banding Problem LOW ISO.

gdanmitchell said:

I'm not big on sharing full size RAW files on the web, but here is a 100% magnification crop

Click to expand...

That is 100% useless for an objective approach. Only a raw file carries enough data to allow specific discussion of the image. See my post below: http://forums.dpreview.com/forums/read_js2.asp?forum=1032&message=32260609

Many photogs talk a lot about the dynamic range of the camera, but most of them don't have much clue re the implications. Following is a 25% ACR rendered version of a shot created specifically for the measurement of noise and DR (underexposed by close to five stops). The raw file can be downloaded from http://www.panopeeper.com/Download/CCC_ISO00100_1221.CR2 , the ACR adjustment parameters are in http://www.panopeeper.com/Download/CCC_ISO00100_1221.xmp

Only such excessive adjustments as shown in the XMP bring out the really deep shadows. Any blacks more than 0 means cutting away the darkest parts - why do you want to have a good camera, if you treate the file like it came from a P&S?

The red, green and blue numbers in the patches show the average raw pixel intensity in that patch. (There is no point to talk about "exposure": the fixed exposure results in different intensities and different noise levels on different patches.) The intensity is measured in EV from clipping, for example 9.26 means, that the red is at the 10.26th stop of the dynamic range (it starts with zero, because the clipping level is 0 EV from the clipping level). The noise in the JPEG is of course the result of the conversion, which applies a multiplication of the pixel values for the WB (and that multiplies the noise as well) and mixes the raw channels. It occurs very seldom, that the three channels have the same intensity.

One can get an idea of the noise pure on a raw channel from the files starting with Canon5DMKII_CCC_ISO0100 in http://www.panopeeper.com/Canon/

Here is the ACR converted image representing the real shadows and the noise:



Finally, here is the result of the measurements on this shot and on the others with different ISOs. If you pick a certain level of noise in terms of standard deviation (as the reciprocate of the SNR), the x axis shows the dynamic range.

However, the graphs do not reflect the ugliness of the pattern noise. It shows, among others, that ISO 200 is practically as good as ISO 100, and that ISO 3200 is useless for raw, as it yields the same noise as 1600, but it cuts off the top EV from the DR.

Note, that these are no "fitted curves", there is no interpolation here; each point represents a measurement, as shown in the Canon5DMkII_CCC_ISO0100_ images.



--
Gabor

http://www.panopeeper.com/panorama/pano.htm

Should I return my camera for another one since it 's few days old ?? After reading a lot useful information I'm a little confused of what the next step should I take ?? I appreciate the time and the effort from everyone to help me find out how the camera works .

Thanks again .

GaborSch said:

gdanmitchell said:

I'm not big on sharing full size RAW files on the web, but here is a 100% magnification crop

Click to expand...

That is 100% useless for an objective approach.

Click to expand...

I beg to differ. An image like the one I shared should be close to 100% useful. It represents well what the 5DII can do at ISO 800 in terms of noise and dynamic range in a typical subject that exceeds the DR of the camera's histogram.

The RAW file is not as significant as you imagine. It is important to my photography in the sense that it must contain the maximum scene data so that I'll have a full range of data to work with, but it will often look pretty awful. It needs sharpening. It will appear flat because no curve adjustments have been applied. The histogram may make it appear that data is lost at the light or dark end, but it can be recovered in a number of ways.

But if you need a RAW file I'm pretty certain that I've seen some posted online. I don't post mine - partly because they are honkin' big files and partly because I don't send my full size original photos out onto the net for reasons that are probably easy to imagine.

By the way, you wrote:

"However, the graphs do not reflect the ugliness of the pattern noise. It shows, among others, that ISO 200 is practically as good as ISO 100, and that ISO 3200 is useless for raw, as it yields the same noise as 1600, but it cuts off the top EV from the DR."

My experience with photographs shot at 100 and 200 and turned into good size prints confirms (and I think I wrote about this earlier in this thread) your statement about those ISOs. I use them almost interchangeably on my 5DII. I rarely shoot at 3200 so I can't comment too much on that - though I have used it in some extreme situations in which the subject was such that I wasn't terribly concerned about the noise issue.

Your point and shoot comment was pretty much uncalled for and more than a bit insulting and pretentious.

I do frequently post crops from full size images to illustrate various posts at my site.

Take care,

Dan

--
---
G Dan Mitchell - SF Bay Area, California, USA
Blog & Gallery: http://www.gdanmitchell.com/
Facebook: http://www.facebook.com/gdanmitchell
Flickr: http://www.flickr.com/photos/gdanmitchell/
Twitter: http://twitter.com/gdanmitchell
IM: gdanmitchell

Gear List: Cup, spoon, chewing gum, old shoe laces, spare change, eyeballs, bag of nuts.

So why would you want to exchange it, there is no reason for that?

If you really are so unsure, go to a camera store that has a 5D mk II on display and shoot a shot with your camera and the one in the store with exactly the same settings.
Then go home and look at the results.

The old 5D or the 1Ds2 or 1Ds3 are better in that respect, but they suck in the High ISO department.
--
Fun stuff - http://adamkozlowski.deviantart.com

Rather than "guessing" it's better to do some tests yourselves

So 160/320/640/1280 is better than 200 400 800 1600?

I thought you previously advised the old 100/200/400 were real and these others were compromised and to be avoided?

I have a 5D mk2 and only get an issue if I underexpose severely, 3 stops otherwise it's a non issue in the studio

Dear John - in your opinion, what is the best camera for high resolution & low noise?

Holscen said:

Dear John - in your opinion, what is the best camera for high resolution & low noise?

Click to expand...

I know you asked John and I'm not John, but...

I like the point of view expressed by Michael Reichmann in an article at his Luminous Landscape site shortly after competing 20MP+ cameras were introduced by Sony, Canon, and Nikon. Basically, his point was that all of these cameras a capable of producing such excellent IQ in comparison to anything in this format (digital or film) that we have had available in the past and that the differences between them are largely trivial in comparison to the high quality images they can produce.

While you can no doubt measure some differences among them and other cameras in the area of image noise, it becomes obsessive to treat this as if it is a significant issue in actual photographs.

Take care,

Dan

--
---
G Dan Mitchell - SF Bay Area, California, USA
Blog & Gallery: http://www.gdanmitchell.com/
Facebook: http://www.facebook.com/gdanmitchell
Flickr: http://www.flickr.com/photos/gdanmitchell/
Twitter: http://twitter.com/gdanmitchell
IM: gdanmitchell

Gear List: Cup, spoon, chewing gum, old shoe laces, spare change, eyeballs, bag of nuts.

gdanmitchell said:

The RAW file is not as significant as you imagine

But if you need a RAW file I'm pretty certain that I've seen some posted online. I don't post mine

Click to expand...

I don't need your raw file; you seem to be in need of explanation of your own raw images.

If you believed that I was asking for a favour in form of a raw file, you thoroughly misunderstood the situation.

Anyway, happy shooting.

--
Gabor

http://www.panopeeper.com/panorama/pano.htm

GaborSch said:

I don't need your raw file; you seem to be in need of explanation of your own raw images.

Click to expand...

Gabor, no, I don't "need explanation of my own raw images." Nor do I need your dismissive and pretentious comments about my knowledge of photography.

Bye.

--
---
G Dan Mitchell - SF Bay Area, California, USA
Blog & Gallery: http://www.gdanmitchell.com/
Facebook: http://www.facebook.com/gdanmitchell
Flickr: http://www.flickr.com/photos/gdanmitchell/
Twitter: http://twitter.com/gdanmitchell
IM: gdanmitchell

Gear List: Cup, spoon, chewing gum, old shoe laces, spare change, eyeballs, bag of nuts.

i though the original 5D had been shown to have both less DR and even more pattern noise in deep shadows at low ISO than the MkII??

Adam Kozlowski said:

The old 5D or the 1Ds2 or 1Ds3 are better in that respect, but they suck in the High ISO department.
--
Fun stuff - http://adamkozlowski.deviantart.com

Click to expand...

gdanmitchell said:

I read a comparison among several of the new 20MP+ cameras recently and I believe someone test the DR at something like 11 stops. The same test test showed that the DR was very similar among the available FF bodies in this class. Sorry, I don't have the article or the URL right now - maybe someone else remembers it and can help?

Click to expand...

The DR you mentioned about was tested by DPreview.
They called it the RAW headroom.

With ACR-best the result was 11.1 EV (stops) for 5DII
http://www.dpreview.com/reviews/canoneos5dmarkii/page25.asp

While Nikon D700 was slightly better with 11.6 EV (stops) using best tweak of ACR
http://www.dpreview.com/reviews/nikond700/page20.asp

The D700 was also better in maintaining color in highlight region due to the late channel clipping.

The 5DII was said to be slightly dominant in shadows area to achieve the 11.1 EV figure.

For some people those differences are huge, for others those differences are nothing.

-

bronxbombers said:

same scale, appears to produce worse deep shadows than my 40D did

Click to expand...

Wow - that's pretty bad. I always thought 5DII had bigger sensels compared to 40D -- and 2 years of improvements in sensor technology. We should have seen some improvements, but...

How did Canon screw it this badly (w. respect to this one issue - so pls don't flame me for this)?

If 1DsIII share the same sensor & how did it manage to escape from this issue?

Thanks a lot,
-Mike.

Point of reference... if the 5DII were really "pretty bad," don't you think that the photographers who use it to make the majority of their work would be posting right and left all over the net about it? Aside from a few threads where one or a few people claim that the 5DII noise is terrible, are you seeing a general sense that this is a big problem with this camera?

When a test tells you one thing and application tells you another, which do you believe? This morning I was thinking about this concept a bit and I thought of several real world stories of situation where careful and supposedly objective testing was followed by real world testing that produced different results. At the moment I'm thinking of the recent story about the need to rework some parts of the design of the new Boeing jetliner after flight tests revealed that there are potential structural problems with the very objectively designed and tested wings. Rather than assuming that "the wings must be fine because our tests say they are," smart engineers are more likely to ask, "what did our tests not tell us and/or how do the tests we contrive differ from real world conditions?"

The tests tell you what the tests tell you. If you have no other information to go on they certainly may be the most important information you have. (Ironic humor intended.)

However, if there is a dissonance between the test results and actual results in the field you need to ask how that is possible. What might it tell you about the supposedly objective nature of your test? Do you believe the controlled test and ignore what you see with field testing? If 100 fine photographers told you that a particular piece of gear was great and they could produce the photographs to demonstrate what the gear could do, and then someone told you that they had test results that made the photographic results you had just seen impossible, who would you believe?

Dan

Michael Anburaj said:

Michael Anburaj said:

same scale, appears to produce worse deep shadows than my 40D did

Click to expand...

Wow - that's pretty bad. I always thought 5DII had bigger sensels compared to 40D -- and 2 years of improvements in sensor technology. We should have seen some improvements, but...

How did Canon screw it this badly (w. respect to this one issue - so pls don't flame me for this)?
-Mike.

Click to expand...

--
---
G Dan Mitchell - SF Bay Area, California, USA
Blog & Gallery: http://www.gdanmitchell.com/
Facebook: http://www.facebook.com/gdanmitchell
Flickr: http://www.flickr.com/photos/gdanmitchell/
Twitter: http://twitter.com/gdanmitchell
IM: gdanmitchell

Gear List: Cup, spoon, chewing gum, old shoe laces, spare change, eyeballs, bag of nuts.

dough, one should expect better.

Holscen said:

So 160/320/640/1280 is better than 200 400 800 1600?

Click to expand...

Not better; just more hassle-free, compared to shooting the latter group with +1/3 EC, as the former work more reliably with converters in general, especially DPP, and the in-camera JPEGs.

Arthur Stanley Jefferson said:

I thought you previously advised the old 100/200/400 were real and these others were compromised and to be avoided?

Click to expand...

I used the terminology "real" and "fake", regrettably, for a very short time a while back (concerning the extended ISOs, I know; I don't know if I called the 30D intermediate ISOs "fake"). The one who continues to dwell on ISO "reality" is Gabor, so maybe you're confusing he and I. I've said from the beginning of the 30D that the 160/320/etc group can be very practical in some situations.

Arthur Stanley Jefferson said:

I have a 5D mk2 and only get an issue if I underexpose severely, 3 stops otherwise it's a non issue in the studio

Click to expand...

Arthur Stanley Jefferson said:

Dear John - in your opinion, what is the best camera for high resolution & low noise?

Click to expand...

For low ISOs, that would be the Nikon D3x. Unlike the other recent Nikons, it has no apparent filtering of the RAW data, and it has extremely low read noise at base ISO.

--
John

evan said:

Should I return my camera for another one since it 's few days old ?? After reading a lot useful information I'm a little confused of what the next step should I take ?? I appreciate the time and the effort from everyone to help me find out how the camera works .

Click to expand...

I see no evidence to suggest anything except that you have a decent 5d2, as far as 5D2s go. Exchanging it is not going to do any good.

--
John

Here's another sample:



ISO 200, no ALO, no HTP. No noise reduction. Pushed 2 stops in DPP via the brightness slider -- that's the only adjustment. Picture was taken in vertical orientation, by the way. 100% crop. Firmware v. 1.1.0.

I am not at all happy with the noise on the neck, for example. And there's banding on the sweatshirt (given image orientation, it's parallel to the short side of the frame).

--
Kaa

n/t

gdanmitchell said:

While you can no doubt measure some differences among them and other cameras in the area of image noise, it becomes obsessive to treat this as if it is a significant issue in actual photographs.

Click to expand...

It's up to each individual to understand the implications of arguments and decide what is relevant to their needs.

Cameras can be different, whether you exploit those differences or not.

At base ISO), the D3x has less read noise than the A900, which has less read noise than the 1Ds3, which has less than the 5D2. The 1Ds3 has mild banding, and the 5D2 has yet more banding.

--
John