photoArne wrote:
Adrian, thank you for posting these. Basically they confirm my own findings. While I too can provoke the criss-cross noise pattern by extreme pushing of the shadows, it will not exactly keep me awake at night... But I agree with Richard that 5D2 noise at base ISO is (a little) higher than 1DsIII and it can call unwarranted attention to itself in the shape of the criss-cross pattern.
Another claim made by Canon for the 5D2 is low noise at higher ISO settings.
So I redeveloped ISO 1600 shots in the same manner and from the same series as above to compare the 5D2 and the 1DsIII.
No noise reduction, no sharpening, same settings in C1.
First 1DsIII then 5D2. ...Show more →
seems like 1dsmkii has a bit more DR at lowest ISOs and 5dmkii has more at the higher isos
i did notice that the 5dmkii had really low read noise at mid and higher isos but a little bit maybe 25-30% worse per pixel than a 40D at iso 100. (of ocurse at the lowest iso you actually
i finally made it appear
took edge of underside of roof and large tree trunks against white sky at iso 100 and then set 0 contrast and no curves and pulled highlight back by 50 and bumped exposure up by 2 stops and pulled shadows up in acr by 8 the underside of the roof did get the same cross hatch.
whether this is new (the cross-hatch and not just horizontal banding) to 5dmkii or not not sure since i usually dont process this way with no contrast and giant 2 step pulls and so on.
after more experimenting i can only sometimes see hints of the cross hatch in the most extreme shots using any sort of PP and shooting style i would normally use so im not sure if it really an issue, i have no other bodies with me at the moment to compare to, maybe i can see what the rebel xt does tomorrow.
now the question is does yours get it much more easily than mine or is it more that i dont normally push things quite that extremely and is it something that the older models really can get away with more before showing vertical pattern noise?
it's not surprising to get the deepest pulled shadows at lowest iso with more pattern
Folks, what you are seeing is pattern noise from the ISO amplifier. It is more apparent at low ISO because amplifier noise is introduced after the ISO amplification; therefore it is more apparent when the signal is *less* amplified (low ISO) since the fixed noise is compared to a less amplified signal (ie lower S/N). All Canons suffer this pattern noise to one extent or another; it is less of an issue in the 1 series where Canon uses higher quality electronic components. Even so, compare the noise at ISO 100 and 1600 for a fixed exposure (same Tv & Av) on my 1D3 (the lower left square of the GMCC chart is about 9 stops down from RAW saturation) . . .
This noise is superposed on any image taken with that camera. Boost the shadows enough and you will see it. Different raw converters will respond to it differently, depending on how their interpolation algorithm is structured. For some, it looks like lines or edges, and to the extent that the algorithm is designed to favor the preservation of lines and edges, the noise will misguide the interpolation. But clearly different algorithm designs will interpret the noise differently.
It is sad that Canon does not expend more effort in dealing with this noise, as it can be easily subtracted out. It is largely coherent across the image, so if a few more pixels were masked off at the edge of the frame, the data would exist for the construction of a template to subtract the line noise across the frame in each row and column. But alas there are not enough masked off pixels in the frame edges that are present in the RAW files of current and past Canon DSLRs to do a statistically reliable job of making an accurate template.
To see what the subtraction could do, have a look at the mouse-over in figure 3 at
ejmartin wrote:
Folks, what you are seeing is pattern noise from the ISO amplifier. [...] All Canons suffer this pattern noise to one extent or another; it is less of an issue in the 1 series where Canon uses higher quality electronic components.
So the inconsistent results between low ISO performance of 5D2's could be (partly) explained due to inconsistent quality in electronic components between different bodies?
Daan B wrote:
So the inconsistent results between low ISO performance of 5D2's could be (partly) explained due to inconsistent quality in electronic components between different bodies?
I'd want to see some evidence that Canon actually uses different (and less capable) components in this situation before I'd accept that statement as a fact rather than a guess.
Update from my end: My 5D 2 had numerous image quality issues (early serial, bought November 28, 2008), including banding ISO 100 in shadows. Just got back from Canon repairs yesterday: CMOS sensor assembly replaced... if you are having IQ problems send it in for warranty repair/replacement.
As you can imagine, the QC on this camera with the rate of production/demand etc. has been questionable IMHO. I would suspect also that current production is fine now that the bulk of the demand has been satisfied.
ejmartin wrote:
It is sad that Canon does not expend more effort in dealing with this noise, as it can be easily subtracted out. It is largely coherent across the image, so if a few more pixels were masked off at the edge of the frame, the data would exist for the construction of a template to subtract the line noise across the frame in each row and column. But alas there are not enough masked off pixels in the frame edges that are present in the RAW files of current and past Canon DSLRs to do a statistically reliable job of making an accurate template.
To see what the subtraction could do, have a look at the mouse-over in figure 3 at
Emil, I always enjoy your informative posts. The link to your template example is pretty impressive. I'm stunned that canon doesn't deal with this issue since it seems like an easy fix.
FashionBoy wrote:
Update from my end: My 5D 2 had numerous image quality issues (early serial, bought November 28, 2008), including banding ISO 100 in shadows. Just got back from Canon repairs yesterday: CMOS sensor assembly replaced... if you are having IQ problems send it in for warranty repair/replacement.
As you can imagine, the QC on this camera with the rate of production/demand etc. has been questionable IMHO. I would suspect also that current production is fine now that the bulk of the demand has been satisfied.
Thanks for posting this... Having had my own share of QC issues with the 5D2 I know what you are talking about. It doesn't surprise me that some early 5D2 may have "crooked" sensors. It would explain the IQ difference between different 5D2 bodies. Kind of disturbing... I am glad I returned mine to the shop for a refund. Maybe I will try again later this year...
BTW How is your 5D2 after the fix... do you notice a big difference?
danmitchell wrote:
I'd want to see some evidence that Canon actually uses different (and less capable) components in this situation before I'd accept that statement as a fact rather than a guess.
Dan
Fair enough. It was an inference based on the fact that my 1D3 has vastly less pattern noise than my 40D. That difference could be hardware or firmware. I don't see how non-fixed pattern noise could be reliably eliminated in firmware given the available data in the masked pixels, but perhaps that's just a failure of imagination on my part . I'm happy to entertain suggestions.
Maybe my eyes are failing me, but I see absolutely no difference between the two. Nothing that would suggest that the 1Ds3 was better in any way. And these results seem to confirm what DXOmark had quantitatively measured: That both cameras show virtually identical performance at ISO 100.
Maybe your eyes or monitor are failing you. I definitely see more chroma noise on the 5D2 shot. It's still not bad though for 2 stop recovery.
Yes I have reason to believe that mine was not the only "bad sensor" out there in Mk2s. Afashionshooter.com went through several cameras and I have been in touch with him... only his recent one has been good, the early ones were problematic.
Daan, I have noticed a huge difference with the new sensor. Noise is much more controlled or eliminated, there is a nice smoothness to the files, and the tonal transitions are now beautiful - basically what I had expected from the Mk2 but was disappointed with in the first sensor.
I was beginning to think I was crazy - I had exchanged emails and images with Chuck Westfall and he didn't seem to see the IQ issues I was seeing, and then when I took the camera in the counter person was looking at me really funny when I explained some of the problems.
I think for those of you who feel the IQ is not quite there, send it in to Canon so they can check to see if it is up to spec.
FashionBoy wrote:
Yes I have reason to believe that mine was not the only "bad sensor" out there in Mk2s. Afashionshooter.com went through several cameras and I have been in touch with him... only his recent one has been good, the early ones were problematic.
Daan, I have noticed a huge difference with the new sensor. Noise is much more controlled or eliminated, there is a nice smoothness to the files, and the tonal transitions are now beautiful - basically what I had expected from the Mk2 but was disappointed with in the first sensor.
I was beginning to think I was crazy - I had exchanged emails and images with Chuck Westfall and he didn't seem to see the IQ issues I was seeing, and then when I took the camera in the counter person was looking at me really funny when I explained some of the problems.
I think for those of you who feel the IQ is not quite there, send it in to Canon so they can check to see if it is up to spec. ...Show more →
That sounds really good... I am glad to hear things got solved for you. I know how these things can drive you mad
At the end of the thread some 5D2 users (and myself using a 1Ds3) have performed some "controlled" noise tests to determine if there are differences between 5D2 bodies (and to a lesser extent between 5D2 and 1Ds3 bodies).
Since you now have a good performing 5D2, would you be so kind to take a few test shots and publish them in that thread... Just for the sake of comparison. The "rules" by which to perform this test are described in the thread (I think page 5 or 6), but if you want more specifics I can PM them to you.
FashionBoy wrote:
Yes I have reason to believe that mine was not the only "bad sensor" out there in Mk2s. Afashionshooter.com went through several cameras and I have been in touch with him... only his recent one has been good, the early ones were problematic.
Daan, I have noticed a huge difference with the new sensor. Noise is much more controlled or eliminated, there is a nice smoothness to the files, and the tonal transitions are now beautiful - basically what I had expected from the Mk2 but was disappointed with in the first sensor.
I was beginning to think I was crazy - I had exchanged emails and images with Chuck Westfall and he didn't seem to see the IQ issues I was seeing, and then when I took the camera in the counter person was looking at me really funny when I explained some of the problems.
I think for those of you who feel the IQ is not quite there, send it in to Canon so they can check to see if it is up to spec. ...Show more →
Yes - fully agree with Daan - I too would be very grateful if you could do the test described on the other thread so that we can compare your fixed camera's performance with the 5D2's tested on that thread. I am very keen to establish whether there are two grades of 5D2 out there, and whether the problems you had are the same as the problems that I am having.
I am sorry for the late reply, I have been tied up shooting jobs the last few weeks. If someone can PM me or explain here the test you want me to do, step by step, I would be happy to do it over the next few days.
Some different processing methods have been carried out, but the one described by brainiac down the page is the one to use.
First process from raw in DPP with brightness +2, then input levels (0 1.00 25) in PS.
. I'm happy to entertain suggestions.