I'll be curious if the effect goes in the direction 20D=5D2<40D<50D<7D. A fixed tolerance for CFA misalignment should cause an effect that varies as roughly the square of the pixel spacing -- on a big pixel, a misalignment yields a leakage of R or B into green that is from a much smaller area relative to the area of the pixel.
It may also be possible to do a least-squares fit of the color charts to see if the effect can be nulled out by subtracting a proportion of R from one G subarray, and another proportion of B from the other G. That is, the hope would be that a fixed proportion (G2+a2 R) and (G1+ a1 B), or the other way around, would get the green channels to match on all patches. The coefficients a1 and a2 would be fit by least squares to minimize the G1-G2 difference across the image. If that works, it would be pretty strong evidence that the issue is the CFA and manufacturing tolerances. It would also suggest a slightly more effective workaround.
ejmartin wrote:
I am very familiar with the discussion at OjoDigital; I am the developer of an advanced demosaic algorithm (AMaZE) which should appear in a future release of PerfectRAW being developed there.
Yep, I know that - but others here might not: I linked to that page so that other people could see it.
It would be much better, if indeed the problem is a manufacturing defect, to know how widespread it is and to get it fixed.
Of course - and that's the point for me.
There has been a huge amount of - can't think of any other way to describe it - mudslinging about what was originally being described, unequivocally, as a fault with the sensor: it might still be, but now at last, people are taking a step back, catching their breath and asking if - just maybe - there's some other explanation: no point expending effort on something if it isn't actually the problem...
At least we can all agree that whatever's going on, it isn't unique to the 7D - and that is important...
Phase One has released v5 of Capture One and it supports the 7D. I've downloaded it and I have been running some of my 7D / 5DII comparison/test images through it. I have my own C1 license, but they have a 30 day trial available also. Really nice results!!! If you are a 7D owner that has some C1 experience, you should give this a try.
ejmartin wrote:
I'll be curious if the effect goes in the direction 20D=5D2<40D<50D<7D. A fixed tolerance for CFA misalignment should cause an effect that varies as roughly the square of the pixel spacing -- on a big pixel, a misalignment yields a leakage of R or B into green that is from a much smaller area relative to the area of the pixel.
It may also be possible to do a least-squares fit of the color charts to see if the effect can be nulled out by subtracting a proportion of R from one G subarray, and another proportion of B from the other G. That is, the hope would be that a fixed proportion (G2+a2 R) and (G1+ a1 B), or the other way around, would get the green channels to match on all patches. The coefficients a1 and a2 would be fit by least squares to minimize the G1-G2 difference across the image. If that works, it would be pretty strong evidence that the issue is the CFA and manufacturing tolerances. It would also suggest a slightly more effective workaround....Show more →
I have to run but I found some colorchecker charts taken under sunlight for 20D,50D,5D2 (the white patch is getting very close to clipping on green on some), if the sun ever shines again I can add the new 7D to the mix:
It would be worth checking if the non-grey patches were close to clipping; that could drive the sensor non-linear and cause the chromatic features that are seen.
ejmartin wrote:
It would be worth checking if the non-grey patches were close to clipping; that could drive the sensor non-linear and cause the chromatic features that are seen.
only the white patch seemed to be, but i can go back again just to really make sure
Today, I was forced to use ISO beyond 3200 due to the ultra low light after sunset.
Here's a few starting with good->bad->ugly, in that order of light and image quality.
All taken with 7D and 400 5.6 hand held or with bike support.
A fair amount of PP has gone in the jackrabbit(?) images, including adding contrast, saturation,blurring and sharpening. Fyi. Exif ,as always, is intact.
#1: In great light, 7D + 400 f/5.6 combo shines...as expected, ISO 800
#2: The light was so low that I underexposed the image by 2/3rd stop(confirming that there weren't any shadow areas in the frame), got a whopping 1/400 SS, ISO 4000, , hand held sitting on the ground.
#4: Yea, I focussed on his ass...but still, I was surprised to even get a keeper considering the light (after sunset), ISO5000, 1/1250. Image could be better...but I like the action, hence sharing.
Kirry: I'm impressed at every level. Sharpness and the amount of detail in those high iso images. Wow. I own a 7D, just got it, and have yet to push its limits.
Jeff, congrats, and thanks for the praise. This camera is simply superb for what it can do, I am just scratching the surface, am sure :-). Expecting some great images from you, good luck.
The banding in low ISO is real enough on the 7D, although you have to push the raw quite a bit in order to see it.
I was just playing around to test the dynamic range of the 7D at ISO 100 and lifting the shadows with Fill Light set to the max, when I found it. Here is a number of crops:
Note that I'm fully aware that I'm pushing the sensor to the extreme and that the photo was merely intended for testing. I would however say, that I never say this kind of banding with my old 5D.
Nobody is denying that this banding exists, but - seriously Stoffer - yours is another perfect example of why it is unlikely to be an issue in the real world for most of us: a 100% crop(?) of an image underexposed by God knows how many stops and then dragged back up at conversion?
Apart from playing with it, that file was only ever destined for the Recycle Bin so it doesn't represent real world usage.
I doubt that any other APS-C camera file would respond particularly well to that kind of treatment either: they might not show the banding, but they'd probably be just as unusable...
keithreeder wrote:
Nobody is denying that this banding exists, but - seriously Stoffer - yours is another perfect example of why it is unlikely to be an issue in the real world for most of us: a 100% crop(?) of an image underexposed by God knows how many stops and then dragged back up at conversion?
Apart from playing with it, that file was only ever destined for the Recycle Bin so it doesn't represent real world usage.
I doubt that any other APS-C camera file would respond particularly well to that kind of treatment either: they might not show the banding, but they'd probably be just as unusable......Show more →
Oh, but I agree with all that, Keith.
I'm not trying to say that the 7D is crap by no means, I would just show that one can actually stumble upon the banding in poorly exposure areas that is pushed well beyond normal usage. I would prefer if it was not there at all, but this is the only case for me yet, so I'm fine with it.
Coming from a 5D it is just a weird thing to have lurking in the back of your head, that all.
yep, I know just where you're coming from - but some folk will use your post as "proof" of the 7D's crapness, and I'm just a big fan of keeping things in proportion..!
I used to see the same type of banding with the 5D (original) and 40D when the image was underexposed too much and pulled back from a RAW file. I think that is normal for all digital SLR's when underexposed too much.
