So I was told by a couple of people who were hanging around the camera store, that sometimes banding can be caused by *memory card* write speed being too slow; and that the 5DII and 7D were especially vulnerable to this. The consensus was that it would be a bigger problem with video. It wasn't clear to me WHY, but this might be a factor. (?)
SooooOOoooo. . . .I guess maybe I'll haul some of my old cards out and make some experiements with my 5DII. I'll try and report back midday-ish, California time.
I know that some cards are more poorly RF shielded and that some banding/non-random noise can be caused by the poor shielding. I suspect that is not the case here but I agree that it might be worth testing a different brand of card to be sure, even if just a "borrowed" card.
My new (to me?) 7D arrived.
I never got that call they promised, instead I get an email this morning to tell me they shipped it. When I checked the tracking it was already out for local delivery to my place.
I took a few test shots from my patio and so far it looks good, really good.....even at 200 ISO.
I am heading out to get some decent shots. In the meantime here are some samples from the original 7D:
original picture reduced to 6mp for smugmug: >click here<
Interesting but I don't think that's the case for me. I'm currently using Lexmark's 233x pro. That's not slow but not the fastest either.
Sorry for the late return; Life (as it so often does) got distracting.
In any event, the clouds didn't cooperate, so I wasn't able to make a complete comparison. I did shoot some images well to the right at ISO 100, with an original technology 128MB card in my 5DII, which had an impressive negative effect on write speed, but I was unable to produce any banding.
So I wandered into my favorite store and asked around the store bums, and it seems that the low-write-speed-causing-banding issue is primarily a video-capture phenomenon, and no-one there had heard of it actually affecting a RAW still capture. So now we know.
edit: clarification of hearsay from the store crowd
Sorry for the late return; Life (as it so often does) got distracting.
In any event, the clouds didn't cooperate, so I wasn't able to make a complete comparison. I did shoot some images well to the right at ISO 100, with an original technology 128MB card in my 5DII, which had an impressive negative effect on write speed, but I was unable to produce any banding.
So I wandered into my favorite store and asked around the store bums, and it seems that the low-write-speed-causing-banding issue is primarily a video-capture phenomenon, and no-one there had heard of it actually affecting a RAW still capture. So now we know.
edit: clarification of hearsay from the store crowd
Have you since done tests in cloudy conditions to see if the banding still happens?
So I was told by a couple of people who were hanging around the camera store, that sometimes banding can be caused by *memory card* write speed being too slow; and that the 5DII and 7D were especially vulnerable to this. The consensus was that it would be a bigger problem with video. It wasn't clear to me WHY, but this might be a factor. (?)
SooooOOoooo. . . .I guess maybe I'll haul some of my old cards out and make some experiements with my 5DII. I'll try and report back midday-ish, California time.
I find that hard (if not impossible) to believe. It's digital out of the camera. I don't believe it's writing anything any different based on the write speed. If that were the case, we'd have to worry about write speeds on hard drives or risk losing data. Sounds like someone trying to get you to buy a more expensive card...
I know I am coming into this conversation a little late, but I have recently been trying to track down this banding issue everyone seems to see in their Canons. I have a 7D and I found it was particularly apparent when I went from using DPP to using LR4, but I believe that is simply because LR4 was allowing me to really highten the contrast in certain areas of my photos, and thus amplifying the problem. It can be in the shadows or it can be in the bright sky. You see it wherever you expect a nice flat image. Your eyes are particular good at processing this kind of "noise," so it can elude those who just focus on a single noise statistic such as standard deviation.
So, being the technical guy I am (physicist during the day), I downloaded some image analysis program that is given out by the National Institutes of Health called ImageJ. It is not really for photographers per se, but for people who want to do things like count cells in a microscope image. One thing it has is the ability to Fourier transform an image. What is that? An FT is basically a spectral analysis of a function, or in this case, an image. It tells you how much of one frequency or another there is in that function. Knowing how to "read" an FT is as hard as wrapping your head around what it is to begin with, but it is an engineer's tool for figuring out where noise is coming from.
Lo and behold, when I did this to a dark image (lens cap), with minimal processing from DPP or from LR, except boosting the brightness (done to increase the signal to noise of the test), I see distinct peaks associated with a 4 pixel modulation across the image, corresponding to vertical banding when the frame is horizontal. These are not "broad" peaks, but very sharp, indicating that the columns are grouped in 4s and every fourth pixel matches the next, but that there is a reproducible variation between them.
This problem suggests that the columns are read in groups of four, with four separate amplifiers. If their offsets or gains are different, then you will see this. It is seen with lower ISOs, because the effect is small, on the order of the pixel to pixel noise. So, you really only see it when the noise is low (low ISOs). Again, your eyes are really good at picking up patterns, so it jumps out at you.
I would think this would all be correctable. Two images are taken, one with low intensity (dark) and one near full well capacity (bright). From those two images, and some image analysis, one ought to be able to adjust the gain and offset for each column (in groups of four).
If I had an infinite amount of free time, I'd consider writing a program to do this, testing the hypothesis, and distributing the result. Now, I just need that time.
Has anyone found if Canon has figured out a fix for this?
I have a 7D and I found it was particularly apparent when I went from using DPP to using LR4, but I believe that is simply because LR4 was allowing me to really highten the contrast in certain areas of my photos, and thus amplifying the problem.
[...]
Has anyone found if Canon has figured out a fix for this?
You mean Adobe, right? You've already determined that DPP shows less banding than ACR. You're assuming that it's because ACR allows you to push the photo further than DPP that it exacerbates banding noise. The flip side of that is that it's allowing you to exceed the capability of the sensor and amplify noise more. Assuming your guess about DPP is correct, then it's doing the right thing. Or maybe the the noise filtering in DPP is just better, since they probably know their sensors better than Adobe. Who knows, it may be doing the kind of correction that you're already suggesting. If you can improve the banding situation in ACR, I would think you'd have a much better chance of selling it to Adobe than you would Canon.
BTW, the OP in this thread determined that his camera was faulty, so this isn't really a solution for him. Banding noise generally happens with Canon sensors when you're trying to push dark areas beyond the capability of the sensor. You can try to reduce banding, but you're still going to have a poor picture. Far better to choose a better exposure or use multiple exposures and blend them.
I know I am coming into this conversation a little late, but I have recently been trying to track down this banding issue everyone seems to see in their Canons.
...
Has anyone found if Canon has figured out a fix for this?
Thom
Please do try using Capture One to see if it helps with both banding and pattern noise issues. ACR does not not seem to be as optimized for these issues in older Canon cameras' RAW files. ACR does much better starting with the Canon 6D and 70D RAW files.
I am saying that it may be a correctable sensor problem. If I can write a program that takes the raw file and "corrects" it using calibration shots, then any raw converter will be fixed. I suspect that different noise reduction techniques between these products do different smudging (for want of a better term) and only partially fix the problem.
As far as exposing pictures correctly, we all know that some scenes have wide dynamic range, and that we do things like dodging and burning to enhance them. Bottom line is, if the dynamic range is supposed to be there, we ought to be able to use it. Lastly, we can't always do HDR, because some subjects move too much.
Unless your program (LR4, DPP, Capture One, ACR, etc.) use calibration data that you can provide for YOUR camera, they can't really fix the problem. It has to do with there being different amplifiers, all which have their own slightly different gains and offsets. The camera company wants these gains and offsets to be the same, but manufacturing control only allows them to do so well. And each camera will be slightly different. That is why some people see it and some don't. Also, why DPP can't just fix it.
In the end, our eyes are darn good at noticing these effects.
I am (slowly) working on this problem. I will report back when I make some progress.
Has anyone found if Canon has figured out a fix for this?
Thom
The problem occurs with low ISO when you underexpose too much and try to push shadows in post. It has to do with the construction of the sensors and the A/D convertors being separate from the sensor array. Canon made a huge leap reducing banding with the 6D. It is almost gone on the 6D, but if you look very carefully there is a small bit left, if you really push hard in post. But much better than what the 5D2 had, and presumably previous bodies.
Of course, if you just expose to the right, and pull exposure down in post, the problem goes away completely. Also some RAW converters, and noise-reduction software are much better at reducing the banding than others. If you are using LR or ACR, make sure you have updated to the latest software versions.
One other thing to be aware of. Check if you have "Highlight Tone Priority" menu setting turned on. If you do, you may want to consider turning it off if you are having problems with shadow noise. It tends to make shadow noise much worse.
Comments
I know that some cards are more poorly RF shielded and that some banding/non-random noise can be caused by the poor shielding. I suspect that is not the case here but I agree that it might be worth testing a different brand of card to be sure, even if just a "borrowed" card.
Moderator of the Cameras and Accessories forums
My new (to me?) 7D arrived.
I never got that call they promised, instead I get an email this morning to tell me they shipped it. When I checked the tracking it was already out for local delivery to my place.
I took a few test shots from my patio and so far it looks good, really good.....even at 200 ISO.
I am heading out to get some decent shots. In the meantime here are some samples from the original 7D:
original picture reduced to 6mp for smugmug:
>click here<
and cropped in:
>click here<
BTW--this is not the worse photo with noise at 100ISO.
My website
Sorry for the late return; Life (as it so often does) got distracting.
In any event, the clouds didn't cooperate, so I wasn't able to make a complete comparison. I did shoot some images well to the right at ISO 100, with an original technology 128MB card in my 5DII, which had an impressive negative effect on write speed, but I was unable to produce any banding.
So I wandered into my favorite store and asked around the store bums, and it seems that the low-write-speed-causing-banding issue is primarily a video-capture phenomenon, and no-one there had heard of it actually affecting a RAW still capture. So now we know.
edit: clarification of hearsay from the store crowd
Have you since done tests in cloudy conditions to see if the banding still happens?
Facebook Fan Page
Blog
So, being the technical guy I am (physicist during the day), I downloaded some image analysis program that is given out by the National Institutes of Health called ImageJ. It is not really for photographers per se, but for people who want to do things like count cells in a microscope image. One thing it has is the ability to Fourier transform an image. What is that? An FT is basically a spectral analysis of a function, or in this case, an image. It tells you how much of one frequency or another there is in that function. Knowing how to "read" an FT is as hard as wrapping your head around what it is to begin with, but it is an engineer's tool for figuring out where noise is coming from.
Lo and behold, when I did this to a dark image (lens cap), with minimal processing from DPP or from LR, except boosting the brightness (done to increase the signal to noise of the test), I see distinct peaks associated with a 4 pixel modulation across the image, corresponding to vertical banding when the frame is horizontal. These are not "broad" peaks, but very sharp, indicating that the columns are grouped in 4s and every fourth pixel matches the next, but that there is a reproducible variation between them.
This problem suggests that the columns are read in groups of four, with four separate amplifiers. If their offsets or gains are different, then you will see this. It is seen with lower ISOs, because the effect is small, on the order of the pixel to pixel noise. So, you really only see it when the noise is low (low ISOs). Again, your eyes are really good at picking up patterns, so it jumps out at you.
I would think this would all be correctable. Two images are taken, one with low intensity (dark) and one near full well capacity (bright). From those two images, and some image analysis, one ought to be able to adjust the gain and offset for each column (in groups of four).
If I had an infinite amount of free time, I'd consider writing a program to do this, testing the hypothesis, and distributing the result. Now, I just need that time.
Has anyone found if Canon has figured out a fix for this?
Thom
BTW, the OP in this thread determined that his camera was faulty, so this isn't really a solution for him. Banding noise generally happens with Canon sensors when you're trying to push dark areas beyond the capability of the sensor. You can try to reduce banding, but you're still going to have a poor picture. Far better to choose a better exposure or use multiple exposures and blend them.
Link to my Smugmug site
Please do try using Capture One to see if it helps with both banding and pattern noise issues. ACR does not not seem to be as optimized for these issues in older Canon cameras' RAW files. ACR does much better starting with the Canon 6D and 70D RAW files.
Moderator of the Cameras and Accessories forums
As far as exposing pictures correctly, we all know that some scenes have wide dynamic range, and that we do things like dodging and burning to enhance them. Bottom line is, if the dynamic range is supposed to be there, we ought to be able to use it. Lastly, we can't always do HDR, because some subjects move too much.
Unless your program (LR4, DPP, Capture One, ACR, etc.) use calibration data that you can provide for YOUR camera, they can't really fix the problem. It has to do with there being different amplifiers, all which have their own slightly different gains and offsets. The camera company wants these gains and offsets to be the same, but manufacturing control only allows them to do so well. And each camera will be slightly different. That is why some people see it and some don't. Also, why DPP can't just fix it.
In the end, our eyes are darn good at noticing these effects.
I am (slowly) working on this problem. I will report back when I make some progress.
Thom
The problem occurs with low ISO when you underexpose too much and try to push shadows in post. It has to do with the construction of the sensors and the A/D convertors being separate from the sensor array. Canon made a huge leap reducing banding with the 6D. It is almost gone on the 6D, but if you look very carefully there is a small bit left, if you really push hard in post. But much better than what the 5D2 had, and presumably previous bodies.
Of course, if you just expose to the right, and pull exposure down in post, the problem goes away completely. Also some RAW converters, and noise-reduction software are much better at reducing the banding than others. If you are using LR or ACR, make sure you have updated to the latest software versions.
One other thing to be aware of. Check if you have "Highlight Tone Priority" menu setting turned on. If you do, you may want to consider turning it off if you are having problems with shadow noise. It tends to make shadow noise much worse.