Canon 5D Mark II: Banding... c'mon, still?

Discussion in 'Canon EOS' started by rishij, Dec 8, 2008.

  1. I have previously brought up the issue of banding in shadows for Canon digital SLRs (http://www.photo.net/canon-eos-digital-camera-forum/00RBTe). Very evident in Rebels, less evident in higher-end cameras like the 5D. But, nonetheless, there.
    Disappointingly, it's still there in the 5D Mark II... you'll see some examples below.
    These are more easily seen by lightening shadows using 'Fill Light' or whatever algorithm you choose. But sometimes they're visible on their own.
    First question: Nikon CCD camera based users: do you see this?
    Now, am I to understand that this banding results from variations in the amplifiers for each row OR from temporal variation in read-outs? And do these banding effects essentially result when these variations in gain or readout exceed the signal:noise ratio (which is much more likely to occur at low signal, i.e. shadows... OR, at high ISO where the gain for each amplifier is increased)?
    Here are some examples from the new Canon 5D Mark II (sadly, yes):
    Here's an example of some horizontal banding (note, the image was significantly brightened in DPP):
    [​IMG]
    Here's a comparison of the same area of a shot, at ISO 50 & at ISO 100, on the 5D Mark II, with Fill Light set to +100 to exaggerate the vertical banding in shadows:
    [​IMG]
    Here's the link to the full-size image.
    And here's the banding seen in a high ISO image (ISO 25600):
    [​IMG]
    One thing that confuses me though -- sometimes it seems like the direction of the banding changes from image to image, by 90 degrees (and no, not because the image was portrait vs. landscape). I'm confused...
    Anyway, the purpose of this post, to really figure out why it's there. Berg previously suggested that it's temporal variation, and if I understand this concept correctly, I don't understand why this temporal variation has to exist if scanners like the Nikon LS-9000, which scans 1 (or 3?) lines at a time, doesn't exhibit this sort of banding in extremely dense areas of Velvia 50 slide scans, as is evidenced below:
    [​IMG]
    Link to Full-Size Image
    And that scan was significantly lightened with the Shadows/Highlights tool in Adobe PS CS3 (Shadows: +100, tonal width=14, black clipping = 0.01%).
    Thanks in advance for your help,
    Rishi
     
  2. Right. If you shoot at high ISO, underexpose, and use raise the fill setting... you will get banding, and noise, and a host of problems.
    There is nothing wrong with the camera in such cases, and if you resolve those exposure and post-processing problems that you have created you won't see these "problems."
     
  3. Rishi, cross posting is against forum rules. If you don't like the flak you take in one forum then you have to at least reword the question to post it again!
     
  4. Oops. I just felt I'd posted it in the wrong section since that that thread also included point & shoots and this is specifically for Canon EOS.
    That's why I reposted.
    Hmm... I dunno what to do about it now.
    Sorry,
    Rishi
     
  5. I'd assume all cameras will show banding if you look hard enough. It's almost certainly due to variations in readout noise as the array is read. I don't know how Canon readout their sensor arrays, but it's probably done on a line by line basis. A typical scheme is a parallel vertical shift register and a serial horizontal shift register, so each horizontal array of pixels is loaded into the horizontal shift register in sequence and read out serially. Tiny temporal differences in gain ("noise") can show up as banding.
    For the Canon CMOS array I think each horizontal row is read out seperately via a multiplexing scheme (i.e. they don't have to be loaded into a shift register), so again there willo likely be very slight variation between the rows, resulting in "banding" if you look very closely at the array.
    The only way around banding would be to read out the pixels one by one in a random fashion, which would presumably be much slower and more difficult to do. That might result in random noise rather than pattern noise.
     
  6. BTW scanners only have one line, so they can be more uniform than sensors which have 4000 or more "lines" of pixels. Expecting every one to be identical is expecting too much. You might still see temporal changes with a scanner, so it appears the temporal component must be small.
     
  7. Don't sweat it, normaly the moderators delete one. This is probably the better spot for it but the other one got more interest and is older, so who knows.
    To put your mind at rest, or start a war, I would ask the question mentioning all the IQ issues you think are there. Something like, Are the real world images comming out of the 5D MkII really affected by the fringing, black spot and banding issues, or are they just pixel peeping non isssues that only show up in excessively post processed badly exposed shots? Please only answer if you own a 5D MkII and have printed your images at (whatever size you are going to print).
    Take care, Scott.
     
  8. Bob, thanks. Someone else had mentioned it's probably more a temporal component, but I wonder how much more it's due to readout noise as you mention or just differential gain between amplifiers, since each column shares one amplifier (I believe, from this: CCDvsCMOS ). The article states that gain (for each column amplifier) was sacrificed to increase uniformity... but that can only help so much (IMHO).
    Is it possible that since on a CCD each pixel charge is read sequentially, that any non-uniformity would be less visible (and more randomly dispersed)?
    Rishi
     
  9. You'd have to ask Canon. I don't know the details of their sensor readout scheme for the EOS 5D MkII
    I suspect this is mostly a non-issue except for those who really like to shoot at high ISO settings and post-process the &#@^ out the the resulting images.
    If you look closely enough at pretty much anything, you'll find defects. Canon never claim perfection and you shouldn't assume it. It's a camera, not a precision photometer with every pixel calibrated (and corrected) to NIST standards. Maybe if they cooled it with liquid nitrogen they could do better.
    If you want to do some experimental science (rather than photography), take multiple images and see if the patterning is consistent. If it isn't, it's probably mostly temporal. If it is, it's probably associated with some sort of difference between rows, either due to amplifier gains or some other chip level differences in layout or connection.
     
  10. You're right in that perhaps I just shouldn't worry about it.
    It's just that, to me, it's hard to stomach given that that LS-9000 scan of Velvia 50 received a +100 in the Shadows/Highlights tool (Tonal Range to which it was applied: 14 on a scale of 100), and still showed no dark non-uniformity.
    The Imacon 848 I use, though, shows lots of dark non-uniformity, kind of like on par with what I'm seeing from certain Canon sensors... ok maybe that's a bit unfair. The Imacon (this one anyway) is probably considerably worse. Here you go:
    [​IMG]
    Link to Full-Size Image
    Again, same Shadows/Highlights adjustment to the Imacon scan. And apparently the Imacon's CCD is peltier-element actively cooled.
    Some really cool results of cooling a Canon DSLR sensor here: http://www.pbase.com/terrylovejoy/image/65814409
    Still don't know if it'd get rid of non-uniform noise in the form of banding due to differential gain of amps though. Not that active cooling is practical (or necessary?) in DSLRs for practical use anyway...
    Cheers,
    Rishi
     
  11. Hi Rishi , where is the Nikon 21mp CCD to compare?
    Cheers, Manfred
     
  12. Though I'm hardcore Canon, I have a theory that Nikon cameras don't show this banding. Not their CCD cameras anyway.
    I'd love for someone from the Canon camp to challenge me. I haven't seen this banding in the limited set of Nikon CCD camera images I've attempted to bastardize via Fill Light.
    I don't have a Nikon. Open to debate. Maybe I should post this over in Nikon forums to see what they say.
    Rishi
     
  13. "Maybe if they cooled it with liquid nitrogen they could do better."
    Actually, doing this would lead to a disaster: at 77K, the microlens array will crack and fall apart, and the chip will fail completely due to carrier freeze-out. Cooling mainly helps to reduce dark current which is very well controlled in modern image sensors.
     
  14. OK, so if dark current is well controlled, why in the name of $%&* is column amp variation not?
    I know $2700 for a body may be nothing for you folk in the industry, but it's a lot for a hobbyist. It was hard enough for me to shell out $3000 for the 'holy trinity' of L-series lenses, and $750 for a Gitzo tripod + Manfrotto head capable of holding a 35mm camera still enough for longer exposure shots.
    Before I shell out thousands more for a digital body, I want to know that it's worth more than the $18/roll of Velvia I pay. Wait a minute, that doesn't account for the countless # of hours I spend on a $12k Imacon that can't even function correctly...
    So I guess the moral of the story is: current consumer technology clearly does not satisfy me. Go figure [I'm a self-proclaimed pixel peeper].
    -Rishi
     
  15. Rishi,
    Yes, banding noise is the result of temporal and sometimes permanent variations in the row/column sense amplifiers, at high ISO since the gain is increased this non-uniformity is more visible in the form of vertical or horizontal stripes. This is also true about deep shadows, there is very little signal in the shadows so you're just exploring the noise floor of the sense amps just above the dark current. There are a few points you need to consider about 5d MKII though, the ISO 12,800 and 25,600 are not real ISO's they are derived from 3200 values by sw multiplication of intensity levels by the corresponding factor. This amplifies any non-uniformity and noise. The ISO 50 image you posted doesn't show any banding, all I see is posterization, ISO 100 shows some but I think this is due to the fact that it has been heavily raised ve it's shadow levels in post processing.
    Also regarding scanners, a scanner is much slower than a 5DMKII which can shoot at 4fps, thus temporal fluctuations due to white noise are averaged out, so this is expected.
    Banding is not just an issue with CMOS, in fact the most acute from of banding was observed in Nikon D200 which had a CCD and a particular issue with the readout circuit, Nikon provided a hardware fix for that. Also no Nikon camera currently in production uses a CCD, they are all CMOS.
     
  16. <p>"&nbsp;Actually, doing this would lead to a disaster: at 77K, the micro lens array will crack and fall apart, and the
    chip will fail completely due to carrier freeze-out. Cooling mainly helps to reduce dark current which is very well contro
    lled in modern image sensors."
    <br>
    Brag, the patterned oxide layer (microlens) is epitaxially grown and will survive all the way down to 100mK
    , If temperature is reduced gradually. A Silicon transistor will perform much better at lower temperature due to red
    uced phonon scattering. You can increase the clock speed of your CPU by a factor of 2 or more if you cool it down to
    LN2 tempratures. A silicon photo detector will also perform much better at low temprature due to&nbsp;much re
    duced SRH recombination of&nbsp;optically generated EHPs&nbsp;(silicon being indirect&nbsp;gap material) and trap&n
    bsp;assisted recombination as well as reduction in dark current and&nbsp;Johnson noise.&nbsp;At 4K sensor will operate
    close to quantum noise limit, although you might see some weird&nbsp;coherence effects.&nbsp;&nbsp;I routine
    ly test IR&nbsp;detectors at 4K in our lab to beat noise. A moderately doped (&gt;1E18 cm^-3) silicon wafer will co
    nduct nicely at 77K despite&nbsp;the fact that it&nbsp;has lo
    wer carri
    er conce
     
  17. Rishi -- I would bet this camera produces superb 8x10 inch and larger prints. Why the super-pixel peeking fuss? Or, is that your hobby -- transistor gains and stuff?
     
  18. I never understand you people here... while most people are trying to reduce noise in their images, you guys are trying to exagerate it as much as possible.
    From all i've read, the 5D is the best camera for noise ever made... are you guys still complaining? Just go take some damn pictures and enjoy it.
     
  19. Well said David....I find these forums amazing..inhabited by pixel peepers and 'photographers' with way too much spare time on their hands...why aren't they out taking pictures?
    Way too much fixation on one aspect of the photographic chain and way too much time looking at images at 100% instead of normal viewing distance.
    Get a life folks!
     
  20. Well said David....I find these forums amazing..inhabited by pixel peepers and 'photographers' with way too much spare time on their hands...why aren't they out taking pictures?
    Way too much fixation on one aspect of the photographic chain and way too much time looking at images at 100% instead of normal viewing distance.
    Get a life folks!
     
  21. Again, well said David!
    If you deliberately underexpose a high ISO image and then brighten it up in PP you're asking for banding. Why even bother doing that? Would you underexpose your work in really life? The answer is likely "NO". If you expose your images correctly and there's still banding, now we have a serious problem.
     
  22. sometimes if the images are oversaturated I'll notice this but if saturation is lowered and no sharpening is applied I havent really noticed any. I'll go home and try to replicate.
     
  23. Hey Charles I agree, but for the stock photographers out there, we've been breed to examine the pics ad the image reviewers do so we could get them accepted and have less images rejected. Banding will get an image rejected, so will excessive noise and artifacting.
     
  24. Arash - Microlenses are typically formed in patternable organic material like photoresist. I've never heard of any epitaxial layer used for a non-crystalline material, e.g., silicon dioxide cannot be epitaxially grown.
    Sure, a transistor will work at cryogenic temperature, but the photodiode region in the pixel has a lightly doped region where the carriers will freeze out at low temperature. 1E18 is actually a highly doped material. CMOS sensors are typically fabricated on 5E14 doped epi silicon. IR detectors hybridized to silicon readout multiplexers work better when cooled but commercial monolithic silicon detectors won't. Modern sensors have pinned photodiodes with very low surface recombination velocities, yielding high sensitivity and low dark current.
     
  25. Thanks Felix... exactly my point.
    Anyway, for those of you who think I was posting 'unreasonable' examples, note that this image where I showed banding by adding some Fill Light:
    [​IMG]
    ... comes from the lower left corner of this image:
    [​IMG]
    In such scenes encompassing a wide dynamic range, unless you're shooting negative film, you tend to expose for the highlights to retain their color. In post-processing, perhaps I want to bring some information out of the shadows. But BAM I get struck by banding.
    Not pleasant.
    -Rishi
     
  26. Yeah, well thats digital for you.
    Me? I'm staying with film for a while yet on my fine art photos and scanning the negs at 4000dpi.
     
  27. Stephen,
    Yeah, the ironic thing about me saying that you made a fanboy outta me is that, well, uh, I think I am gonna remain a fanboy of Velvia for now :)
    Although, I could just get a Fuji S5 and expose for the shadows.
    Cheers,
    Rishi
     
  28. Berg,
    You had previously linked here to show how to get rid of fixed column (row?) amp noise.
    Looks nice, but how do you go about doing it? Does Canon/RAW converters attempt this themselves? Or is it something we can implement by analyzing time-invariant noise from dark (black) exposures?
    Rishi
     
  29. Brag,
    you can certainly grow oxide by epi, it is going to be single crystal oxide which is basically glass, It has even been done with MBE. microlense in not organic or photoresist, photoresist is light sensitive and will change over time.
    Carriers in a photo diode are generated optically and not by doping, Freezout is not an issue in ANY CMOS device, 1E18 is low doping typical N-doping today is in the order of 1E20/cm^3. Even with a 5E14 doping conductivity is much better at 77K than 300K, it is simple physics. At 300K with Nd=5E14 you get 4.93E14 electrons/cm^3 while at 77K you get 3.11E14 electrons/cm^3 while mobility is increased by 2 order of magnitude, just look at a smiple electrical conductivity vs temprature plot for bulk silicon in any basic semiconductor book.
    The dominating factor in any detector, solar cells etc. is bulk SRH recombinatio at the junction which improves at low T. Any ways.
     
  30. Arash, your technical knowledge is beyond my questioning, sometimes, but it staggers me that somebody that researches this field in the USA can't spell temperature, ever!
    As I said in the other post with the same question, the worst camera I have used for this is the 1D, (it has a CCD sensor) but it is very easy to allow for, the exposure latitude is comparable to Velvia, anybody used to exposing for slide film would never have an issue with this stuff.
    Anyway back over to you academics, I have to go do some more processing.
    Take care, Scott.
     
  31. Brag,
    FYI, here is an abstract (public) from U.S. Patent 6001540 about how microlens array is fabricated
    "A process is described for forming a microlens, either directly on a substrate or as part of a process to manufacture an optical imaging array. The process starts with the deposition of a layer of silicon oxide over the substrate, said layer being the determinant of the lens to substrate distance. This is followed by layers of polysilicon and silicon nitride. The latter is patterned to form a mask which protects the poly, except for a small circular opening, during its oxidation (under the same conditions as used for LOCOS). The oxide body that is formed is lens shaped, extending above the poly surface by about the same amount as below it, and just contacting the oxide layer. After the silicon nitride and all poly have been removed, the result is a biconvex microlens. In a second embodiment, a coating of SOG is provided that has a thickness equal to half the microlens thickness, thereby converting the latter to a plano-convex lens. "
    Also interesting enough, I found an online process for making microlens array by etching oxide:http://www.mems-exchange.org/catalog/P3433/
    I think the organic material you are refering to is used in the CFA (color filter array), which is a seperate layer on top of microlens array.

    00RkWA-96267584.jpg
     
  32. Scott,
    I see I dropped an "e" from temperature-I am sorry-I sent that message from my blackberry with its tiny keyboard, I apologize for the typo, you have to excuse people for typos on internet forums. Also, I know very little about many things, CMOS devices and low tempErature characterization just happen to be my area of research. I look forward to learning more about photography from all the experts on this web site.
    Thank you for pointing this out.
     
  33. Who's 'Brag' anyway?
    And while I love academic talk, I'm still wondering if there's any practical way to get rid of the banding in software... Berg had given me hope earlier that there is. Can anyone else chip in as to the actual process? What software? What algorithm... etc.?
    Rishi
     
  34. Arash,
    You're right that the banding is non-evident in the ISO 50 image -- because ISO 50 is in fact ISO 100 overexposed by one stop... which essentially means that there were not deep shadows in the exposure. Hence no banding.
    So, since you're an expert on CMOS... is it just a quality control issue that some cameras show this defect more than others?
    Is the only way, in the end, to reduce this problem by using lower gains on amps and making higher sensitivity photocells? If I understand correctly, they lowered the amp gains in the beginning when introducing CMOS to these cameras in order to reduce non-uniformity in dark areas.
    Rishi
     
  35. Not seeing this on Nikon.
     
  36. "So, since you're an expert on CMOS... is it just a quality control issue that some cameras show this defect more than others?"

    Dear Rishi, I am not an expert in commercial image sensors and I don't work for Canon, I just know a little b it...
    Rishi, all cameras have banding because there is always mismatch between the sense amplifiers whic h will be evident when you amplify such a small signal from darkness, even if there is no mismatch and you use unity gain the signal level will still be slightly different because you are operating close to the noise floor of the sense amps. The only way you can take better pictures in darkness is larger or more efficient pixels to raise the sig nal level.
    Some of the banding also comes from cross talk and mismatch between signal pathways that connect the sense amplifiers to buffers and ADC. I know that in DRAM for example this issue is corrected by equaliz ing signal paths by using a folded structure, however this is not doable in an image sensor. I am sure the re are many other circuit tricks I am unaware of. I think Sony/Nikon use a column parallel design which places ADCs at each column and thus shortens the analog pathway which might result in lower banding. You can also average for longer like your scanner but then you wi ll have a very slow camera.
    Like I said I don't see banding an issue with any of the cameras on the market today at least up to ISO 3200 (the max I use) At 12,800 or 25,600 images from all these cameras are plain noisy a nd useless with or without banding. At low ISOs I haven't seen much banding from 5DII files yet, can you post a raw file with shadows that you think might exhibit significant banding? here is a crop form a 5DII image po sted in
    00RkZ1-96289884.jpg
     
  37. hum looks like photo.net truncated my post! but I was trying to say that the above crop at ISO 3200 pushed by 2.0EV shows a lot of noise but no banding.
     
  38. Banding on the 5D and probably the 5D MkII has to do with the lenses used on it. The 50mm 1.4 motor caused banding on the 5D. (Along with a few micro-ultrasonic lenses).
     
  39. Arash,
    Here's a link to an ISO 100 RAW image from photographyblog.com that exhibits it (same image I posted above). Increase the exposure or brighten the shadows, and you'll see the banding in the building/windows.
    http://img.photographyblog.com/revi..._ii/sample_images/canon_eos_5d_mark_ii_07.cr2
    Carnagex: That sounds like a rather implausible explanation to me, but, it'd be easy to test: just take some images of black with the lens off (assuming that's possible!).
    -Rishi
     
  40. From what I've seen thus far, banding starts in at around 1600 and gets ridiculous by 6400 (IQ = barely usable, if at all).
     
  41. Arash - I can tell that you are not very familiar with image sensor technology. Microlenses are formed on top of the color filter array, not under it. Glass is simply not a crystalline material. Another practical problem: The various alumimum layers used for interconnects in a CMOS chip will melt at the temperatures required to clean the wafer surface prior to epitaxial growth...

    Rishi - Sorry for the distraction... The banding you're seeing is not the fixed pattern noise caused by slight threshold offsets between the column readout amplifiers discussed in the reference I pointed you to. Fixed pattern noise correction is simple and automatically performed in the camera FPGA
     
  42. Berg -- if it's not fixed pattern noise, then why does Igor see this when layering 16 different Canon 20D exposures and averaging (i.e. with one exposure it wasn't evident, but averaging 16, it became evident):
    http://www.flickr.com/photos/98957482@N00/3095714410/
    Igor, are you here yet?
    Thanks,
    Rishi
     
  43. Berg,
    Glass in your window is amorphous, glass above refers to quartz which is rhombohedral single crystal: http://en.wikipedia.org/wiki /Quartz
    Microlens is aligned to the CCD/CMOS chip post processing so the thermal budget is not altered. Also CMOS sensors (like any other device) use Cu for interconnect no Al, Al has not been used in an y major CMOS device since early 2003, one reference is here http://www.ph ysorg.com/news5184.html
    Any ways, since you are the expert in image sensor technology who is a humble electrical engineer and a physicist to disagree. I go back to my work but you may want to read this excellent article here http:// theor y.uchicago.edu/~ejm/pi x/20d/tests/noise/
    banding noise:
    "In terms of its spatial variation, read noise is not quite white. Upon closer inspection, there are one-dimensional patterns in the fluctuations in Figure 2. Because the human eye is adapted to perceive patterns, this pattern or banding noise can be visually more apparent than white noise, even if it comprises a smaller contribution to the overal l noise. Pattern noise is more problematic in older models such as the Canon 20D; see figure 3. "
    quoted from open source article linked above.

    And because you are an expert I am sure you have access to IEEE database, so I would also recommend reading this paper from one of my great professorrs, Abbas El Gamal, who is the undisputed authority in image sensor technology not only here at Stanford but around the world.
    I quote fro technical paper
    "In addition to temporal noise, image sensors also suffer from FPN, which is the pixel-to-pixel output variation under uniform illumination due to device and interconnect mismatches across the image sensor array. These variations cause two types of FPN: offset FPN, which is independent of pix el signal, and gain FPN or photo response nonuniformity (PRNU), which inc reases with signal level. Offset FPN is fixed In addition to tempor al noise, image sensors also suffer from FPN, which is the pixel-to-pixel out put variation under uniform illumination due to device and interconnect mismatches across the image sensor array. These variations cause two types of FPN: offset FPN, which is independent of pixel signal, and gain FPN or photo response nonuniformity (PRNU), which increases with signal level. Offset F PN is fixedz from frame to frame but varies fro m one sensor array to another. Again, there are more sources of FPN in CMOS image sensor s than CCDs introduced by the active readout circuits. The most serious addi tional source of FPN is the column FPN introduced by the column amplifiers . Such FPN cancause visually objectionable streaks in the image"​
    Source : CMOS image sensors,E l Gamal, A. Eltoukhy H.Stanford University,IEEE circuits and Devices Journal, June 2005, issue 3 pages 6-20

    Quotation from technical journal with author permission.

    I am sorry but this issue is closed.
     
  44. Rishi, I'm here.
    Just wanted to say, no need to repeat the test without the lens, it was tested with an EF 50 1.8 II, so no USM.
    It might be interseting to repeat the test thou, just to see if fixed pattern noise changes with time. The test was done over a year ago.
    At least on 20d in-camera jpeg conversion is realy bad and the banding becomes evident even in relatively bright areas shot at ISO 800.
    For me the problem was removed almost completely just by shooting raw and converting to jpeg out of camera. For example see this picture at equivalent of ISO 25600 .
     
  45. Rishi,
    Please see the references above on banding noise, or streaking artifacts. You should not trust every claim you see in an internet forum without valid citation. also note that temporal fluctuation and random cross talk might cause streaks to appear at different locations in successive images.
    Going back to the RAW file, I converted it using DPP 3.5.1 with +2.0EV EC. you are correct, there is banding artifacts in the shadows more than I expected, it even shows at 50%. I searched through my 40D images and I found two extremely underexposed samples one at ISO 100 and another at 400 to see if I can reproduce banding by +2.0EV EC. To my surprise 40D RAWs did not show banding as much as 5DII (see the examples below). It might be possible that something was wrong with the particular camera you were testing, or if live view had been used extensively it might have raised the internal temperature of the components and aggravating this issue, although this is very unlikely. But I will keep an eye on more samples, the rest of the samples I had seen so far did not exhibit excessive banding.
    00Rklp-96395584.jpg
     
  46. Hmm... the crop seems to be missing?
    Also, this could be an inter-unit variation, and not specifically 40D vs. 5D MII, right? Though probably more a sensor-based issue, so inter-model variation.
    I don't know. I wasn't testing this camera... these were taken from photographyblog.com, just so you know :)
    By the way, your crop of the 5D MII image shows it even better than mine... thanks Arash :)
    Thanks,
    Rishi
     
  47. OK I see it. Nice, not so evident on the 40D...
     
  48. Here's another example from 40D at ISO 400, again underexposed.
    00Rkmc-96403584.jpg
     
  49. and crop, only at ISO 400 it shows a little bit of banding as far as I can see on my screen
    00Rkmd-96403684.jpg
     
  50. I, also, hardly see the banding, Arash, in either of your 40D posts.
    Man... not looking good for the 5D Mark II...
     
  51. Rishi,
    I looked at some of my older photos which were taken with my 20D, interestingly I found that 20D is more prone to this issue, I am posting only one example here, unfortunately I have neither the time nor the energy to post more samples and do exhaustive comparisons, but qualitatively my 20D shows more banding than my 40D. Without quantitative measurements from several cameras of each model it's hard to tell whether this is sample variation or specific design issue related to sensor readout circuit. I am going to the IEDM (International Electron Device Meeting) next week, I will try to talk to the core technical people from Canon who will also attend the conference (this will be difficult as I can't speak Japanese!) and ask them about this issue as well as the black pixel problem. If I succeed and they consent, I will reword and share my findings here.
    Rishi, I am not such a great photographer as you can tell from my gallery but one thing that I have learned from all the workshops and some really talented people here and over on the BPN forum is that if you take a good picture it will be a winner even if it shows some noise or banding. Look at all the grainy photos which won the Pulitzer prize. If you haven't ordered the camera yet just wait a few months and see what happens, if you have be happy, grab it and go take some pictures, worry about banding whenever you see it in your photos.
    00RknK-96411584.jpg
     
  52. 100% crop-good night!
    00RknL-96411684.jpg
     
  53. Agh, this is one thing I hate. I'm extremely suprised its popping up in such a high end camera. It reminds me of the sort of results I used to get with older generation CCD based DSLR's (A100, 7D).
    Give me noise, that's fine, as long as you still get sharp detail and no smudgy NR applied. But banding, oh dear, I don't think you can do much about that in PP. It's just ugly to my mind. This issue obviously isnt going to be an issue with everyones style of shooting, but then if you are paying that much for a camera it just shouldnt be there at all IMO. Some mention it's the users fault that he has underexposed & then pushed the image in PP... well, sadly, you have to do that under some circumstances. Bottom line, I just don't think this "character" should be there at all.
     
  54. Some mention it's the users fault that he has underexposed & then pushed the image in PP... well, sadly, you have to do that under some circumstances.​
    Exactly. Specifically because these cameras have such limited dynamic range. Possibly not as much of a problem with the Fuji S5 since with that camera you can expose for the shadows and not worry about the highlights.
    Arash, cool, see what information you can glean. I agree that this sort of banding, Richard, is hard to get rid of in post-processing... I certainly don't know how. Neat Image & Noise Ninja certainly do not get rid of it, as these programs look for random noise and typically preserve lower frequency detail (the banding falls under 'lower frequency noise'). So, sometimes, the banding is enhanced when you remove all the other higher frequency noise.
    Random noise is not as easily picked out by the eye as regular patterns like banding. The 5D Mark II banding above kinda looks like the banding you get when Epson printer heads are not aligned... and I notice those things immediately. Even on prints made by the Epson R2400 and higher models at the Art Wolfe gallery here in Seattle. To me, it's distracting, and takes away from the 'realism' of a landscape photo.
    Rishi
     
  55. >>I know $2700 for a body may be nothing for you folk in the industry, but it's a lot for a hobbyist.<<

    Well, which is your hobby, taking pictures or pixel peeping? ;P

    Seriously, if can't make a GREAT print using images shot with the 5MKII I would not know what could possibly satisfy you.
     
  56. Thanks Rishi, glad you appreciated my point. Now, i know one can dwell on the bad points of a camera... we should appreciate that this is still a very unique camera with some very interesting features. But the topic is the sensor... and it's noise control... The simple fact is that it shouldnt, be there, should it? I never knew CMOS sensors could band actually, but i do have very limited experiences with such cameas (I get no banding off my A700 at ISO6400)...
    [​IMG]
    "Seriously, if can't make a GREAT print using images shot with the 5MKII I would not know what could possibly satisfy you."
    Likewise, if you need a "5MKII" to take GREAT photos, you have to question what you're satisfying :)
     
  57. WILL ALL YOU PRO SITTING ON HIGH HORSES BE COURTEOUS ENOUGH TO FIRST DEFINE AND EXPLAIN "BANDING." THEN GIVE EMPIRICAL EVIDENCE/EXAMPLES OF BANDING... I THOUGHT PHOTO.NET IS INCLUSIVE AND CONTRIBUTORS HAVE AN ETHICAL RESPONSIBILITY TO DISSEMINATE KNOWLEDGE OF PHOTOGRAPHY TO ALL WHO ARE WILLING TO LEARN & PRACTICE. PERHAPS I AM WRONG & SHOULD MOVE TO SOME OTHER FORUM.
     
  58. Hahaha... that was actually kind of funny.
    Also, why are you shouting?
    Rishi
     
  59. "WILL ALL YOU PROS SITTING ON HIGH HORSES BE COURTEOUS ENOUGH TO FIRST DEFINE AND EXPLAIN 'BANDING'"?
    It's when the photograph contains bands ("stripes") that weren't in the scene photographed.
     
  60. I just redid the test on 20d. The banding looks the same, but the fixed pattern noise I got out of averaged 16 exposures is totaly different than the pattern from a year ago.
    Obviously to reduce banding it only makes sense to subtract average of dark frames from about the same time as the photos were taken.
    One important difference I noticed is noticable amplifier glow which was not present a year ago. Exposure was the shortest possible so it's not from long exposure. I'm not sure what exactly this means, but it seems electronics degrade as time passes. This may affect the photos.
    I might repeat the test tomorrow, to see if it changes from day to day.
     
  61. I've heard of amplifier glow as a camera ages. Not exactly sure what it is, but seems like it's typical.
    Also, the averaged banding may just be something that changes every single time you perform the test. Have you tried doing 16 exposures, merging them, and then immediately doing 16 more exposures, merging them, etc.?
    It may be a pattern that results from an otherwise random process, in which case, it'd change every time
    Rishi
     
  62. Rishi- could you post the RGB histograms for the shot?
     
  63. Out of curiosity I just did the test again. One day later the fixed pattern noise is the same. Thru all the noise I can't spot the difference. But it is totally different than the pattern from a year ago. The fixed pattern is obviously changing but this is a slow process.
    I have two more ideas to try. I will change the lens and redo the test, and then take the battery out and again redo the test tomorrow. I already changed the card so this is obviously not a factor.
     
  64. Nice work Igor. Thanks! I'm glad someone else takes this seriously.
    -Rishi
     
  65. Buy a Nikon?
    Use fill flash?
     
  66. I redid the test as I said. The fixed pattern noise was the same.
    My findings are as follows:
    1. Horizontal bands are random from shot to shot
    2. Vertical bands are random but change very slowly - after two days I can't tell the difference, but a year earlier it was totally different
    3. The lens being used or the changing of batteries don't seem to have an effect on read noise
    The tests were done on a Canon 20d.
     
  67. Wonderful, Igor.
    So that points to the banding resulting from column/row (I forget which one) amplifier static (as opposed to temporal) variation. Amplifier characteristics can probably change slowly over time, but for a day-to-day basis, may remain largely the same.
    So, Berg, or someone -- can't we fix this in software by shooting a black frame(s) prior to our shoots?
    More importantly, can Canon fix it? Even though I doubt they'd really care to given that there are already people in this thread telling me to ignore the banding. Which I find to be a just plain dumb solution. Or lack thereof.
    Rishi
     
  68. Hi I don’t no why this thread is getting so muct response as any digital image that needs that much lifting is scrap way before any banding ... unless you like that look the only way I can see for you is
    1 expose correctly or
    2 sell your digital kit and shoot print film
    I did think that I had never seen banding in any my images from 5D and 20D but then I remember a photo I take inside when I had been using 20D on manual out side on a sunny day A friend I not see for some time came to see me I picked up the 20D and shot him proberly 7 or 8 stops under
    I had to lift it so much that all the shadows went a red banded mess. But I turned it to black and whit and I like the canvas affect
    Dave
    00RnLF-97549784.jpg
     
  69. Hi I don’t no why this thread is getting so muct response as any digital image that needs that much lifting is scrap way before any banding ... unless you like that look the only way I can see for you is
    1 expose correctly or
    2 sell your digital kit and shoot print film
    I did think that I had never seen banding in any my images from 5D and 20D but then I remember a photo I take inside when I had been using 20D on manual out side on a sunny day A friend I not see for some time came to see me I picked up the 20D and shot him proberly 7 or 8 stops under
    I had to lift it so much that all the shadows went a red banded mess. But I turned it to black and whit and I like the canvas affect
    Dave
     
  70. I suspect that this results from Canon's basic technology.
    Perhaps that's why Nikon stayed with CCD's for so long.
    You probably need to prevent this by exposing to the right and fill flash.
    A shoe flash with a head that tilts and turns will let you amplify existing light.
    Many pictures that look like existing light were done with flash.
    A most common source of existing light is a window or door. That's easy to simulate with flash.
     
  71. Thanks Steve. Yeah it's disappointing. Here are what some Nikonians say when they see it in shadows of their D70 (which is really cheap):
    "i just don't think it's as high a quality camera. i'd be willing to bet it's just a mediocre sensor."
    Yet here are Canonites defending seeing the banding on the newest, latest & greatest $2700 price tag Canon 5D Mark II.
    Dave, I don't have a digital setup. I shoot film & was trying to justify switching but am unable to given the latest performance reviews of the 5D MII.
    Rishi
     
  72. FYI the Nikon D200 banding seems entirely different, qualititatively, than what I'm talking about. Also, it doesn't just appear in the shadows for the D200. There was clearly some quality control issue there.
     
  73. The Nikon D200 was a CCD sensor camera.
    At $1600 MSRP it wasn't cheap either.
     
  74. We may be at the limits of what is physically possible. My view is that if I really need 3200 to get an image, no one including the B&G should expect it to look like 100.
     
  75. I found a new problem with my cameras - they dont take photos with the lens cap on. What a horrible defect this is - even a pinhole camera can do that. What is Canon doing? And dont you Canon fanboys come here and try to defend that.
    "Yet here are Canonites defending seeing the banding on the newest, latest & greatest $2700 price tag Canon 5D Mark II."
    Yes - heaven forbid that anyone try to defend a camera that exhibits a problem in a scenario which would virtually NEVER apply in the real world.
    "Dave, I don't have a digital setup. I shoot film & was trying to justify switching but am unable to given the latest performance reviews of the 5D MII."
    So stick to film.... I dont mean to be rude but what is the point of griping about this? If you are trying to find out the cause of the banding, that is one thing. But going on and on about the defects of the 5DMk2 - when it has been pointed out several times to you how contrived the scenarios are, during which these so-called defects become apparent - is a bit pointless. And if you are just not happy about Canon's implementation - email Canon.
    Giampi was spot on the money here. If someone cannot take good photos with a 5DMk2, they should probably consider painting or some other hobby.
    Vandit
     
  76. )))))Dave, I don't have a digital setup. I shoot film & was trying to justify switching but am unable to given the latest performance reviews of the 5D MII.<<<<
    Well I not know that.
    A digital image to me never looks as nice as film.... cannot tell you why Evan when the image has no banding etc I don’t think the reasons for this can be quantified
    If you have a canon eos system maybe if you get a second hand 5Dmk1 to see how you get on with digital, By what I hear/guess, the defiance in IQ between the MK1 and MK11 will only show at hi iso and if you print at bigger than A3
    Dave
     
  77. If you are trying to find out the cause of the banding, that is one thing.​
    You know how to read, Vandit, or did you just wanna jump to giving us your useless opinion such that you missed the premise of this entire post when I wrote in the OP:
    "Anyway, the purpose of this post , to really figure out why it's there . Berg previously suggested that it's temporal variation, and if I understand this concept correctly, I don't understand why this temporal variation has to exist if scanners like the Nikon LS-9000, which scans 1 (or 3?) lines at a time, doesn't exhibit this sort of banding in extremely dense areas of Velvia 50 slide scans."
    That being said, after getting a better idea of what causes, I'd be more comfortable e-mailing Canon about it. It's not like they couldn't use some help pinpointing their own problems, as one has to wonder how the 'black dots' problem ever made it past Q.C.
    Then, 2 posts ago, I was asking if anyone had any ideas on how to get rid of the banding in software.
    So, yeah, clearly I'm here just to gripe. You can also believe in Santa Claus, for all I care.
    -Rishi
     
  78. Rishi, a few ideas, how to remove the horizontal banding:
    1) take 4 or more exposures, average them so that temporal variations cancel out, and subtract an average of a number of dark frames to get rid of fixed noise pattern. I believe most of what will remain in the shadows will be foton noise (at high ISOs).
    Obviously you can only do this with static scenes.
    Actualy, doing that at high iso, is somewhat equivalent to taking a long exposure at low iso, except you can get even more dynamic range by not averaging, but by making a sum of exposures. You'd of course need a sum of the same number of dark frames made at the same ISO to effectively get rid of all non image forming noise.
    2) You could also make only 2 exposures in manual exposure mode one at say ISO 100, the other at ISO 1600. ISO 1600 would be used to extend the dynamic range by 4 stops in the shadow area. This way you'd push the banding 4 stops lower.
    3) Get rid of vertical banding by just subtracting the average of dark frames.
    You also get rid of some horizontal banding this way, but unfortunately you also lose information where there was no banding noise.
    Needless to say, you'd get the best results doing all this in linear 16bit rgb space before you apply other corrections.
    p.s. I don't do any of this, as I said earlier, for my needs out of camera raw to jpeg conversion is good enough.
     
  79. Interesting that the Nikon LS-8000 exhibited banding in shadows.
    http://www.imaging-resource.com/SCAN/CS8K/C8A.HTM
    The article says:

    "the Super Coolscan 8000 ED has a special "super fine" scan mode that uses only one of the three rows of pixels on its CCD line sensor. This supposedly reduces streaking in the image when extreme tonal adjustments are made to the image, apparently by eliminating the slight row-to-row differences between the CCD elements."
    People complained about this on the LS-8000 quite a bit, and it was eliminated in the LS-9000. Imagine that, people actually complaining about banding in shadows of film scans!
    More importantly, they fixed it. I see no banding whatsoever in LS-9000 scans.
    So there's a fix. Albeit much easier perhaps on the LS-9000 since it's just a 3-line CCD. But both the LS-8000 & LS-9000 have 3-line CCDs, yet they fixed the problem.
    This means fixing it must not be impossible right?
    Anyone have any ideas how they fixed it on the LS-9000?
    -Rishi
     
  80. Igor, thanks. Just out of curiosity, what method do you use to average the dark frames?
    I understand subtraction methods using layers in Photoshop because I do this sometimes to remove variations in CCD response of a LS-5000 scanner.
    I'd like to try it with the digital camera... I'm skeptical about whether or not it'll work though. Should be an interesting experiment though!
    -Rishi
     
  81. Basically, banding is the result of trying to capture a photon signal that is less than the electronic noise of the chip.
    Film probably handles this better.
    Consider trying fill flash 3 stops below ambient.
     
  82. You mean that is less than the electronic noise/variation of the row/column amps. What you describe is just 'random noise', which can be easily removed by Neat Image.
    Banding cannot. Not by Neat Image anyway.
    Maybe I haven't seen this banding in shadows in one Nikon image to date because they figured out how to get rid of it on their LS-9000 scanner?
    Just throwin' it out there. As is evidenced by the black dot problem, Canon needs more pixel-peepers on their team.
    -Rishi
     
  83. Rishi, like I said, I don't do any of this, except for occasional test. I do it the only way I know how, by sequentialy merging 50% transparent layers. That's why I merged 16 exposures: 16>8>4>2>1. I stopped at 16 because 32 would be far too much work for me :)
     
  84. This is not banding..this is inherent in any ANY digital device and is the result of physics..gain in the sensor which is naturally necessary to raise the tiny voltage that results from a few photons striking the miniscule photosites. If one takes pictures with an appropriate exposure this effect is invisible. Changing the parameters by emphasizing either the gain in RAW or changing the frequency response of the sensor output by essentially tweaking the normal response of the sensor and amplification of signal is no different than turning up the treble on an audio amplifier ( remember them?) and in the latter case results in hiss from the equipment. We are dealing with physics here and the principles will never go away. Remember or learn for the first time...a CMOS or CCD sensor is an ANALOG device at the photosite. It is converted to digital signals as close to the sensor sites as possible but it starts out voltage dependent and thus there is always a noise component. What a camera company does to ameliorate such noise is unique to each manufacturer and averaging pixels is one way to deal with it as well as secret methodologies. Any digital camera that I have used will show patterning and chroma noise and that includes full frame or smaller sensors and naturaloy the smaller sensors have more noise in general OR lower resolution. Shooting pictures uisng very high ISO settings naturally are noisier unless one has noise reduction used in the exposure which does one of several things but one method is to average a black screen along with the exposure to eliminate electronic and pixel noise by subtraction ( reversing the signal and adding it to the exposed image). I produce as a standard print size..36inch prints from digital images on a daily basis..I do not have banding, I do not have noisy images after appropriate digital processing. I suggest that those that do, are not using their computer correctly in dealing with this technoloy. I have by the way made three foot prints even with the Canon G9 and believe me, I am a stickler for quality. It is all about knowing what you are doing in terms of exposure, minimal ISO, appropriate post processing and so forth. Sure, I can make that little camera show banding..I can also prevent it. On my Canon Dslrs I can also show banding but it requires more effort to ruin the image! The world of digital photography is an incredible improvement over film for me at least. Trying to get a three foot print from a 35mm medium ISO film was a miracle and rarely did it make the cut. I have a long memory for quality and today I am astounded at the quality of the best digital equipment. Surely Canon is in that rarified group of brilliant camera engineers and designers who make such state of the art cameras and lenses.
     
  85. I read Canon's white paper on their CMOS sensors where they explain their noise removal technology for fixed pattern noise. Makes sense, but all this processing goes on (if I read it correctly) prior to gain application by the amplifiers.
    Which, to me, points to the column (or row? I forget) amps being the source of the banding.
    Neil, while I appreciate your explanation, I am still confused as to why I see it more on certain cameras and less on others. I just downloaded a bunch of Nikon samples at full resolution from DPREVEIW, upped the Fill Light to +100, and saw lots of random noise, sure (which can thankfully be removed by Neat Image), but no banding.

    This could just be the sampling that I chose, sure, because some Canon 5D Mark II samples also show no banding at +100 Fill Light; however, more often than not in the samples I've seen, I find it to be common.
    I agree, the level of thought & engineering going into these digital systems is incredible, almost unfathomable. And that white paper only contained a part of the story, right? The resolution of the 5D MII is also incredible from preliminary tests... almost (maybe actually) matching the Sony A900. I imagine that once they fix the black dot problem, I'll probably be picking one up.
    Still, I wonder how Nikon got rid of the banding on the LS-9000. The only way to get rid of it on the LS-8000 was to use 1-line CCD mode (not using the other 2 lines of the CCD).
    Rishi
     
  86. I had a similar problem with only one picture closeup with 5d mark II at ISO 3200. Its banding pattern was same as the one in this forum thread of the picture with the guy with wine bottles. I took the picture to another PC to post it in this forum and send it to canon. But I could not see any banding on the picture. took it back to the other PC, banding seen.
    I think we should also consider our monitors or video cards or drivers as a source of the banding. I am a new to canon equipment as well as this forum. Just a thought.
    Amar
     
  87. Question, I am seeing white and blue drop outs in my frames after 1000 iso. Is this always there for everyone?
     
  88. Floyd, mind posting some examples?
     
  89. I recently bought a Canon 5d markII, and sold my Nikon D300. After only 1 week taking test shot with 5DMII I noticed severe banding, especially in low light situation. I brought it back to Canon service and complained about this. I was told that they will make some adjustments. After I got it back and made some more test shots, I was very very disappointed to see the same problem. Now I am sorry I let go of my Nikon D300. I hope Canon people will take notice of my post. From what I have been reading in the thread there are a lot who are disappointed.
     
  90. This is what i'm getting from my 5dMKII using multiple lenses and numerous ISO/Aperture combinations:
    This particular shot was taken with a 35mm f/1.4L at f1.8, 1/160th, ISO 100
    [​IMG]
    [​IMG]
    [​IMG]
     
  91. The shots at ISO 50 are disturbing but then again, let's get some storytelling composition here can we! The shots appear to be shot incorrect, poor light, wrong settings possibly. I can't believe these shots are from a 5D Mark II...
     
  92. Is this also to confirm, once and for all, that digital camera's are STILL not where the Hasselblad H2 with a film back is today? Digital's great for checking your shot in the field but if this is what I can look forward to when I want to blow up a great landscape shot... FORGET IT! Hopefully someone can play devil's advocate here. I really WANT to like the 5D Mark II enough to buy one.
     
  93. To Joe Casey
    I'm not sure you can say that - there are plenty of cameras that do much better than the 5dII for landscape. Juste google D3x, A900.
     
  94. For the sake of posterity, I want to finally comment that while this was an interesting academic exercise, as many people have pointed out, this is not a real-world problem save for in:
    1. Huge dynamic range scenes
    2. Exposure mistakes
    Not much point in commenting about (2) but as far as (1) is concerned, of course one should use:
    • Graduated Neutral Density filters
    • Multiple exposures at different EV
    • Multiple exposures at same EV, then remove noise in shadows by pixel averaging
    Regarding my previous comments about film scans not showing such banding: I must admit that now, with much more experience with both mediums, the ability to pull detail out of shadows of clean film scans is offset by the lower signal:noise of the film, compared to good dSLRs, to begin with (Roger Clark shows that the signal:noise of Velvia 50 is about equivalent to somewhere between ISO800 & ISO1600 on a Canon 1D Mark II here). Hence you're much more likely to have usable shadow detail in a low ISO digital capture than a Velvia 50 scan, especially given the generally accepted observation of the higher dynamic range of dSLRs compared to slide film.
    One day I'd be curious to see if slide film could do better in the digital era with higher dynamic range scans of slide film, since so much shadow detail is lost because no scanner I've used can see all the detail in the shadows of Velvia (that you yourself can see with your naked eye & a bright light source). But that is for another discussion :)
    I thank everyone for their valuable input on this thread! Funny how hindsight is 20/20. Or, er, 19/20 anyway... I mean you never know ;)
    -Rishi
     
  95. And, to state the (maybe not so) obvious: different cameras/models probably exhibit this sort of banding or fixed pattern noise (correct me if I'm using that term incorrectly) to different degrees.
    For example, here is the fixed pattern noise on my 5D, an average of 16 black exposures at ISO 1600, +4 exposure for viewing ease:
    [​IMG]
    Canon 5D Fixed Pattern Noise, ISO 1600, +4 exposure
    Here it is for my 5D Mark II; same procedure:
    [​IMG]
    Canon 5D Mark II Fixed Pattern Noise, ISO 1600, +4 exposure
    The fixed pattern noise results above seem to indicate lower amounts for my 5D Mark II. Concordantly, in my experience with these two cameras in my hands, my 5D Mark II is more forgiving of Fill Light adjustments, resulting in rare appearances of banding, if ever. With shots from the 5D, I have seen it from time to time but, again, rarely. Both have well controlled noise floors.

    Furthermore, one might consider that the earlier production bodies had more problems, or whatever software was used for the conversions may have played a role.

    But the fact remains that I really only see this as a real-world issue when I try to push the shadows of huge dynamic range scenes shot with my 5D... and of course I remedy this issue by using grad ND filters, HDR, and/or my 5D Mark II, which really rarely ever shows this problem (consistent, again, with the above fixed pattern noise results, in my opinion).

    I'd imagine that your mileage may vary. My results here with my own cameras seem to be better than the posted shots at the beginning of this thread, even with heavy push processing.
    -Rishi
     

Share This Page