Nikon D600 horizontal banding at high ISOs?

[kevin_brown7]

<p>Sigh. Remember how I said you could push the shadows all day long on this camera and get artifact-free results?</p>

<p>I was wrong. I just pushed a test ISO 100 image by 4 stops and I get horizontal banding when I do it, at about 2 stops under neutral (after pushing). The banding becomes more visible if I push by 5 stops (thus bringing the area in question to 1 stop under neutral).</p>

<p>I thought the Exmor sensors weren't supposed to behave like this?</p>

<p> </p>

[kevin_brown7]

<p>I just did a full 13 stop series of shots at ISO 100, working my way from (relative to neutral) +3 to -10, for 13 stops worth of dynamic range coverage. The sensor is supposed to be capable of recording 14 EV of dynamic range, so this is well within the envelope.</p>

<p>Banding is visible in the bottom 3 stops of that range. It's barely visible at -8, and easily visible at -10 and -11.</p>

<p>I expect this is <em>not</em> normal for an Exmor sensor. My understanding is that the shadows at those levels are supposed to be clean, i.e. artifact-free.</p>

 

[kevin_brown7]

<p>Ooops. Make that -9 and -10 for the easily visible banding shots.</p>

 

[ShunCheung]

<p>Kevin, could you post one image sample with banding that other people can see? Without that, I am afraid there is not much we can help.</p>

<p>I am attaching a larger version of my ISO 12800 sample. I see no banding there. Does anybody notice otherwise?</p><div></div>

[kevin_brown7]

<p>Maybe you'll be able to see it in the attached (assuming I'm able to get attachments working). Look in the area of the hair. See how it looks like someone has wiped something from left to right in that area?</p>

<div></div>

[kevin_brown7]

<p>The banding in the image I attached isn't too bad. I presume you're looking for a "real world" image, and not just a sample? I haven't had the camera long, so I don't have much in the way of "real world" images yet, and it's in my best interests to keep the number of shots I take to a minimum now that there's a real possibility that I'll be exchanging the camera (while I've found some banding in high ISO raw files available on the net, I've not found any ISO 100 shots that show banding when pushed by 5 stops like mine does).</p>

 

[kevin_brown7]

<p>For comparison, I'm attaching a version of the same previously-attached image that has been through the debanding program I use (Nik DFine). Compare and contrast that with the previously uploaded image, and the banding in the original may become more obvious to you.</p>

<div></div>

[kevin_brown7]

<p>JPEG compression seems to reduce the banding's visibility. It's more obvious in Lightroom.</p>

<p> </p>

[ShunCheung]

<p>Kevin, I am afraid that you are describing a problem that nobody else can see.</p>

[kevin_brown7]

<p>Hmm....you see no difference at all in the hair between the two images I posted, when viewing those images at 100% (they've already been downsized from the original to 1/4 the resolution)?</p>

 

[ShunCheung]

Kevin, we have given you plenty of opportunities to post some images with the problems you described. If others also

observe those problems, I am sure they will follow up.

[luke_kaven]

<p>Kevin, I can see some behavior that some might consider marginal. f you have a no-questions-asked return policy on the camera, you might give a go at exchanging. Otherwise, you will have a difficult time persuading Nikon to work on it. But I am sympathetic. My D800 does not produce banding at these settings that is intrusive. But I ask you to take an additional factor into account when shooting the D600/800 at ISO6400 and above. </p>

<p>Do you see how the blacks are lifted in your available light portrait profile shot above? The reason for this is background thermal noise from the sensor. At combinations as low as ISO6400 and 1/80th, there is an overall cast of background noise. If I ever shoot at these settings, I need to do a black frame subtraction, not just to fix background levels but also the localized blueish/magenta cast that one gets. </p>

<p>In other words, your shadows are being "lifted" by background noise, and not by signal. After you do a black frame subtraction, you will know better what your true shadow levels and exposure are. I know that this seems arcane, but it is in the nature of the Exmor beast. </p>

<p>Lastly, recall that the camera uses digital gain only at ISO1600 and up. I would not set the camera beyond ISO6400, and even then I'd do it only for preview purposes. You can always throw bits away later rather than let the camera do it for you. You never know when a highlight is going to come in handy.</p>

[kevin_brown7]

<p>I suppose I just see banding more easily than most. I agree that, generally, the banding isn't bad, but it <em>is</em> there.</p>

<p>What surprises me the most is the presence of it at ISO 100 at 9 and 10 stops under neutral. Is that normal for Exmor sensors?</p>

 

[luke_kaven]

<p>If you drop me a note with your email address, I can send you a couple of D800 samples shot at ISO6400+1 of a cityscape. The level of pattern noise there is almost undetectable. So I do have a feeling that yours is right on the margin. </p>

<p>Is there a chance you could compare yours with another D600 on the exact same scene, same time, same settings? The reason I ask is that you would need very good comparative samples that demonstrate the problem in order to get Nikon to consider a replacement. </p>

<p>Remember also that your shadows are being slightly lifted by thermal noise at these settings. </p>

[kevin_brown7]

<p>I'm going to try an exchange through the vendor first, since that is still an option for me. Since I know exactly what to look for and how to find it, it should be easy to determine whether the replacement has the same issue. I'm going to attempt to get the vendor to ensure that the replacement unit has a different enough serial number that it likely came from a different manufacturing batch.</p>

<p>If the replacement exhibits the same behavior at roughly the same levels, then I'll just conclude that it's normal for the line and that'll be that. I just hope the replacement doesn't have some <em>other</em> problem, because this one has been flawless in other regards (autofocus in particular).</p>

 

[kevin_brown7]

<p>I don't suppose it's possible to do a dark frame subtraction within Lightroom itself?</p>

<p>Also, the camera has a "long exposure noise reduction" option that appears to do dark frame subtraction (since it appears to take twice as long to generate the image than the shutter speed is set to, and the latter half of the time is spent with the mirror down). I don't suppose it's possible to get the camera to always do a dark frame subtraction for you, is it? I certainly haven't seen that option.</p>

 

[andylynn]

<p>Kevin, I can see it too. It's very subtle and it's not color banding or the sort of banding a miscalibrated film scanner gives, just a sort of pattern to the noise that resolves as very subtle horizontal lines.</p>

<p>I don't think there's anything wrong with your camera. I think this is an artifact generated in software resulting from raw processing. You're working in very low artificial light here and I think what's happening is the color cast of the light, in combination with the reduced dynamic range of the sensor in very high ISO mode and the presence of Bayer grid, is causing an effect I don't really have a word for so I'm going to call a reverse-photometamerism-ish issue. The raw processor sees the dark areas and is grasping for any light at all in there, and since the Bayer grid's lines have alternating color biases the lines are coming in with alternating slight level differences and spazzing the software's Bayer integration. Those rows are so close together it's giving you a sort of meta-moire. Sorry about half the words in this paragraph being made up.</p>

<p>It's such a subtle effect, I wonder whether it's really going to cause problems. If your monitor isn't calibrated it might be overbright in a way that's exaggerating the effect.</p>

[kevin_brown7]

<p>I can tell you it's not likely to be an error in the demosaicing process. The only way it could be is if the row values stored in the dark region at the periphery of the sensor are somehow being misinterpreted. However, it's notable that even the D800 apparently exhibits <em>some</em> banding at these ISOs at roughly the same tonal range, just not as much. So either the raw converter is doing it wrong for all these cameras, or the sensor or electronics are generating noise during the horizontal readout that is affecting the results. The latter is much more likely. As I said earlier, it's really not until relatively recently, with the advent of the Canon 7D, that high ISO banding disappeared entirely.</p>

<p>Another piece of evidence that the raw converter isn't doing it wrong: the in-camera JPEGs show the same banding. So either Nikon can't get the raw conversion right in-camera (not likely), or the banding is something that can't be directly adjusted for by the raw converter.</p>

<p>It's possible, however, that dark frame subtraction would eliminate the banding. I suspect not (it would depend on the dark frame having the exact same banding pattern as the shot from which it is to be subtracted), but it's possible. I just don't have the right postprocessing software to do it, so I can't test it at the moment.</p>

<p> </p>

[andylynn]

<p>This is a noise processing thing. Don't worry about it - it's only showing up in the high ISOs where you don't expect this good an image anyway, and it's too subtle for even most people on this site, being told it's there and staring at 100% crops, to see it.</p>

[kevin_brown7]

<p>Yeah, I agree that generally, the level of banding at higher ISOs isn't all that high. My concern is primarily the fact that banding is visible in the -9 and -10 stop range at ISO 100, which means it'll be visible in heavily pushed shadows at base ISO. From what I've seen, that is unusual for these sensors. There aren't that many D600 ISO 100 raw shots to play with that have data at those exposure levels, so I haven't been able to determine conclusively whether or not that's to be expected.</p>

<p>Since I'm going through an exchange process, I'll be able to report whether the replacement unit is any better in this regard. If it's not, then I'll conclude that what I'm seeing is just characteristic of the model and move on. It's a Christmas present, after all, so it's not like I have anything to complain about!</p>

<p>Once again, I have to state for the record that this thread is not to complain about the presence of banding, only to ascertain whether it's <em>normal</em> for this camera model.</p>

 

[francisco_salaquanda]

<p>OMG,</p>

<p>Black frame subtraction reversing the photometamerism by spazzing the Bayer integration....</p>

<p>Got it....Thanks...I knew that!</p>

<p>Or...speckles of noise in the hair. </p>

[luke_kaven]

<p>I don't know of a way to do black frame subtraction in Lightroom, if there is one. I do wish Nikon would enable LENR for use at high gain settings, as this would be the best option.</p>

<p>To do it manually -</p>

<p>1) Shoot black frame with same shutter speed and ISO as image. You generally only need one of these. Although the thermal noise is much worse after using live view than it is otherwise.<br>

2) Do capture on image and black frame to 16-bit TIF using the same white balance and gain settings. Do no noise reduction, sharpening, etc on the black frame.<br>

3) Open both in photoshop.<br>

4) While looking at your image, go to Edit->"Apply Image...". Choose "Subtract" from the drop-down box for function. Choose the black frame from the image file drop-down box. <br>

5) You can experiment with changing the percentage to match the effect.<br>

6) Watch the color cast go away and the blacks drop to black levels. </p>

[kevin_brown7]

<p>Well, as it happens, the vendor (B&H) won't ship a replacement until they receive the original. That's understandable, but it means that I can't do controlled side by side testing of the two units. But more importantly, they can't do anything to ensure that the serial number of the replacement is sufficiently different from the unit I have to maximize the chance of them being from different manufacturing batches.</p>

<p>As a result, I'm not going to do an exchange. Instead, I'm going to rent another D600, and possibly take additional shots with other D600s at camera stores around here, and use that to determine if what I'm seeing is normal for the model. If it is, then so be it. If it's not, then I'll have evidence that I can present to Nikon's service center that shows my camera to be faulty. This should prove interesting.</p>

<p>The camera's been flawless otherwise.</p>

 

[ShunCheung]

<p>Kevin, as far as I can tell, not even one person (other than yourself) on this thread is convinced that there is anything wrong with your D600. Therefore, I am afraid that you will have an equally difficult time convincing B&H and/or Nikon.</p>

<p>Don't you have a friend who also has a D600? If you have access to another D600, I would do a quick test. As you know, I have captured a few images with the D600 at ISO 6400, 12800 and 24600 as part of my test routine. I never noticed any banding problem, as shown by the sample I posted. It is probably not worthwhile to invest too much further effort on this issue.</p>

[kevin_brown7]

<p>The difference is that I can't see banding in your sample, but I can see it in mine. That alone is evidence to me that mine may have a problem while yours does not.</p>

<p>However, the banding isn't bad by any stretch. It's just that, coming from a Canon 7D which exhibits no banding at high ISOs whatsoever, it's a bit of a surprise considering the stellar reputation of the Sony Exmor sensors.</p>

<p>I don't intend to spend a big pile of money on the problem or anything. Images from imaging-resource.com do show banding as well, so there is some evidence that what I'm seeing may, indeed, be normal.</p>

<p>Time will tell. I'll certainly report the results back here.</p>