4

Not sure how to explain it, but check out this picture:

enter image description here

So if you look closely on the left strip of the image, the lighting is off (the strip is somewhat brighter than the rest of the image). This doesn't happen in every picture and I can't find more examples of it that quickly, but I can imagine this isn't the only one.

What can I do to find out whether it's something in my body, or something in my software? The lighting changed negligibly during the action of taking the picture, so I doubt it's a rolling shutter effect.

I'm using a Nikon D810 and Capture One Pro 10. I tagged the question very generally because "error" doesn't cover it, and "bug" doesn't seem to exist.

user55789
  • 205
  • 2
  • 6
  • 2
    Does the unprocessed image have that strip on the edge? Are you shooting RAW, or processing JPEG (I assume raw, if you're using Capture One Pro)? Can you export a JPEG from Capture One without any processing, other than creating the export file? If so, how does that compare (does it have the strip?) – scottbb May 09 '17 at 22:44
  • 4
    Try shooting in JPG+RAW and see whether issue appears in both or only one type of files – aaaaa says reinstate Monica May 10 '17 at 02:27
  • Possibly this is the first column read off the chip, and under some circumstances (camera not warmed up, or left on for a long time w/o a readout,e.g.), there's some offset in the readout FETs. – Carl Witthoft May 10 '17 at 11:33
  • 1
    In addition to the above suggestions, take a look at the review screen in your camera to see if the strip occurs in-camera. Zoom in and scroll over to get a close-up view. – BobT May 12 '17 at 18:25
  • You could try shooting pictures of grey cards to see if this is general or a one-off. This pic is low light. It might be you have a sensor that 'leaks' so that an image that takes appreciable time the charge has time to leak off while accumulating. If this is the case, you won't find the problem on exposures faster than some number.

    It can also be a bias voltage that's off. In that case I would expect it to happen on all images.

     – Sherwood Botsford Mar 22 '18 at 20:06
  • I wonder if it's present in all images, but only really noticeable when the left side is dark, as in this scene. – Lowell Montgomery Apr 07 '18 at 06:12
  • You say "This doesn't happen in every picture and I can't find more examples of it that quickly, but I can imagine this isn't the only one. Can you confirm if in places where it does occur, it's exactly the same width? – mattdm Jun 06 '18 at 14:21

2 Answers2

2

This looks like a software or hardware malfunction in the JPEG encoder. Exactly one set of 8x8 superblocks is affected, and in each of the affected superblocks, the top left value (the average) appears to be higher than it should be, but everything else is plausible. This likely indicates either a software bug or a single-bit error in a hardware CODEC chip.

If these came out of the camera this way, I would suggest switching to RAW. The problem will probably go away. Then you can decide whether it is worth getting the camera repaired just for JPEG output or not. (Of course, if it is a hardware fault, it might cause other issues in video shooting mode, so try that out, too.)

If these were shot in RAW mode, then this is probably a bug in Capture One's JPEG encoder.

Either way, the fact that the problem is an 8-pixel-wide strip screams JPEG encoding bug rather than any sort of sensor issue.

dgatwood
  • 2,235
  • 11
  • 15
  • Or possibly a bit flip somewhere, as a now-deleted other answer posits. – mattdm Jun 05 '18 at 19:03
  • Hence the "or a single-bit error" comment in the first paragraph. I would guess that a bit is stuck permanently in a 1 state in some high-speed RAM in the JPEG encoder hardware. The fact that it occurs in the same place every time suggests that the hardware encodes an entire row of superblocks at once, and that the first block of storage is bad. If it were in general-use RAM, I would expect other things to fail randomly, and I wouldn't expect to see a consistent pattern. But that's just a gut feeling; I have no idea what the SW/HW architecture of these things looks like under the hood. – dgatwood Jun 05 '18 at 23:18
  • But not necessarily in the codec hardware — could be the storage or transfer. – mattdm Jun 05 '18 at 23:51
  • It can't be in the mass storage, because that wouldn't affect the first superblock in every row of an image; it would cause either a single bright square or one per logical block (and you'd never get exactly one row per logical block). It has to be in either the codec hardware itself or in whatever dedicated, row-sized buffer that the codec hardware writes the transform output into before it gets run-length encoded. Either way, there's a pretty narrow window of time/hardware in which that corruption could occur without corrupting the entire rest of the file from that point onwards. – dgatwood Jun 06 '18 at 01:08
  • It's unclear to me from the question that this is actually the same in every file. – mattdm Jun 06 '18 at 14:20
  • Fair enough, but it wouldn't be that way in even a single file if the problem were somewhere else. :-) – dgatwood Jun 06 '18 at 16:25
0

That is certainly an odd issue. It almost looks like a sticky shutter issue, but the D810 has a vertical-travel shutter (from top-to-bottom), like most modern digital cameras, so it obviously can't be that. Clearly the issue is too inconsistent to be caused by the lens, and also too uniformly isolated to that band on the left, so I assume that it's a fault in the sensor, as you suspect. I wonder if it appears more often when in certain (cooler / moister) weather conditions? Assuming your camera is still in warranty, I'd suggest getting it serviced. Otherwise, it's probably a pricey fix.

Lowell Montgomery
  • 406
  • 2
  • 11