Personal View site logo
How to provoke banding?
  • karlkarl

    This may sound like a crazy question, but this evening I tried to compare low light results of GOLGOP3 and M6 Sports settings, and one aspect I wanted to test was the "banding issue" (after having read some postings on this phenomenon and watching a recent EOSHD video showing the "banding" problem).

    Now the strange thing is that I failed to reproduce the issue, even when voluntarily shooting very low light scenes at ISO 12800. There was, of course, lots of noise in the clips, but no clearly recognizeable banding.

    So I wonder: What did I do "wrong" that I could not reproduce the banding issue? What settings can you "recommend" to reproduce the banding issue?

  • 6 Replies sorted by
  • Mark_the_HarpMark_the_Harp

    @karl Shoot something whitish, smooth and cylindrical, like a candle. Light it so you get a gradient across the object, and get the ISO/noise as low as you can. Noise will act as dither to reduce banding so in this case you want to avoid it. Also a variable light strength (such as a flickering candle) will show up the banding. You'll see it moving across the surface of the cylinder.

  • Mark_the_HarpMark_the_Harp

    PS Noise can be your friend - because it introduces dither before the signal is quantized, which is probably why your high-ISO shot didn't produce banding. Dither is basically random signal (noise) and in theory, one bit's worth of dither before sampling / quantizing will completely hide the quantizing steps (the quantizing steps we see as banding, are effectively spread between adjacent quantizing levels and are thus traded for noise). I'm basing this on my knowledge of digital audio processing so someone correct me if the same doesn't apply to video.

    If (as normally) therefore you want less banding, you would do better to get a noisy shot and reduce the noise in post because the quantizng steps are traded for noise, which you the reduce afterwards, and the result of all this is that you get less banding than if you'd got a 'clean' shot in the first place.

    So in your case where you are trying to provoke banding, you need the cleanest, least noisy shot possible to start with so the system isn't introducing dither.

  • atticusdatticusd

    @Mark_the_Harp I think @karl is not talking about posterization but about that other nasty band that shows in underexposed high ISO footage. See here: personal-view.com/talks/discussion/693/gh2-showing-fixed-darker-bandstripe-20-pixels-high-across-footage/

  • karlkarl

    @Mark: Thanks for your hints. So now I just shot a cylindrical ornamental batten, made from white painted wood, using 160 to 500 ISO, in medium to very low light. But still, when I look at the recording I cannot find any signs of banding. What I haven't tried yet is the flickering light source, maybe that is an important factor, so I'll try that later.

  • karlkarl

    @atticusd: Yes, that "horizontal band" is what I have been reading about and saw in the EODHD "low light shootout" on vimeo. That is what I tried to reproduce.

    @Mark: If what you meant are "bands" of colors visibly separable instead of smooth gradients - that is not what I meant. They are certainly artifacts of lossy compression, and can best be avoided by reasonably high bit rates and reasonable encoder settings, so I don't consider them a real problem.

  • Mark_the_HarpMark_the_Harp

    @karl I misunderstood what you were referring to.

Add a Comment