2
$\begingroup$

I'm getting an annoying wavy effect on renders in a low poly cycles file. Whats frustrating is that I've rendered several files now with all the same settings without this artifact. The strange thing is that the artifacts seem to appear only in the first room, not so much the second. The differences between whats effected and not might be these things

  • the first room is lit by an area light that the second is not lit by
  • the file is a copied version of another .blend that was then altered and it seems like the objects that existed previously are effected and the objects added after the copy are not i.e. the second room. this is a low poly project and I don't want anti aliasing so blackman harris is set to 0.01 but this went fine with the other files.enter image description here

here you can see the strange thumbprint artifacts

and here is later in the render, where this problem is absent

enter image description here

Let me know any settings that would be relevant. I'm using default denoising and have checked for differences in the affected objects with the unaffected ones in their properties to no avail but I'm not insanely fluent with blender.

enter image description here

Heres another picture clearly showing the problem

$\endgroup$
2
  • 1
    $\begingroup$ To me this simply looks like Moiré pattern. But since you do not provide any information on the material setup it's hard to tell what is causing this. $\endgroup$ Jun 8, 2023 at 5:58
  • $\begingroup$ Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking. $\endgroup$
    – Community Bot
    Jun 8, 2023 at 11:13

1 Answer 1

2
$\begingroup$

Those artifacts are called aliasing - a very wide term, with e.g. the Wagon-wheel effect as a subset of temporal aliasing. Another subset, as Gordon Brinkmann pointed out, is Moiré pattern. The logic is similar to Z-fighting (and perhaps it also qualifies as aliasing, though I didn't find a source claiming so) - two pixel neighbors bifurcate to two significantly different values based on some minimal (unexpectedly significant) input change.

One example of such bifurcation is rounding: you have two consecutive values next to each-other, but one rounds down, and the other rounds up:

>>> round(0.5)
0

>>> round(0.5000000000000001)
1

In case of z-fighting, which might be happening in your case, either of multiple overlapping planes is considered closer to the camera, based on calculation inaccuracies (which depending on the numbers crunched, sometimes will round one plane favorably, and sometimes another). To fix this, ensure no planes are exactly overlapping, and if some are almost overlapping, try decreasing the range of camera clipping (this range is divided into a constant number of subranges, so as you decrease the range, the subranges decrease and so the risk of two planes ending in the same subrange also decreases)

Another possibility is that your shader generates a pattern with cells comparable with, smaller or much smaller with the size of a pixel. While in reality looking from very far on a checkerboard texture results with grey color, when you use a single sample, the Checkerboard texture will only report one of its defined colors, so for a very big scale and 1 sample:

But 100 samples:

Different example:

v=60

v=25

v=10

v=3

v=0.11

v=0.000001

v=0.0000001

There is no general solution to this (other than ANTIaliasing, which you don't want to use), you have to use such scales that boundaries between colors are further apart than 1 pixel.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .