4
$\begingroup$

enter image description here

Here is the node tree, I was following default cube, but his cheeses don't have these grids

zzzzzzzzzzzzzzzzzzz

My blend file is here:

Edit mode:

enter image description here

$\endgroup$
19
  • 7
    $\begingroup$ If you have already shaded smooth, this can sometimes be caused by Normal Map or Bump strength being too high. $\endgroup$ May 4 at 15:01
  • $\begingroup$ please provide blend file $\endgroup$
    – Chris
    May 4 at 15:24
  • $\begingroup$ @AllenSimpson how do I adjust the "normal map" thingy? $\endgroup$
    – Agil GA
    May 4 at 16:31
  • 2
    $\begingroup$ It's only the 'Distance' setting in your Bump map. Either set that lower, (,25?) or, if you want a more exaggerated wobble, you can set the 'Distance' higher, but also give the cheese a level or two of Catmuul-Clark Subdivision. Will answer, later, if no one else does. $\endgroup$ May 8 at 12:12
  • 1
    $\begingroup$ Have you tried sellecting all your faces in edit mode, and doing "Normals - Recalculate outside" ? $\endgroup$
    – arepawpaw
    May 8 at 14:24
9
+25
$\begingroup$

I think the underlying triangulation of the mesh being made visible is the result of an extreme 'Distance' value in the Bump node.

The Bump node has to make an approximation of slopes from heights, by, in some way, sampling heights of adjacent points in the map, and using the differences. It seems that if the differences are too great, and the slopes too steep, the approximation is not convincing, and exposes the underlying triangle normals, even though they are already interpolated by smooth-shading.

enter image description here

The illustration shows 'Distance' settings of: 1 (your setting), 0.5, 0.25, and 0.1. The last chunk shows your setting of 1, but with 1 level of Catmull-Clark Subdivision.

(The code, (at least the OSL version) points to this method of perturbing normals, given heights. What would make a really proper explanation of your artefact, rather than just a fix, would be a more serious mathematician coming along, and describing exactly why the approximation fails. If one comes along, please edit this answer!)

$\endgroup$
6
  • 1
    $\begingroup$ For rasterizers, I'd suspect derivative discontinuities at edges, because derivatives are always calculated within a single face (within a single "fragment"), so bump can be discontinuous at edges. For Cycles, I'm not sure, but this is a case where you could easily have terminator artifacts from self-shadowing, and I'd wonder if terminator artifact fixes were made on the expectation of reasonable bump values (here, less than 0.01 distance.) $\endgroup$
    – Nathan
    May 9 at 16:36
  • $\begingroup$ Hi @Nathan Yes. I wish I / someone could chase down the dX() and dY() calls to their implementation.. The cross and dot-products are executed on N, the interpolated normal, but you could be right. the derivatives may not cross triangle edges. But that's dealt with in the cited paper, AFAIK.. so.. I dunno... $\endgroup$ May 9 at 16:50
  • $\begingroup$ so, is there a solution or something to this? $\endgroup$
    – Agil GA
    May 15 at 21:47
  • $\begingroup$ I can't think of one, other than reducing 'Distance' in Bump, or upping subdiv, as shown. $\endgroup$ May 15 at 21:51
  • $\begingroup$ @RobinBetts I have added Subdiv tho $\endgroup$
    – Agil GA
    Jun 17 at 23:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.