recently I found a strange vray material behavior when using glass shader with imported parts (MOI3D->3ds max).
In some cases, it causes reflection and refraction artifacts. Maybe it cause other artifacts as well, which are not so clearly noticeable.
To reproduce the artifacts:
01. take some imported sheet of something from MOI3D (it comes with explicit normals)
02. select one polygon → inverse → delete. Then copy the polygon in sub-object level and add shell modifier - everything renders fine
03. rotate (not flip) one of the polygon with shell modifier active and you can immediately see the reflection and refraction artifacts. They make bend and tunnel effects. The whole surface also looks more like mirror.
04. if you add the edit normals modifier, then the problem is gone. you can also break the normals. (tried in vray 5, not sure if it works the same with vray next)
UPD to 04 : the problem is still there if you rotate/flip the polygon again or copy the object (tried with vray next). The only solution in vray is to add edit normals modifier → select all normals-> reset (normal vectors should have a blue color)
This problem exists in vray next and latest vray 5 versions.
Glass shader is very simple:
diffuse - black
reflect - white
refract - white
color tint via fog
IOR 1,52
reflect on back side
GGX + Glossy fresnel.
At the same time corona renderer 8 has no problems with handling such geometry at all (tested on legacy material and physical material)
Please look at the attachments and tell me if you need the scenes for investigation.
Yeah that happens a lot with imported geometry from SketchUp and such. Back in the day when I used to model in SU and render in 3dsMax I always had that problem. After a good 2 years I figured I had to reset the normals because they looked green. Figured, green = bad, blue = good.
Once in a while I come across some models which have the same and behave weird as well, resetting most of the times did the trick. No idea what the idea behind it is, didn’t understand the concept of explicit normals and the likes. So if someone cares to explain…
Indeed it is a weird issue. Maybe you can tweak the export settings from MOI3D because I don’t think this is vray problem. What I found is that a smooth modifier also seems to solve the problem.
We use this script for importing geometry from other packages and it’s a real time-saver. 3dsMax plugins for architecture & visualization
Thank you for looking at the scene and for giving the link!
Currently, we are using MOI3D as a bridge between NURBS and POLY worlds. We often get 3d models from manufacturers in original SolidWorks file format, and we are using MOI3D to delete inner parts, unnecessary details etc. It is almost always easy, because we stay at original NURBS level.
At the end, we do NURBS-> POLY export via MOI3d and FBX file format. And we can additionally interpolate different parts with different detail level. The meshes are not ideal (because the conversion algorithm works “island” wise), but they are quads. Such meshes always comes with explicit normals and there is no option to change it.
Of course, for such simple objects like flat glass plane, we do not need any normals or smoothing groups at all. But I find this a nice example. If this is not vray problem, then why corona renderer can render the same parts without any artifacts ?
I also have another example, where I got very strange glass shader behavior in vray. And there are no explicit normals on it. Just simple bended object (spline->extrude->shell ->collapse)
- there are no explicit normals
- no inverted normals
- no coplanar polygons
- no unvelded vers
- no scaling issues
- reset x-form does not help
- deleting/adjusting smoothing groups does not help
corona renderer has also no problem with this part and render it without any artifacts.
Corona renderer supports some vray features as basic vray material and what interesting is, that in corona we can render this part with original vray material without visual artifacts.
Please, look at the attachments.
I have updated the scene in post #4 (Today, 03:18 PM). There is bug 1 and bug 2 inside.
It’s an interesting find though, I never tried to render this faulty geometry with anything else. I always thought it was a model problem not a Vray one. (Maybe it still is, but Corona just ignores it).
It is weird that corona doesn’t have these issues.
The second problem that you mentioned is because of the shader. Please take a look at the attached file.
If you untick Reflect on back side the issue seems to be fixed.
Hi, I always try to compare how the problem behaves in other scenarios and situations - it helps to understand the issues a bit more.
Hi, thanks for looking the scene. Yes, with no back side reflection it renders fine. This can be used as temporary workaround in cost of shading quality, but the overall situation is not normal.
I also have the Issue #3, which I believe has also the same reason somewhere under the hood. It relates to metal shader with Anisotropy. We get same wired lines there. I’ll prepare & post it a bit later
I cannot repro the issue with the most recent beta for v.6.
Would you please check it’s gone on your side too?
p.s.: The shading models between render engines aren’t identical. F.e., in your glass shader you changed the GTR tail falloff, something impossible in Corona.
sorry, but I can not test it with vray 6 beta, because we currently have only vray next and vray 5 installed. (but they are both latest stables)
Is vray 5 support and bug resolving stopped because of vray 6 beta?
I understand that the vray mat in corona is not 100% equal to vray mat in vray - it have tested it with vray mat to show that even this vray mat implementation works fine in corona. (which should be more error-sensitive, I guess).
If we take corona legacy material or corona physical material, then it renders also fine.
Thanks for the information.
So I’m not alone with this problem and it exists for a long time… The issue discussed in your thread is like my first issue here, but my second issue has even no flipped normals and no explicit normals.. (you can find it in the same scene, post 4)
Vlado’s reply was quite clear, though.
As mentioned, the shading models are quite different between applications, and the assumptions they make vary enough that what looks obvious to the eye is not at all so when code is looked at.
V.6 has fixed some old standing issue with refractive shaders, but it took non-trivial amounts of research and important under-the-hood changes.
Besides seeing it’s present in 5.2 and not present in 6.0, we can’t say much more right now.
We’ll investigate the matter, and let you know should there be any news.
EDIT: Here are the results, that particular issue is not fixed yet.
First is regular part, second has elements flipped in Edit Mesh, third has elements flipped in Edit Poly.
