PDA

View Full Version : Roughness bug?



peebeearr
01-21-2019, 12:41 AM
Hi All,

I just noticed some weirdness in the refractive buffer when using roughness on a transparent panel. As the roughness increases you can clearly see a very sharp edge moving over the object. See attached sample scene, preview animation and pictures.
Am I doing something wrong?

143907
143908

peebeearr
01-24-2019, 11:30 PM
Ok it just got even weirder!!

Try to load the attached scene and do an F10 render. It will render 40 frames of absolutely no animation (they are <1s renders). Yet, the refraction goes absolutely berserk on the box in the front.

This box consists of flipped and merged 'air' polys as I cant otherwise create convincing acrylic material. The box in the back is the same box without air polys. That box seems to be of a solid material while it simply consists of individual flat panels...

Also in VPR preview I get lots of weirdness that are hard to reproduce as it seems to happen randomly... but when it happens it goes pretty nuts.
Attached some weirdness that I encountered trying to create a scene where I could reproduce the error (attached although the error happens randomly over the coarse of 40 frames).

There are 2 vases in the scene of which 1 has the same setup with the air polys as the box. The issue does not happen there!?

Can someone help me how too create an acrylic surface that actually works? I thought with the later LW's you wouldn't need air polys anymore?? Does this happen in your LW setup as well?

All done in LW2018.0.7 on PC.

143949
143950
143951
143952

143953

[Edit] Oh, I got attached VPR previews by simply turning on and off the visibility of the object, nothing else changed!

RPSchmidt
01-25-2019, 07:33 AM
I'm not sure what an "air" poly is... but I think it would definitely help if you showed your surface setup.

Tobian
01-25-2019, 07:53 AM
I'm gonna have a look at this, but you absolutely do NOT need an air poly and have not since version 9.5 or there abouts. If anything they will cause more issues than they will help with this version. I am suspecting some broken polygons and incorrect surfacing.

RPSchmidt
01-25-2019, 08:19 AM
I concur with Tobian.

I created a similar setup to yours, just using standard polys.

I created a PBSDF material using white color (255,255,255), 10% roughness, 0% specular, 25% translucency, 35% flatness, and 100% transparency, with Glossy Reflections turned off in the Shading Model tab.

I also set the transmittance color to white.

It's not perfect, but I think with a little more tweaking, it would be usable.

peebeearr
01-27-2019, 04:45 AM
Well, as you can see with the attached scene and object files, things could not be simpler... But still it doesn't give a proper refraction you would expect from a simple object like that.

Surfacing is as simple as can be similar to what Schmidt has done.

I know air polys are a thing of the past but I used the air poly trick just out of curiosity but it actually gives a much more accurate result than without and that made me wonder.

If anyone can create a proper look of an acrylic box, please post it up.

This is the look I'm after:
https://www.themakeupboxshop.com/wp-content/uploads/2017/11/Cosmetic-Cube.png

RPSchmidt
01-28-2019, 09:14 AM
I wish I could provide more suggestions, but in my experimental render, my box looks almost exactly like that one, as far as edges, refraction, level of transparency.

I wish I could post up my experiment here, but unfortunately I can't.

It's funny because in my experiment, I thought my edges shouldn't have been refracting the way that they were; but looking at your example image here, I feel a lot better about my settings.

peebeearr
01-28-2019, 07:43 PM
I wish I could provide more suggestions, but in my experimental render, my box looks almost exactly like that one, as far as edges, refraction, level of transparency.

I wish I could post up my experiment here, but unfortunately I can't.

It's funny because in my experiment, I thought my edges shouldn't have been refracting the way that they were; but looking at your example image here, I feel a lot better about my settings.

Hi RPSchmidt,

If you can't post the experiment could you describe exactly how you got the look of the example box above?

peebeearr
01-28-2019, 08:55 PM
Hi all,

Never mind.............. It must have been the heat. It all seems to work as expected now.
During my vigorous testing I had duplicated a lot of polygons that I thought did not do. That made the whole thing go nuts. As always, assumption (in this case thinking the box was modelled correctly) is the mother of all screw ups.

Next time I'll do more checking before wasting your time.

143985

Tobian
01-29-2019, 04:03 AM
Yeah i had a look last night, it was mostly duplicate coplanar polygons. To get that transmissive glow, you might want to look at sigma2, which is sort of like dielectric, which also has some scattering, i got carried away last night, and got a nice extra slow render (too many recursions, while testing) :D Your lighting was also quite dull, so more area lights or a professional lighting studio HDR environment, with more fill panels in the scene, to make the dielectric glow more.

RPSchmidt
01-29-2019, 05:11 AM
Hi all,

Never mind.............. It must have been the heat. It all seems to work as expected now.
During my vigorous testing I had duplicated a lot of polygons that I thought did not do. That made the whole thing go nuts. As always, assumption (in this case thinking the box was modelled correctly) is the mother of all screw ups.

Next time I'll do more checking before wasting your time.

143985

That looks much closer to what I got. I think the main differences are due to the render settings, but otherwise, it's very similar.

Yup, spare polys can always mess up a perfectly good render, especially when you are working with transparency, reflection, and refraction.

In my render, I used an HDRI and a single environment light with 24 samples.

I also set my refraction and reflection samples to 10 (although a lower number probably would have worked just as well and provided a faster render).

Glad you got a result you could work with!

kolby
01-29-2019, 08:51 AM
I have a noticeboard above my monitors and it says, among other things: Once you finish your model, perform merge points, unify polygons, check statistics panel and remove all unwanted points and polys. :)

peebeearr
01-30-2019, 09:34 PM
... Your lighting was also quite dull, so more area lights or a professional lighting studio HDR environment, with more fill panels in the scene, to make the dielectric glow more.

Hahaha, don't worry, this whole setup was specifically made to prove a point, which you guys pointed me that wasn't the point at all in the end..... I should have merged polys and points. Anyways, I just wanted to get my point across.
At least I did't use a point light but point taken;)