Jump to content

Increasing samples for mesh lights in PBR?


Recommended Posts

Hi,

 

I'm new to Mantra. I've used Blender Cycles, LuxRender and Indigo Renderer before.

In order to learn the new engine well (instead of skipping problems), I lit a free scene from 3Drender.com in Cycles and now I'm recreating it in Mantra.

 

So far, I came across two problems I can't easily solve.

 

- There's a lot of noise from mesh lights, while the spot lights start to look good quite early.

Is there a way to fight the mesh light noise without increasing overall pixel samples?

I thought that Diffuse Quality would help me with that but it doesn't seem to change much.

 

- Glass (mantra surface) seems to treat 1mm thin glass objects like they were solid, not just double-sided shells.

What can cause the problem? Maybe normals get inverted automatically?

 

the scene:

LightingChallenges_Science02.jpg

 

cycles vs mantra:

http://oskarswierad.com/public/3d/LightingChallenges_Science02_VS.jpg

 

1x1 pixel samples:

http://oskarswierad.com/public/3d/LightingChallenges_Science02_1sample2.jpg

 

render settings:

http://oskarswierad.com/public/3d/LightingChallenges_Science02_settings.jpg

Edited by OskarSwierad
Link to comment
Share on other sites

are you using Geometry lights for them or just emissive material?

if you are using Geometry Lights (which you should) then it already by default should produce less noise than just emissive materials

on top of that there is Sampling Quality setting on the light itself

and you can as well play with pointcloud settings on geometry lights to get hopefully even smoother results

Edited by anim
Link to comment
Share on other sites

lights: there is a Sampling quality parameter on light object that can reduce noise in light's shadows. for reducing specular or diffuse noise you have to crank up either ray samples or pixel samples. not sure about your quality parameters on mantra node. I assume that it is something you added in as it is not included in default setup and I think these are supposed to drive local shading for materials so it might not be doing what you expect it to do. I recommend you working with default mantra node as it comes with houdini unless you are really familiar with how it works. houdini often allows you to make crazy settings that doesn't make any sense so don't suppose something is working just because you didn't get a warning message :)  

 

glass: hard to tell, normals might possibly be the problem.

Link to comment
Share on other sites

Thanks for your replies.

(here's the scene: http://oskarswierad.com/public/3d/LightingChallenges_Science02.zip- HDR may be missing, but the rest should be fine.)

 

@anim: I used just emissive materials... So that's the first thing to try, thanks!

 

@davpe: this Diffuse Quality may be something that was added in H15. I can't see it in docs for H14. I'll keep in mind your advise though.

Edited by OskarSwierad
Link to comment
Share on other sites

Geometry lights work perfectly (instead of emission). Great thing.

It seems that raytracing bias actually helped with glass, plus I had an internal shell missing for 1 object.

Now I have indirect light 'lamp'. I forgot about it the last time.

 

Now it takes 2 hours to render. Noise level 0.01, 9x9 pixel samples. (The glass ball has problems but it's my fault :D)

ZxzhrhE.png

Without refractive materials and with slightly lower quality (noise lvl 0.05, 8x8 pixel samples) it takes 8 minutes. Glass makes a massive difference :/

qmi3aPz.png

I also had to turn off HDR image in the environment light. The light (with portals) generated a ton of noise when it used HDR map. Flat color is super smooth. What can I do about it?

I thought about blurring it while keeping the sharp version for reflections but maybe there are simpler things to try?

Edited by OskarSwierad
Link to comment
Share on other sites

not sure how you build your scene but with 2 hours for that image you have something wrong there or you are totally oversampling something. i suspect that changing the raytracing bias might be the thing. you might want to change it if you are rendering some extreme scale scenes (either too small or too large) but personaly I have never had to change it. it may have a massive impact on raytracing speed sometimes and it can introduce/reduce some scale related artifacts. that might explain why your render is so much faster when you turn off refractions. it might be the same problem with your hdr light too. environment hdr light is generally very cheap to render. blurring the image for diffuse rays helps you get a smoother result for sure but it is not making any significant speed difference in most cases. this is just my guess here but my advice is to put the raytracing bias back where it was originally and work from there. if changing it helped you with refractions then check your scene scale if its not something totally off. or maybe try to model simple sphere in houdini and render it with refractions. maybe there is something wrong with your geometry. my first guess would be that one side of your glass has inverted normals (or it has no thickness at all) bcs it would look exactly like that. or maybe something in the shader... hard to tell, refractions work just fine for me. no catch there... 

Link to comment
Share on other sites

edit: out of curiosity I opened the scene you posted before and the problem with refractions is simply that the glass does not have any wall thickness. so the glass sphere on the left hand side is rendered as a solid glass object of course and refraction you are getting is technically correct. solution is either to properly model the object to have actual thickness as in real world or you can check "Thin film refraction" option in mantra surface material. That makes treat any model as hollow even if you have no thickness (like a soap bubble). that will give you more of what you are looking for but of course it will look a bit odd on something that actually should have some visible thickness.

Link to comment
Share on other sites

I did a lot of test thanks to your posts.

Raytracing bias really affected the speed by as much as 25% - and it wasn't necessary. I remodelled the glass object's back faces (to make sure normals are ok). I given them more thickness and added some space between the glass and the fluid - to avoid the need for raytracing bias.

I think the problem was with a lot of reflections and refractions stacking up. I ended with "glass" made of reflections + opacity with falloff, no refractions. Plus the fluid with refraction but no reflections. It allowed me to lower the refractions limit to 3 (from 6).

Here's the empty scene with HDR and new materials: http://oskarswierad.com/public/3d/LightingChallenges_Houdini_ScienceGlassOnly_HDR.zip

Still quite long - 30 seconds on an empty scene?

Thank you davpe for your help so far!

DDibx2W.png

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...