Luxcore renders light through glass very slowly

Use this forum for general user support and related questions.
Forum rules
Please upload a testscene that allows developers to reproduce the problem, and attach some images.
Post Reply
User avatar
FarbigeWelt
Donor
Donor
Posts: 1046
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Luxcore renders light through glass very slowly

Post by FarbigeWelt »

One of my friends made an update of a 3D robot with a green laser for engraving wood, plastics or leather. It works well based on diffrent open source drivers and software. He sent me a picture showing the beam and its reflections through transparent, coated glasses. Give me a try, I said, I can simulate a similar picture with the very great luxcorerender blender addon.

I've tried different approaches but run always into the same issue. Luxcore does render lights through glasses resp. mirrors them very slowly. Because one cannot see a ray in clear air, I've set up the whole scene with a homogen, slightly scattering volume. The scattered beam is good visible with BiDir only.

Why is a scattered beam not visible with the path render? Why does sun light not pass through glass but diffuse HDRI or sky light passes with path render? Do this questions have a connected reason?

Decided to use BiDir for my test setup, I placed a glass cuboid between scattered beam and camera. It took a few hundereds samples to get a visible scattered beam but even 1600 samples haven't been enough to get a straight, unbroken line. Maybe, I thought, homogen volume for the whole scene isn't appropriate.

I've setup the scene attached after lots of different tests. The scene, clear volume, contains a cuboid made from Null with homogen, slightly scattering volume. In the background there is a mirror in place. In the foreground there is a glass covering parts of the scene. And there is also a cuboid made from matte placed behing the glass. All materials but the matte have a faint, different color to make them visible. The color is faint enough to avoid much absorption of the scattered beam. The glass under this observation has a reflection of 0.0.

The attached scene shows within the first hundred samples an object made from matte is rendered fast if seen through an object made from glass or mirrored from an object made from mirror, but the scattered beam takes hundreds of samples.
I've tried sobol, metropolis and random, all leading to similar performance.

Is there any good reason for the slow rendering of scattered beams seen through glass or mirrored?
Renders Glass very slowly.zip
Test Scene, Scattered Beam and Matte in Glass and Mirror.
(90.87 KiB) Downloaded 191 times
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
MacBook Air with M1
Fox
Posts: 437
Joined: Sat Mar 31, 2018 11:17 am

Re: Luxcore renders light through glass very slowly

Post by Fox »

I have perhaps related issue with slow sampling through archi glass and non archi glass.
I set the sun size 10x and sun gain 10x and 60W light bulb inside room is overpowering the sun with linear tonemapper.
Sun comes in from another openings by reflecting from another surfaces, but not through glass.

This light that is directly visible to camera from HDR map, comes through fast.
Even BiDirVMcpu 16000 + MLT 36768 + DLC can not find sun that is not directly visible to camera.
Random path is more effective especially in homogeneous volume, but the samples from 200 poly mesh emitter 60W bulb in room, are overweighting sun samples through archi glass which at some gain levels is recommending clamping values in 100 digits.
User avatar
FarbigeWelt
Donor
Donor
Posts: 1046
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Sobol hides sun light through glass

Post by FarbigeWelt »

With CPU BiDir, Sobol sun light does not pass through glass in some circumstances. To find the source in the scence leading to this effect took a while to be found. With CPU BiDir, Metropolis sun light passes through glass in the same circumstances. Also interesting CPU Path, Metropolis works too, but GPU doesn't.
CPU BiDir Sobol
CPU BiDir Sobol
CPU BiDir Metropolis
CPU BiDir Metropolis
Source of effect and abstracted scene following.
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
MacBook Air with M1
patrickawalz
Supporting Users
Posts: 35
Joined: Tue Dec 05, 2017 1:45 pm

Re: Luxcore renders light through glass very slowly

Post by patrickawalz »

sds paths are notoriously hard to render, and metropolis light transport is a very timed algorithm for this very situation.
User avatar
rickyx
Posts: 26
Joined: Sun Apr 29, 2018 5:23 pm

Re: Luxcore renders light through glass very slowly

Post by rickyx »

Can you post an image of what you want to reach?

I got this:
Laser.png
In interior design often I totally remove the window glasses. Often I don't need reflections or I just mix transparent with glossy using a fresnel.
User avatar
FarbigeWelt
Donor
Donor
Posts: 1046
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Sobol hides sun light through glass

Post by FarbigeWelt »

EDIT: Replaced pictures and scene with colored surrounding, added examples for Path.

In the scene the effect started with a plane surrounding the room outside.
2018-09-22 CPU BiDir Sobol (2)
2018-09-22 CPU BiDir Sobol (2)
2018-09-22 CPU BiDir Metropolis (2)
2018-09-22 CPU BiDir Metropolis (2)
In the abstracted scene there still passes some light through the windows with Sobol but in the wooden room scene not.
Glas Test.blend
Glas Test.blend
(663.8 KiB) Downloaded 203 times
My solution for the moment is to switch windows' glass to architectural. But I rather would like to know if this behaviour of Sobol and Random is usual or even expected. The glass windows have all an offset to show the border of the glass. There is also a different light pass through the thin gap below the glasses, which have a different thickness. Why is there a difference between for the different thickness?
2018-09-22 GPU Path Sobol
2018-09-22 GPU Path Sobol
2018-09-22 GPU Path Metropolis
2018-09-22 GPU Path Metropolis
Again, there is the difference of CPU and GPU Path for Metropolis. Would be nice if GPU und CPU Path behaved the same way.
2018-09-22 CPU Path Metropolis
2018-09-22 CPU Path Metropolis
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
MacBook Air with M1
User avatar
FarbigeWelt
Donor
Donor
Posts: 1046
Joined: Sun Jul 01, 2018 12:07 pm
Location: Switzerland
Contact:

Re: at rickyx, reason for investigation

Post by FarbigeWelt »

rickyx wrote: Sat Sep 22, 2018 8:08 am Can you post an image of what you want to reach?

I got this:
Laser.png

In interior design often I totally remove the window glasses. Often I don't need reflections or I just mix transparent with glossy using a fresnel.
Well, my idea was to setup a fog chamber for tracing beams. But the beams haven't been visible through the glass of the chamber. Then I started to investigate this behaviour viewtopic.php?f=2&t=474&start=120#p5737 what lead finally to the test scene "Renders Glass very slowly".

I think the glass windows issue is related to my earlier observations. This is why I resumed this thread.
Light and Word designing Creator - www.farbigewelt.ch - aka quantenkristall || #luxcorerender
MacBook Air with M1
kintuX
Posts: 809
Joined: Wed Jan 10, 2018 2:37 am

Re: Luxcore renders light through glass very slowly

Post by kintuX »

The issue on the window shelf is caused by geometry.
Tracing through glass is painfully long everywhere, that's just the nature of these engines.
This is where RTXs are supposed to fill the gap. Or PoverVR previously, but nobody cared... or did Apple...?

Personally, i favored SPPM above all (in my tests even VCM & UPBP didn't came close). It had everything going really well except, it was unintuitive to set up, as right settings are totally dependent on the scene (each scene needs different setting, which today could be solved with the help of Ai). But still, even if too complex for an average mind, it was ported to GPU (in academia/university)... and then forgotten, abandoned... and I can only imagine what would become of it with ingenious minds at work with today's tech advancements... :(

& Here's 10min render on 9 year old machine - 2x xeon X5650 (22 threads @ 2.67 GHz)
Number of glass panels (3mm thick) - from left to right: 0, 1, 2, 3.
sppm.png
PS. disregard time (dinner & chit-chat in between) ;)
Post Reply