Page 4 of 7

Re: DLSC test

Posted: Mon Oct 01, 2018 1:21 pm
by Dade
B.Y.O.B. wrote: Mon Oct 01, 2018 10:51 am Maybe the DLSC building phase could also use the visibility map, if it was computed?
He does but visibility map is one single global result (it would need a local entry like DLSC: one map for each point). A local visibility map would be far more effective in general but it would require a LOT of ram.

Re: DLSC test

Posted: Sun Oct 14, 2018 10:13 am
by lacilaci
Are there some performance improvenents planned for dlcs calculation? In a scene with several lightbulbs (thick glass) I lowered entry count 10x and raised entry radius to 60cm and yet it takes forever to start remdering. I shut down blender after 4 minutes.

Re: DLSC test

Posted: Sun Oct 14, 2018 10:37 am
by Sharlybg
lacilaci wrote: Sun Oct 14, 2018 10:13 am Are there some performance improvenents planned for dlcs calculation? In a scene with several lightbulbs (thick glass) I lowered entry count 10x and raised entry radius to 60cm and yet it takes forever to start remdering. I shut down blender after 4 minutes.
For your thick glass use architectural glass. And be aware that DLCS is suitable for night render.

Re: DLSC test

Posted: Sun Oct 14, 2018 11:16 am
by lacilaci
Sharlybg wrote: Sun Oct 14, 2018 10:37 am
lacilaci wrote: Sun Oct 14, 2018 10:13 am Are there some performance improvenents planned for dlcs calculation? In a scene with several lightbulbs (thick glass) I lowered entry count 10x and raised entry radius to 60cm and yet it takes forever to start remdering. I shut down blender after 4 minutes.
For your thick glass use architectural glass. And be aware that DLCS is suitable for night render.
Architectural looks wrong.. This goes back to the missing feature hiding objects and lights from diffuse/reflection/camera...

Re: DLSC test

Posted: Sun Oct 14, 2018 1:47 pm
by Sharlybg
Architectural looks wrong.. This goes back to the missing feature hiding objects and lights from diffuse/reflection/camera...
DLCS is for direct light. With emitter mesh enclosed in glass with high IOR will give hard to compute indirect light and caustics rays for simple path tracer. So switch to bidirectional engine.
But for me arch glass + path can easily solve the problem. Even transparent light bulb in real life use architectural glass ( just increase reflection to match your need)
https://www.google.com/search?q=transpa ... 0M_ioeoBDM

Re: DLSC test

Posted: Sun Oct 14, 2018 2:00 pm
by lacilaci
think about this. Being able to disable diffuse/specular/glossy/shadow and camera visibility for meshes and lights, I could create lightbulb that would have it's internals glowing and being reflected but not emiting light(disabled diffuse)
and then a simplified mesh around the bulb that would have light material and disabled camera visibility and shadow and reflection visibility.
This way I could use many lights and benefit from dlcs but it would also be very optimized.

This is a very powerful tool that allows not only for great in camera artistic controls but also allows incredible optimizations that cannot be recognized as "fake" without actual reference.


Right now it seems while dlcs can help with many lights. It still struggles with many complex lights. I can try to find out if there i ls something specific about my scene setup that causes the issue, but I feel like I'm fighting the renderer, instead of the renderer just doing what I want.

Re: DLSC test

Posted: Sun Oct 14, 2018 2:02 pm
by lacilaci
Sharlybg wrote: Sun Oct 14, 2018 1:47 pm
Architectural looks wrong.. This goes back to the missing feature hiding objects and lights from diffuse/reflection/camera...
DLCS is for direct light. With emitter mesh enclosed in glass with high IOR will give hard to compute indirect light and caustics rays for simple path tracer. So switch to bidirectional engine.
But for me arch glass + path can easily solve the problem. Even transparent light bulb in real life use architectural glass ( just increase reflection to match your need)
https://www.google.com/search?q=transpa ... 0M_ioeoBDM
I'll see if archglass will actually help and get back to this..

Re: DLSC test

Posted: Sun Oct 14, 2018 8:15 pm
by Sharlybg
Please show me an image reference of what you want to achieve + your actual render.

Re: DLSC test

Posted: Mon Oct 15, 2018 5:08 am
by lacilaci
Sharlybg wrote: Sun Oct 14, 2018 8:15 pm Please show me an image reference of what you want to achieve + your actual render.
What do you mean? I want lightbulbs to light the scene(or part of it)

In attachment is the geometry(hidden glass mesh) You can see that since it's dense mesh the light count is high, that's ok...
And sure, if I want thick glass DLCS won't go through the glass cause it cannot be hidden for diffuse rays.
However even if I use archglass and dlcs works it is very slow even for one lightbulb and a plane...

So my question is what I need to do to use this light bulb(several of them) as a ligtsource

Re: DLSC test

Posted: Mon Oct 15, 2018 9:30 am
by Sharlybg
I mean your final Goal. wich kind of render output you want to achieve.
V-Ray_photo reference_3D render.jpg