IES + Glass = Noise

[vicnaum]

I don't know if I should post it here, or in Bugs section... But let me describe the problem.

Everyone knows that IES lights give much noise in Corona now. And the noise is un-even, clear in front, and noise on sides as you can see:


But what's more problematic is that the noise increases dramatically if IES is hitting glass:


Or even more glass => even more noise:


Guys here faced this problem on a museum render, with high glass boxes everywhere. IESes there are helly noisy after hours!

What can be done? Any advice?

[tomislavn]

Are you using "Thin" checkbox in glass material?

[vicnaum]

Nope, not using it. Because it's making it very unreal.
Just an ordinary 0 Diff, 1 Relf, 1 Refr glass.

[tomislavn]

Hmm not really sure what to do other then that. There was one post last month from Ondra where he said that they are working on improving IES/Noise. Might still not be done in A7.1. Probably for 1.0 release then.

I am usually trying to avoid excessive noise by using "Thin" glass without caustics, or maybe using IES without presets.

Maybe some more experienced users here could help you with it, I am not making a lot of glass renders.

[fire3d]

Try Light Portals, don't know if it works, theoreticaly it should :)

[vicnaum]

Nope, portals don't help (or I don't know how to put em?). I put one plane after the glass (i.e. from the wall side).

[quadrays]

are caustics turned off ? you should trace is as transparent, not glass object, simple glass plane is only architectural solution with no caustics.

even you can override global illumination in corona ray switch shader

[Siahpoosh]

try using rayswitch material and put your glass material in all slot except global illumination , u will lost some shadow from glasses but it will clean noise

[snakebox]

So def a problem, but next time you want to compare something make sure you have the basics right.

I'm sure you are correct but using your images as a reference is nearly useless as none of them have the same amount of passes.. 11 , 9 and 3 ? how are we meant to compare that?

[vicnaum]

I'm sure you are correct but using your images as a reference is nearly useless as none of them have the same amount of passes.. 11 , 9 and 3 ? how are we meant to compare that?

Simple - they have the same amount of time - 1 minute :-) Caustics turned off.

Yes, RaySwitch helped, the noise is less, and similar to such without glass:


But if the glass is colored, frosted, or any other than clear - it won't work.

[Ludvik Koutny]

Most of the noise is not from shadow, but from reflective caustics. Glass reflects that hotspot on the wall, and GI is sampling the glass reflection on the wall where light shines. It's reflective caustics basically.