Reflection Glossiness less than 1.0 causes lights to not reflect properly?

[James]

Not sure what I'm seeing here, but the scenario was - using maps in the reflection glossiness of a mirror material (for very delicate fingerprints/scratches etc) was causing the reflections of opaque lights in the scene to reflect as hard spheres/shape of the light itself. But a pure mirror material worked ok.

So to try and work out what was going on - in the simple scene here theres a single sphere light inside the opaque glass light. Mirror on the left is a pure simple mirror material (fresnel 999, reflection @ 1.0 & 1.0). Mirror on the right is the exact same material but with the reflection glossiness set to 0.99.

Causes the the sphere light to reflect with a hard edge and even with a faint halo around it.

Am I doing something wrong here?, Cant see that I am, but as it is I cant see a way of adding very fine maps to the reflection glossiness without causing this issue to happen.

Using Dailybuild Jan 16 (Afraid I cant test on a dif version right now - no admin rights on this PC)

[TomG]

Does the scene set up include anything along the lines of "Cast shadows" being disabled? See the "Light transport issues" section on this page, to see if the scene meets any of those known limitations: https://coronarenderer.freshdesk.com/support/solutions/articles/5000516180-known-bugs-and-limitations

[maru]

That looks odd. How is the translucent lampshade material set up, and as Tom asked - is there some "fake" involved like something not casting shadows, or light with no visibility in reflections/refractions?

[James]

-Appearance of refractive materials changes when reflection glossiness other than 1 is used (even if it's set to 0,999). The workaround here is to enable caustics for the refractive material.


Ah thanks, so this has fixed it. Turned on caustics in the opaque glass material.

No fakes involved like turning off casting shadows etc. That's why I built a quick test scene from scratch to see what was going on. Screenshot of the material attached.

Ok so heres the next question then. With caustics turn on it actually renders faster?? even though its always said "(slow)" - so Ive never gone near it. :-D

Rendering for 2mins. Caustics turned ON = 45 passes. Turned OFF = 35 passes

[alexyork]

-Appearance of refractive materials changes when reflection glossiness other than 1 is used (even if it's set to 0,999). The workaround here is to enable caustics for the refractive material.

We have come across this a lot lately. Are there any plans to fix this? Seems that it ought to work out of the box, really, for any refractive material with a light source behind it.

[maru]

This is happening because the hybrid glass (glass with caustics disabled, casting transparent shadows) is essentially a fake effect.
Corona then has troubles with what should be "shown" in the reflections/refractions.

There is however a fix to it (which may not work in 100% cases):
-duplicate your glass material
-create a rayswitch material
-plug the glass material with caustics off into all rayswitch slots
-then plug the glass material with caustics on into the "reflect override" slot of the rayswitch

I am attaching an example "fixed" scene.

[James]

Thanks thats an interesting workaround.

Any idea why it actually renders faster with caustics turned on?

I ran the same 2 min test with your rayswitch material setup, caustics on and caustics off - and a smudges etc map in the glossiness to replicate setting glossiness to less than 1.0.

Slowest is caustics off - only 34 passes. Caustics on - 43 passes. Rayswitch glass - 44 passes

So the rayswitch setup works and renders the fastest, but the reflection doesn't fully match the original object, so even though its a bit quicker I think just ticking caustics works best - as its almost as quick and looks better plus you don't have to spend the time building a rayswitch material.

[maru]

Any idea why it actually renders faster with caustics turned on?

I am not really sure, but maybe with caustics off the shadows are taking more time to calculate in this case, as they are essentially "fake"?

So the rayswitch setup works and renders the fastest, but the reflection doesn't fully match the original object, so even though its a bit quicker I think just ticking caustics works best - as its almost as quick and looks better plus you don't have to spend the time building a rayswitch material.

If enabling caustics works best for you here, then great, but keep in mind that in a more complex scene caustics can actually give you some more headaches. That is, in a scene where the caustics would be actually visible, for example on some diffuse surface. Maybe in this scene it won't be a problem.

[James]

Yep, not suggesting that caustics slowing down rendering is totally unfounded. I was only talking specifically about opaque glass lights such as this example and window glass for interior scenes (also tested that on a scene and caustics switched on on the window glass does seem to speed it up, although it cut down on the amount of light entering obviously)