Self Reflect Refract on Caustics

[davius]

Not sure if this has been posted before but here it goes:

Yesterday, after 12 hours of rendering caustics on corona I noticed that the light generated by the caustics are not visible in the reflections/refractions of other objects on the scene. Unfortunately, I don't have the image here to show the problem - the PC crashed while rendering.

Worth noting is that I used the VCM engine and, even after this 12 long hours, I could not get rid of noise/fireflies on the caustics bright spots. I'll make another render later on a lower resolution so I can post the image here.

Cheers!

[Ondra]

reflected caustics are notoriously hard to render, there is only few algorithms capable of rendering them. You cannot get them correctly unless you use progressive photon mapping, or Bidir/VCM in VCM mode. Please post the images and your settings.

[davius]

Thanks for the quick reply sir! I'm going to re-render the image and post all the settings I used, but basically I just changed the engine to BDir/VCM.

My PC is a little old (i7 920 18 GB Ram Win7-64 bits) and I used Max 2013 for the scene. I'll let it render for a couple of hours and post here.

Cheers!

[davius]

Hey! Here's the image (1 hour and 11 minutes or rendering)

And here's the settings:


As you can see there's indeed reflection of the caustics on the chrome ball so setting the engine to BDir/VCM and the mode to VCM does help. To bad the render time goes through the roof...

Still, this is one of my first tests with Corona and I must say I'm pretty impressed! It's very easy to set up and the results you get by using the Progressive mode are awesome!

Gonna keep on learning it a bit more. I know Photon Tracing/Mapping is somewhat old tech, but it's fast and works very well, specially when rendering caustics (at least that's my experience with VRay and MRay).

Cheers!

[Ondra]

the fireflies on the chrome ball are really weird, can you post the scene?

[davius]

Yeah, sure, but the scene is kinda big ATM (I've got some thousands small rock pieces scatter around it, they're hidden in the posted image though).

I'll delete some stuff in order to make it lighter and post it here. About the caustics themselves - any way to make them "converge" faster? As it is now, even after 12 hours I could not get rid of the pointillism in the hard caustics area...

Cheers!

[davius]

I just sent the scene to you via PM. Sorry, but I can't share it on the forum... hope you understand :)

[Ondra]

note to self: peebles.zip

[Ondra]

ok, I went over the scene, and it turned out the source of the problem is the bump map applied to the reflective materials. It is too weak to be visible in most areas, but where it causes the normal to randomly face the sun, the resulting mirror reflection of the sun is visible as a bright dot. So what you got was actually the correct solution. You can get more plausible result by disabling the bump map or disabling the sun visibility in reflections (which unfortunately does not currnetly work with bidir/VCM as those are experimental/toy engines)