Coating+transparent substrate

[thesquirell]

Hi everybody! (Hi Dr.Nick!)

I was just playing around with some of the effects I would love to see on the glass, and the irregular glass rainbow pattern came into my mind. It is possible to achieve such an effect with coating material and glossy transparent substrate, but not with the specular one. Using arch. glass is also impossible in this system. But, using this Coat+GT system takes a lot of time to clear up, with possibility that it will never clean up (30k samples, pic1 "CGT"). So, I was forced to try something different, and that was blending specular material (arch. glass) with coating+diffuse mat. This solution was faster, but there is something that is troubling me. When viewing it from the "outside" it produces nice effect, and the caustics show up (pic3 "ext"). But when viewing it from the "inside" the caustics fail to converge (pic2 "int"). If someone has the idea why, or has an idea how to achieve this effect without post-pro, please share your mind, and your opinion on this subjects.

[CTZn]

As to why, trivially, this is a structural weakness in renderers, possibly specific to unbiased renderers but I don't know that.

My intuitive way to put it is that specular and diffuse light diffusion types are antagonistic. A diffuse surface will make difficult(?) the transport of caustics to another specular (as in real life somehow, though IRL we are dealing with a virtually infinite sample count). Another trivial name for the situation could be "difficult sampling paths". It's not a bug though.

Pretty much an uneducated answer but a beginning still

[thesquirell]

Thanks for that explanation! What worries me here is that it seems impossible to create, for instance, a dirty window view from the inside, because every blend system with specular, or glossy transparent material seems to kill caustics viewed from the inside. The ones visible outside the object itself are perfectly visible and strong, but the inside is...troublesome.

[CTZn]

try switching the floor to a phong, then controlling its exponent (glossiness) should help with secondary caustics.

If I'm correct in the first place it's the in-between diffuse that is messing with the light paths.

[thesquirell]

Not sure to what value am i supposed to adjust the phong settings, and set it up, but with phong IOR 1.6, exponent from 10, to 1000, it doesn't help at all. And it seems that sample number is not the problem. Getting really curious to find out what is going on here... haha!

P.S
If someone encounters the same problem in their workflow while creating this effect or would love to use some dirt on their glass, I would suggest using something I called a double mesh system, where your coat of dirt is actually an another mesh, slightly in front of your glass, with a blending system that can include coatings, bumps, etc., and is made transparent by blending it with a null material. Or to put it in more simple way, just imitate the nature, and Indigo will do the rest for you. Using this system we can even take advantage of arch. glass specular material. Have fun! (all other pics, except the Glossy Transparent one from the first post, are actually under 4 minute renders with very low samplings, but with impressive level of clarity!)

[Oscar J]

But... the last image has the rainbow effects. It works, or what am I missing?

[thesquirell]

You didn't miss a thing, dear Oscar J! The last image works because we modeled it in the accordance with the real life. The other ones didn't, because we wanted to cheat it by blending transparent materials (specular,glossy transparent) with the others, which kinda killed our secondary caustics,or dimmed it a lot. With the system explained in the last post, we have one solid mesh of specular arch.glass, and one in front of it, which would be the case if we had some dirt on our real windows. We would have some mesh on top of our glass material, and that is exactly what we did, and it works, as long as we don't use any specular or glossy transparent materials blended with others in that outer mesh. That is why I needed to point out that this coating was made transparent by using a blend with null material.

[CTZn]

Well idk wether my idea of the problem was relevant at all, I got mixed up here I think.

IIRC that issue with caustics and blended null was already reported in the forum (I for one did).