OTOY Forums

Stromberg90 wrote:The new renders is interesting, i did one test myself with the meshes intersecting and didnt get any artifacts...
So i dont know what the problem is, and i have so many diffrent versions about this that i dont know wich one is true either

But please keep on testing

so You did intersections wrong Void got it right . When surface are intersecting I got same effect like first glass of voidmonster.

Void- Good job with those tests by the way

Well i doubt i got it wrong, cause i have done it like this on every single glass - liquid render i have done
when it works for me and not for you then you are doing something wrong

Still hope to see some more test on this, dont want to get in a discussion about if i have done it wrong, i only did it the way i learnt it, and i works fine... for me atleast

I just tried one using a model setup for Vray, so it's a solid glass object with an interpenetrating solid liquid (normals pointing out), and interpenetrating ice cubes. Seems to work okay on this one. Seems odd that the results are so variable doesn't it?

I'm still waiting for MLT and refraction attenuation before going too far with these kind of images, but I was curious to see how it would look.

b

Just to go slightly backwards in the history of this thread, a re-render of the inciting image using the technique I figured out with the simpler experiments. Even though it's not currently possible to do an accurate comparison to Maxwell due to the absorption effect in Maxwell's dialectric simulation, this is still kind of interesting.

The Octane render using flipped normals looks much, much better to me. I'd render with that and be happy about it. Enough so that I'm going to go ahead and finish up the image I had in mind at the very beginning of this little excursion.

The key differences I notice between Maxwell and Octane.

First a great big DUH on my part. Botched the light intensity and I didn't try to match the tone curve. Plus the difference in the floor material, so there's an overall large difference in the light distribution.

Second, the base of the glass still renders dark in Octane. I don't know why -- I've tried rendering it so that it's floating above my ground plane but that didn't really help. Seems to be something to do with fresnel falloff. The effect of that is present in the interreflections within the glass too. Things there are generally darker.

It looks to me like there's a distinct difference in the IOR between the two images. I was really careful to set that, including doing the IOR averaging that PhilBo suggests. Dunno what's up there. Still, it's within the tolerance for general 'look' of glass+liquid. I'm happy with it.

The last thing is antialiasing, but we all know about that and how to fix it.

With that done, it's time to render out one last test. In it I will show conclusively why the interpenetrating geometry method (or VRay technique) should only be used as a last resort.

keep this scene handy, and maybe also can you zip it up and PM it to me?
this is good material to test MLT with.

Radiance

I see what you mean, i also think the one with the flipped normals look better.

radiance wrote:keep this scene handy, and maybe also can you zip it up and PM it to me?
this is good material to test MLT with.

Sure thing! Do you want it with or without all the textures/HDR maps? The only texture that's particularly special is the one I'm using on the glass for the micro-imperfections. The HDR is from one of the public free HDR sites and the ground plane is an Arroway texture.

And here's a nice example of why rendering with interpenetrating geometry should only be used as a last resort.

First and foremost, it's slower to render clearly and generally raises the noise level in the scene. This image took about an hour to render up to 16,000 samples. The interpenetrating glass is VERY noisy.

Second it's further from correct. There should be apparent reflection happening between the liquid and glass interface. Instead the liquid appears to fill the volume of the glass out to the very edge in a way that actual glass does not. See attached photo reference.

Note, my test glasses aren't modelled similarly enough to the physical shot glass to really see how the light interacts between layers. So, of course, I am currently modeling a 0.01mm accurate version of my shot glass.

I am a gigantic dork, but at least this time I've got good bourbon on board.

Where the edges are thin it's supposed to look like the liquid is filling the glass completly to the edge.
So in you latest test the left one looks more correct, but in the way of noise and rendertime the right one will be more effictient.

Here is a vray tutorial refrence.