[Kabe]

It seems (given the new explanations) Maxwell is doing things more correctly than I previously thought.

Yes, and the solution is compatible to any renderer worth it's salt

Given how thorough Maxwell already is, I suspect any improvement will be in the form of a trivial tweek in the paradigm somewhere.

I would like to see NDs below 1 for underwater scenes... It's a cheap and very useful trick to render underwater air bubbles in most renderers

Also, the "tolerance" idea seems interesting (in a pair of facing surfaces A and B, exclude surface A from reflections in a space less than epsilon .... or something to this effect.)

Actually, I don't like that idea too much, because it will create edge artifacts for shure, and it's very hard to keep the geometries consistent with such an approach.

Kabe

[Axe]

Thanks Thomas! Excellent solution. The only problem I see is that the capillary meniscus around the edge is given a glass material instead of the liquid. I guess there is no workaround for this, as if the upper surface were liquid and the lower were glass it wouldn't work right.

[Thomas An.]

Thanks Thomas! Excellent solution. The only problem I see is that the capillary meniscus around the edge is given a glass material instead of the liquid. I guess there is no workaround for this, as if the upper surface were liquid and the lower were glass it wouldn't work right.

Well, that is trivial. Just pull your liquid around the edges to follow the shape of the glass meniscus while maintaining the infinitescimal offset.

-

[Kabe]

I think it should be possible to drag the liquid into the meniskus as well, as long as it stays inside the glass. This was a minor details I was thinking of, but neglected to mention in the greater scheme of things.

Kabe

[Thomas An.]

I would like to see NDs below 1 for underwater scenes... It's a cheap and very useful trick to render underwater air bubbles in most renderers

Agreed.

Also, the "tolerance" idea seems interesting (in a pair of facing surfaces A and B, exclude surface A from reflections in a space less than epsilon .... or something to this effect.)

Actually, I don't like that idea too much, because it will create edge artifacts for shure, and it's very hard to keep the geometries consistent with such an approach.

Kabe

Yeah, I am sure there are better much more elegan solutions (the tolerance thing seems more like a duct-tape solution).

[Thomas An.]

The only drawback of the approach of Thomas An. is that it's a bit harder to model, but optically such a setup is the cleanest way to render it - probably this is true for most available renderers.

Kabe,

What modeler do you use ?
This thing is ultra trivial to do in Rhino.

[Mihai]

Kabe, can you explain a bit more why you would need an IOR below 1 to render airbubbles in water? Do you mean only if the camera is below the water surface or also if you were to see bubbles in a glass of water for example?

[kingpin]

First, Thomas...
I really appreciate what you have done. But I have a question about your scheme.

1. If the infinitesimally small gap between surfaces do not affect IOR, then tom's model and Mihai's model should give the same results... But they didn't...

2. The difference between your model and Mihai's model is a top surface of liquid (air->liquid), but when we evaluated Mihai's model, we thought that there were problems with refraction at (liquid->galss and glass->air). And yours still inherit the same problem.

I still think... the problem is that maxwell does not bend a ray when a polygon is hit from behind (just place a camera inside a liquid... you won't see any distortion you would expect)... From you and others' work, now I know that maxwell does bend caustic hit from behind, but none of these experiment has shown that maxwell actually bend a ray hit from behind....

Kingpin
PS. You know... what we REALLY need... simple few words from NL... I know that they were and are watching these threads...

[Thomas An.]

First, Thomas...
I really appreciate what you have done. But I have a question about your scheme.

1. If the infinitesimally small gap between surfaces do not affect IOR, then tom's model and Mihai's model should give the same results... But they didn't...

It is not the same. In Mihai's model the surfaces are facing and reflecting each other (please see discussion above)

2. The difference between your model and Mihai's model is a top surface of liquid (air->liquid), but when we evaluated Mihai's model, we thought that there were problems with refraction at (liquid->galss and glass->air). And yours still inherit the same problem.

Again, not the same. (please see previous explanations)

I still think... the problem is that maxwell does not bend a ray when a polygon is hit from behind

It does (as the experiment revealed). Please try to follow the diagrams and the experient. I believe it covers a lot of ground.

[Mihai]

For animations then, if we don't have doublesided materials or some sort of ray tolerance, how do we render that correctly? As a basis I'm thinking that you would run the simulation and insure that no part of the resulting mesh penetrates the glass, because either you have the whole mesh penetrate, or none of it.

[DELETED]

DELETED

[kingpin]

Sorry about my post.
Now I think I know... you need to model an object so that every surfces of object are paired from every angle to be correct (and there were missing surface for Mihai's model)...

Kingpin
PS. Thanks a lot, and sorry for taking your precious time for waste like my post.

[Kabe]

Kabe, can you explain a bit more why you would need an IOR below 1 to render airbubbles in water? Do you mean only if the camera is below the water surface or also if you were to see bubbles in a glass of water for example?

No, it's just for the case that the camera is submersed in water. If you have air bubbles in a glas, just set them to ND=1 and everything is fine.

I haven't checked in Maxwell yet, but in a number of other renderers you have to build spcial constructions for the camera for underwater scenes, that are cumbersome to use because they change the image angle. So, my usual and illegal trick is to work with faster light

No, seriously: I just take the water as reference for the effective NDs. If you have a water (ND=1.3) - air surface, you can calculate an effective ND for the air bubbles or the sea surface, which is 1/1.3 = ~0.75. So you have to adjust *all* NDs for your scene, but you can work with a normal standard cam.

If the limitation of NDs is just build into the plugin, it can easily be removed. Otherwise the rabbit hole might be a bit deeper.

Kabe

[Kabe]

What modeler do you use ?
This thing is ultra trivial to do in Rhino.

I use Cinema. What I'm looking for is a solution that calculates Glas & Liuid from an "empty glas shape" and a reference surface, and which also whould calculate the meniscus in it's perfect implementation.

I already have an idea... but it's certainly a bit more complex than it is in rhino

Kabe

[Mihai]

But I think this is only relevant to raytracers that trace only from the camera isn't it?

I think Maxwell traces both from lightsources and camera, so if you have a lightsource above the water surface, won't it produce unrealistic results? Couldn't you just instead tell the renderer what medium the camera is sitting inside so that all rays it's sending out know to have the correct IOR? Say it's sitting in water, so it sends out a ray that hits an air bubble, it knows that it's "original" nd was 1.33 and so will bend correctly when it enters the bubble.

Just finished the render using Thomas method, looks good



I have a glass very similar to this one and the bottom reflections look correct. The cappy meni thingy looks too sharp because I modelled the glass in haste and the curve is not well defined. I guess nurbs would be much better for modelling these types of things.

For comparison the old ones: