[simmsimaging]

I'm trying to work out the best way to handle liquids and glass interactions, specifically with condensation/droplets on glass or cans/labels etc.

I'm noticing that caustics and refractions seem to be very different depending on how it's setup, but following the "correct" approach for glass etc does not seem to produce correct looking results (the approved method would be the far right drop in the second row of each set).

See the image below. I tested with a solid block below the drops and with the same block in glass to see how the interfaces affected thigns.

The drop that looks most right to me is the center top one, but it is a full enclosed 1/2 dome that is floating above the surface by a hair. It's physically inaccurate and the gap shows up from many angles of view so it's not a great solution. It also means you need boolean tools to get fully conformed drops to the surface whenever that surface is non-planar.

As soon as I align that 1/2 sphere to the surface and delete the co-planar faces (bottom row, far right) or just align it without deleting co-planar faces (top row, far right) it has very refractive/reflective qualities and the caustics change or seem to disappear.

A direct questions to the guys at NL: what is the correct way to get these looking good and reacting properly to light?


Thanks /b

[Hervé]

well first determine if the surface you plan to put your drops on is hydrophobic or hydrophilic

h/

for hydrophobic surfaces like some leaves, you'll need to keep the water droplet full sphere..
for hydrophilic surfaces like a coke can, you'll need to have the water droplet half spheres

hope this helps..

for instance this is hydrophobic effect

[Bubbaloo]

I would think that for a solid hydrophilic surface, a hollow dome - flush or - intersect / full dome - intersect (makes no difference to Maxwell) would be the correct way. A light ray encounters the droplet's normal, then bounces off of the "coke can's" normal, and back through the droplet's normal, then to the camera.

[Aniki]

http://maxwellrender.com/videos/realflo ... flow05.php

I used boolean cuts on the liquid, as you dont really see caustics through sss volumes anyways;)

I thought the main problem is that caustics through dielectics still pose a problem until you reach some SL above 35...

Sorry for asking, what exactly is meant by "flush" ?;)

[simmsimaging]

Thanks guys.

I don't think the hydrophobic/philic thing is the issue here though. That will create more or less spherical surfaces, but if you look at the drops they are just doing very different things. Below is another render that might make it clearer.

You can see the drop that is solid 1/2 dome and floating above is quite different - that makes sense because the back face is creating more TIR etc.

But if you see the full 1/2 domes that are intersecting or flush you get very strong internal reflections/caustics, but nothing projected out. The hollow domes doing the same get much darker inside, but project some caustics outside. This is what I'm talking about - I'm not sure if either is correct, but I think ideally you would have both going on.

How do you get that to happen?

Aniki: I think caustics should be coming up well before SL35, at least in newer versions of Maxwell, and SSS materials are not really as relevant because they caustics would be smeared so much it's a non-issue, but if there is a darkening problem it would still mean they render darker and duller than they should even with boolean cuts on the back face.

/b

[Bubbaloo]

I see what you mean. I find it very strange that a solid sphere intersecting the surface renders differently than a half (open) sphere intersecting the surface.

[Bubbaloo]

On second look, maybe it's the angle/perspective making the difference. A better test would be to render them one at a time in different configurations so all angles and lighting will be identical.

[simmsimaging]

I looks odd for sure. To me the drop should look *internally* like the top left, but the *outer* area should look like the bottom right.

Some form of intersecting mesh would be the ideal answer though, because if you have drops on a curved surface you will have to custom model them, or boolean them, in order to get perfect alignment. Neither is not realistic with heavy condensation models (my current one is over 7 million poly's when converted to a full mesh so boolean ops are choking on it) and that workflow does not really make for much flexibility in the drop position as client demands change etc.

/b

[simmsimaging]

On second look, maybe it's the angle/perspective making the difference. A better test would be to render them one at a time in different configurations so all angles and lighting will be identical.

I don't think so, but I can try it.

b

[simmsimaging]

Good call Brian - I stand corrected (again), but it's not *totally* that factor



In this one I just moved the solid versions down and the hollow versions up. Much of the differences are gone, but the solid interesecting drop and the flush (co-planar) one (bottom left and right respectively) seem darker inside the drop still.

The hollow and intersecting dome (top left now) still looks different too: the external caustics are not happening, at least not to the same degree. This is the one that is actually relevant to the image I was working on where I cut away the mesh with booleans as per another Aniki's approach on another thread. Here's another view of the same arrangement to confirm: