Caustics: odd behaviour with different shapes

[simmsimaging]

I continue to try and figure out caustics from time to time and still have problems with one thing in particular. see the image below:



My usual problem is trying to get sweat droplets/condensation to have nice caustics and it always fails. In the image I tested a sphere, a cube, and a torus knot. On the left the geo is free-floating. On the right it is penetrating the grey block. Notice the very big change in the square and sphere. I get the same thing with the sphere if I change the shape to a 1/2 dome that penetrates as well.

I can't see how this is correct, but I am not really sure what is going on. Can anyone at Chaos clarify what is happening with internal reflections and caustics - is something odd going on?

Here are my caustic settings, more or less default, and the relevant material settings. Caustics generated by a Max spotlight only, set to only generate caustics.

[PIXELBOX_SRO]

that sphere behaviour is so weird...looks like milky glass....dont think that is what it should look like.

have you tried the progressive method yet?

[Dariusz Makowski (Dadal)]

Are your normal correct? - I dont see any bad behaviour here btw or I'm just stupid

Is ur ray trace depth high enough ?


What is progressive method?

[simmsimaging]

Normals and trace depth are not the problem. Not sure what the progressive method is either - do you mean the directly visible option you can access with that line of code? If so, I have tried it but AFAIK it just makes the process visible, it doesn't do anything differently.

[vlado]

Do you have any actual photos of droplet caustics (I guess I can make some later at home )

Best regards,
Vlado

[simmsimaging]

Here are a few:

[simmsimaging]

Maybe this is totally normal behaviour? I'm looking at the reference I pulled and looking at the render and now I'm thinking it doesn't look as odd as I originally thought .

I know I've never been very happy with the caustics and droplets I've done in the past, but maybe it's more about my expectations being off?

Anyway, I just want to verify (if possible) that Vray is doing the correct thing with penetrating caustic generators like this, then I can figure out what I might be doing wrong.

/b

[vlado]

Well if you look at the references, there are no (or only weak) caustics in the shadows outside of the droplets - most of the caustics effect only appears inside the droplets themselves - which is what I'm seeing my tests here as well.

Best regards,
Vlado

[simmsimaging]

What about what's happening inside the droplet? I can't seem to get any of the more clear or perhaps more subtle, concentration of light in the drops. They just go milky white in this case. Could it be caustic settings? The directly visible method doesn't help in order to experiment/refine, because you can't see caustics inside the drops, and the caustics render element has the same issue (last time I tried it anyway)


b

[Dariusz Makowski (Dadal)]

I think ur sphere is too perfect to give you what you want to see... try modeling it 1st setting it up proper model with layers then see if you like it more...

Remember u need 2 side material since ur object just go into wall where it should bend and touch like it happens in reality...

[simmsimaging]

The shape here is just a primitive to remove variables. I've done more than the odd bit of modeled sweat and I've seen much the same sort of thing. It's not about the geo particularly. The material is two sided and reflect on backside is on etc.

[Dariusz Makowski (Dadal)]

well try bending it next to wall I have feeling that it will look different

[simmsimaging]

It will look different but won't work that way

You can't have co-planar faces, and an air-gap creates totally different effects. The correct way to get the refraction to work is interpenetrating like it is now. The issue (if there is one) is only with caustics.

b

[vlado]

Works fine for me... Will post a few images tomorrow.

Best regards,
Vlado

[simmsimaging]

That would be great. Thanks for looking into it.

b