[m-Que]

Hey guys,

I was studying the problem mentioned in this thread:
http://www.maxwellrender.com/forum/view ... 97&t=37126
And came up to a pretty interesting thing -

So there's basically a glass of water with a hollow bottom and a glass plane w. bump map.
The point is:
- you can actually see the caustics through those materials
- caustics become visible pretty fast at about ~SL11 and become decent at around SL18

BTW, It's pretty noisy, cause I wanted to get a sharper caustics for that plane, so it is just a small emitter.
sorry for the crappy image



Would like to know what you think.
Meanwhile, I'll try to do more tests.

[tom]

Yes, but not having a closed-bottom makes it render wrong because of missing internal reflections (TIR).

[David Solito]

hello Tom,

can you be schematic?

thank you

[tom]

Sure
http://www.youtube.com/watch?v=blEvcGWMUss

[Bubbaloo]

Nice one Tom.

[m-Que]

Yes, but not having a closed-bottom makes it render wrong because of missing internal reflections (TIR).

Sorry Tom, are you talking about a "glass of water with a hollow bottom"?
If so, I meant the glass has that specific form on the bottom, just like the bottom of wine bottle. (I'll do a shot of it)

Meanwhile, here's another one.
Full volume water (without bump map); SL~24,5

[David Solito]

Can someone explain me? I know there is a lot of topics on the forum..but can someone make a graphic to understand?

Thank you!

[itsallgoode9]

Yes, but not having a closed-bottom makes it render wrong because of missing internal reflections (TIR).

But Maxwell has some major limitations in this area sometimes though, right? So either way you're ending up with an "incorrect" render, or render times in the 100's of hrs if you want a "correct" caustic calculation. Sometimes "correct" isn't the best way to do something.

on another note, did not realize bump maps would create caustics! That'll save me some time in my renders.

[m-Que]

Can someone explain me? I know there is a lot of topics on the forum..but can someone make a graphic to understand?

Thank you!

Hey David,

I came up to a certain conclusions recently, so I'll try to wrap things up here.
'Caustics-through-dielectrics' is a known limitation in Maxwell. Now, when I recently saw caustics actually appear through a dielectric material in one of my scenes and at a pretty low SL, I was thrilled to know how it actually happened. So I went 'investigating' and started this thread.
First thoughts where of course, that the thing is material- or geometry-wise.
But it turns out, the truth lies somewhere else - it's the light source.

Here's an example scene - a glass of water.
There're 2 example pictures per case:
- geometry hidden to camera: to show the actual caustics under/behind the glass
- Glass of water: to show how the caustics are displayed through the glass of water

A. Physical sky + sun

1.Geometry hidden to camera + Sun

Solid caustics appear really fast here - at SL2 and also clean really fast; no problem here

2.Glass of water + Sun

Now this one's the most 'problematic' one - basically you will not see any caustics through the glass here.
They'll only start to appear in a form of 'firefly' pixels at around SL21. And as Brian has recently shown, will form into solid caustics at ~SL33.
Please notice, liquid surface and ground surface under the glass both stay dark here.

B. Emitter plane

3.Geometry hidden to camera + Emitter plane

Caustics from an emitter are a totally different story. They first appear later at SL8 - SL10 and form slowly (pretty much like caustics in 1.X did)

4.Glass of water + Emitter plane

Now, this one is the most important -as you can see, here the caustics are visible through the glass of water.
In this case, they started to appear at ~SL8 and finalized at ~SL18; The render was stopped at ~SL22.
Again, notice the liquid surface and ground surface under the glass and compare them with pic.3 and pic.2

Also, in case of emitter lit scene, it's important to know:
- smaller emitter plane = higher SL to finalize caustics, but sharper caustics
- bigger emitter plane = lower SL to finalize caustics, but dull caustics

Another interesting fact -
Caustics from an emitter can be seen in Fire, while caustics from the Sun - cannot; even when geometry is 'hidden to camera'

Well there you have it.
I hope it'll make things more clear now.

[David Solito]

Hello m-Que!

Thank you very much for your post. It makes more sense now.
As tom would say, the best way to have the good answer is to try it.

You make my day.

[tom]

'Caustics-through-dielectrics' is a known limitation in Maxwell.

Well, this is not accurately correct. More correctly, it is just the "inefficiency of caustic convergence on specular surfaces". It's the same thing with caustics on a mirror, behind a glass etc. When the reflective/refractive surface which you expect to see the caustics on/through is specular, it takes a lot of samples to gather the caustics without adding bias/tricks. As previously demonstrated by Brian, Maxwell does not miss these caustics and fully capable of rendering them so, I can't agree it's a limitation.

[m-Que]

I fully agree with you, Tom.
I've mentioned that caustics will actually be visible at higher SL.
I was speaking more of a limitation in terms of time. Since, let's face it, rendering an image up to SL33 takes eternity.

BTW, Tom.
Can you think of any explanation, why that difference between Sun and emitter?

[David Solito]

is it not due to parallel light rays? Or something like that?

[itsallgoode9]

I agree with you Tom, Maxwell can do it, although the time to get it to show up sometimes is SO extreme. I know it's due to the calculations of this type of engine as opposed to Maxwell itself. It's just one of those things that generally isn't feasible in a real working environment due to the time.

After seeing how much SSS was able to be sped up, it makes me consider the current implementation of the caustic calculations in certain scenarios to be a limitation of Maxwell. The problem I run into with this is that there isn't really much of a work around to make it a bit more useable, as opposed to SSS, when it was really slow. You could blur, denoise etc to get SSS to work good enough in some situations, but with some of these extreme caustics situations, that isnt possible due to the nature of caustics needing to stay crisp.

Generally, caustics works well, but in some extreme situations it's a real killer, especially for us killer for us liquid guys.