Speckled Caustics

[thomthom]

I'm sure similar questions has been asked before, but since the search function is currently not working I can't find the posts.

So here it goes; I've been trying, what I thought would be a quickly little scene, to play with liquid looking materials. But also trying to get some caustic effect made everything snowball.

I'm struggling finding an effective setting in order to get a bright and crisp caustic effect. It either turns out splotchy:


Or, it turns out really blurred.



On this last test I had 40000 subdivision set up for the light, and it's still noisy. And render time shoots up.



I've been reading this thread: http://forum.asgvis.com/index.php?topic=248
the V-Ray manual on caustics http://www.spot3d.com/vray/help/150R...s_caustics.htm
and other tuturials such as http://www.vray.info/topics/t0097.asp

But I still don't "get it". I can't get to grips with the combination of settings to make an effective render. Increasing the caustic subdivisions to 50000-60000 to get rid of the last bit of noise seems like brute force overkill. Right?

[dalomar_cg]

Re: Speckled Caustics

First off, its much easier to get a better caustics affect from point/spot lights or very small rectangular lights. From the looks of your scene, it would seam that you have a rather large rectangular light casting the caustics. Its going to be significantly harder to get a smooth result as the nature of the light will lead to blurry caustics.

As to caustics settings, the two main players are Max Photons and Search Distance, and ultimately it comes down to choosing one over the other. V-Ray will keep looking for more photons until either it gets to the Max Photons allowed or that the photons that are being retrieved are beyond the search distance, so its really rather rare that V-Ray takes both into account. In regards to both settings, the smaller they are, the noisier the caustics will be, and the larger they are the smoother the caustics will be. However, its best to understand the words "smaller" and "larger" in relation to the density of caustic photons, which is controlled through subdivisions.

Optimally, you want to take as many photons in consideration as you can, yet limit the ones that are extraneous to that part of the solution. You can control this in a fairly absolute sense with the Search Distance parameter, and although this works well for the sharper areas of your caustics, it may not smooth out the areas where there are less caustics. However, a "good" search distance parameter does not need to scale with photon density, so you don't really have to worry too much about it in regards to increasing the quality of your caustics.

With Max Photons, there is much more consideration that has to be made for density. Keeping them low for tests (maybe at 15-20) may work in low subdivision situations, but when the caustic subdivisions are increased, the same Max Photon values will leave a lot of the photons in your solution unable to actually contribute to the scene. Therefore, you should consider increasing Max Photons with large subdivision numbers to make sure that those extra photons are utilized. The other good thing is that Max Photons is much more adaptive for different areas of the image, so places with lower photon density will have a chance to get the photons that they need, as opposed being cut off with search distance.

So my recommendation is to choose one and adjust that while having the other at a large enough value not to get in the way. Personally I find better control in the Max Photon setting. One thing to note, and this is a very important fact, all of these settings are not tied to the calculation itself, only the interpretation/reconstruction of that solution for the end result. Therefore, if you save your solution, you can play around with these settings to see which are optimal in your situation.

One final note, these suggestions assume the you have a radius of 0 for Max Density. Changing the Max Density value changes the playing field completely and very few of the recommendations above hold true. Best to leave it as 0 most of the time.

[thomthom]

Re: Speckled Caustics

One thing to note, and this is a very important fact, all of these settings are not tied to the calculation itself, only the interpretation/reconstruction of that solution for the end result. Therefore, if you save your solution, you can play around with these settings to see which are optimal in your situation.

Very useful to know.


Thanks for all the info. With this I'll give it another testing round. I had two rectangular lights that where 3-4 times larger than the glass itself. And at the end of my tests I've been using Max Density.

[dalomar_cg]

Re: Speckled Caustics

There is no way to explicitly remove lights from the caustic calculation, BUT if you set the caustic subdivisions of a light to 0 this effectively prohibits it from contributing to the result.

Again, try not to use Max Density if you can, because that changes the structure of the map and complicates things significantly. This is also a setting that CANNOT be adjusted from a saved map.

[thomthom]

Re: Speckled Caustics

Can I reuse the IR and LC map as well?

[dalomar_cg]

Re: Speckled Caustics

Yes

[thomthom]

Re: Speckled Caustics

That will help even more! ;D

[thomthom]

Re: Speckled Caustics

Reducing the lights made a huge difference. Looked like I didn't need that many subdivisions any more. And the caustic was sharper. All though, when I tried to do a test where I ramped the subdivisions for both lights up to 10000 SU crashed. Task Manager showed that it used 1.5GB ram. Wondering if SU ran out of memory. I that the density could prevent that. Will try some more though without turning on density.

[dalomar_cg]

Re: Speckled Caustics

It certainly can help with that...HOWEVER, you will have to rethink the Max Photon/Search Distance control. Basically you can't really depend on Max Photons any more because they will be added together, so you'll have to bring Search Distance into the mix. Also, NEVER have a Max Density radius larger than the Search Distance as this will essentially have only one photon be used.

[thomthom]

Re: Speckled Caustics

I played some more today. Mostly with the light position, and I noticed that the effect was greatly impacted by the size and position. So for the next tests I'll put some more effort into the lights in the scene; something I didn't bother much with in this.

[Micha_cg]

Re: Speckled Caustics

Sidenote: a very intersting caustic thread from Thomas An.:

http://forum.asgvis.com/index.php?topic=248 ... at page 4 Vlado posted a great result.

Maybe it helps. Also it could be intersting, how PPT handle caustics here or Vlado's universal setup:

http://forum.asgvis.com/index.php?topic=438.0

[thomthom]

Re: Speckled Caustics

Yea, I saw that caustic thread. Very interesting. It's because of that I tried the Max Density.


Haven't seen that other link though. But I think I've seen part of the info there somewhere else. But not the whole lot. Nice to be able to read everything. Thanks.

[dalomar_cg]

Re: Speckled Caustics

This thread had me go back and reexamine my caustics workflow. This is something that had stumped me in the past, but I think I've got a solid, repeatable workflow that deals with a number of issues that I'd run into over the years working with caustics. It goes off of 2 main steps (low SubD testing and high Subd refinement) with a specific means of dealing with some of the complexity that can come from the different caustic parameters. I'll see about getting it together and throwing it out there for you guys.

[thomthom]

Re: Speckled Caustics

That'd be fantastic. ;D

[Micha_cg]

Re: Speckled Caustics

I'm curious too, since I avoided to use cautics in the past, but would like to use it in the future.