Caustics: odd behaviour with different shapes

[AJ Jefferies]

Hi Brett,

I've been playing around with trying to get a decent droplet system (for animation) today and thought I'd give caustics a whirl while I was at it.

Here's where I got with it, nothing too fancy in the settings and as you can see that the interior of the droplets looks generally brighter (as per your tests) but I thought I'd post it anyway and see if this is the sort of thing you're after (or not!)



Cheers!

AJ

[simmsimaging]

Hi AJ -
That's pretty much exactly what I am looking for. You are getting the brighter interior, but it is getting concentrated away from the light source and projecting some caustics outside of the droplets. Looks great.

So how are you setting things up?

/b

[MoonDoggie]

AJ, that looks great. Any chance on getting some tips for how you created the droplets?

[AJ Jefferies]

Hi guys, glad you like it. I will post a scene file when I get back to the studio tomorrow.

Cheers!

[MoonDoggie]

Thanks AJ. Great stuff.

[AJ Jefferies]

Here you go!

http://www.mdi-digital.com/Offsite/P...est_AJ_018.zip

It's a Max 2011 file and needs Bercon Metaballs (although it wouldn't look remarkably different if you just used geometry). As you should be able to see, the droplets are created/controlled by a crude PFlow system, I had a couple of Box #2 operators in there but I've taken them out of this file so hopefully it will still work. My particle flow skills are a little rusty so please forgive any irrelevant parameters!

Other than that, GI & photon caustics are both enabled and the light has it's multiplier ramped right up. This scene is cripplingly slow by my usual render times and I'm sure there's some optimising to be done

Any questions give me a shout.

Cheers!

AJ

[simmsimaging]

Thanks AJ - I will check it out today.

b

[MoonDoggie]

Bercon's site doesn't have his metaballs. MASTER BERCON! *Bat Signal*

[vlado]

You can find a recompile here: http://www.spot3d.com/vray/misc/

Best regards,
Vlado

[MoonDoggie]

Thanks Vlado.

[simmsimaging]

Very helpful scene AJ, thanks.

A major component is false expectations, that's for sure. Being able to start with a scene the clearly showed the right kind of effects helped me figure out what was doing what, and why I am not often getting what I want.

thanks again! Hopefully this will save me a lot of hassle cheating in post

b

[simmsimaging]

So I'm still running into oddness with caustics. Using AJ's settings I was getting a caustics map of nearly 2GB. I read over my notes from Vlado and he had suggested that the Max Density should be locked to 1/4 to 1/3 of the search distance. I tried that with the following result: dark droplets.



I tried variations with the Max photons reduced to 0 (again what Vlado suggested) but that did not fix it.

Here's the caustic render element from the 2mm distance version - looks pretty weird:



Anyway, the dark drop result is pretty typical of what I've been getting before. Either super bright or totally dark. I'm wondering what the "correct" settings should be so that I can better fine tune caustics. Any reason why the one's from Vlado would fail in this case (or any case?) Is zero Max Density a better way to go in general?

b

[simmsimaging]

Just to highlight something: in my version of the scene AJ's method isn't actually working all the great for this droplet setup once I started looking at the lower drops. I tried it with the Vraylight from his file, and a spotlight generating the caustics instead. They look different of course, but neither looks good.






So I brought his Pflow into my scene to check it and it looks much the same:




So then I merged mine back into his scene and it works just fine. I played with a few options (like changing the display units to mm from generic, and turning on show visible caustics method) to see if I could break it again, but it seems fine. Any idea what could cause the discrepancy? Here's mine back in AJ's scene:

[vlado]

Can you post your scene then? The search distance/max density is dependent on the scene scale, so if you scaled the original scene on merge, then the settings have to be adjusted accordingly.

Best regards,
Vlado

[simmsimaging]

Can you post your scene then? The search distance/max density is dependent on the scene scale, so if you scaled the original scene on merge, then the settings have to be adjusted accordingly.

Best regards,
Vlado

I did not do any scaling - I just switched the display units from generic to mm in my case, and merging AJ's sphere and Pflow into my scene, and vice versa did not bring up any scaling oddness that I could see. That said, I did try a broad range of values for the search distance and Max density just in case, and nothing was really working.

I've uploaded a version of my scene and AJ's that has the same sphere/pflow in both. I had to delete my version because it relies on Krakatoa and Frost, but you can see the difference in the one set up by AJ (I just modified that one to replace the Bercon metaball with instanced spheres for simplicity). I also lowered the caustic sub-d's on the Vray light to speed up testing, but testing at 10K (where he had it) was not different.

I also have noticed that photon mapped caustics seems to knock out GI caustics - is that correct? With PM caustics off you can see some really nice projections of light through the drops from the GI caustics, but those disappear when PM caustics are on.

Thanks for looking at it - hopefully it is just some odd setting.

/b

http://www.simmsimaging.com/upload/f..._Scene.max.zip (87.23 KB)