Caustics issue - cannot find workable settings...

[simmsimaging]

Here is a set of tests to show the problems I'm having getting something workable with photon mapped caustics. Droplets are too scale, about 3-5mm wide and use a pretty basic vraymaterial with affect shadows off and fully refractive, no fog, no dispersion.
Light source is a vraylight with 3000 caustic sub-d's for the first 3 tests (10K for last one). (I also tried with a Max spotlight with basically the same problem/results)

If the search distance is low enough to get sharp results (see test 3 or 4 in the jpg) I need to put in some very high caustics sub d's to start to get anything reasonable. The final test (test 4) raised the light caustic sub-d's from 3000 to 10,000 and it's a bit better, but not wildly so. It still looks poor and needed more than 5X the render time.

If the search distance is higher to smooth things out then I start to lose any caustics *inside* my refractive surfaces (see test 1 in the jpg). That seems really odd to me?

I cannot find a happy balance after yet another long round of testing and messing around. Is there something amiss here, or is it just a setting issue?


Also: the caustics render element does not show any indirectly visible caustics still. Is that a bug or just not part of the element?

/b

[super gnu]

im guessing you have a big vray light over the whole scene by the look of things? ive always found it difficult to get good caustics from vray lights, partly because the photons are spread over a large area, and partly because its an area light, so will naturally cast very soft diffuse caustics..

the key for me is using carefully placed spotlights with tightly focused cones to constrain the effect exactly where you need it.. i believe you can turn off "affect diffuse" and use a different light for your lighting, and the spots just for caustics.


im waiting for the day when vlado announces an amazing new caustics system in vray thats as effective and easy to use as the rest. it currently stands out like a sore thumb compared to the rest of vray, being a feature people avoid using simply because its so tricky to get a good result from.

[simmsimaging]

I did try a spotlight too and the result is basically identical. It just takes longer with the bigger light from what i could see, but in any case the was no real difference in the problems.

I guess caustics just isnt a high priority item - they have to focus on the areas where demand is highest. Hopefully the upcoming bi-directional system will add something to this, but in the meantime i still need to get this method working better if i can.....

/b

[jacksc02]

Brett, is GI enabled?

[simmsimaging]

Hey Chris - Yes, it's BF/LC

[Paul Oblomov]

Well, here's the scene, you may find useful.

[simmsimaging]

Thanks Paul - I am away for a couple of days but will look at it when I return. Much appreciated.

[simmsimaging]

Thanks for the scene Paul - it did help clear up some things.

1) scale seems to be a factor for sure. I made the drops "real" sized and adjusted the caustics settings down in scale appropriately, but at the larger size of your scene they just seem to be better.

2) Vraysun appears to be a far better caustic generator than regular lights. I tried it with a Max spotlight previously and did comparison here. It not only creates far weaker caustics for a light of about equal intensity, they are also different looking in other ways. See the test image below - in order to get even close I had to up the caustic multiplier on the spotlight to 300 and use 2X the subdivisions (6000 instead of 3000). Even though the light is in exactly the same spot you can see it's very different in effect.

3) Interpenetration of refractive materials and non-transparent objects causes some odd changes. Some shapes lose caustics entirely (the droplet mesh). The sphere is a half-dome made and refracts kinda oddly. The cylinder still refracts when penetrating, but differently. Refections are also different I think.

Why do the droplets seem to lose caustics entirely though? Makes caustics useless for this kind of thing, which is quite a drag - but does explain why I have had so much trouble!

/b

[simmsimaging]

just bumping this one - still trying to figure out if this is a bug or expected behaviour with caustics?

b

[simmsimaging]

A long while later and I am bumping this again. I am still not seeing indirect caustics in the render element, and there is inconsistent results with penetrating/not penetrating geo. Can you shed any light on this Vlado? (no pun intended)

b

[vlado]

2) I tried it with a Max spotlight previously and did comparison here.

You really need inverse square falloff that starts at 1 system unit away from the light. Otherwise spot and point lights will not generate caustics with correct intensity.

3) Interpenetration of refractive materials and non-transparent objects causes some odd changes.

It is not odd at all; it is the same as having liquid in a glass that has air between the liquid surface and the glass. For correct results, you need the penetration.

There is also a way to generate caustics progressively by setting the "caustics_showCalcPhase" option of V-Ray to "true" from MaxScript.

Best regards,
Vlado

[simmsimaging]

Thanks Vlado
I have the falloff on, but might have it set to more than one system unit. I'm still not seeing caustics inside of the objects though - at least not in the Caustic Render Element. Can you confirm that it does not support indirect caustics, or is something else amiss?
b

[vlado]

I'm still not seeing caustics inside of the objects though - at least not in the Caustic Render Element. Can you confirm that it does not support indirect caustics, or is something else amiss?

You need to let the glass material transmit render elements (i.e. set the "Affect channels" option to "All channels").

Best regards,
Vlado

[simmsimaging]

That doesn't appear to affect it. The caustics can be seen in the RGB but not in the caustic element, and for sure it's set to affect all channels. I'm fairly sure I tried that before too.

[vlado]

Worked for me on a simple test just a moment ago - if you have a case where it doesn't, can you post it here?

Best regards,
Vlado