Pathtrace vs PMC caustics question.

[itsallgoode9]

Here are a couple renders side by side and there are no difference between the two kernals whatsoever. In (what i'd consider) a fairly complex specular situation like this, shouldn't I expect to see a different between the two kernals (more caustics going on)? Or when Octane defines PMC purposes as "for complex caustics" does this more mean things like prisms and stuff?

note: I turned spec depth to the max for testing to ensure I wasn't accidentally limiting bounces, so that's why it's crazy high.

[mib2berlin]

Hi, the PMC kernel is better for complex/hard light scenes, one small mesh light for example.
Disable environment light in your scene and add one small mesh emitter to get maximum caustics.
PMC clean up noise much faster then Path.

Cheers, mib

[itsallgoode9]

Ah ok, gotcha. So in this instance, the scene is too simple to benefit from PMC as well as lighting not being setup in a way that PMC would cause a benefit?

So is the end result of path trace and once always the same, it's just that PMC will will clean up the caustics better/faster when you have more complex scene and caustics involved? Or in those more complex caustic scenes, will path trace just completely be missing the more complex stuff caustics that PMC will show?

[calus]

So is the end result of path trace and once always the same, it's just that PMC will will clean up the caustics better/faster when you have more complex scene and caustics involved?

Yes, both converge to the same result (unbiased).
For some extreme case, it's amost imposssible for path tracing to catch some specific rays but "almost" is the keyword here, meaning that with enough time it will converge eventualy to the same result as pmc.

[itsallgoode9]

ok, that's good to know!

I guess after all these years I never fully understood what PMC was for. I'd always used pathtrace for my bottle renders but keep running into issues getting lighting into the punt of these bottles (think, that big bulge inside the bottom of a wine bottle) and I was looking to see if PMC would fix this by calculating better caustics.

I know this is a whole other discussion but i'm convinced that unbiased renders just have a hard time (if not impossible) correctly calculating the light rays in situations like the punt of a glass bottle. I've done tests where I put a wine bottle in a full white environment light, not sitting on a floor or anything, and that base area of the bottle still shows up fully black in areas from certain angles. That samples I posted for example, the surface underneath the base of the bottle is much too dark compared to how it should look in reality.

[whersmy]

itsallgoode9, do you have a reference example of the wine punt, like you would want it to have?
The images you made already look pretty solid.
But yes, just an hdri won`t do the job if you are looking for way more defined caustics. Adding an (hidden) emitter does the job.
And rendertimes go high on pmc....this one took around 5 hours on 2x580 with RR enabled. Though I used an .ies for this one, so it`s still noisy.

[itsallgoode9]

well maybe you prove me wrong :-p

The specific issue I tend to have problem (which is a bit different than your image) with is when light has to enter, exit, reenter, then exit the same liquid/glass volumes...sort of a rounded "W" shape. Here's the example below:



I was actually having some extremely extremely odd results this weekend regarding this but I need to do some more testing on other scenes and make sure I didn't just do something wrong in the setup on my specific scene.