Final Gather and Global Illumination

[SgWRX]

I might be missing something but...

 

I could rarely get a "smooth" GI solution (ie, using photons) enough to not get any noise adding final gather (FG). Until I just tried what I thought was a ridiculous max. sampling radius for the photons, 50ft. I'm doing that with max. number photons per sample of 5000, and avg GI photons per light of 20000. i have two sky portals, mr daylight system lighting a 12x15 room through about a 3ftx8ft window with portal and a 3ftx3ft window with portal (although, no glass just openings in the walls)

 

My FG settings: density 0.2, rays per FG 100, interpolate over 50 FG points. It seems relatively good result. there are no light leaks that I can see. whereas FG alone produces light leaks along the top of the ceiling. and it's pretty devoid of noise.

 

with the addition of AO in the arch&design materials, it seems like i'm getting all the detail i want.

 

like i said, up to now i've never been able to get a smooth enough result from GI to add FG without poor splotchy results.

[SgWRX]

[ATTACH=CONFIG]52842[/ATTACH]

[Francisco Penaloza]

I think the main problem using Mental Ray is understand how Photon Vs FG workflow it is. The latest version of Mental Ray Max introduce IBL that give you other more precise approach to calculate GI but the basics still apply for most of the cases.

This is very long to explain I would strongly recommend to read the manual to get detailed info about it. But in a nut shell I can explain here.

 

GI (Photon mapping, I never understood why the called GI so confusing) Photon mapping is a very old way to calculate GI, not very practical but it has several advantages, one of them is calculate very precise caustics, but is always recommended to be used with FG so you don't require incredible large values to get nice detailed GI.

 

Think of this, each light of photon mapping is a ball of light, that bounce in your room illuminating and collecting color information from different surfaces. When Higher the number, more precise, more data you have. I would recommend to use Average GI photons per light numbers instead using Maximum Number photon per sample. now this number can go easily in to millions, no problems. It is this efficient? well it depend, it is calculated only one time so for animation it can be better to do that, but here is when FG come to the rescue.

The size of the photons is very important, I always recommend to use a size that is not bigger than wall thickness or predominant narrow object, this is to avoid light bleed on surfaces. When smaller the size, more precise GI but more photons you'll need.

When you use large photons you'll think you are getting smooth GI but in reality because interpolation you are averaging a lot of information, so you loose detail on the contact area, AO can solve this but that is not the real answer, The real answer is find the sweet spot of photon size and crank the number until you have a decent coverage, then use Final Gather to smooth the photon mapping and aggregate more details.

So for photon mapping I only use 2 values usually, Average Photons per lights, and Photon size(maximum sampling radius).

 

Final Gather can be used by it self, but when higher values require longer rendering time, sometimes ridiculous rendering time, then is recommended use Photon mapping to speed up things a little.

 

For Final Gather you have 4 main values.

Point of density

Rays

interpolation number

bounces

 

Point of density; is how precise FG is, the higher more precise. default value is too low already.

Rays; this is how many FG ray are shoot, higher number better solution.

Interpolation; well this average the FG rays to speed things up, it should be related with the number, if you make it higher than the rays samples it should smooth things but again it looses precision/ details.

 

Bounces, this are calculated only when you use FG by it self, a good start can be 4 or more, but over 10 does not make much difference really, I think never used more than 6.

 

I did a little tutorial long time ago, you may want to take a look at it, to see with images what I just explained here. Maybe I should update that info, now it seems kind of obsolete

With the introduction of IBL Photon mapping is getting obsolete, the default values of Mental ray are setup to work with IBL, you should concentrate in FG values and sampling before introducing photon mapping.

 

http://www.evermotion.org/tutorials/show/7955/final-gather-tutorial-introduction

 

man this was long explanation sorry.

Fco.

[SgWRX]

thanks. i think the thing i've always run into with FG is that the higher my settings the more noise i get. however, i think that's mainly because of the interpolate value. go too high on that and you loose detail and start to get light leaks. for GI photon mapping, i'm definitely going to take a look at reducing my radius from 50ft to something that might be more reasonable for my scene but also increasing # of photons. i remember the old tutorial that layed out a formula of some sort, something like "half the radius and 4x the number of photons".

 

overall, i tend to like fewer FG bounces, 1 or 2 max. although it depends on the scene and what i need to convey. etc...

 

i'm embarrassed that it's taken me this many years to figure it out, i might just add GI photons more often for interiors.

 

well, let me wrap this up another way, i always followed the tutorials, read the manual etc... and ended up frustrated because it seemed i always ran into a render like you see above taking 20mins+ hahaha. this render took much less time.