RE: FG calculation falloff/scale bugs?

["Halfdan Ingvarsson" <hingvars(at)Softimage.COM>]

Title: Re: FG calculation falloff/scale bugs?

No biggie.

 

The first thing to realise in CG, especially rendering, that everything is basically a hack that is intended to get an effect quicker than would be possible if natural laws were simulated more exactly (compare and contrast with Maxwell render times, which tries to approach full realism).

 

Photon mapping was developed as a response to the immense render times of brute-force monte-carlo sampling (path tracing) and the gigantic calculation and memory requirements of radiosity approaches. For that, photon mapping has its limitations which people should be aware of and which the mental ray manuals do an extremely poor explanation of.

 

Final gathering is another hack that was invented to simulate the *diffuse* interaction component of photon mapping. In all of your example scenes, you were trying to apply final gathering as a specular reflection effect, which is completely not what it was designed for. It is purely intended as a simulation of interaction between diffuse surfaces, of mostly similar intensity, since the colour is only supposed to change smoothly and gradually between final gather points, and not strongly like your examples require.

 

We'll have a look at enabling surfaces to be photon emitting but I suspect that's done in Max/Maya via geometric lights, which I guess we'll then have to finally relent and implement :)

 

 - ½

 

 

---

From: Tim Leydecker [mailto:BauerOink(at)gmx.de]
Sent: Wed 31-Jan-07 19:00
To: XSI(at)Softimage.COM
Cc: Daniel Rind; Halfdan Ingvarsson
Subject: Re: FG calculation falloff/scale bugs?

So,

first off, thanks alot guys!

In response to Daniel´s mail I must admit I had gotten the
conceptual approach for FG wrong, by expecting FG rays
to be emitted based on the intensity of a surfaces´ color.
In a way thinking of FG as a similar approach to GI/photons
emitted by light...

So, to cover one of the problems that come with the sampling
nature of FG, the filterattribute (for being set above 0) cuts in
to "mask out" the increasing splotchiness that comes with the
distance the FG ray travels. Or as Halfdan put it in another email:

"Welcome to the joys of probabilistic sampling and solid angles :)"

After doing numerous renderings I´d humbly say I don´t like the
brute force of the Filter effect and will most likely just keep it turned off.

The second bit, the effect of a compareably large surface getting
exponentially more wheight in the FG result is also graspable by
Daniel´s and Halfdan´s explanation, I may not like it as it isn´t
"real" but I guess I can´t expect a better (indirect) illumination
model any time soon, so I´ll hereby just politely ask (again)
for a photon emitting material shader for a futur XSI/mR version,
hoping that a "light" emission based on intensity may produce better results.

Until then, I´d like to thank you again, for both the calm explanation
and great insights, not to forget the lead to using a seperate Irradiance node,
here´s a scenefile for XSI 6.0 (_watch for lensshader) Modify freely at will:

http://www.hafenlola.com/downloads/ley_irr_gamma.zip

Cheers

tim