PDA

View Full Version : Light Beam vs IOR of Glass



MacGregg
10-07-2006, 02:07 AM
Can anyone explain the attached jpg.

It's a solid glass block with a beam of light shinning down on it with Volume effects on. There is an IOR set to, I think, 1.6. You will notice that the beam of light goes straight through the glass without being bent BUT it comes out of the block is "as if" it were bent because there is a bright spot on the table near the bottom of the block. So, the light comes out of the block where it should, but the beam in the Glass Block is NOT showing the right bent path.

This seems like a bug. Volume effects apperently are not affected by the IOR of Glass.

Any thoughts?

Thanks,
Gregg

Lightwolf
10-07-2006, 02:35 AM
This seems like a bug. Volume effects apperently are not affected by the IOR of Glass.

I'd see this as a limit of the current design ;)
Volumetric lights are a fake in this regard, they are defined as a cone centered at the light (in the case of a spot), emitting away from it. Nothing outside of the volume of the cone is considered as being "volumetric". The cone itself is not bent by refraction, since the cone is defined before any rendering takes place.
Basically, the cone is the container for the volumetric effect.

In your case, imho the only way around it would be a solid volume of "fog" (I.e. ground fog, which is _slow_) which should pick up rays bent due to refraction.

Cheers,
Mike

MacGregg
10-07-2006, 05:33 AM
Yea Lightwolf you're probably right, it just surprised me that all these simulations of physics seem to be separate. Just appears that they are making for more work separating physical phenomenon into different algorithms. Obviously, something "knows" the path of light thru the Glass Block. Doesn't seem like much to tell Volume effects what that path is. Or do it right in the first place and just create a complete model of all the Physics of light in one central place. Then you wont have to compute a refraction effect, or whatever, for each LW feature separately. Like nothing here "knew" that "white light" should make a rainbow when bent thru Glass.

Thanks again Lightwolf,
Gregg

Lightwolf
10-07-2006, 05:47 AM
Yea Lightwolf you're probably right, it just surprised me that all these simulations of physics seem to be separate.
It doesn't surprise me ;)
Look at Maxwell, which simulates light more correctly than any other package out there... and look at the render times involved. I'm not even sure if it does volumetrics at all.
As I mentioned, volume lights are a "hack" as much as just about any other part of the renderer... to properly render volume lighting as in your example, the complete scene would need to be treated as being filled with a volume of "fog" (or whatever, small particles hanging in the air that react to light) ... you can imagine the performance hit due to the work involved ;)

Cheers,
Mike