PDA

View Full Version : Inverse Falloff + Radiosity problem



Netvudu
10-31-2003, 12:04 PM
Hi

I got this kitchen scene with just an area light. Iīve set up the falloff intensity to Inverse Distance^2 and have enabled Monte Carlo Radiosity and...

the render is wrong! The scene is full of bright white points (very big indeed).

Does anyone know about this?

thanks!

toby
10-31-2003, 08:46 PM
So it renders 'correct' when you use a different type of light?

What are your radiosity settings?

Netvudu
11-01-2003, 08:37 AM
It renders fine with linear falloff keeping the same light type. Itīs an area light, so a linear light gaves the same problem for obvious reasons.

Radiosity setting arenīt causing this, because starting from 3*9, the higher I crank it up, the more points I get. Lots of bright coloured dots around.

Itīs precisely the combination of an inverse falloff and radiosity which makes the problem appear. I think an inverse falloff as opposed to inverse falloff^2 reduces the problem but it will still show its ugly face.

toby
11-01-2003, 12:14 PM
Looks like you might-a found a bug, I'll try it - in the meantime you could replace the area light with a luminous polygon - might even render faster

toby
11-01-2003, 12:29 PM
I'm not getting any white spots - using any special shaders?

btw anyone using radiosity should kbnow this trick:
http://www.hades-studios.com/hdritutorial

Netvudu
11-01-2003, 01:33 PM
nope. No special shaders. Itīs really weird.

In fact it isnīt my own scene, but one from an ex-student of the center I teach at. He came with this one and got us all pretty confused because we canīt think on a reason for the problem.

The initial scale was absolutely huge, and we thought that might be the problem because of radiosity and the falloff value, so we rescaled the whole thing including the falloff... to no avail.

Currently the only existing weird factor still there is that some objects are in .3ds format (I donīt know why the student didnīt convert them in modeller to make them native .lwo) so when I come back to work on Monday Iīll try resaving them as .lwo objects, but somehow I doubt that could be the real origin of the dots.


EDIT: By the way, thanks for the link to the tut. I havenīt tried it yet, but it looks most useful.

WizCraker
11-01-2003, 02:36 PM
can you post a image so we can see what it is doing?

also the Inverse Distance^2 is a little different from linear fall off. The cirlce that you see setting up the fall off is for the hot spot from the light. So The larger the diameter of the fall off [like if it surrounds the entire object] it will be too bright as the fall off diameter would need to be turned way down.. Also since you say you are getting white spots and using 3ds format that could be part of the problem. Might be some stray areaes in the model that needed to be cleaned up that is catching those hot spots from the area light and since they are so small they are magnifiying to bright white.

Just a thought.

toby
11-01-2003, 02:42 PM
An easier way to think about what wizcracker means about falloff is that for inverse distance, the falloff radius should be considered the outside surface of the light 'bulb' - however I tried it both ways and still didn't get hot-spots - He could be right about the 3ds model though

Netvudu
11-01-2003, 03:58 PM
I already knew about how inverse falloff works as opposed to linear. The student also knew it (no wonder, seeing how we taught him! :D ).

I will try resaving as .lwo but as I said i doubt the problem lies there. I will try to posts some pics on Monday, when I come back to work.