AWS Thinkbox Discussion Forums

Lights ignore attenuation

Beta 17



Krakatoa seems to ignore the near attenuation of lights. I’m trying to get a point light to look like an area light, so I want the light to cut in 100 units away from the source point. But Krakatoa seems to be ignoring that parameter, and I get light from the point out.


  • Chad

Beta 17



Krakatoa seems to ignore the

near attenuation of lights.

I’m trying to get a point

light to look like an area

light, so I want the light to

cut in 100 units away from the

source point. But Krakatoa

seems to be ignoring that

parameter, and I get light

from the point out.







Logged as defect 4330.

Don’t hold your breath for 1.0 ;o)

No more than I would for area lights.



I AM seeing some nasty lighting bugs over here though. Trying to make a clean sample file now.


  • Chad

Got it.



In this file (max 9 x64) you will see that there appears to be a problem with the edges of shadows on lights. Basically, it seems as though light is “leaking” out the edges where the particle is within the cone of the light, but is not being properly shadowed, so particles near it, which are either in shadow or outside the cone are much darker. This file shows an extreme case, but I’ve seen it show up in very subtle ways in other scenes.


  • Chad

bump



Just wanted to make sure this one didn’t slip through the cracks.


  • Chad

bump



Just wanted to make sure this

one didn’t slip through the

cracks.





Thanks for bumping it, we could repro this - looks like the edge of the attenuation map is missing particles, while the light is still lighting them. We will try to add some padding to make sure no particles slip through the (spotlight’s) cracks. ;o)



Btw, using lower shadow map settings like 128 makes the problem even more visible as more particles fall into that “gray” area of no attenuation.



Darcy is working on a solution right now.

This was fixed for the next build.



It was a problem with performing biliear/bicubic filtering at the edge where half of the data does not even exist. At very low shadow map sizes, the gap would be rather big.

Privacy | Site terms | Cookie preferences