Light bending around objects in Falcon?

[Vorsos]

I have some point lights in clear cases (solid color material; render tag: no shadow, etc) beneath larger geometry.


They look as expected in Cheetah renderer, but Falcon seems to bend light upward around the outsides of the crane arms in center frame. Any idea why?
Hiding the light cases doesn't change anything, so they are working as intended.

 

[Vorsos]

Sigh. I figured it out. The crane arm lights had a softness of 5 for some reason. Nonetheless, that shouldn't make the result physically incorrect...

 

[frank beckmann]

Just a small reminder regarding IES-light sources:

I built a lowpoly cage and put an IES light source inside - grouped&placed:

... gives some interesting lighting methinks.

Cheers
Frank

 

[Martin]

Sigh. I figured it out. The crane arm lights had a softness of 5 for some reason. Nonetheless, that shouldn't make the result physically incorrect...

Do the normals of the mesh possibly point inwards?

 

[frank beckmann]

Maybe wrong smooth angle of the mesh?

Cheers
Frank

 

[Martin]

Maybe wrong smooth angle of the mesh?

Maybe, but I'm not sure if that explains that effect completely. But I might be wrong.

 

[frank beckmann]

Mmmh - I´d say I don´t see anything unexpected:

Cheers
Frank

 

[Martin]

Mmmh - I´d say I don´t see anything unexpected:

Yes, those renderings look absolutely fine.

 

[Vorsos]

All meshes are facing out. All objects' smooth is Constraint and 45°, but thanks for the example—now I know what the angle does! The light cage is interesting too; when determining what areas to focus modeling time on, they cast a long shadow...


Here are some more tests with my point lights. The first render still shows soft lights wrapping around the outside.
I definitely have a mesh issue based on the shadow artifacts with 0 softness lights, which 5 softness lights either don't show or blur enough to hide. The crane arms have an X length of 72 and are pretty simple; adding sections reduces artifacts.

I should probably switch to area lights anyway for the free color-matching geometry and Falcon's free soft calculation. Although area lights have similar trouble with this mesh:



Here, point lights 0.15m away from a relatively simple 8x4x12 box (1 section per side) cast similarly odd shadows. It has a boolean subtract (triangulate) for the door cutout. Adding sections or a subdivision modifier makes the shadow artifacts smaller, but not gone. I could just be ignorant of best practice modeling with regards to lights near large polys.

 

[frank beckmann]

Right now I have more questions than answers:

Don´t judge lighting in 3d view because OpenGL uses per-vertex lighting:

Can you probably copy and paste the problematic areas (just polygons + light sources) into an empty scene/polygon object, compress and upload?
I wonder if your light sources get somehow scaled as they´re part of a stacked hierarchy where transforming will take over.

Cheers
Frank

 

[Vorsos]

Your guesses are better than mine. Examples above are from the 'crane arms' and 'office' objects & their child lights. Crane arms have both my original point lights in poly cases, and area lights for testing.

I'll be disappointed if this model shows up in a Disney/Pixar background. With my lack of formal training, it's Dreamworks quality at best.

 

[frank beckmann]

Can you please lower the Smooth angle of the "Cargo crane/crane arms" to 30° - render - and see if that´s more what you´re after please?

Cheers
Frank

 

[Vorsos]

Can you please lower the Smooth angle of the "Cargo crane/crane arms" to 30° - render - and see if that´s more what you´re after please?

That actually helped, though I'm not sure why. A 30° smooth angle eliminated shadow artifacts on the crane arms and office wall.

 

[frank beckmann]

Because all your boxes have bevelled edges the default 45° Smoothing Angle is wrong - as I guessed.
For sure Martin as a more precise technically explanation why smoothing angles are essential.
From my humble understanding the rays hitting meshes with a wrong (too high) smoothing angle will be diverted:

Cheers
Frank

 

[Martin]

Sigh. I figured it out. The crane arm lights had a softness of 5 for some reason. Nonetheless, that shouldn't make the result physically incorrect...

Ops, I miss-read your original solution to the light bending. Your solution actually fully explains why the light was bending around the object.

I've checked out your scene and the light softness of 5.0 in "Light.14" and "Light.15" is definitely too high. When using soft shadows with point or spot lights the sampling points of the ray origin is displaced as if the light was a disc shaped area light pointing towards the surface shading point. That's why some light samples are visible on the sides of the "crown nest". Reducing the softness value below half the width of the "cron next" should avoid that effect.

 

[frank beckmann]

Nonetheless - pretty much every smoothing angle is set too high:

Cheers
Frank

 

[Vorsos]

Yeah, I'd never really thought about the smooth angle before. Definitely a property to pay attention to going forward. Ironically, I added some bevels because they made lights look better in the OpenGL preview—the exact thing you said to ignore!

When using soft shadows with point or spot lights the sampling points of the ray origin is displaced as if the light was a disc shaped area light pointing towards the surface shading point. That's why some light samples are visible on the sides of the "crown nest". Reducing the softness value below half the width of the "cron next" should avoid that effect.

That makes total sense. The point light must "grow in size" to calculate soft shadows. With the softness value corresponding to meters, it could be useful in special effects, like a single light producing a diffuse glow throughout the insides of a hot engine.

Thanks for the help, Martin and Frank. This kind of support is well worth the software price.