Subsurface scatering does not indirectly affect diffuse. #111136

New Issue

Khana · 2023-08-15T15:25:08+02:00

Khana commented

2023-08-15 15:25:08 +02:00

Hello,
I've noticed an oddity: materials with SSS are considered totally opaque, but only for the diffuse aspect, but remain visible for glossy surfaces ;

It looks like a limitation, but it doesn't seem to be listed anywhere :

-https://projects.blender.org/blender/blender/commit/0df9b2c715
-https://docs.blender.org/manual/en/latest/render/shader_nodes/shader/sss.html

Materials, such as ivory, jade, marble, rubber or skin, will produce variable results.

If this is a limitation, it's strange that it's only on the diffuse component and not the glossy one.

So it looks like a bug.

Thanks for reading, and sorry about my bad English.

System Information
Operating system: Windows 11
Graphics card: nvidia 2080ti

Blender Version
Broken: 3.6
Worked: none ?

Exact steps for others to reproduce the error

Load SssPasstrought.blend , render

Hello, I've noticed an oddity: materials with SSS are considered totally opaque, but only for the diffuse aspect, but remain visible for glossy surfaces ; ![SSS_anomaly.png](/attachments/87af39be-f733-4ced-b1b2-ec12309cbc01) It looks like a limitation, but it doesn't seem to be listed anywhere : -https://projects.blender.org/blender/blender/commit/0df9b2c715 -https://docs.blender.org/manual/en/latest/render/shader_nodes/shader/sss.html Materials, such as ivory, jade, marble, rubber or skin, will produce variable results. If this is a limitation, it's strange that it's only on the diffuse component and not the glossy one. So it looks like a bug. Thanks for reading, and sorry about my bad English. **System Information** Operating system: Windows 11 Graphics card: nvidia 2080ti **Blender Version** Broken: 3.6 Worked: none ? **Exact steps for others to reproduce the error** Load SssPasstrought.blend , render

SssPasstrought.blend

1.5 MiB

SSS_anomaly.png

1.1 MiB

Khana added the

labels 2023-08-15 15:25:09 +02:00

Evan Wilson commented

2023-08-15 18:34:30 +02:00

Not a full triaging due to time constraints, but noting that shadow calculation is involved, as unchecking Cast Shadow under the light settings will make the diffuse material squares reflect the light that passes through the SSS rectangle.

Not a full triaging due to time constraints, but noting that shadow calculation is involved, as unchecking `Cast Shadow` under the light settings will make the diffuse material squares reflect the light that passes through the SSS rectangle.

YimingWu commented

2023-08-16 04:57:00 +02:00

Can confirm. Being like this since at least 3.3 so it's probably a limitation.

YimingWu added

Module

Render & Cycles

Status

Needs Info from Developers

and removed

Status

Needs Triage

labels 2023-08-16 04:57:14 +02:00

Khana commented

2023-08-16 17:30:47 +02:00

Not a full triaging due to time constraints, but noting that shadow calculation is involved, as unchecking Cast Shadow under the light settings will make the diffuse material squares reflect the light that passes through the SSS rectangle.

I've just tried it, and the result is astonishing:

A modification of the sss'radius shows that the problem persists :

And the glossy samples the SSS well, it's only the diffuse that misses.

> Not a full triaging due to time constraints, but noting that shadow calculation is involved, as unchecking `Cast Shadow` under the light settings will make the diffuse material squares reflect the light that passes through the SSS rectangle. I've just tried it, and the result is astonishing: ![SSS_ano cast shadowlight.png](/attachments/e922d1e2-63df-4831-9a9b-b42426c0cef1) A modification of the sss'radius shows that the problem persists : ![SSS_radius_cast shadowlight.png](/attachments/048306ce-91f6-49fc-b92e-b9f228aa1577) And the glossy samples the SSS well, it's only the diffuse that misses.

SSS_ano cast shadowlight.png

1.2 MiB

SSS_radius_cast shadowlight.png

1.2 MiB

Khana commented

2023-09-15 15:27:47 +02:00

Hello, has anyone had any feedback on this problem?

Lukas Stockner commented

2023-10-05 01:29:22 +02:00

This is due to an optimization that turns Subsurface closures into Diffuse if the path has already encountered a previous diffuse bounce.

The logic behind it is that usually, the indirect diffuse light is not affected by the difference, and it saves a lot of render time. In this case, however, there is a big difference, since all the light comes from scattering through the thin subsurface object.

I'm not really sure whether to consider this a bug or a limitation to be honest, I don't really see a good way to fix it. Eventually (4.1), the Thin Sheet mode on the Principled BSDF should address this kind of use case. @brecht, any opinion?

This is due to an optimization that turns Subsurface closures into Diffuse if the path has already encountered a previous diffuse bounce. The logic behind it is that usually, the indirect diffuse light is not affected by the difference, and it saves a lot of render time. In this case, however, there is a big difference, since all the light comes from scattering through the thin subsurface object. I'm not really sure whether to consider this a bug or a limitation to be honest, I don't really see a good way to fix it. Eventually (4.1), the Thin Sheet mode on the Principled BSDF should address this kind of use case. @brecht, any opinion?

Brecht Van Lommel commented

2023-10-05 15:58:33 +02:00

I would consider it a known limitation. We could have a setting to control this, but thin sheet would indeed be more efficient to render.

Brecht Van Lommel added

and removed

Needs Info from Developers

Type

Report

labels 2023-10-05 15:58:47 +02:00

Khana commented

2023-10-05 18:22:18 +02:00

Hello, thank you for your answers.
I don't know what to think: on the one hand, it's a clever optimization, but on the other, it implies that the object doesn't vary too much in thickness and that it remains thick enough so that the problem isn't noticeable.

How would "thin sheet" be used?
If I have a surface with a large variation in thickness, would I have to use a closure with a painted texture to mix the two?

Hello, thank you for your answers. I don't know what to think: on the one hand, it's a clever optimization, but on the other, it implies that the object doesn't vary too much in thickness and that it remains thick enough so that the problem isn't noticeable. How would "thin sheet" be used? If I have a surface with a large variation in thickness, would I have to use a closure with a painted texture to mix the two?

Lukas Stockner commented

2023-10-06 01:45:30 +02:00

Hello, thank you for your answers.
I don't know what to think: on the one hand, it's a clever optimization, but on the other, it implies that the object doesn't vary too much in thickness and that it remains thick enough so that the problem isn't noticeable.

How would "thin sheet" be used?
If I have a surface with a large variation in thickness, would I have to use a closure with a painted texture to mix the two?

The main issue here is that Subsurface Scattering in general is an approximation to volumetric scattering that is optimized for dense media. Therefore, the assumption when implementing it generally is that it will be used to add some detail to "diffuse-style" objects. The classic use cases (skin, marble, rubber, ...) all match this. Light bleed through the object is supported because it adds some detail, but in the cases for which SSS is intended, it doesn't really contribute much to the overall illumination.

The test scene above, meanwhile, goes in a completely different direction. If you have a more complex object and expect significant amounts of light to pass through it and illuminate whatever is behind it, the best approach probably is to use proper volumetric scattering instead.

The thin sheet mode would essentially behave similar to the current Translucent BSDF: Light passes through it and is diffused directionally, but doesn't scatter to a different location on the object.

> Hello, thank you for your answers. > I don't know what to think: on the one hand, it's a clever optimization, but on the other, it implies that the object doesn't vary too much in thickness and that it remains thick enough so that the problem isn't noticeable. > > How would "thin sheet" be used? > If I have a surface with a large variation in thickness, would I have to use a closure with a painted texture to mix the two? > The main issue here is that Subsurface Scattering in general is an approximation to volumetric scattering that is optimized for dense media. Therefore, the assumption when implementing it generally is that it will be used to add some detail to "diffuse-style" objects. The classic use cases (skin, marble, rubber, ...) all match this. Light bleed through the object is supported because it adds some detail, but in the cases for which SSS is intended, it doesn't really contribute much to the overall illumination. The test scene above, meanwhile, goes in a completely different direction. If you have a more complex object and expect significant amounts of light to pass through it and illuminate whatever is behind it, the best approach probably is to use proper volumetric scattering instead. The thin sheet mode would essentially behave similar to the current Translucent BSDF: Light passes through it and is diffused directionally, but doesn't scatter to a different location on the object.

Khana commented

2023-10-06 14:16:43 +02:00

I understand, thank you for your answers.

One last question:
I've just looked in arnold, there's a type of "thin sheet" mode but the light also passes through in SSS mode !

I wonder, does it automatically switch SSS to "thin sheet" mode depending on the meanfreepath and/or ray length?
Or is it the sampling strategy that makes the difference?

The image below is a conversion in 3dsmax

I understand, thank you for your answers. One last question: I've just looked in arnold, there's a type of "thin sheet" mode but the light also passes through in SSS mode ! I wonder, does it automatically switch SSS to "thin sheet" mode depending on the meanfreepath and/or ray length? Or is it the sampling strategy that makes the difference? The image below is a conversion in 3dsmax

SSS_Arnold.png

140 KiB

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Subsurface scatering does not indirectly affect diffuse. #111136