OTOY Forums

Hello,

There is an unusual artifact with subsurface scattering, please see the attachment renderComparison.png.

On the left side of the image, is a test geometry with an SSS material applied, please see the attachment matSSS.png for the settings.

On the right, is the same test geometry with a transmissive material applied, see the matRefract.png for the settings.

Neither of these two geometries are subdivided, see the attachment wireframe.png for what the topology looks like.

Looking at the left side of the render comparison, an unusual artifact on the mesh with the SSS material can be observed, the wireframe can be somewhat perceived, in a way that is not desirable.

As a comparison, I've sought to replicate the look using a transmissive material using transmission depth, it is close, and no true SSS, but there is no visible wireframe artifact.

Why would there be a wireframe artifact on the SSS material, but not the transmissive material?

Is it possible to eliminate the wireframe artifact on the SSS material?

The smoothness achieved with the transmissive material is desirable for the SSS material. Unfortunately, the noise on the transmissive material takes much longer to resolve than the SSS material and makes it unusable in production.

Using SSS would be preferred, as it resolves more quickly and with less noise.

Subdivision of the mesh does help, but it is not an answer to the SSS artifact, as you have to subdivide a lot to resolve it and even then it is still present upon closer inspection.

Is this SSS artifact known? Is there any setting or technique to resolve this artifact without subdividing the mesh?

Thank you for your time.

Tighe

I don't think it's the wireframe you are seeing, it's the normal angle. I haven't used SSS enough to be fluent with it, but off the top of my head I'd check the scale of your scene to make sure it's reasonable, and check your SSS settings for any extremely high or low values.

Hi, this is most likely as frankmci said. Try normal smoothing and see if that helps.

Hello, thank you both for your response, Normal smoothing unfortunately does not help.

Normal smoothing was on in the render in the original test.

Please see the following comparisons: normalSmoothingOff.png and normalSmoothingOn.png

The SSS artifact is even more present with smoothing off, and is still present with smoothing on.

In another test, I've overdriven the geometric vertex normals to have a cusp angle of 180 degrees, for super smooth normals, and still the issue persists. See geoNormalSmoothed180.png

The artifact is also still present on geometric vertex normals with cusp angles of 0 and 180 degrees. See geoNormalSmoothed0.png and geoNormalSmoothed60.png. Interestingly, the SSS artifact does not become smoother from a cusp angle of 60 to 180, could there be some angle limitation?

The behavior of the transmissive material, the pig head on the right side of the render, does not show this artifact on any of the cusp angles. This would be the desired behavior for SSS for all cusp angles as well.

Could it be that some component of the SSS computation is not smoothing the normals? Or that there is some normal angle limitation?

Thank you for your time and feedback!

Hi tigher, could you please PM/share that test scene so I can investigate. Thanks!

Sure thing, I have sent you the file via PM, thank you

It seems that this has been an issue for years.
viewtopic.php?f=9&t=74760&p=382684&hili ... ng#p382684