hi,
how can i get a glossy transparency in cycles ? i´m lost…what do i have to link into alpha or transmission?
the "Refraction bsdf " is ok, but how to mesh that in my principled Bsdf material?
thanks for help!
thomes
you have to set the roughness and then the transmission weight value… obviously you can use textures to control them.
2 Likes
Sometimes it’s just better to do things without relying on Principled. I may mix in diffuse and translucency with the refraction part, and may go for a PBR oriented fresnel if rough glossy is a thing, but here is what I tend to start with:
2 Likes
thanks a lot !, that easy 
, in any way i always screwed with wrong values…
thanks carl, i tried it out and it looks nearly the same. for me, most visible difference is the dark edge. what exactly differences the “lightpath node” in your material; what means…“if shading is executed for a shadow ray” ?
thanks for help!
blurred refraction.blend (138.1 KB)
The problems I found with your setup:
- Normals on the glass are the wrong way, so you get a fresnel effect similar to Snell’s window.
- The size is humongous (like 0.6m thick).
- Scale are not applied to anything.
- The ground plane is clipping through the glass.
- You’ve set the shadow (Transparency shader) to black making it completely opaque. Should be white (if caustics are disabled), or some bright color for explanatory effect. Read more about the reasoning behind this trick in the manual. Although I personally don’t mix in Reflection ray (or Diffuse Ray) unless tests against ground truth deem it necessary. You don’t need this trick if you rely completely on caustics for light transport; this will be more accurate but also more noisy and thus costly to render.
- I’d use some volume absorption with a green tint for (most common) glass types. To handle total internal reflections in a glass material with volume absorption (to look like this edge), you also need to up your total/reflection/and transmission max bounce count. I recommend minimum 32.
- Indirect clamping was enabled. It’s one of them weird defaults I’ll never understand.
- The world has a color to it, so it makes it harder to judge what is actually going on. You’ll want to examine the response with that single light source to see if things are behaving.
- Reason I just can’t handle the Principled for glass are due to its unified values; in many cases in the real world (not always), rough/translucent glass have sharp reflection and blurry refraction components to it.
Major point here is; if you’re going to audit how materials behave, the scene has to be plausible as well.
Here I have scaled everything down with a factor of 0.01, so the glass is here 6.06mm thick, applied the scales, and fixed a few other things. Note how - with all caustics disabled - the top yellow cube isn’t lit at all through the glass, and both the reflection and refraction is blurry. The lower glass uses a slightly blue tinted Transparency shader (should be white in this case) which allow the floor behind the glass to be tinted blue. You’d set the shade of this so that it matches the end result of ground truth found by using a much higher number of samples, and then combine it with refraction caustics but at a more sensible number of samples. Note the settings I’m using; no caustics, and a high number of relevant bounces. The high blurriness is why it doesn’t look as nice as the edge above, plus the cutting edge would likely be less perfectly smooth as well.
1 Like
hi carl, thanks for looking in. it was a 5 second scene only to check what difference your material makes to the standard principled bsdf.so no eye on world colour or glas material tint.
-normals are correct as i can see with “face orientation”.
-these are the both initial material renders
your material
-principled
-object “apply scales” changes the refraction. with your material ior1.45 it does not look good.
-with IOR value of 1.3 in your material this becomes good again.
-this is the render for the Principled material ior 1.45
-thickness of the glas does only change the strength of the blurryness.this is correct and should be like that
-groundclipping does not change a thing with both materials
-black transparency does not change something in this scene
-with indirect clamping you mean the value of 10 in “light paths/clamping” ?
dont know what it does. keep looking into this. but if i set it to 0 it also does not change a thing.
…maybe with your material you have more screw options, but the overall look in this scene is not that obviously different to the Principled material in my opinion.
That doesn’t make sense. My material used 1.55, but in BOTH refraction and fresnel, these have to match. I examined from many angles and that effect didn’t show up. My material - the one for objects with thickness is the same as regular glass but I switch to Transparency shader for shadow rays. It adds the benefit of having Glossy roughness separate (as well as you can use different colors and normals if you want). Principled (and Glass shader) unifies these inputs to affect both components, making it less flexible.
Other than the error, it’s hard to say it “looks the same”, because you don’t have any objects clearly reflected in front of the glass.
Not on thinwalled glass, no. But on glass with thickness you’ll see the floor in there instead of the bottom internal glass surface which on the inside should have a very strong level of total internal reflections (at that angle anyway).
I didn’t check it, but I suspect relying on caustics for light transport would get clipped with your extremely high powered light source, which I assume stems from the massive scene size.
The main problem is the coupling of roughness. Usually dealing with frosted glass in say a bathroom or office environment, you’d normally want the glossy to remain sharp. I think the only time I’d expect both components to be rough in these circumstances, would be for glass with a visible bump structure to it (instead of material roughness), or when using film and you’re watching it from the applied side. If film is embedded in two sheets of glass, you’d have sharp reflections from both sides.
I’d like to see where you’re at with this setup at this time. So, share again?
Edit: My setup for thick glass is faulty. Obvious in retrospect, but I didn’t spot it. It only deals with shadow rays for the refraction part, meaning that total internal reflections will still obstruct light transport. You have to move it until the end of evaluation, and if you want to have “fresnel shadows”, that have to be faked; either 1-fresnel but accounting for backfaces, or a simpler 1-(layer weight → power 5 → add 0.05) where backfacing fresnel isn’t an issue.
thanks , if i find the time today, i try to understand what you mean with that. tomorrow my vacation begins… 
…maybe you could post a blenderfile to see the difference.