[seghier]

using backface material with nd = 1 is not a final solution ; maybe it can improved or not

[seghier]

comparaison between renders with " hdr and a plane " in the background
i notice that when quality in fire increased to (>5) studio crashed

[Eric Nixon 20170610100005]

Seghier these tests are a bit simple to judge properly, but it does look problematic, I doesn't seem logical either so yeah I think its just breaking the engine.

I found a good enough workaround for headlights without uv mapping;

8 WHEELER 2017 flexible v2h.png

No need to adjust the normal-map file, uses cubic mapping with no tiling, offset 50 % and moved mapping across so that the rear of the object receives the map, front is blank.

[Eric Nixon 20170610100005]

Fernandos suggestion works Its a little bit fiddly - C4D shell modifier is a bit quirky with sharp corners, and I manually edited to avoid coplanar faces - does that matter when using nested?

2.png

[Eric Nixon 20170610100005]

I had a question earler about coplanar faces vs slight intersections, It seems that coplanar faces are ok with nested dielectric (it seems to render the average value), which is very helpful . The chevron in this render is made of instanced 'lenses' with are coplanar, also the dust on the outside of windscreen, and the bump on the inside of the headlights, are using the cubic offset mapping trick.

kj.png

Here is a thread about the retroreflector geometry technique; viewtopic.php?f=48&t=44874

We really need a better solution for mapping onto transparent surfaces, these are really horrible with plenty of situations where they wont work.. Cant rely on UV mapping as an option either.

Maybe someone has a bright idea? how about depth clipping in Z for flat-mapping... hey thats not bad.

[Nasok]

In fact I think we should rely on UVs as selecting polys and assigning a different material - will and should break things up .. if it wasn't doing it before than .. well it was wrong ..

Especially for glass objects .. as you know you, for proper glass you gotta have both walls .. front and back .. and selecting polys on front wall and changing material - confuses the system as it turns out that when light ray enters it receives one information and when it exits it receives different one.

So instead off selecting custom polys (which i also do sometimes) use UVs. as poly selection works fine only on straight lines or loops. you can map textures to pretty much any parameter that makes you alter the material the way you want .. that includes normal mapping, nd, k, attenuation, colour and etc. ..

Ups are made specifically for that - so I think engine itself behaves as it should.

[Eric Nixon 20170610100005]

Nasok, whether we use UV's or not doesn't affect the issue of rays starting with one value and then recieving another. Which is something maxwell can cope with quite well anyhow.

The point about UV's is that its often not an option to use them, due to the nature of the geometry, which may have been spat out of a cad program with the resulting wacky topology.

Maybe someone has a bright idea? how about depth clipping in Z for flat-mapping... hey thats not bad.

A simpler way might be, to set the mapping to front faces only? NL is that more doable?

[Nasok]

Well, CAD systems produce weird UVs generally unless we're talking about pure planar surfaces, because when you cut, stretch, mirror and boolean in CAD you don't think about UVs at all that's CAD

But that is natural for all transparent and "see through" objects to give a visual feedback based on a back wall. without it you can't estimate how thick is the surface - for starters. So you can't really calculate the IOR, attenuation distance and so on. break up in spec and normal maps - are the least important things here. That technique would work fine for an opaque objects. Where light ray bounces of the surface and provides a visual feedback.

T.

[Eric Nixon 20170610100005]

Just realized that front faces only wouldn't work if you see the object through itself (like a wine glass), so forget that idea. Is the depth limit a possiblilty?

[Eric Nixon 20170610100005]

I wanted to raise/bump this issue again because I have a scene which uses voroni fracturing.

If for example I fracture a transparent/sss vase object, I would like to add roughness to the broken side walls of the vase. There is no way to this, I have to use poly-selctions in maxwell2

NL THIS NEEDS TO BE FIXED ASAP.

[Eric Nixon 20170610100005]

This relates to Seghiers question about decal on glass;

Thin-sss glossy stickers on the inside of glass surfaces- 'window stickers';

The decal channel is mapped only to the inside of the glass surface (using offset non-tiling cubic mapping)

The thin-sss decal here is 1 normal solid layer + 1 additive layer with thin-sss bsdf + gloss bsdf.



I think this technique maybe ILLEGAL because it fails if the sticker is on the outside of the glass relative to the dominant lightsource, here is fail render with light dircetly behind the yellow+pink sticker, on slightly green glass, looks like infinite refraction from that area of the glass, oh well

[Eric Nixon 20170610100005]

Oops, I forgot that the car windows were hidden to GI, here is corrected render (low sl).

Still think the stickers shadow is the wrong colour (green rather than yellow) so I do suspect there is infinite refraction, but its not a visual concern, so this still appears to be the best way to get see-through stickers on glass. Maybe modelling the sticker is sensible sometimes, seems like hassle though.

[Forum Moderator]

@seguier
So, you when you have two materials, you duplicate the material of the second dielectric (the one applied to the triangle selection), change Nd to 1 and apply this material as backface of the whole object?

[seghier]

@seguier
So, you when you have two materials, you duplicate the material of the second dielectric (the one applied to the triangle selection), change Nd to 1 and apply this material as backface of the whole object?

i don't remember ; i will search if the scene is still in my disk

[PA3K]

I came to the problem with selecting part of surface with different dielectric material few years ago, when i tried to simulate glass plate with rough edges illuminated by small (LED) emitters. This is usual situation these days in furniture designs and also used in car industry to illuminate hands on instrument panels or speedometer. Rough surface is such a delicate (0,00x mm), that it can not be modeled as a geometry with depth. I tried some workarounds back then (1. selecting just edge with rough glass, 2. duplicate edge and offset inside with inverted normals, 3. same as 2. without inverted normals, 4. extruded edge (0,001 mm) to make it closed volume with rough glass, and glass plate (closed volume) with zero roughness, 5. SSS material on edge, 6. SSSS material on edge, and after all not very good and real results i made 7. emitter on extruded edge ).
I had reference in the front of me and also on photos, but I really can not replicate it. Here is small rendered image. If someone have some good solution ... will be welcome .
Patrik