Some glass panels display black instead of see through in MSFS

[jgarnold]

I have created a balustrade consisting of vertical posts, top and bottom rails using mesh cube and a glass panel using mesh panel.
I have applied the same material to each panel set as MSFS Glass yet some panels display see through in MSFS and others black.
Any ideas what could be causing this?
The material I applied to each is MSFS Glass with a light shade of blue color, the alpha channel set to 0.3, metallic factor set to 1.0, roughness factor set to 0, Glass reflection factor set to 1.

As I used a mesh plane for the glass panels I also tried setting 'double sided' on but that made ALL of them display black in MSFS.

Could the use of the modifier array cause problems?

 

[mamu]

I have created a balustrade consisting of vertical posts, top and bottom rails using mesh cube and a glass panel using mesh panel.
I have applied the same material to each panel set as MSFS Glass yet some panels display see through in MSFS and others black.
Any ideas what could be causing this?
The material I applied to each is MSFS Glass with a light shade of blue color, the alpha channel set to 0.3, metallic factor set to 1.0, roughness factor set to 0, Glass reflection factor set to 1.

As I used a mesh plane for the glass panels I also tried setting 'double sided' on but that made ALL of them display black in MSFS.

Could the use of the modifier array cause problems?
View attachment 86019

View attachment 86020

Hi John,
Full metallic is a recipe for a mirror,
I would use zero metallic ,
0.02 to 0.1 alpha

And 0.02 to 0.2 max roughness

And a subtle touch of colour, very very slightly darker than full white

I would also avoid double sided, better duplicate the face and rotate 180,/invert normals

The array should not affect the glass shading

 

[jgarnold]

Hello mamu

Thanks for the advice. I just thought that the higher the value for metallic the greater the reflections not even considering a mirror - now I know better.
Alpha for transparency so a low value for higher transparency?
And roughness to control translucency?
Yes I have now set the colour (slight blue) at R 0.8, G 0.8 B 1.0. The problem I have here is that in Blender the colors don't display as 'strongly' as in MSFS.

Also I have decided to use a mesh 'cube' instead of a plane for the glass panels in an effort to try and fix my problem. I only started using a mesh plane as there are less 'faces' but I am silly as the entire model is only around 1600 faces so not worth the trouble.

 

[=rk=]

I just thought that the higher the value for metallic the greater the reflections not even considering a mirror - now I know better.
Alpha for transparency so a low value for higher transparency?
And roughness to control translucency?

I think of "translucency" to be the level of cloudiness, like plastic or stained glass and that's controlled with color/alpha. Roughness affects the gloss "shine." You'll need some level to identify it as glass, full roughness would be like sandblasted glass without the blur, in which case a dirt smudged blended texture might be even better.
Metallic provides the glare, the sun flash and I always set metallic very high, although above 90% whiteness it begins to become mirror, as mamu suggests. For roughness I go very low, full black or zero for new glass, some slight level of roughness for weathered glass - and I always use single layer normals facing outward, because I encounter too many issues with multiple layers of transparency, again, mamu is correct on this one. MSFS seems very forgiving in terms of orientation, compared to FSX/P3D, still unmatched faces show through reflections, shadows, etc.

 

[rhumbaflappy]

Also I have decided to use a mesh 'cube' instead of a plane for the glass panels in an effort to try and fix my problem. I only started using a mesh plane as there are less 'faces' but I am silly as the entire model is only around 1600 faces so not worth the trouble.

Perfect. A plane is only 2 dimensional. When it's seen from the 'other' side, it simply doesn't exist. A thin cube should solve the problem. The normals all point outwardly. With a plane, the normals all point just one way. As Mamu wrote, you could copy and rotate the plane and join the 2 to get outwardly pointing normals. Easier to just use the cube mesh.

 

[=rk=]

Perfect. A plane is only 2 dimensional. When it's seen from the 'other' side, it simply doesn't exist.

Not entirely true, MSFS supports backside materials. The model below was created in Sketchup. The railing visible in the lower right corner is a default Sketchup "fence" material, applied to the "front," white, or normal facing outward side and we are looking through the back, or blue side of the single sided polygon(s).

Edit: Thinking back, the model may have texture applied to the back side, I forget, the result is the same, back faces "clearly" are visible. I do know that if a different transparent texture is applied to both sides, each side shows only the transparency that is applied to it.

 

[jgarnold]

Thanks Rick and Dick for your input and opinions. I remember some time ago having some 'problem' using a plane because the rear face didn't 'show' (or something similar) and it was suggested I 'turn on' double sided which fixed that problem but it didn't seem to fix the problem raised here, As mentioned earlier I only used a plane to reduce the number of faces (from that of a thin cube) as I have 14 glass panels as I am always trying to reduce faces to a minimum but it was a bit silly as I only reduced faces by around 70 (14 x 5).
I will (later) adjust the MSFS glass parameters as suggested. And yes (as Flying Theston points out) the alpha channel controls translucency not roughness.

 

[rhumbaflappy]

Here's 2 pictures showing the differences between a plane with a texture applied, a plane with a double-sided material, and 2 single-sided material planes with one rotated around (separated slightly and the 2 planes joined).

Front view.. all is good:

Rear view. The nearest object is the joined 2-plane object, each with a texture. The middle is single-sided material plane. The far is a plane with a double-sided material applied... not quite right (probably an issue with the normals).

 

[jgarnold]

Thanks Dick
I decided to replace the mesh planes with mesh cubes for the glass panels and set material as MSFS Glass, Alpha channel 0.2, base color R 0.8, G 0.8, B 0.8, metallic factor 0.69, roughness 0 and I am happy with the result.

 

[mamu]

Here's 2 pictures showing the differences between a plane with a texture applied, a plane with a double-sided material, and 2 single-sided material planes with one rotated around (separated slightly and the 2 planes joined).

Front view.. all is good:
View attachment 86032

Rear view. The nearest object is the joined 2-plane object, each with a texture. The middle is single-sided material plane. The far is a plane with a double-sided material applied... not quite right (probably an issue with the normals).
View attachment 86033

Thanks Dick for taking the time to explain why the double sided materials should be avoided,
The light shading is apllied to the geometry normal,
The only real face that model (double sided material) has,
is pointing to the sunlight, so the engine is apllying the same shading information (the only he knows) to the "false" face too!

A hint about double sided: the blender Asobo exporter is setting double sided materials by default

 

[jgarnold]

I for one will (try) not to use mesh planes again.

A hint about double sided: the blender Asobo exporter is setting double sided materials by default

Interesting.

 

[=rk=]

Here's 2 pictures showing the differences between a plane with a texture applied, a plane with a double-sided material, and 2 single-sided material planes with one rotated around (separated slightly and the 2 planes joined).

Front view.. all is good:


Rear view. The nearest object is the joined 2-plane object, each with a texture. The middle is single-sided material plane. The far is a plane with a double-sided material applied... not quite right (probably an issue with the normals).
View attachment 86033

Thank you for this demonstration, although I'd point out, you are using opaque materials to clarify a point about transparency and I'd like to share my own interpretation to consider.

The nearest panel, is how we would want a sheet of non transparent material to stand in the sun, the farthest, is how we would want a translucent panel, like a sailboat sail, or a plastic roof skylight, to show sunlight from the back side. I'm not claiming the visual effect is intentional, however, that is not a transparent layer, as are the glass panels.

The only real face that model (double sided material) has,
is pointing to the sunlight, so the engine is apllying the same shading information (the only he knows) to the "false" face too!

Therefore, if the layer is transparent, this shading information is not seen, only shading/glare on an applied texture would be seen. Material color, tinting, would be unchanged. So in this case, a clear polygon with backside material, would look, well, like glass and pretty much exactly like two front side polygons set back to back. In fact, you could probably spend some time trying to find circumstances where the difference mattered.

We are talking about scenery models and if it assists the immersion to add two polygons per pane to remove invisible sun glare from the back side texture, I say most definitely, go for it. If you are suggesting that scenery models should have double sided glass so that windshield wiper textures are properly rendered from the inside, you are probably correct.

 

[rhumbaflappy]

Using alpha materials... Front:

and back:

Using single-sided material on a thin cube works too.

 

[=rk=]

I feel like we are playing a duet, or doing stand up.

Here is my double sided materials model:

I did my best to give the cloudy model the greatest advantage, by looking almost directly into sunlight. I hope you will agree that there is no glare cloudiness and it is my belief that you will find that the greatest difference between your model and mine, is that I have set roughness to zero, following my tried and true formula of .8 metallic and 0 roughness, which I feel gives those dramatic sun flashes.

In fact, you could probably spend some time trying to find circumstances where the difference mattered.

The cube model has the right side that is facing the view removed. Double sided materials have more clarity, than do two front sided polygons, back to back and we can see that comparing the two images. Looking at the trees of the double front side image, there is a cloudiness that is not present in the grass of the cube image.

 

[rhumbaflappy]

We were discussing a plane, not a 3D object. The normals on a cube (or flipped and joined planes) will look good, as we are on the outside... so in that case it really doesn't matter is you are using double-sided materials.
Your tint box example is single-sided. I converted it to double-sided, but as I suspected, the view of the cubes makes no difference, as it is a 3D object, and we're viewing it from the outside.
I think Jon's solution was perfect. Make panes of glass a thin 3D cube mesh and don't worry about single-sided or double-sided materials.

 

[rhumbaflappy]

https://www.fsdeveloper.com/forum/t...he-trees-impossible-to-fix.456589/post-911469 shows an example of cross-plane trees using double-sided materials. It just doesn't work right.

 

[=rk=]

We were discussing a plane, not a 3D object. The normals on a cube (or flipped and joined planes) will look good, as we are on the outside... so in that case it really doesn't matter is you are using double-sided materials.
Your tint box example is single-sided. I converted it to double-sided, but as I suspected, the view of the cubes makes no difference, as it is a 3D object, and we're viewing it from the outside.
I think Jon's solution was perfect. Make panes of glass a thin 3D cube mesh and don't worry about single-sided or double-sided materials.

I feel like you are spoofing me. The distinction about a plane is arbitrary, or maybe I did not state clearly enough

The cube model has the right side that is facing the view removed.

This means that we are looking through the cube, unobstructed, at the inside backside.

If it is not double sided, fine, the tint still shows without glare and it is still clearer than “double pane.”

This afternoon I will post an image of a single, naked polygon, both textured and tinted, that is absent cloudiness, glare, etc. if by then you do not agree that my formula works, I’ll conclude you have your mind made up, at least all this is available for inquiring minds to explore. I mean, you already had that model in your editor, what else can I do.

 

[jgarnold]

Make panes of glass a thin 3D cube mesh and don't worry about single-sided or double-sided materials.

Yes that is what I will do in the future.
As mentioned before as I am ever conscious of keeping models as low poly as possible I initially chose to use mesh plane for my glass panels to save some 'faces' but 'looking outside the box' (as the saying goes and not a pun ) I would save very few 'faces'. When imitating glass it is MUCH easier using a cube IMHO.
I will continue to use mesh pane but only for objects such as signs which are always opaque or with an image of a real sign uv wrapped onto it.

All this had indeed created some 'lively' discussion and we can only be the better for it (learning - I certainly have learned from it).

 

[rhumbaflappy]

Your tint box has faces both inside and out. 10 faces, 20 triangles. It's not the same as a Blender panel, such as Jon displayed for his fence. A blender panel is a single face, and 'points' only one way. That was the original problem.

Your tint box has side faces both inside and out, unlike Jon's original glass panel that has only one face (impossible in the real world), which he made in Blender, using a double-sided material. Jon then made a thin box in Blender, applied a material and fixed the problem, as the glass now has enough faces pointing outward. What I'm saying, is the sim doesn't handle double sided materials as we were used to them in FSX or P3D, and should be avoided. Cross-plane trees will look weird, for example. Better to make them cross-box trees, or 2 panels facing outwardly and joined, or solidify the plane... then it doesn't really matter if you use single-side or double-sided materials.
Here's the better way to solidify a plane in Blender:

 

[GaryGB]

Interesting complexities (but a bit mind-numbing as well); reminiscent of issues previously mentioned in this thread:

https://www.fsdeveloper.com/forum/threads/msfs-glass-problems.450326/


KatsBits offers some interesting explanations in this tutorial:

https://www.katsbits.com/codex/double-sided-materials/


However, I'll have to do some more studying to discern what KatsBits meant by this statement in that same tutorial:

"Ordinarily to work around this problem objects would need to be hard or physically duplicated and set up to use their own unique properties, materials and UV’s before being inverted (Mesh » Normals » Flip) to produce a model comprising two meshes, one with surfaces pointing outwards, the other with surfaces pointing inwards."

Hard ? ...Like "Hard to understand" ?

UPDATE:

OK, It took some searching, but apparently "Hard" in 'Blender-ese' ...refers to object (sur-)faces that do not (yet ?) use SMOOTHING:

https://www.katsbits.com/tutorials/blender/character-2-mesh-editing-making.php

https://www.makeuseof.com/what-is-hard-surface-modeling/


Oops ...almost forgot to clarify that in Blender "Solidify" is distinct from making a surface "Hard".


Apparently even a shallow extrusion can make a Face a "Manifold" (air/water-tight) 3D object.

IIUC, the "Manifold" attribute can change rendering display attributes for Z-sorting, Light transmission and Shadowing.


But, I wonder if "Solidify" actually will change all the above display attributes for a (sur-)face that is subdivided ?

AFAIK, an extruded (sur-)face that is subdivided, will not be truly "Manifold".

And, IIUC, if SMOOTHING is used, the (sur-)face is subdivided, Edge lines are hidden, and the object is no longer "Hard"


https://docs.blender.org/manual/en/latest/modeling/meshes/editing/face/solidify_faces.html

"Solidify Faces
Reference
Mode: Edit Mode
Menu: Face ‣ Solidify Faces

This takes a selection of faces and solidifies them by extruding them uniformly to give volume to a Non-manifold surface. This is also available as a Modifier. After using the tool, you can set the offset distance in the Adjust Last Operation panel."


https://www.google.com/search?q=Ble...gAQLIARTAAQHaAQYIARABGAg&sclient=gws-wiz-serp


PS: Some pertinent sections of the MSFS SDK Docs on these subjects:

https://docs.flightsimulator.com/ht...ax_Plugin/Materials.htm?rhhlterm=double-sided

"The following is also worth noting:

• double-sided materials do not modify the normals when viewing the back-face of materials (i.e. the normals will always be point outwards)
• texture sampler information is ignored (wrapping method is always CLAMP_TO_EDGE)
• animation sampler interpolation method is ignored (interpolation method is always LINEAR)."

https://docs.flightsimulator.com/ht.../General_Principles.htm?rhhlterm=double-sided

"Double Sided

When checked, back-face culling is disabled and double-sided lighting is enabled.

Be aware that the normal of the back-side face is not flipped.

If correct lighting is required, use double-sided geometry instead."


BTW: Dick has previously made at least 2 downloadable worked examples for 'transparency' in the past for FS2Kx.

It would be great to see his example above, as yet another downloadable worked example for 'transparency' in MSFS.

GaryGB