3D Shading of Trees

[algerdes]

I have been looking and looking for what is causing this, and have finally come to the conclusion that there is some sort of shading going on in the 3D layout depending on the view.

In the attached screen shot #1, our ground tree seems to be only 1/2 there in the layout.  Notice where the arrow is pointed. This half of the tree is "missing".  The tree is above the grid.  (We even moved it way up above the grid, and nothing changed.)

I did move/rotate the grid so that I was looking directly down on the display (screen shot #2).  When I did this, the shading on the ground tree actually swapped sides and the other megatrees showed as if they were only half there.

It is with this observation that it is believed that there must be some sort of shade being applied to the view.

Can you please confirm this observation?

Also, if this is factual, is there a way to turn it off?

Thanks

[Gilrock]

I don't even know how to create shading so unless someone added something I didn't see I have no idea.  The OpenGL drawing routines do try to calculate which objects are in front of other objects and a tree is not one object every pixel in the tree is an object.  It would be better to show a video so we can see what this does when you zoom in and out and pan around the objects.  Sure there might be special positions you can orient the view where the pixels in front collide with the pixels in behind them in a tree and cause weirdness.

[algerdes]

I finally was able to get some time and use a video capture program (new to me).
The attached file shows the problem.

Note at the beginning how far the whole layout is above the grid.
When I change the view from head on, to straight down, you can see the legs of the ground tree change from bright on the left half to bright on the right.  Also, note the megatrees and arches in the view.  They do something very similar.

I ran just a few seconds of a sequence to show you how it looks with data going to it.  You can clearly see how it affects the playback. 

If you have any ideas, I would love to hear them.  I've reached the limit of what I know.  (Didn't take long, for sure.)

[Gilrock]

Someone else posted something similar on a GitHub issue and I believe Dan already made a code change that alters how the model blending works.  Have you updated to the latest release to see if that fixes the problem?

[algerdes]

Just upgraded.  2020.31  - No change.   

I've been through every setting I can find.  Nothing.
Note that this happens on different computers.   Some connected to my in-house network, some totally stand-a-lone.  Same.

[Ebuechner]

I suggest you make a small test sequence to show the problem and then do a package sequence and post it.
That will give us all the files we need to take a look.

[dkulp]

No, my fix was in the effect rendering.   This is in the OpenGL stuff which I'm really not planning on touching at this point.   Right now, we don't make any attempt to Z-sort the points which is what is likely causing this.   Changing this would require massive changes to the opengl rendering code which is not something I see as worth spending the time on right now.    Keith and I have talked briefly about moving entirely from OpenGL to Vulkan (I keep expecting OpenGL to be removed from OSX at some point, it's been deprecated for years), although I'm more likely to completely replace all the OpenGL code with some other sort of 3D engine that would map to Metal on OSX and DX11 or similar on Windows.   

In any case, I wouldn't expect any changes in this area for this year.   

[algerdes]

Thanks all!
This is not something that is needed right now.  It is a pain sometimes as the legs "on the dark side" sometimes don't show up as well when viewing a sequence.  I'm living with it.