Physics issue, abnormal land impact.

[Gwyddion]

Hi, I am just uploading a simple arch, (mesh.png) which I built in Blender, which I have done many times before, although with different meshes.

I would expect with the simple shape and simpler physics I am using for this to be a maximum of 3 or 4 prims, but more likely to be 1. 

 

 

Instead it is 9,000+!! Physics shape in phy2 image, with a double skin to prevent avatar walking through either side of the thick wall. I tried rotating one side and it's the same.

see 9,000+ in the edit window when I switch to prim shape in physics.

I have found a workaround, which brings it back to 1 LI but it is not the shape I need exactly.  

Any help appreciated.

This image shows the workaround physics shape, which is only 1 prim after upload.

 

 

Edited August 26, 2021 by Gwyddion
organising pictures

[ChinRey]

A wall with a doorway/archway is one of the few occasions when you need analyzed physics. One cube on either side of the opening is all you need.

The reason is that Havok doesn't like triangle /that is unanaluyzed) physics for very thin objects. There's a lto of posts about it already but I supose there's no need to go too much in detail. Just remember:

• One cube on either side of the opening
• Analyze physics

[Gwyddion]

Great thank you so much ChinRey, I will remember that 

[Wulfie Reanimator]

I believe your specific problem in the first image is the use of flat planes with gaps between them.

You're not creating a simple (and solid) hull that might be easy for physics.

You've created four distinct physical planes that allow other physical objects to pass between them. I imagine there are some serious performance implications about that.

Here's a similar object, 10x20 meters in size, 0.6 LI with only 0.36 physics (I think that's the absolute minimum physics cost you can have).

The physics shape is a single, solid object. I've just simplified the geometry a bit from the original.

Edited August 26, 2021 by Wulfie Reanimator

[Aquila Kytori]

5 hours ago, Wulfie Reanimator said:

I believe your specific problem in the first image is the use of flat planes with gaps between them.

My guess is that such a high Physics weight of +9000 is caused by the bug that can appear when you are using triangle based physics that consist of two parallel collision surfaces where one or both are not perfectly flat.

An example :   This is my visual Arch model :

 

I had intended to force this bug to show  itself but by chance when I uploaded my first test physics model the bug was already there:

Also I forgot to check the normals of my first physics model so this one had the back wall normals facing the wrong way. The Physics cost was 8837 ! 

 

Uploaded again but this time with corrected normals. The Physics cost was now a whopping 29709.037 :

 

 

Usually this very high physics cost can be corrected by making sure that both the front wall and back wall faces of the physics mesh are perfectly flat, doing this by selecting each surface in turn and scaling to zero along the Y axis, (S Y 0) : Now the Physics weight is a correct 0.500  :

 

Back in 20213 @Drongle McMahon   tried to track down the exact cause of these occasionally very high physics weights when using triangle based physics:

Episode 1 :

Episode 2 :

Episode 3 :

Episode 4 :

Unfortunately the images are missing from Episodes 1,  3 and 4 but still intact in Episode 2 where you can get some idea of how much time and  effort he used to put into his investigations.

 

Back to the Arch Physics ..........

To me something has changed in the last few of years in the way the Physics weights are calculated for triangle based physics models like this.

For something like this arch or a floor with an opening you used to be able to add 1 or 2 "thin" triangles that would join one surface to the other or be used to increase the bounding box dimensions and not have the physics weight increase significantly but now its just not practical.  An example below shows how the Physics weight increases dramatically when adding collision surfaces that join the front and back walls at the arch opening. First adding one quad (two triangles) increase the weight from 0.500 to 7.373 !

 

And adding 2 quads increases from 0.500 to 18.510 !

 

 

Another thing

5 hours ago, Wulfie Reanimator said:

Here's a similar object, 10x20 meters in size, 0.6 LI with only 0.36 physics (I think that's the absolute minimum physics cost you can have).

Your screenshot is showing a physics weight of 0.360 for the Base Hull. That would be the cost if you left the Physics Shape Type at the default Convex Hull in the Features tab of the Edit floater. The arch would be closed.

When the Physics Shape Type is changed to Prim (archway open) then the weight would be equal to the Mesh value of  2.178 as indicated in the little Physics panel of the uploader window.

 

Edited August 27, 2021 by Aquila Kytori

[Wulfie Reanimator]

22 minutes ago, Aquila Kytori said:

Another thing

5 hours ago, Wulfie Reanimator said:

Here's a similar object, 10x20 meters in size, 0.6 LI with only 0.36 physics (I think that's the absolute minimum physics cost you can have).

Your screenshot is showing a physics weight of 0.360 for the Base Hull. That would be the weight if you left the Physics Shape Type at the default Convex Hull in the Features tab of the Edit floater. The arch would be closed.

When the Physics Shape Type is changed to Prim (archway open) then the weight would be equal to the Mesh value of  2.178 as indicated in the little Physics panel of the uploader window.

Thanks for pointing that out!

Also thanks for linking those threads. I'm familiar with other things he's investigated, so I know they will be an interesting read.

Edited August 27, 2021 by Wulfie Reanimator