PDA

View Full Version : Any special demands on collision-objects for particles?



CharlieL
10-01-2010, 08:37 AM
I have been studying the rainsplash.mov video (Vaughan video on particles)
on how to manage particles to spash when they meet a collision-surface.
But now I am a bit confused.

I tried to apply it on a rotating nozzle that shoot water under high pressure
inside a tank. What I would like to achieve is a continous splash where the
beam hits the inner surface of the tank.

What confuses me is that when I use my object (inner roof of the tank) as a
collision object, the beam is erazed long before it meets the surface. With a simple
collision-plane it will work as expected and erase the water-beam at the plane.
But not when I change to my own object (polygons as collision-objects).
And yes, the Radius/Level is zero for the Collision-object..

Are there any special demands on the geometry in order to get it to work as a
collision-object?
In this film Vaughan talks about an umbrella as collision-object.
But, does it have to be designed in a very specific way to work?

Anyone having experience of collision-objects?

Lightwolf
10-01-2010, 10:19 AM
Since I actually used them recently, here's a few pointers:

* Triple the mesh (just in case) - quads are o.k. as well
* If you want to collide inside of an object... model it with the normals going out and then tell the collision to happen inside. Modellinh with the normals going in cause nothing but problems here
* Make sure your mesh is closed and modelled water tight. You can even model a lower res proxy for collisions to speed things up.

Oh, and feel free to post a scene.

Cheers,
Mike

CharlieL
10-02-2010, 11:35 AM
Since I actually used them recently, here's a few pointers:

* Triple the mesh (just in case) - quads are o.k. as well
* If you want to collide inside of an object... model it with the normals going out and then tell the collision to happen inside. Modellinh with the normals going in cause nothing but problems here
* Make sure your mesh is closed and modelled water tight. You can even model a lower res proxy for collisions to speed things up.

Oh, and feel free to post a scene.

Cheers,
Mike

Thank you Mike for interesting advice! But I am a bit confused about "the normals going out".
Because I have to render the inside of the tank. Do you mean I should make the walls
double-sided with normal outwards?
I can see the problems with normals going in.

The scene is a bit complex with corrugated walls in the tank with lots
of extra stuff as it is a part of a ship. I would need to do some cleaning
before I can send it, otherwise it will be just confusing.

Regards,
CharlieL

Lightwolf
10-02-2010, 11:54 AM
Thank you Mike for interesting advice! But I am a bit confused about "the normals going out".
Because I have to render the inside of the tank. Do you mean I should make the walls double-sided with normal outwards?
I can see the problems with normals going in.

Create a solid mesh (just for the collision) that basically is the volume within the tank. In a way, model the empty volume, if you get what I mean.

Then tell the collision object to collide Inside.

I was surprised as well, but it does seem to make quite a difference, especially concerning "leaking" particles.

Cheers,
Mike

CharlieL
10-03-2010, 09:40 AM
Thanks for an interesting idea!
I think I get your point. Kind of a an inner-bolean part.

I am curious if I can involve the girders running under the top of the tank.
And if they will affect how the water-beam splashes. I have to try it out,
It would be just nice if I could get it to work as expected.

3D is so fun when it works, but a very often a demanding and tiring challenge.
Thank you Mike for your suggestions!

Lightwolf
10-03-2010, 09:50 AM
I am curious if I can involve the girders running under the top of the tank.
And if they will affect how the water-beam splashes.
Without seeing your mesh I'd assume yes. But you should boolean and make sure that your collision object is a single mesh.

Cheers,
Mike