Static RBD vs Source Volume, Preferred method for Flip Collision?

[Cuckon]

Hi all,

I found there's (at least) two way to make geo collision with flip fluid:

1. Converting geo as static RBD
   
2. Using Collide with objects shelf button and houdini would create a Source Volume DOP to perform the collision
   

Though it looks like the officially preferred way of making collision is using Source Volume DOP way, I find it more accuracy to deal with it using the first method, please have a look at my attachment, in which the RBD version of the rotating box have a nice collision detect obviously.

If this is true then does it mean I should always use RBD to perform collision, and leave Source Volume Dop way for other effects such as pumping or something like that?

collision.hip

Edited June 26, 2013 by Cuckon

[JuriBryan]

check your volume... make sure that your volume has enough detail to build the sharp edges of the cube.

And the volume collision method is better in my opinion since its easy to feed noise into it as velocity plus with VBD its extremely fast to build complex collision geo.

[Cuckon]

check your volume... make sure that your volume has enough detail to build the sharp edges of the cube.

And the volume collision method is better in my opinion since its easy to feed noise into it as velocity plus with VBD its extremely fast to build complex collision geo.

Thanks Juri, I have checked the resolution, and found that the Volume Source have a higher res ( 46 against RBD's 30), so I really doubt the accuracy of the Volume source method.

[JuriBryan]

well easy solution. you have a collision volume with points inside of it... that really can't work can it now.

Try to start with the collision objects out side of the flip sim, that helps a lot.

Plus the volume source can be as accurate as possible, if the particle separation and with that the collision volume calculated to hit the flip sim is low the collision will be off with both ways.

Turn on the collision guys geometry on the flip tank and then change the particle separation, you will see what I mean

[anim]

I prefer RBD for colliders as well (for flip and pyro or anything), since it doesn't need to recompute SDF unless it's deforming and it can as well use point or RBD velocity for collisions

it's simply faster easier and more accurate

in your case you are using deforming geo even though you don't have to, but even in this case it's faster to use RBD since you don't have to create velocity volumes

[JuriBryan]

Okay, so I just ran a test with a high poly Bust sculpture with custom velocity values applied to and a normal flip particle flow running over it.

And you guys are right, static RBD sims are much more accurate... using the volume collision method I got a lot of points inside the the geo over with a collision super sampling value of 6 and a grid scale of .5.

also it was quit a bit slower...

But since you would cache out both collision volumes, if it comes from source volume or RBD, before running the final sim that does not bother me to much.

Its really the collision that seams to be very inaccurate especially with fast moving fluids and even under high sampling values.

Thanks for pointing this one out to me guys since I was about to run a big scale production shot using the volume collision method:)

Cheers,

Juri

[Solitude]

I almost always build my collision volumes in sops, cache it to disk, and use the static object with proxy volume to read the volume from sops.

[Skybar]

I usually use RBD with "Use volume based collision detection" turned off, depending on what I want to do. It's usually the case that with it turned on, I have to crank up the divisions quite a lot and that quickly gets insanely slow, especially if I have an animated static RBD.

[Cuckon]

Well, things get clear now, thanks all you people, and so glad to make lil help to Juri.