Julien FAVINI

Hello everyone,

I know this is a recurring topic, but I need any advice.
I have a bowl, which I scaled to about 1m radius, I thickened the bowl.
I am currently at a grid scale of 1.5 and I have a particle seperation of 0.005 / 0.004 (between 8 and 13million points)

I initialize all my rbd hero at frame 1, some on the surface and others further away so that they gradually fall into the bowl.

The rbd solver is set in bullet mode and has 25substeps
The flip solver has a feedback of 0.17 and 1/4 substeps

The first surface collisions go well, then after x impacts the water level begins to drop drastically.



I tested a lot of things,
- lower the grid scale even lower
- different substeps for the 1/10 or 2/3 flip for example
- set the rbd in rbd mode and have 4/4 substeps
- change the collision detection (particle / moving)
- add a little viscosity
- change the bowl with a cylinder
- or even limit too heavy or fast rbd by taking them out of the flip


I also saw on the forums a technique based on divergence, but in my tests it only blew the flip.

I even tried pushing with a collider from below, or emitting from the bottom, but that doesn't work besides adding time to the simulation.

Also concerning the rbd from time to time some go crazy: they react strongly to other rbd / flip

Any advice is welcome
Thanks!

1. Do you have a sound card or specific audio drivers / softwares ?
Because it was my case months ago, when I was dealing with a lot of crashs, and everything was related to my asus xonar drivers ( its a specific driver used by most of the sounds card ) ( you can find old topics about it )

2. Are you using windows 10 ?
I got some problems with an update month ago with a “ransomware protection”.

Great !
Thank you