Well, this is the actual cluster I wanna simulate:

It contains 1.700 coins.

There are three coins flitting into it, so they have to start moving everyone at another frame. This means I can't activate them all at the same time.

I tried to unpin them from top to bottom, but that looks very unnatural. Adding weak constraints didn't work neither. And I think there should be a way to make entirely flat objects just lie on top of each other, or is there still no solution for that?

Attachments:
cluster1.jpg (131.7 KB)

Here is an example hip file... I cracked open the rbd bullet solver and in the RBD Packed Object node, changed the geo representation to "cylinder," which is much more efficient and will allow you to reduce padding to 0. Then turn the global substeps up to two (or higher), which will prevent the intial settling bounce.

After the assemble node, you may want to put in a point jitter (with 0 in Y) and jitter the position of the coins in the stack slightly to get a more interesting look.

Image Not Found

Attachments:
rbd.png (24.0 KB)
stackcoinsMachineFX.hipnc (1.9 MB)

Attachments:
RBD-Stack.zip (1.2 MB)

Attachments:
poker-chips26_machineFX.hipnc (3.4 MB)

Thanks, scaling it up helps to stabilize the scene. I would have preferred using Bullet World Scale (which I thaught would do the same effect) cause my Geometries' sizes are all adjusted to each other, but it doesn't seem to work that way.
However scaling geometries up individually will do.

The sceond issue though confuses me a lot.
When I look at the coins very closely in orthographic view I can see there are spaces everywhere in between.

To me the gaps seem to be very small but distinct. I made them this small so the coins don't have to fall down on each other in the beginning of the simulation. With the collision representation objects it is the same. Am I missing some intersecting meshes here? Or can there be a "too close" to each other?

I might easyly increase the distances cause I created the whole cluster procedural. Do I have to choose a higher value here?


EDIT: I scaled the scene up by 10 million now and it works great without a single substep. Thanks a lot man! Though I haven't found any difference caused by @found_overlap.

What still slows the simulation down is the amount of coins flying out of the scene remaining active. I don't need them anymore, so is there a way to delete them or at least make them sleep inside the solver based on their position with some kind of "container" as flip simulations use?
I found the popkill-node to approach this... everything's working fine now

Attachments:
space_between.jpg (117.2 KB)

mrCatfish

To me the gaps seem to be very small but distinct. I made them this small so the coins don't have to fall down on each other in the beginning of the simulation. With the collision representation objects it is the same. Am I missing some intersecting meshes here? Or can there be a "too close" to each other?

Make sure you always look at the bullet representation of the collider, not the geo going into the sim.

Of course, that's what I meant writing collision representation objects.
I see gaps between the coins everywhere... have I actually overlooked any intersections?

Attachments:
space_between_bullet.jpg (47.8 KB)