Help Dynamic particle attachment question

[MEPETAMINALS]

Hi -- I've read through the manual and tried a number of things, but I'm encountering an issue with particle attachments that I'm hoping to get some insight on.

You can see in the video that the corner of the sail (this is a dynamic particle attachment to a rigidbody at the end of the rope) tends to jitter considerably.

I'd thought that this was a matter of solver settings, but have been unable to make any meaningful improvements -- even cranking iterations up to goofy amounts (20+).

Regarding settings, collision is off, both on a cloth blueprint/particle level, and for the obi rigidbodies involved. The mass of the attached rigidbody and the attaching particles are close (I've tried varying these as well. The only change is that if I increase the mass of the particles to be considerably higher than the attached rb, the issue reverses and the rigidbody starts flickering.)

Worth noting that the issue remains even if the attachment is set to static, and the attached rb is kinematic.

This issue is unchanged on CPU/GPU solving.
This issue is unchanged on RB Interpolation ON/OFF (For both the solver and hierarchy of rigidbodies).

I'm inclined to think that it's due to being a child of a rigidbody, but from my understanding that should be permissable. (Essentially the ship is a rigidbody, which is a parent to all other dynamic elements.)

Suggestions appreciated.

https://www.youtube.com/watch?v=ya3jV_JDYSU

[josemendez]

Regarding settings, collision is off, both on a cloth blueprint/particle level, and for the obi rigidbodies involved. The mass of the attached rigidbody and the attaching particles are close (I've tried varying these as well. The only change is that if I increase the mass of the particles to be considerably higher than the attached rb, the issue reverses and the rigidbody starts flickering.)

Worth noting that the issue remains even if the attachment is set to static, and the attached rb is kinematic.

This issue is unchanged on CPU/GPU solving.
This issue is unchanged on RB Interpolation ON/OFF (For both the solver and hierarchy of rigidbodies).

I've been unable to reproduce this (using Obi 7.1, synchronous fixed solver synchronization mode). Would it be possible for you to share a minimal setup that causes this issue so that I can take a closer look at what's happening?

I'm inclined to think that it's due to being a child of a rigidbody, but from my understanding that should be permissable. (Essentially the ship is a rigidbody, which is a parent to all other dynamic elements.)

Yes, this is not only permissible but the expected setup in cases such as yours. Just keep in mind that the ship's ObiRigidbody (if any) *must* be marked as "kinematic for particles", otherwise the inertial reference frame in which the ropes are simulated (the ship) would be affected by the simulation itself resulting in an unphysical feedback loop.

kind regards,

[MEPETAMINALS]

I've been unable to reproduce this (using Obi 7.1, synchronous fixed solver synchronization mode). Would it be possible for you to share a minimal setup that causes this issue so that I can take a closer look at what's happening?


Yes, this is not only permissible but the expected setup in cases such as yours. Just keep in mind that the ship's ObiRigidbody (if any) *must* be marked as "kinematic for particles", otherwise the inertial reference frame in which the ropes are simulated (the ship) would be affected by the simulation itself resulting in an unphysical feedback loop.

kind regards,

Thank you for the response - your comment on mode seems to indicate the problem. I had solver mode set to asynchronous, clearly misunderstanding the reasons for it.

Synchronous fixed does indeed solve the issue, though unfortunately it takes performance from 90fps to 30, so perhaps best left to a toggle option for those with beefier pc's. (All constraints disabled except distance, collision, and pin -- 3 substeps, constraints between 1-3)

Help much appreciated.

[josemendez]

Thank you for the response - your comment on mode seems to indicate the problem. I had solver mode set to asynchronous, clearly misunderstanding the reasons for it.

Asynchronous mode "frees" the solver from Unity's FixedUpdate() calls, kickstarting its update during a frame and making sure it's completed during the next. This allows it to be updated concurrently with everything else - including rendering.

However this means interaction with rigidbodies is less accurate, as it suffers from lag. This may or may not be acceptable depending on your specific goal. For cases where accurate coupling with rigidbodies is needed, you must use synchronous fixed (which starts and completes simulation in FixedUpdate) or at least synchronous (which starts and completes in the same frame).

Here's an in depth explanation of how physics update works in general and how Obi is bolted onto Unity's update scheme when using all of the 3 synchronization modes:
https://obi.virtualmethodstudio.com/manu...eloop.html

let me know if I can be of further help,

kind regards