Forum
Welcome, Guest
Username: Password: Remember me
  • Page:
  • 1

TOPIC: Question about minimum number of iterations with DOPnet substeps

Question about minimum number of iterations 2 weeks 19 hours ago #322

  • JK
  • JK's Avatar Topic Author
  • Offline
  • Junior Member
  • Junior Member
  • Posts: 11
  • Thank you received: 0
Hi,
Our artists are adding substeps in the dopnet and want reduce the number of iterations accordingly (all related to slow mo shots).
Original iterations 40 per frame, they're trying to have 10 substeps and have 4 iterations only on the solver.
I was wondering if we could lower the minimum number of iterations on the carbon simulation dop node or if you guys think this is not a good idea.
Thanks for all you great support.

Please Log in or Create an account to join the conversation.

Question about minimum number of iterations with DOPnet substeps 2 weeks 18 hours ago #323

  • admin
  • admin's Avatar
  • Offline
  • Administrator
  • Administrator
  • Posts: 109
  • Thank you received: 16
Hi JK,
We would not recommend using less than 10 iterations because this is a question of convergence towards a solution at each solver update which is executed at each simulation substep. So if you are adding substeps / subframes on the dopnet, assuming that your frame rate is 24Hz, then you are calling a simulation update more than once per frame. So if you use 10 subframes on the dopnet, then the physics will already be called 10 times per frame at a frequency of 240Hz but with the advantage of the animated data also being evaluated at that frequency. So, if you set the physics simulation with 10 substeps, what actually happens is the physics is substepping on top of your dopnet substeps and you end up with the solver running at 2400Hz. If you are substepping on the dopnet, what you can do is reduce the number of simulation substeps while keeping the number of iterations.

To clarify the difference:
When you use substeps on the dopnet, you are calling the simulation update as many times as you have subframes. So, you benefit from the animated data used by the physics being evaluated more precisely than if you have no substeps and only update the simulation once per frame.
When you use simulation substeps, it slices the time step during a simulation update and linearly interpolate the animated parameters in order to evolve the solve with smaller increments in time which produce better collision detection and better sampling of visco-elastic problems. It is not the same thing as the dopnet substeps and is cumulative to it. If you have dopnet substeps, you simply don't need many simulation substeps. But you need iterations to converge to a solution at every solver update.

Hope this helps,
JC
The following user(s) said Thank You: JK

Please Log in or Create an account to join the conversation.

Last edit: by admin.

Question about minimum number of iterations with DOPnet substeps 2 weeks 12 hours ago #324

  • JK
  • JK's Avatar Topic Author
  • Offline
  • Junior Member
  • Junior Member
  • Posts: 11
  • Thank you received: 0
Thanks for the explanation.
Your explanation makes complete sense.

Please Log in or Create an account to join the conversation.

  • Page:
  • 1
Time to create page: 0.155 seconds