+- tyFlow Forum (https://forum.tyflow.com)
+-- Forum: tyFlow Discussion (https://forum.tyflow.com/forum-1.html)
+--- Forum: General Discussion (https://forum.tyflow.com/forum-2.html)
+--- Thread: Retiming Issue (/thread-1044.html)
Retiming Issue - insertmesh - 08-23-2019
Hey there,
first time doing some extreme retiming (simple physx explosion) and wondering how to do it properly...
I´ve already upped the steps per frame to 1/32, but I still get a bit wonky interpolations on the very slow retiming part. Spin looks ok, but positional changes seem to ease in /out between the simulated positions.
I realize that I might be reaching the limits of how much I can retime a simulation, just wondering what those limits are and how I could work my way around it (and if using more steps per frame is actually helping the retiming at all or just adding to my simulation time...
Using the latest version (16043).
RE: Retiming Issue - tyFlow - 08-23-2019
The retimer uses linear interpolations, not cubic interpolations, so if you retime things very, very slowly you'll see the linear trajectories between steps.
Also the cacher caches on whole frames only, so reducing substeps will have no effect on the retimer fidelity.
The interpolation itself has a flat trajectory, so you should not be seeing ease-in, ease-out within the subframe values...but due to the nature of the interpolation you may be noticing accelerations happening across multiple frames.
RE: Retiming Issue - insertmesh - 09-02-2019
(08-23-2019, 03:23 PM)tyFlow Wrote: The retimer uses linear interpolations, not cubic interpolations, so if you retime things very, very slowly you'll see the linear trajectories between steps.
Also the cacher caches on whole frames only, so reducing substeps will have no effect on the retimer fidelity.
The interpolation itself has a flat trajectory, so you should not be seeing ease-in, ease-out within the subframe values...but due to the nature of the interpolation you may be noticing accelerations happening across multiple frames.
Yeah, thats probably what I´m seeing. Looks ok in most cases though, but good to know that substeps don´t matter.