Hi!
This is a question best answered by NL or an expert I guess.
I've been advised when doing hybrido-sims in RF12, to scale my scenes in maya instead of using the scene-scale parameter inside of RF, according to LuisM this would allow the hybrido-solver to compute more efficiently.
Is this the case still in RF13? Will scaling the scene in RF impact my sim-times and results?
Scene scale in 2013?
-
LuisMiguel
- Posts: 528
- Joined: Fri Jun 04, 2010 2:37 pm
Re: Scene scale in 2013?
Hey,
The scale is important because is not the same to have pool of 3 meters than another one of 3000 meters. So, the fluid won't have the same behaviour for both situations. So, for this reason, I advice to use the real scale, if not, you'll need to change the daemons force values if you want to have a good behaviour working with huge scales. Basically, the scale you work in RealFlow is not important whenever you change the daemons force values to have a right movement for your fluid.
I hope it helps.
luisM.
The scale is important because is not the same to have pool of 3 meters than another one of 3000 meters. So, the fluid won't have the same behaviour for both situations. So, for this reason, I advice to use the real scale, if not, you'll need to change the daemons force values if you want to have a good behaviour working with huge scales. Basically, the scale you work in RealFlow is not important whenever you change the daemons force values to have a right movement for your fluid.
I hope it helps.
luisM.
Re: Scene scale in 2013?
Hey Luis,
Thanks for your answer, yes I'm aware of this, let me go a little more into detail.
The way the modelling on the film I'm working on is done, I usually want to scale my scene to 10%.
Whenever I can, I group my objects in my maya-scene and scale to 0.1 before exporting to realflow, but sometimes when there are animated characters it's virtually impossible to do it because the rigs will break.
When I cannot scale the scene in maya, I have two options in realflow.
1) Keep the geo-scale at 1 but increase the daemon force to a factor of 10.
2) Set the geo-scale to 0.1
I don't know if it's just me, but compared to some sandboxing scenes I've done at home, which are at scale 1, I find that working in a geometry-scale of 0.1 usually makes it harder to get nice results with the hy_mesh in the end, I often get issues with dimples in the mesh. And setting the daemon factor to 10 sometimes makes my grid fluid jittery when I playblast, but maybe this is not going to show up in the mesh anyway...
Have you experienced a difference in meshing in scale 0.1 vs scale 1, and have you experienced jittery gfd-behaviour when working with daemon scale set to 10?
I guess what I'm ultimately asking you is: would you rather scale or increase the daemon factor, given my situation?
Thanks for your answer, yes I'm aware of this, let me go a little more into detail.
The way the modelling on the film I'm working on is done, I usually want to scale my scene to 10%.
Whenever I can, I group my objects in my maya-scene and scale to 0.1 before exporting to realflow, but sometimes when there are animated characters it's virtually impossible to do it because the rigs will break.
When I cannot scale the scene in maya, I have two options in realflow.
1) Keep the geo-scale at 1 but increase the daemon force to a factor of 10.
2) Set the geo-scale to 0.1
I don't know if it's just me, but compared to some sandboxing scenes I've done at home, which are at scale 1, I find that working in a geometry-scale of 0.1 usually makes it harder to get nice results with the hy_mesh in the end, I often get issues with dimples in the mesh. And setting the daemon factor to 10 sometimes makes my grid fluid jittery when I playblast, but maybe this is not going to show up in the mesh anyway...
Have you experienced a difference in meshing in scale 0.1 vs scale 1, and have you experienced jittery gfd-behaviour when working with daemon scale set to 10?
I guess what I'm ultimately asking you is: would you rather scale or increase the daemon factor, given my situation?
Re: Scene scale in 2013?
I know about the Character Problem...
I wouldn't recommend using the global force scale and hack in a 10 there. You will be having more control when adopting each Deamon individually as scaling by 10 might work very good for some deamons but bad for others.
Another idea - Have you tried exporting the animated characters as Alembic and then scale the alembic inside Maya? That should enable you to scale the characters in a way they work for you inside RF.
I wouldn't recommend using the global force scale and hack in a 10 there. You will be having more control when adopting each Deamon individually as scaling by 10 might work very good for some deamons but bad for others.
Another idea - Have you tried exporting the animated characters as Alembic and then scale the alembic inside Maya? That should enable you to scale the characters in a way they work for you inside RF.
Re: Scene scale in 2013?
Good points Florian, thanks!
I have thought of the alembic double-cache workaround, the only thing that got in my way of actually trying it out (other than not taking the time to do so) is that I've been wondering a little bit about substeps and alembic?
Early on in the production I would open a mayafile with alembic files and export SDs from that, but then realflow didn't support alembic so I don't know if it's still an issue, but back then they only had a substep of 1 so they would cause all sorts of behaviour, do I need to recache our alembics with more substeps? How many are a good going value, I know the new HY2 usually manages to get good results with 1-3 substep but SPH-sims often requires substeps in the hundreds, this would cause quite large files... Or am I missing the point here?
What's your workflow with alembic->RF?
I have thought of the alembic double-cache workaround, the only thing that got in my way of actually trying it out (other than not taking the time to do so) is that I've been wondering a little bit about substeps and alembic?
Early on in the production I would open a mayafile with alembic files and export SDs from that, but then realflow didn't support alembic so I don't know if it's still an issue, but back then they only had a substep of 1 so they would cause all sorts of behaviour, do I need to recache our alembics with more substeps? How many are a good going value, I know the new HY2 usually manages to get good results with 1-3 substep but SPH-sims often requires substeps in the hundreds, this would cause quite large files... Or am I missing the point here?
What's your workflow with alembic->RF?
Re: Scene scale in 2013?
SPH and FLIP (Hyb2) are different technologies, so you cannot compare the substeps you need for SPH with the ones for FLIP.
Usually, when working with FLIP 1-2 substeps should be sufficient, unless you see weird things happen. You might want to slightly increase substeps when fast moving objects are interacting with the fluid.
For SPH there are many factors to consider, as one of the major drawbacks of SPH is the well- known exploding particles. I always try finding out an optimal substeps setting tht works best for my scene (sim time wise).
Keep in mind, even though substeps affect the fluid behavior it will not look better or more realistic with increased substeps, just different. Substeps are most important with regard to simulation time and stability.
Usually, when working with FLIP 1-2 substeps should be sufficient, unless you see weird things happen. You might want to slightly increase substeps when fast moving objects are interacting with the fluid.
For SPH there are many factors to consider, as one of the major drawbacks of SPH is the well- known exploding particles. I always try finding out an optimal substeps setting tht works best for my scene (sim time wise).
Keep in mind, even though substeps affect the fluid behavior it will not look better or more realistic with increased substeps, just different. Substeps are most important with regard to simulation time and stability.