dForce - Are subdivision levels not used for the simulation?

linvanchene

Tested with DAZ Studio 4.10.0.123

 

​Update: tldr version:

​It has been confirmed that

​- for the clothing objects simulated with dForce only their Base Resolution is considered.

​- the subdivision level of other scene objects is used for collision detection

- - -

 

Related information and terms to understand this post:

To make any plane visible by the simulation apply a dForce dynamic clothing modifier

- select plane in scene tab

​- in the simulation tab options select dForce / add dForce modifier: dynamic surface

compare:

dForce - Start Here

https://www.daz3d.com/forums/discussion/203081/dforce-start-here

Some basics about subdivision levels are explained here:

https://www.daz3d.com/forums/discussion/180391/guide-working-with-hd-addons-and-subdivision-levels

Basically when you subdivide a mesh you multiply the point count by four.

Example:

low resolution: low point count, low vertex count

high resolution: high point count,  high vertex count

 

When I use a high polygon count clothing object (100'000 vertices or 1'000'000 vertices)  dForce simulations take longer but also yield drastically more detailed results.

The basic idea was:

The more vertices (points) a clothing object has the more points are calculated by the simulation.

But currently it seems that the higher point count of subdivided meshes is completly ignored by the dForce simulation.

​To me it seems the dForce simulation is only using the "base resolution" of any objects.

Is this working as intended or should I submit a bug report?

 

- - -

For all images the same simulation settings were used.

Gravity -0.40

Initialization time 2

Simulation range 1-30

Collision Mode: Best Continous

default settings for the rest.

 

If you compare the simulation results of simple planes with a high or low  vertex (point) count you can get detailed results like these:

plane primitive with around 1 million vertices

simulation time ca 15 minutes (with a GTX 1080)

 

plane primitive 90000 vertices

​simulation time ca 3 minutes

-> For objects without subdivison levels everything seems to work like I expected.

The higher the point count of the mesh the more detailed the simulation results will be.

- - -

However if you raise the subdivision levels to get a higher point count this does not seem to make any difference:

Subdivision level 1 - 4225 vertices

Subdivision level 3 - 66049 vertices

Subjective Commentary:

​To me it seems the simulation is just using the base object at 1089 vertices and is ignoring the higher point count of the subdivision levels.

At least in these tries I was not able to see any noticeable difference in the simulation by raising the subdivision levels.

Has anyone found a forum thread in which this was discussed?

Is this working as intended or is this a bug?

- - -

What I expected how it would work:

My basic idea was to raise the subdivision level for dForce clothing to get a higher point count version of the same clothing that allows for more detailed folds in simulations.

If the subdivision levels are not actually used for the simulation I currently fail to understand how I am supposed to get higher quality simulation results with dForce clothing sold in the store?

 

-> low resolution clothing will not be able to show very detailed folds no matter which settings you try

-> a high resolution clothing version would be needed to simulate finer clothing folds

- - -

Am I missing an obvious solution or workaround or is this just how it works for now -

dForce is only able to run the simulation with the base resolution of objects?

- - -

Richard Haseltine

I believe this is the correct behaviour. It is possible to use the SubD verts for collision detection, but not for actual deformation.

Edit: this has been confirmed.

linvanchene

Richard Haseltine said:

I believe this is the correct behaviour. It is possible to use the SubD verts for collision detection, but not for actual deformation.

Edit: this has been confirmed.

Thank you for confirming.

One reason why I was under the wrong impression that subdivision levels might have any influence on dForce simulations was this official video:

dForce Mesh Modification

While trying to get rid of the distortions at the edges of a plane viewers are told that

"a first fix is to adjust the mesh resolution"

it is then concluded that "this has not done much to help with the issue".

In hindsight I realize that the speaker was trying to fix distortions AFTER the simulation was run.

Therefore:

-> Changing the resolution levels or changing to a different SubDivision Algorithm might still help to marginaly change the look of your simulation results AFTER the simulation has run.

BUT as now has been confirmed

During the simulation

SubDivision levels of  other scene objects are used for collision detection with the simulated clothing.

But for the simulated clothing itself just the base resolution is considered in the simulation.

- - -