ncloth on running character

[chiax2]

mhm..
Anyone?
screen shot my problem .
These 2 pictures are frame 151 and frame 152.
Before that character has a very fast move swinging.. jumping... which is so stable and nice simulation but when he starts to turn around and run... after the second frame he turns.. this problem turn out.no matter what I increase I even increate the substeps to more than 50. and it still runs . when he hits this frame it will start coming out with weird result.
I have tried traped check. i even make the input mesh attract to 2 and above .

I did a test with just full input mesh attract without painting the sleeves. the hand came out when he runs to the end of the shot.

I have no idea why is this happening?

any idea?
Or do you need more info to help me out wiht this?
If i have a solution I will definately post it here to share it.

hope someone could help.

cheers,
Chiax2

[Duncan]

Perhaps the thickness for collisions is a little on the low side, although it is hard to tell.
You really shouldn't need any input attract. If you post a simplified scene with the problem somewhere I'll take a look, otherwise it is hard to tell what the issue is.

Duncan

[chiax2]

thanks.
The input attract is painted on the body.
I wanted the body to follow 100% the skined cloth.
I have tried us constrain but it's not the result I wanted.

I try to clean up the file so i can upload to here.
but unfortunately I can't publish the file here.
can i send you through your email?

cheers,
Chia

[mascio]

Try to don't use 1.0 as maximum input attract value.
Use 0.9 for example and a little bit of Input Attract Damp.

...[mascio]...

[chiax2]

hi,
Thanks,
I have tried using value starting from .55 step up slowly to 2.

The best value that gives me is 1.35 I don;t know why.
but this value doesn't distort my cloth it preserve the shape.
any value that lower than this will distort my cloth.
I wonder isn't it that when it's painted white for the inputmesh it should follow 100%?
like in softbody?

When he runs to a certain distance the cloth starting to distord and weired result coming out.

When he is standing it works pretty good even in a very fast move.

[chiax2]

If I dont put inputmesh attraction in high value how do I make the ncloth follow the skinned cloth 100%?
Is that the reason the mess up the cloth when it turns at frame 150 ?

Chia

[Castius]

Personally it just looks like you need more substeps in your nucleus. Nothing to do with the input attract.

[Duncan]

The input mesh attract is when you want it to match the skinned animation, however then the solution is no longer a dynamic one, but simply your deformer animation (but useful to work around problems with sudden moves). The input mesh attract range is 0-inf, although values around 1.3 or so should be enough to lock the motion. Don't use values much over 1.5 or things may get slow( the iterations for this attribute go up by the 5th power of the value ). Also note that the map only scales the value of the input attract and map values outside the 0-1 range should be avoided.

At anyrate you should be able to get the motion to look good with no input attract at all. You can send me the file (duncan.brinsmead@autodesk.com (http://forums.cgsociety.org/duncan.brinsmead@autodesk.com)) and I'll take a look at it. (The simpler the scene file the better)

Duncan

[chiax2]

I have send you the file.
Let me know if you have received it. and let me know your feed back .
really appreciate it.

we are now trying to test on syflex see if it's more stable then ncloth...hope will solve our problem then we dont need to worry about getting in other plugins.

i have been staying over nite a few night to get it looks normal like a cloth... and it seems getting harder and harder...

Really hope after you seeing the file will help us resolve the problems.

Thanks.

Chia

[Duncan]

I've sent you a file with fixes. The main problem was that stretch/compression resistance was not high enough and that the stubsteps were too low.
For a real world character that is moving relatively quickly stretch values of 500 or more are sometimes required. The substeps should be at least 1/10th or so of the stretch value and the maxCollisionIterations can be about 1.5 times the substeps. I noticed you had keyframed the substeps, which is OK, but beware that if the substeps is too low then the material will stretch more than it should. Also if the low substeps causes some collision failure in parts of the mesh, then this failure could still propagate even after the substeps are raised. I would keep the base substeps at least around 30 for this simulation and keyframe possibly as high as 150 for extreme motion frames.

A second problem is that you only used vertex edge self collisions, which was not sufficient for your simulation. One can use vertexFace if there are no self colliding open edges (like on a sphere, for example. Full surface is required if you have border edges that can self collide. However a second approach if you don't need fine detail in the self collisions and have a relatively uniformly tessellated model is to use vertex self collision with a large selfCollideWidthScale( so the vertices overlap a little ). Self collisions with "full surface" would be exceedingly slow with a large selfCollideWidth scale, but are reasonably fast when using "vertex". Another benefit of using the fat vertex self collision is that it tends to simulate a natural bend resistance( one can turn off the bend resistance when doing this ).

Additionally I set your spaceScale (on the nucleus node) to 0.1, which would result in the right gravity if your character were modeled with each unit being 0.1 meters. In your scene the gravity was too low(spaceScale was 1), which tends to result in the cloth flying all over as the character moves.

I also increased the drag and tangential drag a bit... generally there will be some tangential drag on cloth. As well I made the stretch damp 0.5 to make the cloth less bouncy. You might play with the mass value to simulate the fabric you are after... lighter fabrics will have a lower mass which results in more drag basically.

Duncan

[chiax2]

After increasing the substeps to 30 and 50 onwards and lower the spacescale to 0.1
It does gives me a more stable simulations. I think from now one I will put my space scale to 0.1.

thanks Duncan. Could not have done without u.
thanks.

cheers,
Chia

[dim1984kimo]

I've sent you a file with fixes. The main problem was that stretch/compression resistance was not high enough and that the stubsteps were too low.
For a real world character that is moving relatively quickly stretch values of 500 or more are sometimes required. The substeps should be at least 1/10th or so of the stretch value and the maxCollisionIterations can be about 1.5 times the substeps. I noticed you had keyframed the substeps,
Duncan

I really love you , Duncan, AWESOME!

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.