"Peter Popov" <pet### [at] vipbg> wrote in message
news:c9vv4us4j8706qsvtgem59flufjv5qudhc@4ax.com...

> <fcu### [at] yahoocom> wrote:
>
> >Algorithmically not very different, but used in another context.
>
> >I hope you like it, and I'll be glad to receive your comments and
> >suggestions.
>
> It's *way* too elastic, like raw rubber or latex.

Yes, I also think it is too elastic, and that's why I gave it that plastic
texture, so it does not look like a cloth.

> Cloth elasticity comes from the fact that it is made of interwoven
> threads, be they elastic or not. Tensile forces 'tighten up' the
> threads and make the cloth thinner, but the threads themselves don't
> elongate, or at least not relevantly much.

That's right. The threads are very inelastic. But the equations describing
that must be very complicated. You can simplify your model if you ignore the
"tightening" of the threads and compensate with a bit of more elasticity to
the cloth.

I'm now trying to reach an alternative way to solve this, but I still have
to work many hours :)

Thanks for your comments,

Sbogom!

Fernando

Post a reply to this message

<fcu### [at] yahoocom> wrote:

>That's right. The threads are very inelastic. But the equations describing
>that must be very complicated. You can simplify your model if you ignore the
>"tightening" of the threads and compensate with a bit of more elasticity to
>the cloth.

You can impose a penalty on elasticity if strain goes above a certain
value. Not cap it altogether but penalize it strongly so that the
cloth tends to preserve its original shape without being too elastic.

I would use a strongly non-linear stress-strain relationship if I were
you. Try logarithmic for a start, or exp(-x*t). This way for higher
stresses there'll be almost no increase in strain.

And then you may consider material failure :)


Peter Popov ICQ : 15002700
Personal e-mail : pet### [at] vipbg
TAG      e-mail : pet### [at] tagpovrayorg

Post a reply to this message

These cloth animations of yours are truly amazing!
This one, as many others have said, is a tad too stretchy, but the way it
folds when it falls and how it behaves when the corners are released is very
realistic (too slow, but nonetheless realistic- like a slo-mo of it).


news:3c4e36a8@news.povray.org...
> Algorithmically not very different, but used in another context.
>
> I hope you like it, and I'll be glad to receive your comments and
> suggestions.
>

> Mexico City.
>

Post a reply to this message

Thank you! I'm glad you liked it!

I'm right now working on an entirely different algorithm, to avoid all these
problems, but it is somewhat complicated, so I don't know if I will able to
implement it correctly.

Thanks again,

Fernando.


"Mahalis" <don### [at] fakeycom> wrote in message
news:3c50a0b3$1@news.povray.org...
> These cloth animations of yours are truly amazing!
> This one, as many others have said, is a tad too stretchy, but the way it
> folds when it falls and how it behaves when the corners are released is
very
> realistic (too slow, but nonetheless realistic- like a slo-mo of it).
>

> news:3c4e36a8@news.povray.org...
> > Algorithmically not very different, but used in another context.
> >
> > I hope you like it, and I'll be glad to receive your comments and
> > suggestions.
> >

> > Mexico City.
> >
>
>

Post a reply to this message

I do the same, but i need to change time step each time I change the
"accuracy" of the mesh, and I don't like tries and error algorithm :o(.

I've been driven to a site where there are some studies about cloth
stability with large step calculus, but I did not succeed to get something
out of this stuff yet.( http://www-2.cs.cmu.edu/~baraff/papers/).

Chaps.


news:3c503d0f@news.povray.org...
>
> "Chaps" <cha### [at] yahoocom> wrote in message
> news:3c5009ef@news.povray.org...
> > I like it very much.
> >
> > I  am trying, from time to time, to build an external tool wich could
> > generate clothes as mesh2 files from a pov scene, but I have to fight
with
> > instability. do you? how do you handle it?
> >
> > Chaps.
>
> Hi Chaps. I'm still trying to have a more robust algorithm, but meanwhile,
> my only solution is to take smaller time steps.
>
> Fernando.
>

Post a reply to this message

"Chaps" <cha### [at] yahoocom> wrote in message
news:3c514dd3@news.povray.org...
> I do the same, but i need to change time step each time I change the
> "accuracy" of the mesh, and I don't like tries and error algorithm :o(.
>
> I've been driven to a site where there are some studies about cloth
> stability with large step calculus, but I did not succeed to get something
> out of this stuff yet.( http://www-2.cs.cmu.edu/~baraff/papers/).

Hi Chaps!

What a coincidence! I spent nearly all day yesterday to try to understand
that document. Actually I did understand most of it, but I got badly stuck
with the C(x) condition vector... I believe there must be some errors, or at
least it is not very well explained, because in the Stretch Forces section
(4.2), C(x) is defined as a 2-dimensional vector:

C(x) = a (|wu(x)| - bu , |wv(x)| - bv )

However, in the shear and bend forces section (4.3), C(x) is defined as

C(x) = a * wu(x)^T * wv(x)

Which, I believe it is 1-dimensional, that is, a scalar.

And then, in the Forces and Force Derivatives section (4.1) it says that the
force can be computed like this:

f_i = -k dC(x)/dx_i * C(x)

But f_i is a 3-dimensional vector, so I believe there must be an error
because there is no dimensional agreement. I may be wrong, but I'm stuck.

Did you have the same problem?

Fernando.

Post a reply to this message

On Fri, 25 Jan 2002 13:21:46 +0100, "Chaps" <cha### [at] yahoocom> wrote:

>I've been driven to a site where there are some studies about cloth
>stability with large step calculus, but I did not succeed to get something
>out of this stuff yet.( http://www-2.cs.cmu.edu/~baraff/papers/).

Regarding cloths, you can find more of David Baraff & Andrew Wittkin's
papers and course notes online. They are the guys that are behind
Pixar's cloth simulation engine. Xavier Provot is also a
well-recognized name in the field. I believe Demetri Terzopolous has
recently been focusing on cloths rather than solids and he has some
interesting ideas like using dynamic NURBS (D-NURBS), or neural
networks, or state-space theory to simulate elastic dynamics.


Peter Popov ICQ : 15002700
Personal e-mail : pet### [at] vipbg
TAG      e-mail : pet### [at] tagpovrayorg

Post a reply to this message

Well! What can I say?
Its absolutely fantastic! =)
Congratulations.

Best Regards.



news:3c4e36a8@news.povray.org...
> Algorithmically not very different, but used in another context.
>
> I hope you like it, and I'll be glad to receive your comments and
> suggestions.
>

> Mexico City.
>

Post a reply to this message

Thank you JPGargoyle!

It's nice to know that you liked it!

Fernando.


"JPGargoyle" <jpg### [at] clixpt> wrote in message
news:3c533601$1@news.povray.org...
> Well! What can I say?
> Its absolutely fantastic! =)
> Congratulations.
>
> Best Regards.

Post a reply to this message

I should have some time before end of week, and I'll take a look at that.

May be I'll get illuminated.

Chaps


news:3c518884@news.povray.org...
> Hi Chaps!
>
> What a coincidence! I spent nearly all day yesterday to try to understand
> that document. Actually I did understand most of it, but I got badly stuck
> with the C(x) condition vector... I believe there must be some errors, or
at
> least it is not very well explained, because in the Stretch Forces section
> (4.2), C(x) is defined as a 2-dimensional vector:
>
> C(x) = a (|wu(x)| - bu , |wv(x)| - bv )
>
> However, in the shear and bend forces section (4.3), C(x) is defined as
>
> C(x) = a * wu(x)^T * wv(x)
>
> Which, I believe it is 1-dimensional, that is, a scalar.
>
> And then, in the Forces and Force Derivatives section (4.1) it says that
the
> force can be computed like this:
>
> f_i = -k dC(x)/dx_i * C(x)
>
> But f_i is a 3-dimensional vector, so I believe there must be an error
> because there is no dimensional agreement. I may be wrong, but I'm stuck.
>
> Did you have the same problem?
>
> Fernando.
>

Post a reply to this message