I just began studying cloth and find out that 2 samples from Help work incorrectly.

1) $HFS/houdini/help/examples/nodes/dop/clothsolver/ClothSpringsBalls.otl

houdini freezes on the first frame , loading the CPU for 100% and writes in the StatusBar: "integrating cloth over time interval "

2) $HFS/houdini/help/examples/nodes/dop/clothsolver/PinnedClothWind.otl

the cloth is behaving so as there is no links between it's points (see attached image)


houdini 9.5.210 , vista x64

Yeah, I've been having troubles too - if you build this from scratch with new nodes(there's been changes lately to cloth), it behaves the same. I've found wind to be very unpredictable lately in DOPs. I thought ‘well, let’s try a fan' - it works better but you need to pull the force *way* down from the default, it starts out fine and stays that way, but for some reason it starts getting slower and slower to compute as it carries along(just rippling in the fan force). That seems wrong too.


SUSE 10.2, H9.5.211 64 bit


Cheers,

J.C.

This could be related.. I was having trouble gettnig the SBD Spring Constraint (Cloth Spring Constraint) to do anything.

These examples need to be updated. The wind example will work fine if you reduce “Scale Force” to something much smaller, say 0.001.

On closer inspection, I found a bug in the one of the cloth assets. This bug resulted in the cloth getting an incorrect mass (too small). This is fixed in tomorrow's build of Houdini 9.5. The examples that are mentioned above will work in this build. This also explains why you would have had trouble getting the spring constraints to work.

Michiel

I just wanna share this because, as discussed before cloth is “not That ready” in houdini, sure you can do simple grids stuff but when it comes to real production stuff like a tshirt things get ugly.

http://www.blender.org/features-gallery/feature-videos/?video=realtime_cloth [blender.org]

I'll like to do some stuff like this in houdini, that's awesome and I'm gonna stress test Blender as soon as I got some time, as I said, with a real project.

just to end I love houdini

I've had problems with the cloth too.

When I run the attached sim on any of my linux boxes, the constraints all turn red and don't hold.

I've tried this in many different builds, including 9.5.212 and it still doesn't work.

Interestingly, it works just fine on my Mac. Unfortunately, my Mac is not on our production network, where I need to run all my real sims that have the same problems as this file.

Dave

Thank you for the files Dave.
This is a known issue. So far I've managed to reproduce this problem only for a 32-bit linux build that is installed on a 64-bit machine.
All other configurations work fine for me.
So if you happened to be using a 32-bit version of Houdini on a 64-bit machine, then you may want to try installing a 64-bit linux version of Houdini 9.5.

If this doesn't help you, I have a workaround: in the examples you posted, you don't really need a stitch constraint. You can just use a pin constraint on your point group. I verified that this workaround works on the machine that reproduces your problem.

Michiel

Thanks Michiel,

That works great.

Just as an academic exercise, should my example files work the way I have them set up? Or is this just a bad way to accomplish this?

Dave

Hi Dave,

In your example, you're constraining a group of points to their initial locations. It's better to use a pin constraint for that.
Pin constraints simulate faster than stitch constraints.
This may make a noticeable difference for large point groups.

Michiel

http://temp.roxik.com/datas/cloth/index.html [temp.roxik.com]

varomix
http://www.blender.org/features-gallery/feature-videos/?video=realtime_cloth [blender.org]

I've not had much luck with cloth either… seems to blow up for no reason quite often, (and yes, especially with wind).

Blender is very impressive. And it's open source…. maybe SESI could borrow the code, (and improve it of course)!

Sean

Sean,

Can you please post the cloth hip files that are giving you trouble? We are making a lot of progress, often under hood, these days, and you may find that something that did not work the day before is fine today. There is light at the end of the tunnel.

Hi Sean,

A bug that was causing bad results with wind was fixed last week
(builds 212 and after). This was the same bug that caused the two examples to break.

If you install the latest build of 9.5, then you should find that the cloth behaves as it should in the presence of wind forces.

One tip: If you use really large wind forces, you may set a higher “stretch stiffness” value under Cloth/Material to prevent the cloth from stretching too much in the wind.

Please let me know if this works for you.

Michiel

hi

As Michiel said, cloth had MAYOR fixes in the recent builds, he said that had to be fixed for 9.5 and yes they did, I can now interact with cloth live in the viewport as solid and “realtime” as the video I posted on Blender.

The nice nice thing here is this, we made a point, cloth needed some fixes, and SESI hears you, takes notes, and improves the product, just as you guys said, this means only one thing. SESI ROCKS and houdini is your SONG.

I wanna say thank you, once again to everybody at SESI.

keep it ROCKING!!!

Finally got a chance to upgrade to 9.5.219 and look at cloth again, much better. Cloth behaves with more stability and predictability for the basic stuff, at least.

I'm finding when rebuilding a brand new version of the original example that aliax posted that I need to punch the wind to very high numbers, however, to get something akin to a pinned cloth being ‘filled out’ by the wind. In the 80000 range for the scaling factor, which seems high to me, especially since the noise amplitude just needs to be 1 to be noticable. Is it possible there's a scaling issue with the wind? Actually, I just tested with the fan, which in the older ‘broken’ version got a nice head of steam up at much lower values, behaves the same. I'm assuming this is because cloth is defaulting to much denser, but do those values seem reasonable?

Also, with the same example, near those corner pins it seems to have individual points that drag behind the others occasionally, creating a crimp.

Cheers,

J.C.

Hi,

Whether those values are reasonable really depends on the size of your cloth. By default, the cloth mass is computed using a density. This means, the bigger you make your cloth the heavier it will become and the more wind force you will need to apply to it to get the same effect.

If you don't like this automatic mass calculation, then you can specify a cloth mass directly (disable the “Compute Mass” parameter on the cloth object).

About the points near the pins:
You should verify that the cloth thickness is significantly smaller than the lengths of the cloth edges. You may also want to experiment with higher “bend stiffness” values.

Michiel

OK, got it!

Cheers,

J.C.

Strange stuff happens in the 64bit linux build of 220 (apprentice).
I try the same very basic test on XP 32 SP2 and ubuntu 7.10 64-bit (with 64-bit Houdini): a polygon sphere as a Static RBD object and a grid above it as Cloth.
On XP32 the simulation runs well. On Ubuntu 64, if I open the file saved from XP then during the sim I get a huge memory leak and a crash. If I build the sim from scratch, few frames after the two objects collide, the cloth just dissapears.

Dragos

Hi Dragos,

Thank you for reporting this.
I'll take a look at these issues.
Do you still have the hip files?
It would be very helpful if you could post them.

Michiel