Jump to content

Viscous Teapots


goldleaf

Recommended Posts

The results of my exploration of FLIP Fluids (many thanks to those who already lent a hand!), by doing a Viscosity test like Igor's Elephants (http://vimeo.com/29045955).

Here is a render, of the three shaders for the surfaced liquids. One thing that's been hard to nail down has been getting the surfacing to be smooth. The white teapot on the right shows it most; that bumpiness is in the geometry. Anyone have tips on surfacing between 500,000 and 2,000,000 points from a fluid?

Animation/flipbook coming soon....

Thanks!

post-3767-133110317984_thumb.png

Edited by goldleaf
  • Like 2
Link to comment
Share on other sites

CeeGee: Thanks Igor! Better/larger ones are en route, once I get some sims finishes :) Thank you for the elephants! That was such a nice looking test, technically and artistically. BTW, did you render that with mantra? Was the meshing done in NAIAD, I assume?

dynamoanders: Thank you! Yeah, I used the Viscosity shelf tool to turn on viscosity. Then on the flipfluidobject, under the Physical tab, the Viscosity parameter is at the bottom.

Some observations so far:

- Turns out that the density strongly affects the viscosity settings. For example, these teapots are about 1 meter square, and their density is at the default of 1000. This meant that my viscosity settings for the three teapots you see here are 1000, 36642, and 72285. I'm resimming at a lower viscosity, 9500 (that's why there is only the really low "1990's" resolution render attached :) That was something I wasn't aware of before! So it was really cool to learn.

- I'm using a Phenom II X6 to sim/render these, and for FLIP Fluids, there seems to be lots of time that only some cores are being used, which is where the i5/i7 CPUs seem to be stronger; but mantra really uses all cores well; I never really saw the usage drop from 100%

- Using some of the tips found here, I re-rendered the large still. Interestingly, I noticed the SSS noise decreased dramatically, but the noise on the plain white shader increased (I did tweak the ground shader a bit; perhaps that caused it?)

Thanks for the encouragement!

post-3767-133114438575_thumb.png

viscous_teapots_small.mov

Edited by goldleaf
Link to comment
Share on other sites

I'm using a Phenom II X6 to sim/render these, and for FLIP Fluids, there seems to be lots of time that only some cores are being used, which is where the i5/i7 CPUs seem to be stronger; but mantra really uses all cores well; I never really saw the usage drop from 100%

Nice! BTW, did you see the Viscosity Tips here:

http://www.sidefx.com/docs/houdini12.0/dyno/liquids#viscosity

In particular it looks like you might still have Treat As Ballistic for underrresolved particles.

You might also try turning off Use Preconditioner on the Solver tab, which basically trades having to take more solver iterations for multithreading. Depending on your processor(s) it might speed up the FLIP solve.

Link to comment
Share on other sites

Sweet! Thanks for those tips Johner! I totally missed that section of the doc page. I've got another set of wedges running now (with more reasonable viscosity settings), and I've turned off Teat As Ballistic, and turned off Use Preconditioner. I also turned off reseeding, and it appears to be simming noticably faster (and the CPU graph seems to be spending much more time with busy cores, than idles). Thank you!

Here is a decent sized flipbook (nearly filled my ram! need more!). I resimmed with viscosity at 9500, and put that one in the middle, and moved the previous middle (36642) to the right.

I'm confused by the way the first set of teapot liquids all flatten before they hit the ground. Hrrmmm....... Is it air resistance? It's so flat though...

teapotsA2.mov

Link to comment
Share on other sites

Hi goldleaf,

I did render with Maya and Vray, because i already did one render setup for my first viscosity test, mesh is done in Naiad.

2h ago i start same test with Houdini. This is resault.

Thanks to johner and you for Flip simulation tips.

Scene file ( if someone want to test more) and render.

Igor

elephant.rar

elephant_VIS_v001.MOV

Edited by CeeGee
Link to comment
Share on other sites

That is so nice.

And I really appreciate that you can tell differences between Naiad and H12.

About Sim time, Meshing Time, Ram usage, etc.

This time we got H12, so I am really wondering if this is done with Naiad??

I should but I dont get Naiad installed right now so if you dont mind.

I think as sim get larger, Naiad get better, but not sure.

I guess lot a people wanna know about this.

Thanks!!

Hi goldleaf,

I did render with Maya and Vray, because i already did one render setup for my first viscosity test, mesh is done in Naiad.

2h ago i start same test with Houdini. This is resault.

Thanks to johner and you for Flip simulation tips.

Scene file ( if someone want to test more) and render.

Igor

Link to comment
Share on other sites

Wow! Thanks Igor! That's really generous of you to put that up, thanks! Can't wait to dig in!

Here is a new set of sims, with the teapots simmed at 1000, 14000, 27000, and 40000 (left to right), and re-rendered.

BTW, on Fedora 14, I installed ffmpeg so I could export *.mp4 files from mplay, and setting the image format to PNG resulted in an error, but JPG works just fine. I tried mencoder, but that resulted in random blue lines across the frames (on png).

Thanks!

teapotsB_anotated.mov

Link to comment
Share on other sites

Hi, nice test guys!

Just wondering, why your first row of teapots squeeze in the air ? :)

btw, same is happening with Igor test - just frame before ground it looks like colliding with invisible plane

  • Like 1
Link to comment
Share on other sites

Wow! Thanks Igor! That's really generous of you to put that up, thanks! Can't wait to dig in!

Here is a new set of sims, with the teapots simmed at 1000, 14000, 27000, and 40000 (left to right), and re-rendered.

BTW, on Fedora 14, I installed ffmpeg so I could export *.mp4 files from mplay, and setting the image format to PNG resulted in an error, but JPG works just fine. I tried mencoder, but that resulted in random blue lines across the frames (on png).

Thanks!

I have pretty much given up on making good mp4 or mov straight out of Houdini and just export pictures and then run ffmpeg from the terminal instead to create the vid from the sequence. Nice simulations. I need to sign up for the AUP Apprentice HD soon.

Link to comment
Share on other sites

Has anyone experienced the flattening flip fluids before?

A few things to try to help alleviate that problem:

1) Decrease the Surface Extrapolation parameter under the Collisions tab, closer to 0.1 or so. This controls the amount the fluid is extrapolated into the collision, in voxels, so decreasing it might make the collisions take effect slightly later.

2) Enable Collision Separation on the FLIP Object and set it to the same value as your Particle Separation. This will give you a higher-res collision field and possibly more accurate collisions.

3) Set up your file like Igor's elephant test, where the closed bottom of the simulation volume is used as the ground plane, so collisions don't enter into the simulation at all (I'd try this first, myself).

BTW, to Igor's very well set up file, I'd only add that for very high viscosity settings, this blurb from the Viscosity Tips in the documentation is relevant:

The FLIP Solver computes the number of substeps to calculate based on velocity. While the viscosity solver should always be stable, it will be more accurate with smaller substeps. A slow moving, highly viscous sim might need more substeps than the 1 that the FLIP Solver will give it based on its velocity. Increase Minimum Substeps until you get sufficient viscosity.

Link to comment
Share on other sites

A few things to try to help alleviate that problem:

1) Decrease the Surface Extrapolation parameter under the Collisions tab, closer to 0.1 or so. This controls the amount the fluid is extrapolated into the collision, in voxels, so decreasing it might make the collisions take effect slightly later.

2) Enable Collision Separation on the FLIP Object and set it to the same value as your Particle Separation. This will give you a higher-res collision field and possibly more accurate collisions.

3) Set up your file like Igor's elephant test, where the closed bottom of the simulation volume is used as the ground plane, so collisions don't enter into the simulation at all (I'd try this first, myself).

Actually, I'm not sure if this is the same flattening problem, as the teapots all start to flatten out 6 frames before impact.

In case the above suggestions did not do the trick, I would be curious to know whether toggling off "Dynamically Resize Fields" or adding to the "Max Cells to Extrapolate" are of any help.

Link to comment
Share on other sites

  • 1 month later...

Actually, I'm not sure if this is the same flattening problem, as the teapots all start to flatten out 6 frames before impact.

In case the above suggestions did not do the trick, I would be curious to know whether toggling off "Dynamically Resize Fields" or adding to the "Max Cells to Extrapolate" are of any help.

I also had some flattening in my tests. I got rid of those by increasing the substeps in the Flip solver.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...