retime simulations

   41620   29   3
User Avatar
Member
230 posts
Joined: Oct. 2009
Offline
I did few test in order to find the best way to retime RBD simulation: increasing fps or decrease scale time on the dopnet node gives the look of slow motion but screw the all forces and you and up with a very weird simulation. Then I tried timeshift SOP which gives gaps on the animation with paused frames. The best way seems to be using CHOPS, you import the animated geo and scale as you want. Now I have few questions:

1. There are too much details when you bring the geometry to CHOPS and I have to wait 2 minutes to load the curves and keep closed the Motion View, otherwise it will keep freezing in every change. The good this is that once you load it the animation goes fast but is there anyway to avoid this delay?

2. So far I use the simple stretch node to retime the simulation, is there anyway to give different values to different areas of the simulation and all of them to be attached and keep running as one piece?

3. Is there anyway to retime fluid simulations, like smoke and liquids? I tried to retime with CHOPs a pyrofx fireball simulation but without success.


thanks in advance for any help or info.
User Avatar
Member
390 posts
Joined: Jan. 2012
Offline
1. You can resample the channels in chops and reduce their detail.

2. Not really sure I understand.

3. Try adjusting the timescale parameter on the Simulation tab of the pyro solver.


sorry its not much. hopefully enough to dig around a bit until you get better answers
.
User Avatar
Member
230 posts
Joined: Oct. 2009
Offline
zdimaria
1. You can resample the channels in chops and reduce their detail.

2. Not really sure I understand.

3. Try adjusting the timescale parameter on the Simulation tab of the pyro solver.


sorry its not much. hopefully enough to dig around a bit until you get better answers

thanks for the reply zdimaria

1. I tried resamle already but does not make much of deference on the delay because either way you have to load the original geometry and loosing details of the animation.

3. I am afraid you are right, I did few test with FLIP and pyro and only timescale works but as usual effects the look of the simulation

2. Now I use stretch and I scale up or down the whole simulation. But lets say you want to start with normal speed, then gradually slow down and keep it for few frames and then return to normal speed. I hope this makes more sense.
User Avatar
Staff
2540 posts
Joined: July 2005
Offline
I would give the Timeblend - Timeshift SOP combo another try. It should do a linear blend between the two volume primitives.

If you have dynamic resizing of containers on in the Pyro simulation, you will definitely see strobing and pulsing due to the volume primitives changing dimensions.

One way to keep resizing post-sim is to use a new Volume SOP set to pick up the final container settings from the Pyro Object and then use a Volume Mix to copy the dynamic container in to the new container.

This is a linear interpolation so you will start to see the hits at whole frames at about a 5 to 6 times slowdown.

See the example file.

Note that you will need to do a Volume SOP and volume Mix for each volume primitive you want to re-time.

Btw repopulating your high res sim in to lower res volume containers with this technique is something that you should know. You can set the Volume SOP to be a frustum based volume from the current render camera as well.

Attachments:
retime_billowy_smoke_in_sops.hip (1.4 MB)

There's at least one school like the old school!
User Avatar
Member
230 posts
Joined: Oct. 2009
Offline
jeff
I would give the Timeblend - Timeshift SOP combo another try.

thanks for the reply and the file Jeff, I will give it a try now. Is this works with FLIP simulation? CHOPs it doesn't, I tried it.
User Avatar
Member
230 posts
Joined: Oct. 2009
Offline
jeff
I would give the Timeblend - Timeshift SOP combo another try. It should do a linear blend between the two volume primitives.

If you have dynamic resizing of containers on in the Pyro simulation, you will definitely see strobing and pulsing due to the volume primitives changing dimensions.

One way to keep resizing post-sim is to use a new Volume SOP set to pick up the final container settings from the Pyro Object and then use a Volume Mix to copy the dynamic container in to the new container.

This is a linear interpolation so you will start to see the hits at whole frames at about a 5 to 6 times slowdown.

See the example file.

Note that you will need to do a Volume SOP and volume Mix for each volume primitive you want to re-time.

Btw repopulating your high res sim in to lower res volume containers with this technique is something that you should know. You can set the Volume SOP to be a frustum based volume from the current render camera as well.

PYRO
Jeff this method works great. But when I use explosion for example from the pyrofx shelf the fire part missing from the rendering and is everything like a smoke. I thought that may the blast be responsible for that by deleting temperature and heat which are used by the shader but even by disabling blast, I get the same result.

FLIP FLUIDS
Also it seems nothing works with FLIP fluids simulation, neither Timeblend - Timeshift SOP or CHOPs, there is some strobing on the particles (the size of the container is constant), I tried to use it before and after the meshing.
User Avatar
Member
390 posts
Joined: Jan. 2012
Offline
FLIP FLUIDS
Also it seems nothing works with FLIP fluids simulation, neither Timeblend - Timeshift SOP or CHOPs, there is some strobing on the particles (the size of the container is constant), I tried to use it before and after the meshing.

There is no reason the timeshift sop wouldn't work with flips, or any geo for that matter. The timeblend however, will probably not work because you probably have reseeding enabled on your flip solver. This changes the number of particles over time, and the timeblend relies on using a consistent number of points to work. This also means that using it on a meshed surface will also not work. If reseeding is turned off, however, the timeblend timeshift method should work just fine.

zak
.
User Avatar
Member
230 posts
Joined: Oct. 2009
Offline
zdimaria
FLIP FLUIDS
Also it seems nothing works with FLIP fluids simulation, neither Timeblend - Timeshift SOP or CHOPs, there is some strobing on the particles (the size of the container is constant), I tried to use it before and after the meshing.

There is no reason the timeshift sop wouldn't work with flips, or any geo for that matter. The timeblend however, will probably not work because you probably have reseeding enabled on your flip solver. This changes the number of particles over time, and the timeblend relies on using a consistent number of points to work. This also means that using it on a meshed surface will also not work. If reseeding is turned off, however, the timeblend timeshift method should work just fine.

zak


well, I tried it also to retime a simple popnetwork and I have the same problem. Also I cant find the reseeding parameter on the FLIP object or solver. I will create an example scene and post it in an minute.

UPDATE
ok I found reseeding parm on the flip solver and it fixed the problem by disabling it. So I cant retime a particles simulation which generates new particles over time?

UPDATE2
as you said the timeblend creates all the problems but if I will turn off reseeding to fix the FLIP fluid. What can I do in the following: I created a system for my project which generates debris from fractured geometry, the amount of debris changes over time, as more fracturing there is, more debris will born. But here I have the same problem with the reseeding, when I am reading the bgeo sequence from disk with the debris simulation, the timeblend creates the strobing, when is turned of I have no linear blend between frames. Is there any solution for this?
User Avatar
Member
230 posts
Joined: Oct. 2009
Offline
I have attached a scene with all possible simulations and there notes if there is any error.

Seems that you can retime either using CHOPs or Timeblend+Timeshift SOPs any kind of simulation (RBD, particles, fluids) but with few exceptions.
When auto-resized container is used by the fluid simulation there is a strobing, Jeff gave a solution to that above.
When you retime FLIP fluids there also a jittering, zdimaria above gave the solution, the reseeding should be turned off.

But when new geometry or points is enters the simulation over time, there is a problem. Here is the example I used above: I created a system for my project which generates debris from fractured geometry, the amount of debris changes over time, as more fracturing there is, more debris will born. But here I have the same problem with the reseeding, when I am reading the bgeo sequence from disk with the debris simulation, the timeblend creates the strobing, when is turned of I have no linear blend between frames.

All the above you can find them inside the attached hip.
Edited by - April 14, 2012 07:52:37

Attachments:
tests_retime.hipnc (2.7 MB)

User Avatar
Member
37 posts
Joined: Oct. 2009
Offline
jeff
See the example file.

Note that you will need to do a Volume SOP and volume Mix for each volume primitive you want to re-time.

This is a cool example. I had a question about retiming the vel field in order to get proper motion blur.

Would I want to break up the vel into 3 scalar volumes using your exact same method? or do I want to create a vector volume to copy the volume mix with? I created both setups in your file.

Thanks

Attachments:
retime_billowy_smoke_in_sops_with vel.hip (1.4 MB)

User Avatar
Member
230 posts
Joined: Oct. 2009
Offline
Loudsubs
Would I want to break up the vel into 3 scalar volumes using your exact same method? or do I want to create a vector volume to copy the volume mix with? I created both setups in your file.

I guess this is just a test setup but if not, in your method 1, you have blast 3 times the vel_x and forgot vel_y and vel_z.
User Avatar
Member
390 posts
Joined: Jan. 2012
Offline
I can't open your file right now, but if you really want to use a timeblend to smooth out the strobing on geo with changing points, you can try adding an id attribute to the points. This wont work with skinned surfaces but with a flip sim or with particles it should work to some degree. When a new point is created you need to add a unique id attribute to it, so that the timeblend has something to follow.

I haven't had a chance to try this out, but when I have a minute I will put together a hip file for you.
.
User Avatar
Member
37 posts
Joined: Oct. 2009
Offline
i9089303
I guess this is just a test setup but if not, in your method 1, you have blast 3 times the vel_x and forgot vel_y and vel_z.

Good catch, its been a long week

Any idea the which is the proper way to retime the velocity field of a volume so that it will match the retime of the density?
User Avatar
Member
230 posts
Joined: Oct. 2009
Offline
zdimaria
I can't open your file right now, but if you really want to use a timeblend to smooth out the strobing on geo with changing points, you can try adding an id attribute to the points. This wont work with skinned surfaces but with a flip sim or with particles it should work to some degree. When a new point is created you need to add a unique id attribute to it, so that the timeblend has something to follow.

I haven't had a chance to try this out, but when I have a minute I will put together a hip file for you.

I have managed to retime almost everything with one or the other way, but this is the only I am missing reading bgeo sequence where new geometry are coming is over time.

The problem is that my method of creating debris uses the velocity of the pieces of the fractured geometry and emmits particles from the interior faces which has vel>5. Every frame new particles are created and using a copy SOP I create unique instances of debris to each particles and pass them to dynamics. In the particles network I use a kill POP, so once the particles will pass from POPs to SOPs to DOPs they will be killed. Can we have this attribute ID on the SOP level in order to be attached to each instance of debris before they go to DOPs? I really hope what I am saying make sense to you.

I hope for your idea to work, it will save project. I will wait your example hip. Thanks for all your help.
User Avatar
Member
390 posts
Joined: Jan. 2012
Offline
Just opened your file. the NEW_GEO_OVERTIME problem can be fixed by rearranging your nodes.. Because you are already creating an id attribute on the particles, and the timeblend is set to look for that attribute, if you just copy your spheres after the timeblend all works well. So, it should go popnet1 feeds into timeblend1 feeds into timeshift1 feeds into copy2.


Is this solution viable for you project?


Ill look at the other errors in a bit.

zak
.
User Avatar
Member
230 posts
Joined: Oct. 2009
Offline
zdimaria
Just opened your file. the NEW_GEO_OVERTIME problem can be fixed by rearranging your nodes.. Because you are already creating an id attribute on the particles, and the timeblend is set to look for that attribute, if you just copy your spheres after the timeblend all works well. So, it should go popnet1 feeds into timeblend1 feeds into timeshift1 feeds into copy2.


Is this solution viable for you project?


Ill look at the other errors in a bit.

zak

well I am afraid is not, the method in example file was just the way to test the debris using particles collision. But the method that I am actually using (as I descripted above) uses POPs only to get the intial velocity but all the simulation happens inside the DOPs. I can't have the slowmotion before the instancing. I will attached a file with this method when i will be home in an hour.

cheers
User Avatar
Member
230 posts
Joined: Oct. 2009
Offline
zdimaria, here is the example file which I fotgot to include with the other examples. As you can see the maingeo has no problem to retime, the same retime nodes applied to the debris creates a strobing on the animation.

Attachments:
innovation2.hipnc (1.0 MB)

User Avatar
Member
230 posts
Joined: Oct. 2009
Offline
any luck?
User Avatar
Member
390 posts
Joined: Jan. 2012
Offline
Not yet. The strobing is strange to me –it seems like they remain the same shape just scale down through the interpolation. I'll keep looking at it though. Although I would still like to figure this out, is there a reason you can't change your sim to give you the results you want, instead of fixing it after?
.
User Avatar
Member
390 posts
Joined: Jan. 2012
Offline
This is the part that I am having trouble understanding-

If underneath facet1, you put a fuse, connectivty, partition, delete, timeblend and then timewarp. Make rule in partition piece_$CLASS. In delete1 make group piece_34 and delete non-selected. Timeblend remove both attributes and then timewarp output range 1 to 1000.

This isolates one piece. The timeblend should have no problem interpolating this as the point numbers stay constant. It does this until it hits the ground. When it does, the strobing starts again, but not in an even pattern. It appears that the piece is rotating and the timewarp isn't following? The number of points is constant… I'm really lost here.
.
  • Quick Links