hello,
I am doing a test for debris where I have a fractured geometry (wall) hitted by a ball. The debris are generated by the sides of the pieces which are moving (group them using a velocity expression) and all the calculation take place in dops. There is 0 bounce to both active and static objects, friction 2.
But as you can see on the attached video they are going crazy.
http://dl.dropbox.com/u/10756448/debris.m2v [dl.dropbox.com]
The wall hitted by the ball is one simulation saved on disk without debris and the I use this in another simulation (dopnet) as static geometry for the debris to collide with.
Full Version: debris in dops goes crazy
i9089303
hello,
But as you can see on the attached video they are going crazy.
Indeed, the roof in on fire!!!
What's the debris made of? if it's just particles, you can do it all inside a popnet with the collide POP and feeding the cached geo from the sim as collision objects.
if you're going geometry instanced to to the particles, that's another hairy beast… Although there's plenty of threads in the forum about it!
cheers
you could try to add an RBD Autofreeze DOP to the end of your simulation in order to freeze objects whose velocity goes under a certain threshold. It also helps speeding up your simulation.
It's really hard to say without seeing your scene, but here's some things that you could check:
- collision guide geometry, does the collision representation of your objects
look correct ? (debris AND floor)
- collision and contact passes in the rbd solver
- substeps
It's really hard to say without seeing your scene, but here's some things that you could check:
- collision guide geometry, does the collision representation of your objects
look correct ? (debris AND floor)
- collision and contact passes in the rbd solver
- substeps
@rafaels
I am doing it with instance and dops because I want the collisions to be accurate. But in my scene I have also a pop version, for preview, which works perfect. I will attached the scene later this evening when I will be back to the lab.
@asnowcappedromance
I will attached the scene later this evening but I am using what your technique that you shared here http://www.sidefx.com/index.php?option=com_forum&Itemid=172&page=viewtopic&t=23195&highlight=debris [sidefx.com]
the only I have changed is to enable “Use Template Point Attributes” in the copysop in order to pass the initial velocity from each particle to instance geometry and then in dops I have enabled the “Inherit Velocity from Point Velocity” of the RBD Fractured Objects.
I tried with autofreeze but does not work, because the most crazy ones they nover go enough slow so as to stop.
I also am using the autofreeze to my first simulation with the ball and the wall but as you can see at the right of the wall there is still some jittering, how can I fix this if the autofreeze fails? I have increased the substeps and collision passes in the rbd solver in both simulations.
I am doing it with instance and dops because I want the collisions to be accurate. But in my scene I have also a pop version, for preview, which works perfect. I will attached the scene later this evening when I will be back to the lab.
@asnowcappedromance
I will attached the scene later this evening but I am using what your technique that you shared here http://www.sidefx.com/index.php?option=com_forum&Itemid=172&page=viewtopic&t=23195&highlight=debris [sidefx.com]
the only I have changed is to enable “Use Template Point Attributes” in the copysop in order to pass the initial velocity from each particle to instance geometry and then in dops I have enabled the “Inherit Velocity from Point Velocity” of the RBD Fractured Objects.
I tried with autofreeze but does not work, because the most crazy ones they nover go enough slow so as to stop.
I also am using the autofreeze to my first simulation with the ball and the wall but as you can see at the right of the wall there is still some jittering, how can I fix this if the autofreeze fails? I have increased the substeps and collision passes in the rbd solver in both simulations.
here is the scene, I hope someone could explane me what cause debris go crazy…
ok there are several problems with your scene:
first of all, the naming convention of your groups. Try to keep naming your objects differently so nothing gets confused by the solver. Your wall fractured groups and your debris are having the same groups (piece*), so I renamed your debris.
Another problem is that you instanced the same debris groups on your particles (i.e. 3 particles got the same debris piece thus being in the same group) which caused the solver to screw up because it calculated the center of mass incorrectly, so I regrouped your debris, check out what I did there.
Also you misunderstood the usage of the autofreeze DOP.
Your enable threshold was at 0 which means your rbd objects never become candidates to actually being frozen. In the details view you can see that there is an entry named cumulativeMovement on each object.
Velocity and angular velocity get added together and only if they reach a certain threshold the autofreeze is testing them against the freezing thresh.
This on the other hand was set to 0.01 in your scene which means some objects cumulative motion never got below that thresh so their active value stayed always one which explains the jittering.
I advice you to read in the help about the rbd autofreeze node to understand what's going on under the hood.
I attached the file, reduced the numbers of fragments to speed it up a little, but that should get you going.
first of all, the naming convention of your groups. Try to keep naming your objects differently so nothing gets confused by the solver. Your wall fractured groups and your debris are having the same groups (piece*), so I renamed your debris.
Another problem is that you instanced the same debris groups on your particles (i.e. 3 particles got the same debris piece thus being in the same group) which caused the solver to screw up because it calculated the center of mass incorrectly, so I regrouped your debris, check out what I did there.
Also you misunderstood the usage of the autofreeze DOP.
Your enable threshold was at 0 which means your rbd objects never become candidates to actually being frozen. In the details view you can see that there is an entry named cumulativeMovement on each object.
Velocity and angular velocity get added together and only if they reach a certain threshold the autofreeze is testing them against the freezing thresh.
This on the other hand was set to 0.01 in your scene which means some objects cumulative motion never got below that thresh so their active value stayed always one which explains the jittering.
I advice you to read in the help about the rbd autofreeze node to understand what's going on under the hood.
I attached the file, reduced the numbers of fragments to speed it up a little, but that should get you going.
Manu thank you very much for your help and especially for the explanation of what was wrong, its important for me to understand what was the problem, because in the majority of the errors the user is the problem, in this case… myself. I am going home and I will soon investigate your attached hip!
cheers
Christos
cheers
Christos
Manu,
the scene works great and thank you again for that, next step dust.
but I have two questions.
In the dopnet_main where I had group with dynamic pieces which I wanted them to step stastic followed by the activevalue node setted to 0. You have disable both of them, is this for a reason, it was causing any problem to the simulation?
Lastly I would like to ask your opinion about accurate simulation and penetrations. Which parameter is more important to be increased, the resovle penetration, collision passes or substeps? Also because my scene, not the one attached… was test, will be in slow motion I have found that you can achieve this by dropping down the value in the scale time on the DOP Network node or by increasing the fps, which you thing is best to use and why? also I have test them both and I have seen that the change a bit the simulation not just slow it down but that's a minor problem.
cheers
the scene works great and thank you again for that, next step dust.
but I have two questions.
In the dopnet_main where I had group with dynamic pieces which I wanted them to step stastic followed by the activevalue node setted to 0. You have disable both of them, is this for a reason, it was causing any problem to the simulation?
Lastly I would like to ask your opinion about accurate simulation and penetrations. Which parameter is more important to be increased, the resovle penetration, collision passes or substeps? Also because my scene, not the one attached… was test, will be in slow motion I have found that you can achieve this by dropping down the value in the scale time on the DOP Network node or by increasing the fps, which you thing is best to use and why? also I have test them both and I have seen that the change a bit the simulation not just slow it down but that's a minor problem.
cheers
I disabled the active value and the grouping in that scene because I changed the voronoi setup to speed up the sim so the piece numbers changed and thus the grouping didn't work anymore.
Well what's more important? Always depending on your scene. In your case I increased resolve penetration because objects were sinking into each other. Substepping is important if you have really fast moving objects in your scene, if it's all average velocity you might get away with low values here. Collision passes is hard to say, depends on how many forces are affecting each object, a value between 5 and 8 works normally for me. Feel free to read about those in the documentation
For slowmotion I'd rather increase the fps instead of scaling the time value. The latter is mostly used to screw with fluid settings as far as I now, it changes the behaviour of your objects quite drastically. Depending on what your going to achieve you might get away with a post-stretch of your dop simulation, either in chops or with the time strech SOP!
Well what's more important? Always depending on your scene. In your case I increased resolve penetration because objects were sinking into each other. Substepping is important if you have really fast moving objects in your scene, if it's all average velocity you might get away with low values here. Collision passes is hard to say, depends on how many forces are affecting each object, a value between 5 and 8 works normally for me. Feel free to read about those in the documentation
For slowmotion I'd rather increase the fps instead of scaling the time value. The latter is mostly used to screw with fluid settings as far as I now, it changes the behaviour of your objects quite drastically. Depending on what your going to achieve you might get away with a post-stretch of your dop simulation, either in chops or with the time strech SOP!
Manu thank you!