How to emit particles from a displaced surface.

[CSN0309]

Hi~guys

Can i distribute particles from the surface that already been displaced by a displacement shader(VEX)? i already applied a displacement shader to my surface, but i want to emit particles from this displaced surface not from the original surface. Any suggestions?

Thank you

[Hazoc]

I suppose scattering points onto your surface and displacing them with vopsop using similar displacement patterns as your displacement shader does. Then using these displaced points as your particle source could do the trick. I can make you an example scene when I'm less busy.

[quinniusmaximus]

Could you use the vopsop as a way to create groups from which to birth your particles?

or conversely to group and delete those particles that are not displced?

[Hazoc]

Could you use the vopsop as a way to create groups from which to birth your particles?

or conversely to group and delete those particles that are not displced?

Was this assigned to me ? Anyhow, sure it's possible and should be done, since all the non emiting points would be just useless geometry. This will also give you a better idea of the shape of the emiter.

[CSN0309]

Thank your both~

Hazoc could you explain it more specifically? i'm waiting for your example,

[petz]

within your shader you could write out a pointcloud of the displaced geometry and use it then as particle source.

petz

[edward]

I'm with petz. The potential problem with displacing at the SOP level is that you won't get all the displacement detail that you get when displacing at render time.

[CSN0309]

petz, can you give me an example about how to use pointcloud from my displacement shader in VEX network? i was thinking about i can use pointcloud to get the SSS effects, but i never thought i can use pointcloud to emit particles.

[petz]

file is attached

petz

disp_particles1.hip

[Hazoc]

I'm with petz. The potential problem with displacing at the SOP level is that you won't get all the displacement detail that you get when displacing at render time.

Sure. The approximation is coarse compared to the pointcloud way. And it's nice to have all in one shader.

BTW is there any neat way to link shop side with vopsop in this case if some one was to use the sop method ? Instead of maintaining two separate yet pretty identical shading networks.

[edward]

BTW is there any neat way to link shop side with vopsop in this case if some one was to use the sop method ? Instead of maintaining two separate yet pretty identical shading networks.

No. If you have code though, some preprocessor tricks could probably be used to minimize the code duplication.

[CSN0309]

Thank you Hazoc, i already made it by the method that you provide to me. but i have to maintain the two separated displacement shader as the same,it's not very convenience so. anyway thank you.

and thank you all to give me your suggestions . thank you for your example petz, i'm studying....

Edited December 30, 2009 by CSN0309

[ranxerox]

hey Petz, very cool example, thanks. I took a look at your scene and was trying to make it work with an animated displacement. I was thinking I could just put a '$F' into the name of the pcwrite1's filename parameter, but it says you cannot have channels which depend on time. So I thought I'd make a parameter to the shader to take the current frame, and then compose the string for the filename out of this.. but I'm a little stuck. I was looking for something like a 'sprintf' VOP but I can't find one. Any ideas ? Or is there a better way ?

thanks

-ranxx

file is attached

petz

[petz]

for using $F you need to feed a parameter vop into the filename of pcwrite. have a look at the attached file.

petz

disp_particles_anim.hip

[ranxerox]

thanks a lot petz !

-ranxx

for using $F you need to feed a parameter vop into the filename of pcwrite. have a look at the attached file.

petz

[pclaes]

interesting use of pcwrite! And nice way of reloading the file sop in the ropnet!

It seems camera dependent though. (So rendering the pointcloud from a better positioned camera - that covers more surface will give you a better pointcloud.)

What would be the best way to get points on the entire surface? Is there perhaps a render property that will force the entire object to be rendered (like render backfaces or something similar)? Currently it seems only the polygons visible from the camera will generate points.

Some other things I was wondering about?

- multiple cameras with multiple pointclouds merging them all together in one big pointcloud and performing some filtering would probably solve most scenarios.

- or perhaps rendering in a different space? like in uv-space?

- what about an adaptive pointcloud? more points where there is more detail.

[mawi]

What would be the best way to get points on the entire surface? Is there perhaps a render property that will force the entire object to be rendered (like render backfaces or something similar)? Currently it seems only the polygons visible from the camera will generate points.

I think disable vm_hidden on the mantra node(add it first) will do this.

[petz]

interesting use of pcwrite! And nice way of reloading the file sop in the ropnet!

It seems camera dependent though. (So rendering the pointcloud from a better positioned camera - that covers more surface will give you a better pointcloud.)

What would be the best way to get points on the entire surface? Is there perhaps a render property that will force the entire object to be rendered (like render backfaces or something similar)? Currently it seems only the polygons visible from the camera will generate points.

Some other things I was wondering about?

- multiple cameras with multiple pointclouds merging them all together in one big pointcloud and performing some filtering would probably solve most scenarios.

- or perhaps rendering in a different space? like in uv-space?

- what about an adaptive pointcloud? more points where there is more detail.

the mantra rop which is used for pointcloud creation has vm_hidden disabled. so it generates a pointcloud for the entire geometry instead of just the visible one. you can find it under the mantra render properties. in the example file you need to increase the shading rate to see the points on the entire surface.

to get an uniform distribution add scanline measuring to your dicing parameters.

but anyway, the pointcloud generation is camera and resolution dependent, since the points are based on micropolygons as long as you are using micropolygon rendering.

- what about an adaptive pointcloud? more points where there is more detail.

one posibility would be reducing the pointcloud based on curvature in an postprocess. the other which i can think about is using irradiance cache for generating the pointcloud. so you would get more points in areas with more detail. if you expose min - and max spacing you have great control how your pointcloud get generated. the problem is that one more step is needed for this.

example file is attached

disp_particles2.hip

Edited January 6, 2010 by petz

[Raymond Chua]

I'm sorry for this. I got lost. How did petz generate the point cloud file?

[pclaes]

I'm sorry for this. I got lost. How did petz generate the point cloud file?

It's done in the shader. He extended it by adding a pcwrite.

So every part of the object that is being diced (every part of the object that is within the render region) will create a point cloud.