Prman Polygon Displacements

[meerkat]

hello,

i am having problems with displacements on polygons in renderman this is what i am getting:

also here is the shader that i am using:

 

displacement plast(
	float Km = 0.1;
	)

{
	vector NN = normalize(N);
	float mag = 0;
	float mag1 = 0;
	float y=10;
	normal Ndiff;

	point PP = transform("object",P);

	/*calculate mag*/
	float i;
	float freq=1;

	for(i=0;i<3;i+=1)
  {
  mag+=(noise(PP*freq)-0.5)*2/freq;
  freq*=2.1;
  }

	mag /= length(vtransform("object",NN));

	/*Displace*/

	Ndiff = normalize(N) - normalize(Ng);
	P=P+mag*Km*NN;
	N=calculatenormal(P);
}

 

 

does anyone know how to fix this problem?

thanks

amir

[AdamJ]

You have to add back the 'normal' difference between your shading normals and your geometric normals back to your final N.

N=normalize(calculatenormal(P)) + Ndiff

[meerkat]

I have another problem when displacing polygons have really hard normals like a cube, It seems to displace every face seperately. Is there any way around this without smoothing the normals?

thanks in advance

amir

[peliosis]

I have another problem when displacing polygons have really hard normals like a cube, It seems to displace every face seperately. Is there any way around this without smoothing the normals?

thanks in advance

amir

25693[/snapback]

Yes because houdini makes edges hard by separating points.

I'm not sure if that's renderman way of dealing with smoothness but it's a good problem to think of.

I'm not sure how other software deals with it, perhaps they use vertex normals not point normals. One point can share several coincident vertices, but it has only one normal and therefore whatever the angle between faces is, they will always be smooth, because they have no separate normals. Cusping mean disconecting points and getting more normals to manage.

Am I on my way?

I don't know the solution to this problem though.

[stu]

It's possible to solve the cracking problem by using two sets of normals: displace using consolidated point normals (soft edges) and shade using unique point normals (hard edges).

[Mcronin]

Yes because houdini makes edges hard by separating points.

I'm not sure if that's renderman way of dealing with smoothness but it's a good problem to think of.

I'm not sure how other software deals with it, perhaps they use vertex normals not point normals. One point can share several coincident vertices, but it has only one normal and therefore whatever the angle between faces is, they will always be smooth, because they have no separate normals. Cusping mean disconecting points and getting more normals to manage.

Am I on my way?

I don't know the solution to this problem though.

25695[/snapback]

You can use the crease SOP to add a creasewieght vertex attribute which will allow you to make hard edges in Renderman without splitting the surface. AFAIK creaseweight does not work with Mantra due to patent issues. The workaround in Houdini when dealing with subdivision surfaces, I guess, is to double up the geometry around edges you want to sharpen, which is what most modelers seem to do in other applications. If you render geometry as is in Houdini with displacements you will get your hard edges with subpixel displacement.

[Mcronin]

It's possible to solve the cracking problem by using two sets of normals: displace using consolidated point normals (soft edges) and shade using unique point normals (hard edges).

25704[/snapback]

That's something that never occurred to me. I'm going to have to try it in Mantra.

[frodules]

I had the same problem with hard edged displacement in houdini

and rendering "Polygons as Subdivision Surfaces" fixed it but seemed to screw up the uv's for the texture - strangely the displacement texture was still fine though?

Also, all t coordinates seem to be inverted so that i'm having to flip texture images vertically so that they orient correctly - not sure whether this is affecting the coordinates when rendering as subdivs?

and i'm getting these error messages in the shell

N02003 Subdivmesh /obj/pip1/pip1 has duplicate edge connecting vertices 431 and 417.: Context "split gprim": Object "/obj/pip1/pip1" (WARNING)

N02003 Subdivmesh /obj/pip1/pip1 has duplicate edge connecting vertices 417 and 423.: Context "split gprim": Object "/obj/pip1/pip1" (WARNING)

N02003 Subdivmesh /obj/pip1/pip1 has duplicate edge connecting vertices 422 and 431.: Context "split gprim": Object "/obj/pip1/pip1" (WARNING)

N02003 Subdivmesh /obj/pip1/pip1 has duplicate edge connecting vertices 423 and 422.: Context "split gprim": Object "/obj/pip1/pip1" (WARNING)

N02011 Subdivmesh /obj/pip1/pip1 is a trivial mesh (created 0 patches).: Context "split gprim": Object "/obj/pip1/pip1" (WARNING)

I tried adding a fuse sop to get rid of duplicate points but doesn't seem to make a difference

seems strange as the texture on the shirt is coming out fine

thanks

[frodules]

there were some duplicate points that the fuse sop wasn't getting rid of or edges or something - anyway - fixed now

[frodules]

would be quite interested to know about the flipped t-coordinate though