I have experimented a issue with the way that Houdini is computing point normals in the next basic setup.
A tube, a box and a CopyToPoints or CopyStamps. When one of both radius values from tube is different and resulting shape is conic, point normals seems take a spinning effect. Usually average should produce a constant value in a regular shape like that, but each normal is rotating over its own axis in different value.
Someone knows how to solve this behaviour? Thanks!
Issue with point normals and spinning effect.
2173 5 1
-
- AlbertoGZ
- Member
- 34 posts
- Joined: Jan. 2017
- Offline
-
- tamte
- Member
- 8539 posts
- Joined: July 2007
- Offline
you are missing up vector v@up attribute
you can create one by plugging Poly Frame SOP after your tube and rename tangentu to up
you can also change style to First Edge to get y axis aligned exactly with one of the edges
you can create one by plugging Poly Frame SOP after your tube and rename tangentu to up
you can also change style to First Edge to get y axis aligned exactly with one of the edges
Edited by tamte - June 19, 2020 14:40:20
Tomas Slancik
FX Supervisor
Method Studios, NY
FX Supervisor
Method Studios, NY
-
- AlbertoGZ
- Member
- 34 posts
- Joined: Jan. 2017
- Offline
Thank you Tamte!
Using up vector it looks better orientation, but still rotate the instances when radius tube changes.
I see also that Polyframe is setting orientation 90 degrees for some instances, even with style from centroid, when radius1 and radius2 is set at same value. If use grids instead boxes as instances to testing, then I see that.
Using up vector it looks better orientation, but still rotate the instances when radius tube changes.
I see also that Polyframe is setting orientation 90 degrees for some instances, even with style from centroid, when radius1 and radius2 is set at same value. If use grids instead boxes as instances to testing, then I see that.
Edited by AlbertoGZ - June 19, 2020 17:34:05
-
- tamte
- Member
- 8539 posts
- Joined: July 2007
- Offline
it does, because that direction is computed from N attribute
as an easy fix for your specific case you can append point wrangle after the polyframe :
EDIT: unless you mean you are still geting some twisting, which may is possible if you didn't switch Style to: First Edge
as an easy fix for your specific case you can append point wrangle after the polyframe :
v@N = normalize(v@N * {1,0,1});
EDIT: unless you mean you are still geting some twisting, which may is possible if you didn't switch Style to: First Edge
Edited by tamte - June 19, 2020 17:49:38
Tomas Slancik
FX Supervisor
Method Studios, NY
FX Supervisor
Method Studios, NY
-
- AlbertoGZ
- Member
- 34 posts
- Joined: Jan. 2017
- Offline
-
- AlbertoGZ
- Member
- 34 posts
- Joined: Jan. 2017
- Offline
-
- Quick Links