A more precise way to model this curved surface than control point editing?


(Esmond Sit) #1

The above is a curved surface that I modeled by dragging the middle control points up. I have tried using many other methods e.g. RailRevolve, NetworkSrf etc as one whole surface or broken up surfaces joined together and it keeps giving me a surface with very subtle “X” shapes pinches from the center to the corners. Nothing could give me the same smooth results as dragging the control points of the surface.

The thickness from the border of the surface to the tangent or the tip of the curved surface needs to be of a specific value. By dragging control points, I can only eyeball it. Is there a more precise way of modelling this surface where I can control all the parameters?

Many thanks!

Edit: The border has to be flat/planar.

(David Cockey) #3

Can you post a .3dm file of the surface? Either “drag” the file to where you type a post, or click on the vertical arrow icon above where you type.

(David Cockey) #4

How precisely does the edge need to be planar compared to the size of the object?


Only a planar circular boundary would give you a surface with uniform curvature. If you are happy with your shape by dragging keypoints, you could possible get the top of the surface (observe the isocurves) so close to the z co-ordinate you are after that it is perfectly ok for manufacturing.

(Esmond Sit) #6

Curved Surface.3dm (102.2 KB)

Hi Dave,

Here you go. I’m not sure if its crucial for it to be planar but the edge is meant to continue into a fillet. I guess as long as its not obvious that the straight edges are not straight?

*The curved surface as shown above is hidden.

Hi Lagom,

You’re right. I can get it close enough to be suitable for manufacturing. I was just wondering if there was another way to it that I may not know.

I just realised after zooming into the border that even as I lift the control points, parts of the border gets lifted up.


Hello -

Here is the method I used to control the height of the dome shape using NetworkSrf. I split the outside shape at the corners so that the surface does not have a “pole” at the center.

NetworkSrf_Model.3dm (5.2 MB)

(David Cockey) #8

The border in @esmond.sit file has G1 continuity between the straight and circular arc segments. A surface (other than a plane) which matches the border exactly will have lines where the curvature is not continuous. That is basic geometry, not a limitation of Rhino. For the surface to have continuous curvature everywhere it will have to deviate from that border.

(Pascal Golay) #9

Hello- One thing to try for this which will keep the border fairly level, is to

1.ChangeDegree on the plane - I used 5 by 5
2. Arrage the arc curves by quadrants.
3. Select these curves and the level border curve.
4. Patch, using the planar degree 5 as a ‘starting surface’, Pull = 0.

Curved Surface_PG.3dm (135.7 KB)

@esmond.sit - here’s a little clip of the process I used.


(David Cockey) #10

Very nice. Deviation of the border from planar is 0.010 mm.

(Esmond Sit) #11

Thanks @pascal. I’m going to have to take some time to process this. :hushed: I actually lost you at 2 onwards.


Another way might be to use cage. Start by rebuilding your surface to a 3-degree 4x4 surface, show points and stretch the center up by some random amount, and then use a 2x2x2 cage. Select the top row of points in the cage & setpt them to the end of the curve that you want to match. The corners will not be perfect, as others have mentioned.

(Esmond Sit) #13

OMG thank you so much for that @pascal! I’ve never used Patch before but that looks really useful.

I really appreciate you taking the time to record a video!

(Esmond Sit) #14


I’ve always wondered if that was the case, that it was geometrically/physically impossible to have a smooth curvature whilst maintaining a planar border.

Thanks for the help!


Thanks for this! This was also one of the ways I’ve tried to achieve the surface. I always wondered if there would be any implications though of having the UV flow in that manner.

(Pascal Golay) #15

Yeah - in general you don’t want the corners to have those co-linear UVs. It makes the normal there disappear (normal is cooked up from angle between the u and v directions and is undefined if they are the same) so things like offsetting (and filleting which uses offsetting) are likely to be messy.



The higher the degree of the surface, eventually one has CPs that allow the sides to converge on a straight curve, a line; it is also perfectly planar. I would think your CP method with a single span surface yields the best curvature homogeneity, and you obtain the required height very accurately.

This squarish flounder here was a humble single span plane in its prior life CP flounder with planar edge.3dm (415.6 KB)