Erm… There should be no problem.
Try joining the cube and bits (except the patch) then choosing the edge faces rather than the curves to make the shape.
Oh… one thing I just noticed I did differently was I drew a circle tangent to the edges of the face then trimmed it rather than a control point curve… Shouldn’t make any difference?
There’s a chance that your curve isn’t coplanar to the face that you were woking on so make sure that you’re working with “project” or at least with the CP set to the face you’re working on.
What I’ve just done is drawn one side with control points and upped the number of sample points and spans - which fixed it:
Why not do a _Sweep (Sweep one rail)?
Select the bottom edge and the two other edges. You get a nice blend:
I generally don’t use the patch command too much because I don’t like the edge flow and it’s difficult to manually repair tangency. I much prefer solutions like this:
Patch should be used as a last resort hack…
I’d go with André here and try to create square topology to close that corner like the last example.
If you have a lot of these blends you should look at TSplines or VSR.
I posted a video on Youtube for creating a nice corner blend. Here’s some good information on NURBS surface continuity. Without understanding the underlying concepts of surface continuity and how to work with it, it’s very difficult to consistently create beautiful, valid surfaces.
@rhinouserx: Nice technique in the video. Thanks for sharing this.
@Cajodo: As mentioned by others, Patch has a number of challenges; however, it can be useful in a few complex or irregular situations. The one setting worth playing with to improve watertightness is the Stiffness. The lower you can go, the better chance of being watertight (but watch for sagging of surface). Try a Patch of 0.10 Stiffness on your model and Analyze Edges. It might work. If not, go lower. Likely the surface will not be satisfactory, but you’ll get the idea of pros and cons with Patch.
I’m not convinced, yet… if the corner is tangent and among three edges and the radii are constant, it is hard to see an advantage of this over simply making a tangent sphere…? and if the radii are not constant, will the curves intersect well enough to use this procedure anyway? I’d like to see a more useful application of this idea, where there is an obvious advantage. i.e. either the real and exact answer is not a tangent sphere, or there is some concrete reason to have the isocurves flowing in a particular way.
I’m no expert, I’ve only been at Rhino for a few months now. Pascal, you really got me thinking about this one. I’m not sure I’m completely convinced anymore either on my exact approach (it certainly doesn’t work across all situations as well as I’d like). But nevertheless there are a number of reasons one might want to layout make the topology follow the edge loop flow (and conversely, a number of reasons to use the standard Rhino tangential spheres).
Curvature Analysis shows that my manual solution is inferior to the tangent sphere in terms of curvature, but it’s still pretty good. I’m not sure the human eye would be able to pick up the differences. The VSR Shape Modeling Multi Blend is 4 clicks/ 4 seconds and blazing fast to execute, but the curvature isn’t quite as good with the default settings (with shading on it looks a bit flat in the centre due to the method for controlling tangency). Depending on how visually/ structurally important the blend is, VSR is most times be my go to solution for this sort of thing. I’ve barely begun to scratch the surface with VSR (and I know one can get better results than what I am getting with it). I’ve been more focused on learning what I can do without plugins.
Here’s some pro’s for what I did and the VSR Multi Blend style of topology for a three-sided blend:
- After converting this type of corner blend to a mesh, the topology is cleaner, more evenly distributed, and symmetrical. Look at the triangles along the edges.
- Texturing will be cleaner (think about where the texture seams are going to be and where the pixels are going to meet). It’s lot easier to layout clean UVs (better aligned UV shells). There is less pixel stretching at UV seams and less texture distortion.
- Am I correct here? The tangent spheres appear to be trimmed surfaces (explode the filleted box and untrim the tangent spheres). There are no trimmed surfaces in my solution.
- What if the result achieved is not the final stage in the modelling process? Starting off with a trimmed surface may not be optimal. Controlling the edge flow/ edge loops in the model can give more control over adding localized detail (and in some cases it may give you the option to use untrimmed surfaces).
- It converts to T-Splines with no issues. It doesn’t work cleanly with a tangent sphere.
- Exporting to other software, having poles and triangles can cause pinching (Zbrush, sub-d surfaces, etc). The topology needs to be as even as possible for subdivision surfaces otherwise pinching will occur and the model won’t sculpt cleanly.
I’ve been trying to work out a somewhat similar method for doing less irregular combinations of surfaces, but it so far I haven’t found a better solution than VSR’s. Some of the methods I’ve been trying have been working pretty well when the surfaces are planar and the input edges have different curvatures. I used a slightly different method for this. The cuts on the three surfaces generated by the one rail sweep need to be made at the centroid of the surface created within the corner points):
It turns out pretty close to the VSR Multi Blend result (but takes significantly longer).
I am just starting to work out a better method for irregular 3-edge non-planar blends. I haven’t quite got it yet but I’m getting there (some issues with tangency), but it’s getting passable (the case in the image, the edges all have different curvatures and concave/ convex).
More or less I’m trying to work out a manual way of doing some of what VSR Shape Modeling does. I own VSR, but I just want to understand more clearly how to create Class A surfaces so my workflow become more intuitive.
Thanks for asking good questions Pascal. It’d be great to get some feedback (or links) from other users. I’ve seen a few good posts and here but I’d still like more information. I hope what I’ve posted is of use/ interest for someone.
cornerBlend.3dm (891.1 KB)
Hi Andre- I think the overall arrangement is perfectly fine - I can understand one might need untrimmed surfaces, UV texturing requirements, etc. And where the 'right; answer is not an obvious and exact thing like a sphere, I like this arrangement BUT there is a lot of fussing around locating that one point that you need in the middle- I’ve messed a little with trying to find reliable ways to locate that point, I’ll see if I can dredge up a file.
Indeed it is, as you are experimenting with core design issues to be solved and you are demonstrating that each design can require a different approach depending on the situation/results, both functional and desired if they diverge.
Keep up the good work! Non “standard” edge “closures”, where common tools such a 4 edge or blends are not appropriate, remains a hassle with nurbs in certain situations. Conversely, “closures” with tools like t-splines, which will created a “beautiful” surface with ease, may not easily create the exact surface you want or need, something nurbs excels at.
This all looks cool but you’ve all moved away from the original posted geometry, the faces don’t have square corners so a tangential sphere would not work in that way?
What are your suggestions if you move back to the original:
I’m writing down all your steps so far though
I still fiddling around with this and so far I have not come up with an elegant manual solution, but I’m working on it. I have a few ideas and I’ll post something when I figure a few things out. There is a similar thread in the Rhino Windows forum but I don’t think there has been much luck over there either.
See if this post is of use.