Boundary surface: A-B spline surface could not be created with the given definition

brendan_mitchell525

Hi all! 
I'm dipping my toes in the water with surface modeling by designing an ergonomic attachment for my mouse. I have some profiles sketched out that I want to use to make a boundary surface, but I'm getting errors that I can't find documented anywhere. The error is:
A-B spline surface could not be created with the given definition
The error shows up when I try to select the final edge of the boundary surface as shown here:

I've done the same thing with (seemingly) similar areas and gotten exactly the result I want. Any ideas?
Here's the CAD:
https://cad.onshape.com/documents/3bc0a79571b63a8b30eb7fb2/w/29ba987c49223e396f417b1f/e/98aedc8fc8f32dd1c32ea235

S1mon

It looks like you already fixed your issue.

glen_dewsbury

Try using a loft as can be seen in this example. Part 3 tab, I used the separate boundary surfaces in Part 3 copy but did not look as good. Turn on zebra strips and you'll see the difference.
https://cad.onshape.com/documents/2d2cf3d0b732941e8f78c63a/w/55b6fa827615a9b0b8bfddba/e/ddc7dc09dd6fc3ab239d09c2

EvanReese

in your screen shot you need another V profile. You need 2 U and 2 V

brendan_mitchell525

@Evan_Reese yes I know, the screenshot is showing the edges I want to select. When I select the 4th one I get the error described and the surface does not generate. 

brendan_mitchell525

@glen_dewsbury Thanks for the example, I agree this is better. I'm still curious about the meaning and reason for the error. It seems like my setup should have been valid for a boundary surface.

S1mon

3D Fit Spline generally produces crappy curves. I only use it in limited situations. Boundary surface is very fussy about having the U and V curves very exactly connect at the corners. It will trim the curves if, for example, the U goes beyond the extends of the V curves, but if they don't connect within a very tight tolerance it will fail. Pierce relationships are often better than coincident, even when you would think that they should produce identical results. Your cross curves (sketch 2 and sketch 4) may need to have pierce relationships with the 3D Fit Spline, not the underlying context curves. The fit spline and the context curves are not the same - the fit spline is an approximation of the context curves. The context curves from the imported CAD are not continuous, so trying to use composite (which doesn't rebuild the curves) won't work.

The other thing that I've been finding with Boundary is that it doesn't want to deal with overly complex inputs. If fitting to the inputs would require a much higher degree curve than it wants to use, it will just sometimes fail with very little useful info, or in at least one case I ran into, it will create a surface which doesn't match one of the input curves at all. 

If you have very clean inputs, Boundary will often do more intelligent things than Loft or Fill, but with crappy inputs (curves/surfaces/tangency constraints), Boundary is much more fussy.

In this particular case, I might try to use the curves from your context by using Composite (one at a time to copy) rather than 3D Fit Spline with a bunch of edges (which approximates using a very dense degree-3 curve). I would also make your cross curves line up with the breaks in the original model. So Sketch 2 should move to where these two edges join, but don't quite match. 

Fill will happily take a lot more crap and try to do something useful, but unfortunately in this case with all the cross curves as precise guides it ends up very lumpy. 

That said, if you move the cross curves around, or even just use the sampled mode with the guides, you might get something decent.

brendan_mitchell525

@S1mon wow, thank you! This was super informative. I appreciate the attention to detail!