@Pascal - Is there a hidden way to change the number of sample points used by Rebuild?
I don’t believe the difficulty Rebuild has with accurately rebuilding a degree 3 curve with 4 control points to a degree 3 curve with 4 control points is related to the number of points used for sampling. Rather I believe the fundamental cause is that Rebuild distributes control points evenly. I’ve had similar problems attempting to rebuild more complex curves with unevenly distributed control points.
Exactly - that was my point above.
-Pascal
So what did you mean by:
Did you mean use more control points in Rebuild? What does that have to do with the intervals Rebuild samples at?
Hi David - the more points you ask for, the smaller the spacing between samples in Rebuild. As far as I recall, there is no more smarts than this - sampling and interpolating - in rebuild unless it is to maintain the end tangents on degree 3 output.
-Pascal
Thanks for the reply. Based on a bit of experimenting it looks like Rebuild just samples the curve at the ends a set of interior points equal to requested degree-2, and then interpolates. Not as sophisticated as I thought it was but that explains some of the unexpected results I’ve encounted using Rebuild. I was assuming a fit based on minimization of a quasi-integral or integral deviation function.
That’s not what the OP struggles with. He suffered from what at Mitsubishi (early 90s) we called “pointilism” - massively over-complex curves that also sport the wrong degree.
He should simply draw his curves as single span, making it easy for himself.
Glad to hear there is a better way to do this. What is a single-span curve? It sounds self evident but it is not. What is a patch layout?
What I have done since the original post, in order to keep the project moving along, is to use FitCrv, per David’s suggestion. I also tried Rebuild non-uniform, but it was not too clear how to make it work in this instance.
FitCrv immediately created a curve very close to my original. I tested it with Explode and found it could not be Exploded. I also used the Symmetry command to recover tangency at the nose.
FitCrv was so quick and easy there must be some penalty for using it, but it is not clear what that might be.
Thank you for your insights.
Michael
Hi Michael -
If the fitted curve is dense, as is more than likely, then the tangency match will be, potentially, only ‘local’, that is, since only the first two points on the curve get changed to get the tangency match, it’s always possible that the tangency match is more pro forma than actually useful. I would check with the CurvatureGraph and make sure you have a clean graph. One advantage of simple curves is that the curvature graph is that much more likely to be clean and progressive.
A single span curve or surface is one that has the smallest number of control points for a given degree, that is, Degree +1 control points. It’s ‘single span’ because there are no internal knots.
-Pascal
See this example. A patch layout is the network of surfaces needed to model a form, by building to common boundaries, hypotheticals or by way of overbuilding and then trimming. Regarding curves, spans and quality/simplicity, maybe have a look here as it applies to all NURBS modellers. You’ll become a better Rhino user soon that way.
My first example is pure single span. A single span - nurbs has one span, meaning it has one Bezier Segment, making them de facto a Bezier curve. Single span curves do always have a degree of cp count -1.
This ensures a good curvature flow no matter which control point is modified. FitCrv does not produce single span, so the result is not helpful. But as seen in the missing reaction on this thread again, McNeel an Others do have another opinion on that. So its up to you to decide if using single span shapes or not. In Automotive people tend to use single span, whereas other industries such as naval does have other philosophies here.