I am trying to (re)create a 3D surface starting from a very old 2D paper drawing (scanned). I want to end up with a low CV-count surface (say 6x6) so that it is more easily manipulated, yet is as close as practical to the original drawing. See attached image.
frame 1 is the scanned drawing with (assumed) evenly spaced “hand drawn” sections representing sections on the surface.
frame 2 is a view of 11 degree-5, single span curves drawn over the original 11 section lines.
frame 3 is a perspective view of these 11 degree-5 curves after they were moved lengthwise into position.
frame 4 is a loft (using normal) of the section curves with CV points turned on.
frame 5 is a rebuild of the surface to degree 5x5 single span each way with points turned on. And isocurve line density set to 5.
frame 6 shows contours made from the rebuilt surface with the contours spaced to match the original drawing section spacing.
Observations:
The contours shown in frame 6 end up pretty close to the original section curves, but not perfect. Maybe the original drawing is not as fair as it could be, or the sections were not taken at exactly the right spacing, or the paper changed shape, or maybe the 3D model still needs some tweaking. The isocurves are nicely spaced/shaped. The ends of the surface still need editing, so that is a known issue to be worked on.
But, the CVs in the rebuilt surface are located all over the place, at least the ones along the v-direction. Adjacent CVs oscillate well above and the well below the surface. My intention is to end up with a relatively low CV-count surface (say 6x6) and then nudge some CVs by hand to see if I could correct some of the mismatches in the contour lines. But the CVs ended up all over the place and I cannot keep them sorted out on the display.
So, is there anything in my workflow that I can do differently to end up with a more orderly arrangement of CVs so I have a more easily edited surface.
All of these could be causes but it very likely that degree 5 with 6 control points is not sufficient to accurately represent the input curves. More control points are probably needed to accurately represent the input curves. And possibly using degree 3 rather than degree 5. Yes, the resulting surface will not be single span but a single span surface may not suitable.
Rebuild the original lofted surface using degree 3, not degree 5. Experiement with the number of control points to find the minimum number in each direction which still matches the input close enough. The higher the degree of the rebuilt curve or surface the more likely is oscillations of the control points.
Thank you for those comments. That gave me a way forward, I think.
One thing I discovered after posting earlier is a Rhino command that permits rebuilding the surface in one direction only, which prevents my CV’s from being jumbled up. This command only creates degree 3 in the chosen direction, so there are 3 spans in that direction when I specify using 6 points. And some loss of surface continuity that direction, which may not be too important. The section contour match with the original curves is a somewhat worse than the version with the jumbled up CVs. So, as you say, I need to experiment some to see what gives acceptable results.
