Hi all.
With a simple 5 faces Sub-D subd.3dm (92.1 KB)
Zebra analysis shows good continuity:
But the slider to set Sub-D meshing is not used, so there is no control for better zebra quality.
Converting the Sub-D to mesh gives lets you properly use the slider and the result is a nice high quality mesh:
Converting the Sub-D into Nurbs gives this instead: (see near the center of the object)
G0
G1
G2 is even worse.
So it seems that from Sub-D to Nurbs tangency continuity is lost…?
I am doing something wrong?
Creating more cases is simple and seeing it on screen it seems a… defect on the shape.
I went back in thinking i miscliked something, instead it was just the conversion…
test subd.3dm (819.2 KB)
Visible on the attached pic, but much more visible on rhino while zooming near the “extraordinary vertex” (obviously)…
Tried with latest 7 Wip, 7 Evaluation and 7 Evaluation service release candidate, all same results.
Edit: I’ve just now seen Dale Lear’s comment on mcneel myjetbrains … lol … he was right!
Will, asking for smooth zebra stripes, backfire at something else?
Is this currently re-adressed?
I’ll surely buy Rhino7 update, and that mainly for SubD use… as many others, probably.
This seems a really important matter, to me…
The YouTrack issue I Iinked to has indeed been fixed. In your model, it improves the quality of the edge tangency on the edge pointing up in your image. Compare these two images:
The top one is the old default (G1 extraordinary vertices), the arrow pointing to the bottom of the second stripe of the right-hand side surface is facing about the bottom of the fourth stripe in the left-hand side surface. The bottom one is the new default (G1x extraordinary vertices), the arrow is still aimed at the bottom of the second stripe of the right-hand side surface and is now facing about the bottom of the thrird stripe in the left-hand side surface. So tangency is better on that edge. The others are unchanged.
The YT issue has more details and quantitative tangency measurements: edges that were mis-matched by more than 1° now have improved tangency. Some images are private but it should not remove too much meaning.
A good way to measure these deviations is to use the _EdgeContinuity command. On your model it shows that the maximum angle deviation on the edge pointing up is reduced from 2.05° (G1 option) to 0.83° (G1X option).
I agree that this does not achieve the results you are looking for yet, but we are planning on improving edge matching at extraordinary in ToNURBS incrementally. It will take some time before all cases can be made as good as possible, and we do not know yet how good this is going to be. It might not be possible to have perfect edge matching without dramatically increasing the CV count (such as what NX is doing, as is explained in the linked YT issue).
