Modeling the tip of a cone-shaped tip

lawrenceyy · October 20, 2016, 10:48pm

What is the best way to finish off this shape:

Sweep2 from one of the edges to a tip does not produce good shapes. I want tangency between all surfaces at the very least.

161020 Temp.3dm (165.4 KB)

pascal · October 20, 2016, 10:58pm

Hi Lawrence - this is a perfect place to use a loose loft - see the attached file.

Split the shape curves at the top of the existing surface.
Turn on points for the split curves. They both have four points, and are degree three.
Copy the edge curve of the existing surface and scale the copies to hit the points of the curves.
Loft the curves to a point , the point is where all your curves converge. Set the loft style to ‘Loose’.

161020 Temp_LooseLoft.3dm (246.8 KB)

The trick here is that loose Loft with the Loose option exactly follows a degree 3 (uniform) curve if the loft inputs are on the matching pairs of curve points.

-Pascal

lawrenceyy · October 20, 2016, 11:15pm

Interesting, I did not realize this is how loose loft works. One problem though, there seems to be continuity issues between the original surfaces and the new one. Matchsrf does not fix this problem for some reason.

pascal · October 20, 2016, 11:21pm

Hi Lawrence - hmmm - vertical stripes and a finer analysis mesh looks ok here. CurvatureAnalysis is also pretty good…Horizontal stripes is off though as you say… MatchSrf for Curvatutre, with Refine checked. Direction = Automatic.

-Pascal

Stratosfear · October 20, 2016, 11:25pm

Here’s two more topology options.

lopacki · October 20, 2016, 11:27pm

Lawrence,
I just split your curves at the bottom and then used network surface. Not sure how accurate you need this but it comes out looking just fine. An image and the file.
Danny

finalTemp.3dm (240.6 KB)

lawrenceyy · October 20, 2016, 11:28pm

Good call. How do you keep matchsrf from breaking the continuity of other edges? Now when I mirror the surface, there an obvious mismatch.

pascal · October 20, 2016, 11:36pm

Hi Lawrence, before spending too much time at the tip, I’d make sure the existing surfaces are also nicely matched, which they are not, quite - match them all the way around for Tangency, with Average set. Then do the matching at the top:Tangency in the ‘vertical’ direction, tip surface to base surface, no Average, then for Tangency between the new segments for Tangency with Average, all the way around. Any better?
161020 Temp_LooseLoft.3dm (225.2 KB)

-Pascal

jeff_hammond · October 20, 2016, 11:44pm

you could RailRevolve the whole shape in one go:

… you might not like the point at the tip’s tip… might be ok for this though.

lawrenceyy · October 20, 2016, 11:57pm

Thanks pascal. It worked.

My assumption was that if the surface was drawn (sweep2 using long verticals as rail) from tangent sections that the surface would also be tangent- therefore I would not need to matchsrf. Is that not true? Shouldn’t that also guarantee G2 continuity since the curve sections are symmetrical/mirrored?

Robb · October 21, 2016, 9:33pm

try edge curves. I’ve attached a file that uses that method along with a windows snip of the completed polysurface.161020 Temp.3dm (6.9 MB)

Robb · October 21, 2016, 9:42pm

timglobal · November 4, 2016, 11:53am

Hi, guys, I hope I can revive this topic once again. Could you help me with this one? I tried Network surface but I feel that this surface might need some work first. I always have problems with non-uniform(or what I’m trying to say not round or oval) shaped tips that need to be capped.

Handle.3dm (1.2 MB)

I hope patching can be avoided, and something smooth and nice could be achievable.

jim · November 4, 2016, 4:05pm

Here is one possibility using BlendSrf

BlendSrf_Handle.3dm (1.2 MB)

pascal · November 4, 2016, 4:06pm

Hi Tim - I’d still use a loft here, just set up a little differently -

Set a Cplane to the planar opening
DupEdge the edge there
Mark the AreaCentroid of the edge curve
Scale2d copies of the curve inward.
Move the middle curve and the point together off the plane - vertically that is.
Move the second curve up, but less/
Using History, Loft to the point, using the Loose loft style.
Always keeping the point and the middle curve in plane with each other, use Gumball to move and scale the input curves until it looks nice, and has reasonable continuity with the main shape
MatchSrf for Curvature to clean up - make sure Preserve other end is set to Tangency. (Breaks history so get it close first)

HANDLE_PG_11-4-16.3dm (1.3 MB)

-Pascal

timglobal · November 4, 2016, 6:09pm

@jim Hi jim, thank you for reply, I tried to replicate the result. You blended original surface first, then trimmed out the ugly generated top to cap it… here I’m a bit confused. Blue curves try to match the curvature as close as possible but not precisely, right?

@pascal, gotcha. Faith in loft restored XD. Here is a third try, I guess the “feel” comes with practice. I assume your example wasn’t the result of a magical first try )) Gonna keep trying, thanks for this awesome technique.

pascal · November 4, 2016, 6:15pm

Hi Tim - make sure to adjust the mesh in Zebra - set Minimum initial grid quads to say, 5000. Or more in this case. Setting max edge length might be good in this case as well since one surface is so much larger than the other.

-Pascal

timglobal · November 4, 2016, 7:07pm

Ok, I will note that. I cranked the slider all the way to the right on simple controls, I think it was just a poor loft that time. I’ll keep trying. Thank you, @pascal.

timglobal · November 5, 2016, 10:31am

Slightly better, it takes a lot of time and “feeling” to make it right.