Phantom Discontinuity At Mirrored Surface Seam

matthew_stacy

Why are faces generally split at the seam when a surface is mirrored?

In this example (https://cad.onshape.com/documents/51b9043cee7e3dbb60c1192c/w/31a742765743fe1fd962a776/e/7ad5ec31da1120d731ab044c) a lofted surface is shaped with guide curves.  In one Part Studio the guide curves are sketched "full-span", symmetric about the front plane.  The result is a single smooth contiguous surface.  In a second Part Studio only half of each guide curve is sketched.  When this surface is mirrored there is a visible seam at the plane of symmetry.  Turning on curvature visualization indicates that the underlying geometry appears to be identical in both Part Studios, but one has a seam.

Unless I am missing something these "half-span" guide curves are identical to their full-span brethren, with exactly the same tangency and curvature constraints where they intersect the plane of symmetry.  So why is the mirrored face split?  And is there a way to merge them into a seamless whole?

NeilCooke

@matthew_stacy - extruded and revolved surfaces are analytical geometry i.e. their shape can be described with simple math. A surface created by splines is a whole different ball game and @Evan_Reese is right - it is not possible (or not easy at least) for Parasolid to recreate the math for the whole mirrored surface with perfect accuracy. 

Evan_Reese

Interesting point, Neil. I hadn't fully thought about what's happening when it merges those simpler surfaces. Is it making a new surface and replacing the old one (including IDs and all that)?

Either way, I don't personally think I'd get a lot of value out of making symmetrical surfaces with no seam any more than I would by removing tangent edges of fillets. The only time I would want to do something like that would be to have a clean single surface to drive an Attractor Pattern feature with. To map a pattern across multiple surfaces I may have to spend a decent amount of time making that clean base surface that's similar enough to the parts I want to texture.

Evan_Reese

Just because they are curvature continuous doesn't mean they are the same face. Each surface has a UV grid (kind of like an XY grid, but warped to the shape of the surface). Let's say that each surface has a UV grid that goes from 0 to 1 in U and in V. That would let you find any point on the surface by using UV coordinates. In your first example, the surface grid looks like this, and goes all the way across the form. the middle split line would fall on the 0.5 coordinate of the U direction (Assuming U runs right to left).

In your second example, you made a surface that's half the size and it's UV curves look like this. see how the UVs are all smushed together?
In this layout, your "split line" falls on the 1.0 coordinate of the U direction. Just because you mirror it, doesn't create a whole new surface with coordinates spanning the whole form. You get two of this half-surface. While your two example parts are, for all intents, the "same" they aren't mathematically the same.

To merge them into a whole, you'd have to recreate a similar (but not mathematically identical) surface to replace them with. I actually don't recommend this and prefer to have a split line where I want true symmetry, because it is possible for the part to look symmetrical, but not actually be symmetrical, which can cause other issues down the line. The split line is a clue that it's perfectly symmetrical.

matthew_stacy

@Evan_Reese, adding UV curves (Face Curves FS) is a really great visualization tool.  Good idea.  I appreciate your response and respect your opinion.

However I would argue that you need to reduce the U-curve count by 50% because the length along each V-curves is precisely halved.  Then the grid looks identical, not "smushed". 

Suppose that you are modeling a smooth formed panel, with no crease or visible edge at the plane of symmetry (e.g. automotive air dam, hood, roof, trunk lid, or spoiler).  To me, the face boundary is a graphic artifact of our CAD software.  If I saw that line on the real part I would likely soil my knickers and be wondering if we could stop payment on the stamping dies.

Symmetry is a powerful tool that can make the modeling process vastly more efficient.  It seems unfortunate to have to take the mirroring/patterning workflow off the table just to mitigate a graphic anomaly.  This issue seems to be limited to a certain subset of SURFACE models.  Extruded or revolved solids and surfaces mirror seamlessly as shown in the screenshot below (unless the design features a slope or curvature discontinuity at the plane of symmetry).  What is it about the geometry in my original example that results in two faces?  I haven't yet been able to discern the rules governing when Onshape splits a mirrored surface.  Perhaps the software just presumes the possibility of a discontinuity in the 3rd order derivative?

Is it unreasonable to expect that a face boundary indicates a surface discontinuity or an intentional split?  At the least it would be nice to be able to boolean faces together (if they have full position/tangency/curvature continuity).  I tried to delete/heal face but that didn't work.

NeilCooke

@matthew_stacy - extruded and revolved surfaces are analytical geometry i.e. their shape can be described with simple math. A surface created by splines is a whole different ball game and @Evan_Reese is right - it is not possible (or not easy at least) for Parasolid to recreate the math for the whole mirrored surface with perfect accuracy. 

Evan_Reese

Interesting point, Neil. I hadn't fully thought about what's happening when it merges those simpler surfaces. Is it making a new surface and replacing the old one (including IDs and all that)?

Either way, I don't personally think I'd get a lot of value out of making symmetrical surfaces with no seam any more than I would by removing tangent edges of fillets. The only time I would want to do something like that would be to have a clean single surface to drive an Attractor Pattern feature with. To map a pattern across multiple surfaces I may have to spend a decent amount of time making that clean base surface that's similar enough to the parts I want to texture.

matthew_stacy

@NeilCooke and @Evan_Reese thank you for sharing these insights.  I truly appreciate it.  This forum is a huge asset to the Onshape community.

Evan_Reese

matthew_stacy said:

This forum is a huge asset to the Onshape community.

Hear hear!