Phantom polygons:

Phantom polygons:

* For a model of Antoni Gaudi´s Sagrada I use Booleans to construct the hyperboloids he uses extensively in the vaulting.
* The two elements of the Boolean are:
1 The “plain” hyperboloid, a lathed linear spine.
2 A “star-shaped” box to generate the somewhat complex bounding perimeter which interacts with adjacent hyperboloids and the angled columns.
* This box is just a simple cuboid primitive with suitable sections. I select specific edges and perform normal moves or transform the position along the X- / Z-axis. Needless to say, these are trivial operations.
* Occasionally, by no means consistently, these operations generate phantom polygons, typically "folded" triangles where a quadrilateral should have been. Somewhat strangely, the poly can not be selected, unless the “valid” part of the face is clicked on. However, upon selection the entire poly is highlighted. When switching the poly-selection to edges, only the proper edges are shown. Indeed, there is no edge (in edge mode / wireframe mode) on the outer side of the phantom triangle.
* I dimly remember a thread on this issue some months ago but I can´t find it.

* This bug has no effect on the Boolean, so it is irrelevant to my model. Clearly, however, it may cause problems when the distorted box is used as a base for further “straight” modelling. Unfortunately - upon deletion - both fill hole and create polygon simply reestablish the phantom triangle, even after all four points of the quadrilateral have been selected. The incorrect “doubled over” geometry is just recreated.

* Any idea what is going on here?
PS The problem persists in V7.
Servus from Vienna

Well I´m not going to build the Sagrada just to test what you´re describing - but is the twisting of the centered object somehow necessary?
Can´t you just upload that portion or an example file?

Cheers
Frank

* My current model is in V7, partly because of the superior Booleans (drool :wink.
* I have uploaded just the edited box for the Boolean, as the hyperboloid / the Boolean is irrelevant to the bug (?).
* The wire mesh is is the original primitive object, a cuboid with 4 sections in X and Z.
* The shaded object is the edited element which serves as a constructor for the Boolean.
* Note the marked folded over triangles, which should be quadrilaterals.
* I could model the hyperboloids by other means, e.g. by a lathed hyperbolic curve, but it would still require the Boolean functions to "cut out" the bottom "brim" where required. Doing this manually would be an option, but there are dozens (I think I have 18, but more are required) of tailored shell-templates which need to be flexible for further corrections. Parts of the vaulting are impossible to record on camera, as the columns obstruct significant details when shot from the floor. Climbing up the stairs adjacent to the choir does not help. I guess a drone plus remotely controlled camera / lens would help but it may not be legal. I can control my camera via my iPad. but I doubt if this works when the camera is 45m up in the air and I have to steer the drone as well.


* The Sagrada won´t be completed until 2026, so there is no hurry :tongue:

Just to indicate the complexity of Gaudi´s concept, a view of an old model of the hyperboloid vaults in the nave / aisles / transept / crossing / apse and ambulatorium.

That is the same glitch as here: http://www.cheetah3d.com/forum/showthread.php?t=11609
Blender and Silo do have the same difficulties with displaying such extrem "negative transposed" polygons:
You need triangles at those spots.

Cheers
Frank

Thank you, Frank.
Some fine tuning seems to solve the problem.

Just in the case you need solid quad topology and don´t mind the inner structure you can go with this for example:

Cheers
Frank

* For anybody with a similar problem:
* The "simplest" solution in my case was an inner extrude on the top & bottom quads (selection by area in front view, standard parameters, preserve group).
* I can then move the edges as required by normal move / transform without any glitches of "negative transposals" of polygons.
* Seems to work fine in V6 / V7 :smile:

Helmut said:

* For anybody with a similar problem:
* The "simplest" solution in my case was an inner extrude on the top & bottom quads

Click to expand...

Really?

Cheers
Frank

* The solution (or not) seems to depend on when the inner extrude (IX) is applied.
* I have IXed the original box (see grid of the example) and this seems to work.
* You seem to have IXed the box after pulling out some edges where it clearly does not work.

* I assume there is a proper replicable logic behind this. If so, I don´t understand it :frown:
Cheers from Vienna:roll:

I simply opened your file - made the box invisible - select top+bottom polygons -> inner extrude - no change.
What ever you might think to have as a workaround you´re not going to win the first price in the world´s championship of superior class modeling style. with it
Still want to know why the "twisting"?

Cheers
Frank

* A hyperboloid is a twisted ruled surface. Ruled surface means that it can be generated by rotation (C3D = lathe) of a straight line. Typical ruled surfaces are barrel vaults, hyperboloids and hyperbolic paraboloids.
* Unlike constructing a dome, you (well, the builders) can use straight / linear components in constructing the “mould” (aka formwork) for a hyperboloid and can use straight reinforcement steel bars in the concrete shell.

PS Your evaluation of my skills is entirely uncalled for. This forum is not the stage for for ad hominem attacks.

@ Frank B:
* Even if I replaced the hyperboloid (as a lathed linear angled spline) with a cylinder plus a bulge (actually pinch) modifier, I would still require the Boolean box constructor to generate the zig-zag / star-shaped geometry of the bottom edge of the pseudo hyperboloid (see attached screen shot of the vaulting in the nave from an "old" model). Clearly, this can´t be a solution.
* As stated below (or above, depending on your sorting of postings), the hyperboloid is not the problem. As also stated, the glitch in editing the Boolean box-constructor is not a problem either, as it has no effect on the Boolean result unless the two elements are poorly positioned.

* This was primarily a note on a dysfunction which affects the C3D geometry of a simple cuboid primitive element when it is edited by frequently used standard tools.
* It is perfectly possible that alternative 3D applications exhibit a similar glitch. It is also possible that “my solution” only functions when tools are applied in a specific sequence.

Servus von Wien nach Hamburg

Hi Helmut

Impressive! I'm keen to see how your model of the Sagrada Familia progresses. Hopefully, yours will be completed well before 2026!

Cheers

Chris