Cone geometry (problems)

[rophys]

Hi folks,

I had the following problem: I tried to do a 90 cone (R1=0 mm and R2=30 mm) but seems to me that its not perfect this geometry.
I exported it and I confirmed that this geometry is not correct.

To show what's in going wrong I tried to upload a screen shot (.jpg) but this web site tell me "image is not valid" =/

Somebody can help me ?

Many thanks.

[NormandC]

First things first:

1 - What is your version of FreeCAD (Help --> About FreeCAD menu), version number and revision number
2 - What is your operating system

jpg attachment is broken. http://www.google.com/#hl=fr&cp=11&gs_i ... 24&bih=916

Best is to attach your actual FreeCAD (.fcstd) file. That works.

And by the way, this should have been posted in the Help forum, not the open discussion forum.

[rophys]

Hi normandc,

Sorry to post in the wrong place.

1) My FreeCAD version is 0.13
release date: 2012/02/05

2) My operating system is Linux (Ubuntu 10.04)

Please find attached my cad file.

Thanks.

[NormandC]

I confirm that your file has a problem.

Out of curiosity, I modeled your shape in one go from revolving a sketch using the Part Design workbench (I hardly use booleans and primitives anymore, it's just not as efficient as working with sketches).

Here's the tricky part: the revolution seems to produces a perfect 360º cone, but if I revolve it only 90º then the tip end has the same surface problem as your object.

So I made the revolution 360º and cut through it to keep only a 1/4 of the cone: then the cone tip gets wrinkled again!

I've changed the shape view accuracy in the preferences, and it realy does not seem to be only a view tesselation problem, the shape itself is wrong.

[NormandC]

New test: if you finish the cone with a 0.01 radius instead of zero, then the 1/4 cone is perfect.

So the problem is with a zero radius at the tip.

[rophys]

Hi normandic,

I also tried everything to correct the sketch but nothing works properly. I agree with you in the sense that the problem is in the R1=0, and I've tried to do a cone with R1=0.01 and it is correct as you wrote in you latest post. Unfortunately, in my case I need a zero degree radius cone and if somebody know how to solve this problem, please let me know.

Cheers

[NormandC]

Hi,

In my opinion this is a bug in the software, we need to report it here: https://sourceforge.net/apps/mantisbt/f ... t_page.php

In the meantime, I would try to make a full cone with the Part Loft feature (Part workbench, Part --> Loft... menu) see my post here for more info: viewtopic.php?f=3&t=2206#p16294

Then you cut the cone so you get only 1/4 of it. Since the loft feature uses a different algorithm, maybe the resulting 1/4 cone won't have the same problem as with a primitive or a revolved sketch.

[NormandC]

A lofted cone seems to have the same geometry problem once it's cut in a quarter. This is very puzzling.


Cone made by lofting a circle to a vertex, looks just fine


The cone is cut to keep 1/4 of it, notice the crease on top.

[NormandC]

I exported it and I confirmed that this geometry is not correct.

So far I had not questioned your affirmation that the underlying geometry of the 3D model was at fault, but how exactly did you do what you say? In what format did you export, and which program did you use to verify?

Because I exported 2 of my files to STEP, and had a look in Solid Edge, a commercial CAD program.

Here's how they appear:


Revolved cone


Lofted cone

As you can see, they are perfectly fine.

My conclusion is that there is nothing wrong with the underlying geometry. It is "only" a visualization issue, although I admit it's slightly annoying.

[jriegel]

Ok, some words to cones with sharp tip.

Although mathematically well defined leads such a cone to some problems in Brep data
structures. A surface come together in one infinite small point at the top. Such a geometry will
always leads to problems. the mesher, which produce that visual artifacts you see, e.g can not
decide at the tip what surface normal he have to use for the triangles at that point. also quit sure
some modeling algorithms will fail at that tip.
So.....

don't do it!
Use a small plane at the top or think of something else....