Importing a geometry fails with STEP but works with BREP


I use FreeCAD 0.18.4 and have a case of a STEP-file which causes an error while importing.

I created a certain mixer geometry and saved it as a STEP-file. The import in FreeCAD fails and creates strange shapes. This does not happen, when I save and import it as a BREP-file.

When I save the geometry from FreeCAD as STEP-file and try to re-import the geometry, the same error happens.

AutoCAD Fusion 360 has no problem with reading the STEP-file correctly. The same goes for the shaper module of Salome. Only the old geometry module of Salome has the same problem.

I am not sure, if that indicates more a problem with the used OpenCascade or with FreeCAD itself. For that reason I included the geometry files and pictures of the results.

I really hope this example can help to trace down the cause of this error!

John Bijnens's picture

I think the problem is caused by the model accuracy of your STEP AP214 file.
The units are mm and the model accuracy = 0.0001 mm
I have changed the model accuracy by using PTC Creo 6 to 0.01 mm and exported it again to STEP AP203.
I have attached the resulting STEP file to this message.
I was able to import this STEP file in FreeCAD 0.19

jens.acker_155503's picture

Thank you for your answer.

What I have problems to understand is: In which way can a higher precision cause the model to be partially inverted? And why does it work for the BREP?

These cylindrical cutouts in the discs have to go precisely from the outer diameter to the inner diameter (of the inner cylinder along the Z-axis). How can I hit that precisely, if the precision is turned down so much?

John Bijnens's picture

Model accuracy is the smallest distance between two points where those two points can still be distinguished as two separate points. As such increasing the accuracy can lead to additional errors. E.g. you have created a model with an accuracy of 0.01 mm and you have two points that are on a distance of 0.0005 mm of each other. In this case the two points will be considered as coincident. If you export the model with an accuracy of 0.0001 mm the two points that were previously coincident will now be considered as two separate, individual points. So by increasing the accuracy you have introduced new errors within your model. In this particular model some circular holes are located very near the outer cylindrical surface and are as such very prune to errors. I'm not that familiar with the BREP format. Perhaps someone else can jump in and give his or her thoughts about this matter.