I agree it is curious that one works but not the other. None-the-less, I think this is an issue with the cad kernel FreeCAD uses (OCCT). In my experience fillets almost never work if there is no remaining face after the fillet operation. I.e the fact that you managed to make two 0.5 mm fillets on a 1 mm plate is the anomaly. Not saying that it would not be nice if it worked.
I am sure you know, but for the record you could workaround this issue by having the rounded edge directly in the first sketch.
just do 0.4999 and visual it is close enough. on the other hand would it be a low hanging fruit to make that somebody implement kind of roundoff setting automically. that 4, 5, ... digit behind the seperator add or remove 1 so it fits, when visual display is 2 or 3 digits it shows the value correctly...
Jee-Bee wrote: ↑Tue Feb 09, 2021 11:14 am
just do 0.4999 and visual it is close enough. on the other hand would it be a low hanging fruit to make that somebody implement kind of roundoff setting automically. that 4, 5, ... digit behind the seperator add or remove 1 so it fits, when visual display is 2 or 3 digits it shows the value correctly...
-1 on this.
That would not be right. IMO, We should not assume what precision is needed by the user.
In my view this is not a severe issue in fillet. Because when you round the edge fully like that, it is no longer a fillet. Even if we fix fillet so that it allows 0.5 mm fillets, the user could still try to apply a 0.6 mm fillet, and it would fail anyway.
Every failcase that is eliminated (and so questions / complains) is one. Offcourse if you fix one use case another pops up. Fillets that are exacly the availeble space are quite often. of course it is dependendend on the scale of the user what the ideal number of digits should be, but for general case (70% - 80% of the users) there should be some kind of general optimum. Now the user have to think about it and try if the twelf 9's work out...
to response on your reasons:
2 fillets with a flat surface of 0.0002 in between is better as any machined part i have ever seen.
two fillets of 0.6 on a sheet of 1 thick never works and so not related to fillets
Jee-Bee wrote: ↑Tue Feb 09, 2021 1:48 pm
Every failcase that is eliminated (and so questions / complains) is one. Offcourse if you fix one use case another pops up. Fillets that are exacly the availeble space are quite often. of course it is dependendend on the scale of the user what the ideal number of digits should be, but for general case (70% - 80% of the users) there should be some kind of general optimum. Now the user have to think about it and try if the twelf 9's work out...
to response on your reasons:
2 fillets with a flat surface of 0.0002 in between is better as any machined part i have ever seen.
two fillets of 0.6 on a sheet of 1 thick never works and so not related to fillets
I am all for improving the UX, even when it comes at the cost of additional complexity in the source code. And I also think we should be pragmatic. But my willingness stops when you make the geometry incorrect. We can agree to disagree on that one. An additional point is that the proposed method would result in a degenerate surface (of 0.0002 mm width, or whatever). There is no guarantee that this model will work with all versions of OCC, I can also see that there might be problems with STEP export, with FE meshing, etc. In my opinion it is a hack, and not a proper solution.
everything FC does related to the OCC kernel but is not part of the OCC kernel itself is a fix by definition! This include workarounds for the fillet!
In current state the user have to do what i proposed manually or remodel if possible. And that is what told in help forum too (at least as long as i have seen it so somewhere OCC 6.x to current one).
So basically nothing changed except it goes automatic...
By my personal experience i never had any problems with step files.
With FEM i have always trouble related to sheet metal, just because the thickness is low and surface is large (with lack of mesh type good for sheetmetal, and lack of multi mesh type) what result easy in non positive jacobian. I think for FEM in general remove all unnecessary features including holes radii so it don't matter much for this situation
I second David's view. Until now we like to claim that FreeCAD models are precise as opposed to e.g. stl files. We would give this up, if we just add additional edges and faces. If we can reliably detect such situations we could show an appropriate message telling the user about his options, or even give him some choices which then can be carried out automatically.
And I am pretty sure that there are not fiew cases where not the precision itself, but the additional edges will cause problems.
If I meet such situations I rethink my model if I have other possibilities to reach my task, e.g. by filleting inside of a sketch, changing the order of features, ...
The very last resort is to change the fillet radius in the way discussed here.
I use Jee-Bee's approach frequently, because it works. I also agree with David and Chris that this a not a stellar approach. It is certainly not user-friendly for beginners and others who do not know the "hacks".
