just for entertainment:
12 Likes
The center of the arcs are going to be perpendicularly aligned to the top surface because both arcs are tangent to that surface at that same point. The vertical wall may need draft added when this part is manufactured, that may move the point where both arcs line up, that would also change the radius of the big arc slightly. That explains why the blue print doesnât define the radius of the arc, but leaves it up to the CAD modeler to figure what radius works.
I did not say it would be rejected on every CAD model. I said it may be rejected on the vast majority of models used to supply parts to major industries. Manufacturers want true arc fillets because it makes things like adding draft and wall thickness much easier. True arcs can also be accurately measured to determine if the conform to specifications. The whole idea that you wonât supply a true arc solution because your CAD program doesnât have a way to do it would strike any major manufacturer as hilarious.
The main reason good CAD programs give their customers an easy way to make true fillets is because they know if they did not many of their customers would just leave as many Rhino users have abandoned Rhino for its lack of a convenient way to make good fillets.
1 Like
Dear Moderators,
I think Gijs won this thread, we can close it now.
Have a great weekend.
G
6 Likes
Quite nice trick. Too bad that it only works on such simple planar models. 
1 Like
Not sure if anybody mentioned it, but actually you can choose the surface quality of the patch in Fusion 360 on two occasions.
The file I uploaded last time was G1.
Fillet in Solid Tab
Re-patch in Surface Tab
2 Likes
No way!
Thatâs crazy⌠
1 Like
Absolutely beautiful, great approach and masterfully crafted, far better than any of the patch solutions. But as you say only viable if not being true to the constant radii is acceptable.
2 Likes
Yes, mechanical parts stick to ball fillets due to the need to be easily modified in parametric CAD programs. Or, simply because the designers and engineers donât want (or are not allowed by the boss) to spend extra time in perfecting the transitions, unless itâs about an exterior part thatâs going to be highly reflective and must be appealing to the consumer.
1 Like
using a back and for _offsetSrf is a fancy workaround.
it s a pitty because it shows, that the basic algorithms in rhino would be capable to solve this kind of corners directlyâŚ
3 Likes
What most Rhino users fail to grasp is thatRhinoâs Filletsrf command makes world class fillets.
All these alternative methods of producing fillets create sloppy, inaccurate fillets.
The alternate methods are dead-end solutions because the sloppiness and inaccuracy will cause downstream problems.
In the example weâve been discussing, if you make the fillets as G2 blends and want to create offsetsrf or shell with 4mm thickness you will probably end up with garbage. And that is with planes as the starting surfaces, make the base surfaces more complicated and the the resulting mess of offsetting will be even greater. The same thing applies if you want to add more features and fillets on top of what you have so far. The downstream operations are just going to get harder and harder to deal with.
If you export your G2 blends to other programs they will not be recognized as fillets which means they cannot be edit by the recipient. You are exporting dead-end geometry.
If you make your fillets in Rhino using Filletsrf you will be able to export the geometry and it will be recognized as a true fillet by other CAD programs and they will be able to edit it just like it was native geometry. If you make your fillets using Filletsrf then offsetting and shelling will work without problems. Adding more fillets on top of existing fillets will also be easy and accurate.
Even on the rare occasions when Filletedge does work the fillets that Filletedge makes are more often than not also inaccurate and sloppy. That means you may have the same downstream export and modeling problems.
All this is to explain why I think automating the one fillet solution Rhino has that is not a dead-end solution is the only sensible way forward.
4 Likes
Superior fillets:
Iâm grateful you shared this. It looks very much like what I think of as what a âtrue filletâ is.
Edge 5, 1, 3 are further smoothed. Even edge 7 looks smoothed.
This type of geometry would provide molecules the highest resistance to fracture â imo.
And the dialogue demonstrates the user has quite a bit of control.
Too bad the price point is bad still. And too bad Fusion is on the cloud.
I agree 
The whole âtrue to constant radiiâ, if said to be âtrue filletâ, then I say is nonsense.
Itâs many orders of magnitude more helpful if molecules are given the smoothest surface possible.
As apposed to âkinksâ:
âmolecular weakspotsâ i.e. fracture points.
1 Like
And when ppl donât know better, they think ârolling ballâ ideology is the best and the only way it ever needs to be â never changing or evolving.
Later I plan to share how ârolling ballâ can be interpreted much differently than ppl think.
This ability is irrelevant imo.
I remember there was a market for separate programs that were designed specifically for translating data between different CADâs â CAD Exchanger for example.
I remember the good ones were like $15k-$20k.
I think mostly your having a translation error, when your other CADâs donât recognize a âfilletâ.
Not to mention the fact what a âfilletâ is, is up for interpretation.
The true nomenclature for âfilletâ, is âouter/inner roundâ. And even that, I donât find relevant to what âroundingâ and ârollingâ really means.
From a molecular standpoint, molecules will have weakspots for separation. Hence, a product will fail due to fracture regardless of what someone thinks a âtrue filletâ is, if thereâs kinks where fracture can form.
Those are perfectly G1 points, just like any other point on those edges
I would not dare, but that kind of connection above with XNurbs is done in a minute.
1 Like
Heh heh heh, Today you are the defender of molecules. What grand cause will you be defending tomorrow?
Your latest rant against true arc fillet reminds me of a story a guy who works in QC dept. of a major US foundry told me:
So, the head of the Quality Inspection Department of a major US foundry gave a new hire, who claimed to have expertise in CMM operation, the job of finding the edge of fillet on a casting that the foundry produced. He gave this assignment as a joke ( he knew the recruit would not be able to accurately find the edge of a fillet).
The new hire spent all day with the CMM ( Coordinate Measuring Machine) trying to measure where the edge fillets were on casting. The best he could do was locate the edge of a fillet by +/- 1 mm.
The CMM he was using was able to accurately find the location of a point on the casting to an accuracy better than .001mm but the guy who spent all day measuring castings could not locate where the edge of a fillet twas by better than 1 mm.
Iâm absolutely positive that the above story about about how manufactured objects are not the same thing as CAD objects will not register with your brain.
1 Like
At a molecular level those points are cusps.
Down the line those will propagate a grain structure in things like meshes, surface topology etc.
Surely those cusps will be lost in translation from one format to another, but one thing can always lead to another.
I agree theyâre âG1â points, but âG1â doesnât mean âbestest smoothnessâ.
Should we see those âG1â topologies from different angles?
1 Like
I could find it with a $40 handheld 20x magnifier loupe.
I could probably find fractures, flakes, and other anomalies too.
It would take me 2 min per sq ft for a rough anlysis per say.
Using a several hundred dollar inspection devise, I could capture photos into a computer â as proof probably.
I always thought CMMâs were really silly. Although I know of ppl who think theyâre amazing.
I could even scan those âcastingsâ with a 3D scanner, and use maybe about $20k-$70k technology, and prove using reverse engineering software accurate to approx 150k measurements per square inch, where all the edges are and anomalies are.
This method would take 1-5 days of scanning, depending on size of object. And 1-4 weeks of initial processing, depending on quality analysis requisites. The data could be further used going forward, for many other things, even for correcting design flaws that were discovered and improve the castings of the future.
Fractures however, should use some form of x-ray tech or permutation thereof â for best results.
