Unable to fillet this edge but why?

stages are simple,

  1. create cylinder with sloping top, use sweep2 for slope,
  2. then fillet the edges.
  3. I then booleanSplit the two crossing cylinders into it,
  4. Fillet the four edges radiating from centre,
    so far so good, then unable to soften up the 4 edged leaf shape. Wish for quite a rounded edge, at least 1mm rad if poss. but even a small fillet of 0.3 fails.

Prior to this I had another of these with not so soft a top fillet in stage 2, a fillet of 0 .5 at tips and 1 elsewhere worked., but 1 all round failed. I get fillets sticking out like unfinished roads into space.
I made the edges softer with a larger fillet hoping the 4 leaf fillet would work better but its even failing on 0.3mm now.



1 Like

In Rhino if you want your fillets to fail you use FilletEdge command.
If you want fillets to succeed use the FilletSrf command.
Takes about 2 minutes for your example.
FILLETSRF.3dm (427.2 KB)

1 Like

New to this. Can you please explain to me how you made FilletSrf work in this example. For me it totally failed using that command. What command prompt settings did you use? I tried them all. Could a tolerance of .001 generally have something to do with it? Thanks

First of all you need to understand the topology of the geometry. and what it is you are trying to accomplish.

In the example given you have two sets of tangent surfaces that are separated by a crease. The goal of filleting here is to round the crease between the 2 sets of tangent surfaces so that when you are done all the surfaces are tangent (no crease remains).

The process of making fillets with FilletSrf involves clicking on 2 surfaces (one on each side of the crease that you want to round).
The user has to make each individual surface by clicking twice.

After you have made one fillet surface you will notice that where that fillet ends is where it reaches the boundary with the next tangent surface.

The ends of the fillet you just made tell you where to click to make the next fillet surface in the chain of fillets. So the process involves 2 clicks for each surface in the chain of fillet surfaces and the last fillet created tells you where to click for the next fillet. . You have to make two clicks for each and every surface all the way around the chain.

Of course, it would not be difficult for McNeel to automate this process. After you make the first two clicks the software could make all the fillets in the chain until either it ran out of tangent surfaces or ended up back where it started. (as in this case). Each fillet created defines where the next 2 clicks go for the next fillet so its ridiculous that the user has to keep making all those clicks.

It is unfathomable why McNeel has spent so much development time on trying to improve filletEdge with no success when it would be so simple to improve filletSrf command which makes fillets that are as good as or better than any other CAD program.

Thanks so much for taking the time to reply in such detail. I’m hoping I
can follow your process when I get home from work tonight. I’ve been a
bench handmade Jeweler for the last 30 years and I’m finding it very hard
to achieve the nice continues flow to my Rhino generated models that I had
then. I’m wondering what you mean by double clicking. Is that away to get
more then two sections of a FilletSrf to fillet at one time. The command
asks me to click once for one of two adjacent surfaces. That’s were my
problems start. Would it help to explode the surface first? Guess I’ll find
out later tonight. Thanks again.

[quote=“Et_Rec, post:5, topic:29257”]…
I’ve been a bench handmade Jeweler for the last 30 years and I’m finding it very hard to achieve the nice continues flow to my Rhino generated models that I had then.

The transitions between surfaces in your handmade jewelry were probably essentially continuous curvature (G2) or better, particularly for polished surfaces, unless you worked to maintain a crisp edge. Fillets have a circular arc cross section and have tangency continuity (G1) which is not as smooth as curvature continuity. Level of continuity has a major effect on how smooth transitions look and feel. BlendSrf and BlendEdge create curvature continuous transitions.

Would either of those blend commands work on the example problem given in
this case without having to go through all the steps of splitting by
isocurve>trim and then blending. The amount of time spent on a job is
crucial in the real world. Thanks for the reply. I really appreciate all
the help I can get. PS. Pardon my amateur ignorance but in making
complicated shapes I’ve often found myself wishing for a 3 rail sweep
command. Probably NetworkSrf is its replacement but a 3 rail sweep would
save lots of time and maybe create a simpler surface.

It asks you to pick one surface and after you pick one then it asks you to pick another. If you pick so that the selection menu doesn’t pop up you should be making just 2 clicks for each surface that you create. Each fillet surface connects 2 surfaces on opposite sides of the crease. The surfaces don’t need to be touching as long as the fillet surface that you are creating will reach both.

You don’t need to explode the surfaces. I made the loop of fillet surfaces and then joined them and then used DupBorder on the joinrd loop to get two curves to split and trim with.

I couldn’t make FiletSurf work on his model either. Could you explain further?

Thanks. Can’t wait to try it out tonight

The enclosed file shows one fillet and the 2 places I picked on the base surface to create that one fillet. There are 36 fillet surfaces in the entire loop - 2 picks for each fillet.

The ends of that first fillet tell where the adjacent fillet will begin. It is a good idea to pick on the surfaces in such a way that the selection menu doesn’t pop-up because contending with the selection menu will really slow you down. Avoiding the selection menu takes a bit of practice.

Of course all of this would be much easier if the software did it for you. It is trivial for the software to figure out the boundary where each fillet ends and whether the next surface is tangent at that boundary.

That is, all Rhino has to do is determine if both surface normals are the same (within tolerance) at the two corner points on the end of the last fillet and if the normals are equal then it creates the next fillet. The whole loop would be made in an instant

FILLETSRF2.3dm (195.5 KB)

Thanks David… Is there a way of bringing up a popup when doing Fillet work
that gives the user choice as to curvature, tangency, G3 etc. like when
using Blend?

A fillet has a circular arc cross section so the curvature of the fillet is fixed by radius of the fillet, and the fillet is tangent to the adjacent surfaces. That means the continuity will be tangency (G1). Higher level of continuity is not possible and it remain a “fillet”.

BlendSrf, BlendEdge or VariableBlendSrf can provide G2 and above continuity.

I think the problem filleting this set of surfs is that the top planar surface ends up out of alignment by a small margin with the scalloped surfaces (see image). If you delete the planar part and extend the adjacent surfaces then split and rejoin, the fillet works. I suspect this method may be acceptable in this example, although the replacement for the planar bit may not be identical.

As jim managed to fillet my edges here I think easy, yet I fail, as do others, how did jim manage it wihout any removal and recreation of the planar surface, and I spotted that offset also.

I am also mentally rusted up having been away from Rhino a few months or more, so cannot even remember right now how to delete the planar part and extend the adjacent surfaces !

I select a radius similar to jim, (not sure what jim used) and select jims area, I get a fillet, I then select the next pair inboard of that and get a fillet, but it doesnt follow on, ‘end to end’ to the last one. Its as if we get fillets but they dont understand they are supposed to abut each other.

A picture speaks a thousand words, and this sort of thing is best seen, I would love to see a video of this being correctly done, three of us struggling here. From sorting out that little offset to establishing greatest radius one can get away with.

Furthermore and crucially, on that last note, how does one establish just how big a radius can be used here to maximise the roundness of the edges, I would like to see 1mm rad, but with things kicking out into space I dont know if its the radius being too much, or another reason. To keep doing the entire shape , then undoing the process trying for a different size, then undoing it, I even tried turning on record history, hoping that adjustment of the handles afterwards would see the fillet update and find the best combination that way, maybe less radius at the outer tips of the gulleys. I couldnt find out how to get the handles to show after fillet was made.


That tells you you missed a fillet in between. You need to do this work in a parallel viewport because you will easily miss things if you try to do it in perspective view,. If you zoom in you will see which 2 surfaces that you missed connecting with a fillet. When you click on those two you should get the narrow fillet piece that you missed.

The largest fillets that can be supported are determined by the size of the fillets you used previously. Both of the previous fillets have a radius of 1 so a filet of .99 will probably be as large as you want to go . If you had used a larger fillet then the subsequent fillets could be larger.
Enclosed is a file with the fillets at .99.

FILLETSRF3.3dm (371.1 KB)

Rhino’s fillets really struggle to deal with surface edges under certain conditions. Sliver edges, edges almost intersecting etc. You either learn to work around this or use another cad package to do your fillets. This takes years of experience to figure out and yes it sucks. When someone posts a fillet failure file. The first thing I look for are surface edges. Often times they can be fixed easily with the adjust closed surface seem command.

Your model is overly complicated. All you need to do is rebuild your cylinder cap surface as a simple revolve. That will eliminate all the problem surface edges that were causing the fillets to fail.

1 Like

As another bench jeweller of more than 30 years my advice would be learn Rhino but add T-Splines to your bag of tools. The two work extremely well together for jewellery.

T-splines can give you sharp edges but defaults more towards nice filleted edges and flowing shapes. Rhino defaults more to sharp edges IMO which often turns into a nightmare filleting. I work mostly by eye on the bench and in t-splines I do the same. That said if you want or need to work by numbers then t-splines will do that as well.

We are very lucky as jewellers in that despite working on small items we don`t need to be that accurate or curvature spot on etc. For years when we were doing most of our shaping by hand we were working no where near the accuracy that is native in CAD.

What I mean here is a small dip or crease from poor modelling in Rhino or creases from a star point in T-splines are often easily fixed with a file, emery, polish etc. during finishing. In fact many of those flaws will fail to be visible when you print the model from my experience and if they are then usually they are easily fixed during hand finishing.

So work to improve your modelling but don`t let that little flaw stop you from getting the job done.

Thank Sochin…yes I agree that a file or emery paper not only fixes a crease or dip, which is necessary anyway, since, by the time a .3dm file becomes s .stl file, grown in 3d printer and then Lost-Wax casting process…the accuracy of Rhino is long gone. Still, I would like to be the one who decides the direction my work in Rhino goes and not visa-versa. PS It’s also a matter of time spent on a job. A recent Gypsy style mans ring took me 5 hours to get right in Rhino and I could have carved it in wax by hand in 3. Certainly, the curves would never have been as even, but I’ve very rarely seen a customer who looks at the product the same way as the craftsman.

thanks, the first to point this out, …in my initial post one can see what was a simple cylinder and top planar disc with sloping sides after fillets made on the ‘edges’, then the booleanSplit of the two cross cylinders. In my mind two simple commands and I assumed all was well, even such simple moves can dirty up a structure it would seem, would have been ok maybe until fillet was used which is somewhat tricky at times.
You say to take the capping before the two cylinders are cut across it and use revolve to reform it, then booleansplit and progress from there. I would reach for extract isocurve, trim to get the top part and revolve that, am I right ?
I am sure many would have not stopped at that stage having no foresight of what might be to come, this is where experience counts. I bet though we all try it on, then backtrack if things go squiffy, or should one tidy up and simplify at each stage , and how many users do that ?

I have posted elsewhere about a best practice video collection, one to go into that I reckon.

Sochin and Et_Rec

I also agree, one dances to a finer tune than is required at times, and fine tolerances often deny fast progress, throwing up issues and a need to work at great care and magnification, knowing that emery, other crude methods and a need for far less accuracy are to be used on the end result. I work on some subjects where 1mm doesnt matter, yet lead a merry dance to 0.01mm tolerance which causes problems to delay me by hours at times. I wish for a command called ‘make up the difference’ !..aka ‘you can see what I am trying to do, dont be so fussy and complete it for me’ !