Issue with fillets

What am I doing wrong in trying to create these seemingly simple fillets?

Fillet issue.3dm (147.4 KB)

Hi Zews - I’d make that inner surface a single surface trimmed out of one cylinder - then you don’t get the tangent edge between the two and filleting can negotiate that more easily:


Pascal is correct. You need to keep that inner surface continuous & one piece. Next, you need to stage or sequence your fillets. Bigger fillets first, smaller fillets later.

Here is my painful and weird analogy. Smaller fillets are designed to ‘drive around’ the larger fillets. So, think of a truck and road; build the road first. The truck must be smaller to fit on the road. Otherwise, it will crash through the railings and explode in ball of fire. (Nice visual!)

Here is a image to explain


This might make a nice video… if someone asks nicely.

Let me see if I did this right.

First you trimmed away the inner surfaces using the cylinder surface as the cutting object. Then you used the ring as cutting object to trim away three sections of the cylinder surface. Then you flipped the direction of the resulting surface. Then you joined.

I tried something else, which also seems to work. This is however a much less elegant solution and requires more planning. After seeing your solution I realize my solution is actually very clumsy.

Originally I created the two parts including the center hole, and then did a boolean union before trying to do the fillets,which failed

When I make the two part without doing any boolean difference. Do a boolean union on the two parts and then do a boolean difference for the center hole, this works also.

I have to learn to use Trim and Split more and try to rely on Booleans less.

That is an excellent observation. Most people gravitate towards the booleans when they are beginning, since they are definitely easier and faster. The problem is that you get some ‘patchiness’ of the resulting surfaces as opposed to be making them cleaner or one piece.

I’ve got two free lynda videos you can check out: Booleans and Dave’s Golden Strategies. All of my six lynda course are listed here. If you are not a member, it’s only $25 per month, or try my free 10-day link. Sweet!

Hi Zews - I ExtractSrf-ed and deleted the existing inner surfaces (I did not check if Unrtim on one would get me the entire cylinder, that might also work) and made a new cylinder of the same radius- then Split that with the rest of the ring, deleted the bits I did not want and Joined the surface I wanted to the ring.


I understand how to make the inner surface continuous. I do not understand how you make your fillets look the way they do.

I am able to create the fillets, but they do not look like yours in one place. I also do not understand a=6 , b=6, c=4 although I think it may have something to do with how you set your grid lines differently than I have mine set.

This is how my fillets turn out. The section next to the left arrow is wider than the section next to the right arrow. In your example they are even.

You mention the make the fillets in 3 separate passes. When I do that, all kind of weirdness happens.

BTW, I have taken your excellent Lynda course Rhino 4.0 Essential Training about a year ago. I learned so much from that

Aha! Good catch. I used the Solid→Fillet edge option called ‘RailType = distance to rails’ That will keep the distance from varying and just looks better.

It all depends on your units. I was just trying to show relative sizes. The only focus should be big fillets first, smaller afterwards. Use different numbers for your file, just make sure they ‘fit’.

:slight_smile: I am a goldsmith. My attention to detail is very precise by training and from more than 40 years experience.

This is what is available in Rhino Mac.

When I click Chain Edges, I get this:

Which one do I pick? No matter which Chain Continuity I choose, Position, Tangency or Curvature, the result is the same, i.e. uneven in that location.

I just visited your site,
great work, it combines sensuality with geometry in an visually attractive way.


Yup … the old “features are missing from my beta software” problem. I am always nervous to run beta software for anything EXCEPT personal experimentation. You might want to run the PC version in emulation… that would ALSO let you render with V-Ray.

My understanding was that RhinoMac is no longer Beta, but RC1. I did not realize that this is still considered beta.

I wonder if McNeel intends to include this fillet option (‘RailType = distance to rails’ ) during the further development of RhinoMac. There are probably several (possibly many) other options here and there inside commands that are not (yet) implemented in the Mac version. I would not know, since I never used the Windows version.

I am glad that through this topic I found out about this particular option missing inside the fillet command, since your result looks so much better than my result, even though most people might not even notice it. So I request that it will be included in a future update.

Even though I still run a couple of very old programs on Windows XP on my Mac via Parallels Desktop, I am no longer interested in installing anything in this very old version of Windows. I switched in 2007 because of the “daily indignities” Windows was throwing at me at the time, and have not looked back. (Not that I don’t have some gripes about OS X). The Windows 3D program I used for so many years is now in the dust bin.

The funny thing is, I took over the license for trueSpace in September of 1997 from a guy who had used it for some time. He told me that he was switching to this new “very promising” 3D program that was about to be published, he told me that it was called … Rhinoceros 3D.

Well, with RC1 and RC3 we are in a grey-area, but mostly you are correct in your initial assumption…we are moving out of beta.

There are many. We are tracking commands differences here, but there may be many command options that are not getting accounted for. We need to fix that, but the best way is likely logging individual feature requests…which brings me to…

Got it on the list (MR-2048). Thanks.