This is a simplified extract from something I'm designing:

bodyWidth = 25;    
bodyHeight = 15;  
slotWidth = 6; 
slotHeight = 10;

$fn = 50;

module slot()
{
    rotate( [-90, -90, 0] )
    union()
    {
        cylinder(d = slotWidth, h = bodyWidth + 2);
        translate( [0, -slotWidth/2, 0] )
            cube( [bodyHeight - slotHeight, slotWidth, bodyWidth + 2] );
        translate( [bodyHeight - slotHeight, 0, 0] )
            cylinder(d = slotWidth, h = bodyWidth + 2);
    }
}

difference()
{
    cube(bodyWidth);
    translate( [bodyWidth/2, -1, bodyWidth/2] )
        #slot();
}

This results in a cube with an oval hole through it (see attachment)

What I would like to do is to put a nice rounded edge around the hole at both ends. The trouble is, I don't know where to start! I would appreciate it if someone could give me some suggested techniques, helpful hints, etc. I'm not expecting anyone to actually do the work - just give me a clue about how to do it!

Note - this is something I've often wanted to do, but shied away from due to a lack of knowledge.

TIA,
David

Edited 2 time(s). Last edit at 09/25/2019 05:02PM by David J.

I presume you want this chamfer removed from the cube not the slot.

That is a rounded chamfer, and is very annoying do in openscad. (This is one of openscads biggest issues in my view)

You need to apply the chamfer to each component part, ie two cylinders and the square.

What I do is a 45 degree chamfer. Ie create some 90 degree cones at each end and remove how much you want
Do the same on the straight parts with a 45 degree rotated cube.
This is much easier to print, you don't get overhangs over 45 degrees.

Edited 4 time(s). Last edit at 09/26/2019 05:14AM by Dust.

But if you insist on rounded chamfers. Here is a cylinder for cut out with round chamfer (I just wrote this)

module Chamfered_Cylinder(CylinderDiameter,CylinderHeight,ChamferRadius,ChamferFacits) {
    difference () {
        union() {
            //Central Cyclinder
            cylinder(d=CylinderDiameter,h=CylinderHeight);
            //Bottom Chamfer cylinder
            translate ([0,0,0]) cylinder(d=CylinderDiameter+ChamferRadius,h=ChamferRadius/2);
            //Top Chamfer cylinder
            translate ([0,0,CylinderHeight-ChamferRadius/2]) cylinder(d=CylinderDiameter+ChamferRadius,h=ChamferRadius/2);
        }

        //Generate Bottom chamfer 
        for (i = [0:1:ChamferFacits-1]) {
            rotate([90,0,i*360/ChamferFacits+90]) translate([CylinderDiameter/2+ChamferRadius/2,ChamferRadius/2,-ChamferRadius/2]) cylinder(d=ChamferRadius, h=ChamferRadius);
        }

        //generate Top chamfer
        for (i = [0:1:ChamferFacits-1]) {
            rotate([90,0,i*360/ChamferFacits+90]) translate([CylinderDiameter/2+ChamferRadius/2,CylinderHeight-ChamferRadius/2,-ChamferRadius/2]) cylinder(d=ChamferRadius, h=ChamferRadius);
        }
    }
}


CylinderDiameter = 15;
CylinderHeight = 40;
ChamferRadius = 5;
ChamferFacits = 50;
//Its best to have the cylinder facits and the Chamfer facits the same value.
$fn=ChamferFacits;  

//Extra cylinder on bottom to appease the scad gods
translate ([0,0,-ChamferRadius]) cylinder(d=CylinderDiameter+ChamferRadius,h=ChamferRadius);
//Extra cylinder on top to appease the scad gods
translate ([0,0,CylinderHeight]) cylinder(d=CylinderDiameter+ChamferRadius,h=ChamferRadius);
//The Cylinder with chamfer
Chamfered_Cylinder(CylinderDiameter,CylinderHeight,ChamferRadius,ChamferFacits);

Edited 3 time(s). Last edit at 09/26/2019 12:50AM by Dust.

If you can live with the outside edges of the cube being rounded, then Minkowski is the answer. However, this is very slow operation. Put minkowski() {... sphere(1); } around your main block.

minkowski() {
difference() {
    cube(bodyWidth);
    translate( [bodyWidth/2, -1, bodyWidth/2] )
        slot();
}
sphere(1);
}

Hmm - I need to study your answers! Especially that code example...

I hadn't thought about using minkowski when the object has the slots - in the real project I had applied it to the body prior to 'cutting' the slots. I shall have to try that out. However it means re-arranging my code as I have included screw holes in the same part as the slots (not a major crisis).

My other random thought last night was about making the slot itself a better shape - making the cube and two cylinders with 'fat ends' so that a difference with a larger object results in a rounded exit. I suspect that's what the code example is doing, but I'll confirm that later today...

Thanks for all your help - I'll see how it works out.

Right - I tried using minkowski... life is too short... and all of my dimensions were screwed up (as expected); it would be a PITA to make size adjustments to compensate. So, good idea, but too painful.

I've been trying to develop my own idea, and here's how far I've got:



UPDATE 1: I've just look at the code written by 'Dust' - mine follows a similar technique. I didn't copy yours, honest!
UPDATE 2: Dust's approach is better than mine...

Obviously it doesn't yet achieve my original objective, but it's a start. I won't show the code just yet as it's messy and horrible, but I do plan to make it into a library so I'll put it up here for review when it's in decent shape (and when I'm not too ashamed to show it in public!).

Cheers,
David

Edited 2 time(s). Last edit at 09/26/2019 08:16AM by David J.

My code is freeware! do with it what you wish, no strings attached.

Quote
Dust
My code is freeware! do with it what you wish, no strings attached.

Thank you! If I do create a library then you will be credited in the header...

Interesting problem...

Here my approach, which is a bit different. Instead of subtracting cylinders to get the chamfer, I construct the cylinder in slices:

// chamfercyl - create a cylinder with round chamfered ends
module chamfercyl(
   r,              // cylinder radius
   h,              // cylinder height
   b=0,            // bottom chamfer radius (=0 none, >0 outside, <0 inside)
   t=0,            // top chamfer radius (=0 none, >0 outside, <0 inside)
   offset=[[0,0]], // optional offsets in X and Y to create
                   // convex hulls at slice level
   slices=10,      // number of slices used for chamfering
   eps=0.01,       // tiny overlap of slices
){
    astep=90/slices;
    hull()for(o = offset)
       translate([o[0],o[1],abs(b)-eps])cylinder(r=r,h=h-abs(b)-abs(t)+2*eps);
    if(b)for(a=[0:astep:89.999])hull()for(o = offset)
       translate([o[0],o[1],abs(b)-abs(b)*sin(a+astep)-eps])
          cylinder(r2=r+(1-cos(a))*b,r1=r+(1-cos(a+astep))*b,h=(sin(a+astep)-sin(a))*abs(b)+2*eps);
    if(t)for(a=[0:astep:89.999])hull()for(o = offset)
       translate([o[0],o[1],h-abs(t)+abs(t)*sin(a)-eps])
          cylinder(r1=r+(1-cos(a))*t,r2=r+(1-cos(a+astep))*t,h=(sin(a+astep)-sin(a))*abs(t)+2*eps);
}

// now build David's example, the cube with the chamfered hole (viewed from below to make things easy...)
$fn=36;
difference(){
   translate([-12.5,-12.5,0])cube(25);
   chamfercyl(3,25,3,3,[[-2,0],[2,0]]);
}

The module chamfercyl() takes as parameters the radius r and height h of the cylinder and in addition the chamfer radii b for bottom and t for top, where a positive chamfer radius goes to the outside and a negative to the inside of the main cylinder. The slices parameter gives the number of slices for rounding the chamfer. The optional offset array can be used to generate a convex hull at the slice level (cannot be done at the object level because the final object is not convex if outside chamfering is used). In David's object I use the offsets to make a slot instead of a round hole. Finally, the parameter eps can be used to add a tiny overlap of the slices.

Here some more simple examples to show the basic working of the module chamfercyl():

translate([-60,0,0])chamfercyl(10,50,0,12);
translate([-30,0,0])chamfercyl(10,50,-3,-6);
translate([0,0,0])chamfercyl(10,50,7,-3);
translate([30,0,0])chamfercyl(10,50,-9,2);
translate([60,0,0])chamfercyl(10,50,4,4);

gives the following cylinders:

@enif

Darn that is a good way to do it!!

@Dust

Thanks!

Quote
David J
Right - I tried using minkowski... life is too short... and all of my dimensions were screwed up (as expected); it would be a PITA to make size adjustments to compensate. So, good idea, but too painful.

Right... I'd forgotten that Minkowski with a sphere moves the surface by the radius of the sphere. For sphere of radius 1, you need to subtract 1 from each of your width and height variables.

Quote
frankvdh

Quote
David J
Right - I tried using minkowski... life is too short... and all of my dimensions were screwed up (as expected); it would be a PITA to make size adjustments to compensate. So, good idea, but too painful.

Right... I'd forgotten that Minkowski with a sphere moves the surface by the radius of the sphere. For sphere of radius 1, you need to subtract 1 from each of your width and height variables.

Correct - and it's not a significant problem if you're dealing with the external measurements of a solid. However, both sides of the holes are also reduced by the radius of the sphere - the calculations just get too complicated. But really the compilation time is the major issue when you have more than 1 hole in the object.

Wow, that's tough to read!
Clever though... I think. I'll have to get studying.

Quote
enif
Interesting problem...

Here my approach, which is a bit different. Instead of subtracting cylinders to get the chamfer, I construct the cylinder in slices:

-- snip for brevity ---

Edited 2 time(s). Last edit at 09/26/2019 04:03PM by David J.

Well, enif's code works very nicely - here's what I did with it:

Desk cable organiser

Enif - have a look at what I did with your code, and let me know if you're happy with what I've done. Also, if you want a more detailed credit in the code and on Thingiverse then please PM me with your details.

Many thanks to all,
David

@David J

This is perfect - thanks for giving credit

Since I am also quite active on Thingiverse, you could -but only if you like!- underlay the word "enif" with a link to my thingiverse account, which is www.thingiverse.com/enif .

I will do! Thanks for the help.

Here's a filleted cylinder tool like enif's, but should run faster:

module fillet_cylinder(
 r,      // cylinder radius
 h,      // cylinder height
 b=0,    // bottom chamfer radius (=0 none, >0 outside, <0 inside)
 t=0,    // top chamfer radius (=0 none, >0 outside, <0 inside,
 deg=10  // degrees per rib of fillet
)
 rotate_extrude()
  polygon(concat([[0,h],[0,0]],
   [for(a=[0:deg:90]) [r-b*(sin(a)-1), abs(b)*(1-cos(a))]],       //bottom fillet
   [for(a=[90:-deg:0]) [r-t*(sin(a)-1), h-abs(t)*(1-cos(a))]]));  //top fillet

As an easter egg, if you also pass in $fn=4, you'll get a filletted square prism.

Just commented on your Thingiverse design, very useful. thank you.

Looks like a nice solution to a serious headache problem in openscad, until I tried creating a pin from a filleted cylinder with rounded tip and flared base. Then in a "for" loop to create a sideways elongated tab from it, to be used in a difference from a cube. This is a very standard locating pin style as used in injection molding machines.
I had to force a quit on openscad after maybe 30 minutes of doing nothing and then it started to affect the rest of the computer.
I take your word for it that it is faster then minkowski (which I avoid like the plague for being too slow to be useful) but IMHO this is still not useful either. My code was only 17 lines long! (which I had not saved before the crash)
I might try something else later, for which I got the idea looking at your solution. Split the pin I try to make in 3 parts, base, center and tip.
base is a large cylinder with a torus taken off it so it creates the bottom flare. center is just a cylinder, top is just a sphere.
Now translate to the length of the tab I need to make, and hull each of the three parts separately. Then put them in a module and use them to difference out from a cube.....need to try this to see if it will be just as slow.

So I still love Openscad to bits, but until something works better then these solutions my designs will resemble an mine-craft landscape, mostly anyway. Thanks to chamfer.scad though, just takes the sharp edges off this problem a little ;-)

As mentioned earlier, I said I would have to try my suggestion. Still not fast but I feel way faster then minkowsky on any day.

Just a pig to rewrite all these modules in subroutines to be applied in different parts of the design, and again, if you have many it will be slow, but probably faster then minkowsky. Where the previous solution crashed on my pc, this reviewed in about 1 minute.

$fn=64;
based=40;
baseh=10;
bodyd=20;
bodyh=10;
trans=50;
//calling parts
for(a=[0:1:trans])
translate([a,0,0])pin_base();
hull(){
pin_body();
translate([trans,0,0])pin_body();
}
hull(){
pin_top();
translate([trans,0,0])pin_top();
}
//base
module torus_cut(){
rotate_extrude(angle=360) translate([based/2, 0, 0])circle(baseh);
}
module pin_base(){
difference(){
cylinder(h=baseh,d=based,center=true);
translate([0,0,baseh/2])torus_cut();
}
}
//body
module pin_body(){
translate([0,0,baseh/2+bodyh/2])cylinder(h=baseh,d=bodyd,center=true);
}
//top
module pin_top(){
translate([0,0,baseh/2+bodyh])sphere(d=bodyd);
}

Another approach is to use Torleif Ceder's very handy unionRound() module, available on his github page.
It allows you to join any two arbitrary objects with a radius fillet.

In this case I join two cylinders using unionRound() and then subtract this from a cube. Easy to extend for the oval'ed slot case.


module unionRound(r, detail = 8) {
epsilon = 1e-6;
children(0);
children(1);
step = 90 / detail;
for (i = [0: detail-1]) {
{
x = r - sin(i * step ) * r;
y = r - cos(i * step ) * r;
xi = r - sin((i * step + step) ) * r;
yi = r - cos((i * step + step) ) * r;
color(rands(0, 1, 3, i))
hull() {
intersection() {
// shell(epsilon)
clad(x) children(0);
// shell(epsilon)
clad(y) children(1);
}
intersection() {
// shell(epsilon)
clad(xi) children(0);
// shell(epsilon)
clad(yi) children(1);
}
}
}
}
}
// unionRound helper expand by r
module clad(r) {
minkowski() {
children();
// icosphere(r,2);
isosphere(r,70);
}
}
// unionRound helper
module shell(r) {
difference() {
clad(r) children();
children();
}
}
/*
// The following is a sphere with some equidistant properties.
// Not strictly necessary

Kogan, Jonathan (2017) "A New Computationally Efficient Method for Spacing n Points on a Sphere," Rose-Hulman Undergraduate Mathematics Journal: Vol. 18 : Iss. 2 , Article 5.
Available at: [scholar.rose-hulman.edu] */
function sphericalcoordinate(x,y)= [cos(x )*cos(y ), sin(x )*cos(y ), sin(y )];
function Nn,x)=
let(toDeg=57.2958,PI=acos(-1)/toDeg,
start=(-1.+1./(n-1.)),increment=(2.-2./(n-1.))/(n-1.) )
[ for (j= [0:n-1])let (s=start+j*increment )
sphericalcoordinate( s*x*toDeg, PI/2.* sign(s)*(1.-sqrt(1.-abs(s)))*toDeg)];
function generatepoints(n)= Nn,0.1+1.2*n);
module isosphere(r,detail){
a= generatepoints(detail);
scale(r)hull()polyhedron(a,[[for(i=[0:len(a)-1])i]]);
}


$fn=60;

difference()
{

cube([20,20,10],center=true);

unionRound(4, detail = 10)
{
translate([0,0,10])
cylinder(h=10,d=20,center=true);
cylinder(h=20,d=5,center=true);
}

}