(Note: I tried to embed a Mastodon video here but couldn't do it right. Oh well! Here's the video I was trying to link, and here's another good one from Emily Velasco!
The IBM Selectric typewriter uses an instantly-recognizable "ball" of type instead of a fanned-out array of arms like a conventional typewriter. These typeballs could be swapped out, meaning you could easily write documents with different fonts, font sizes, special characters, or different writing systems altogether.
IBM and some other vendors made lots of different typeballs, but most of them are 40-50 years old and no one has been making new typeballs for a long time. 3D printing is a natural fit for making new typeballs, but most printers still don't have the ability to produce the sharp details necessary on a sufficiently-resilient material. So it's understandable that no one made a really functional 3D-printable typeball. Until now, that is!
The blank typeball used in this design is based on 1944GPW's typeball on Thingiverse, which is released under a Creative Commons-Attribution license. I suspect that my project wouldn't exist if it weren't for this one. I had to change most of their typeball dimensions, and there are major issues with the way their characters are generated, but I sure as heck would have made those same errors myself, so I'm infinitely grateful for the people before me who documented their processes!
Another project that deserves a lot of credit is The Sincerity Machine by Jesse England. Jesse is a delight and a constant source of creative inspiration for me, and it brings me great joy to watch this project evolve with him.
How I got the sides of the letters to look right, finally
Getting those nice lofted edges on each glyph was not easy! I came up with a rather elaborate way that worked pretty well, but then got an excellent suggestion from Kris Slyka that works so, so much better. This is truly a community effort. I'd be very curious to hear how you would have approached it!!
The OpenSCAD Typeball by 1944GPW attempts to loft the shapes by scaling the letterform. This was my first thought, as well, but since the letters scale around a specific point, it means the lofted sides won't all spread out evenly. Instead you get unhelpful overhangs:
Instead of a scaled extrusion, we need the base to be a puffed-out offset version of the original letterform. It's not trivial to do that in an automated fashion, especially if there are small holes that will close up (like in the middle of the # character). The method I came up with uses a command-line script to extrude the letter in OpenSCAD and then displace one layer of vertices by their normal vectors. This means the vertex will always be pushed out away from the body of the STL.
If you look at the underside of each letter, you'll see a that this method leaves a lot of topological scars from attempts to re-mesh this self-intersecting shape. But it turned out to be pretty printable, so I released the project with this approach.
After seeing the discussion about typeballs on Mastodon, Kris Slyka approached me to suggest using OpenSCAD's
minkowski() function. It has a reputation for being agonizingly slow, and trying to generate all the glyphs at once would exhaust my PC's memory, but with the batch-script, one-letter-at-a-time method, it can work! I highly recommend getting a recent build of OpenSCAD nightly; the fast-csg setting makes
minkowski() operations so, so much faster.
I put some typeballs on Printables! They are also available on Shapeways (links in the Printables descriptions) if you don't want to print them yourself.
To affix the typeball to the typewriter, you will need a bent wire or a small clip such as this one from Dave Hayden. Thank you Dave for encouraging me to pick up the Selectric project again!
I had been testing on a not-fully-functional Selectric I, which I think was set to 12 characters per inch. Knowing characters per inch is important...Read more »
Just saw this, wow I'm really happy my project inspired you to take the idea further!
Especially congratulations on figuring out a good way of lofting the letters. I realised my technique of projecting them onto the ball surface wasn't very good, but relied on the filament smooshing (is that a word?) out and filling some of the overhang space. This is almost guaranteed to happen with a cheap 3D printer like mine. Using a resin printer looks to give the results this project deserves.
I like your clip design, saves having the screw heads too close to the top surface periphery.
You've done well with solving this. You should try recreating a Daisy Wheel next :)
Steve / 1944GPW / Mr Name Required