"Printed threads: when they work and when to give up"
6 September 2026
Three generators on this site cut threads now. The jar does big friendly lid threads, the PET cap mates with the molded neck of a soda bottle, and the newest one turns out metric bolts and nuts down to M3. Between them I have re-learned everything I thought I knew about printed threads, usually by printing the counterexample first.
Coarse wins, every time
A thread is a wedge wrapped around a cylinder, and a printer draws that wedge out of 0.2 mm layers. The fewer layers per turn, the cruder the wedge. My jar lids run a 6 mm pitch with a 1.8 mm tooth: thirty layers per turn, and the result feels like a peanut butter jar, smooth and forgiving. An M8 bolt at the standard 1.25 mm pitch gets a 0.75 mm tooth, less than four layers per flank, and it still works because there is a whole stack of turns sharing the load.
Then comes M3. At a 0.5 mm pitch the tooth is 0.3 mm deep, which is smaller than the nozzle drawing it. It prints, technically. The bolt generator will make you one, and the README politely calls it light duty. This is the give-up line for a 0.4 mm nozzle: below M6 or so, print the hole plain and let a steel bolt cut its own way in, or size up. Anything finer than a 0.5 mm pitch is not a thread, it is a texture.
There is a ratio hiding in all three generators: tooth depth around half the pitch, never more than 40 percent for the round jar profile. Push deeper and the profile collapses into overhang. The jar generator refuses outright, with the limit in the error message.
Clearance depends on who made the other half
This is the number people copy from the wrong project. It is not one number.
When both halves come off the same printer, the errors partly cancel, and a small gap does it: 0.25 mm radial on the bolts, 0.35 mm on jar lids with their taller teeth. When your print mates with an injection-molded part, the molded half is perfect and yours is not, so all the slop has to live on your side: the PET cap uses 0.5 mm against a bottle neck, twice the bolt number, and at 0.35 it jams on the second turn.
The M3 preset taught me the funny edge case. A 0.25 mm clearance against a 0.3 mm tooth erases the entire thread, so the small presets carry their own tighter clearances, 0.15 mm for M3, and the validator now refuses any combination where the gap eats the tooth. If a fresh nut binds slightly, run it on and off twice. The bolt is its own tap.
Three tricks that took me too long
Print threads vertical. A bolt standing on its head prints each turn of the helix as a clean stack of rings, and the 60 degree flanks stay within overhang limits by construction. Laid flat, the top of the cylinder is a sagging bridge and the thread comes out oval. Every part my generators export already stands the right way up.
Steal the multi-start idea from milk bottles. The 38 mm juice standard is a 3-start thread: three parallel helices, so the cap climbs three pitches per turn and pops off in a third of a twist. Big printed lids deserve the same. The jar generator does it with one field.
And the rule that makes two printed halves mesh at all: the nut's crest is the bolt's groove, half a pitch out of phase. I got the phase wrong exactly once, in the first jar prototype, and produced a lid that seated with a satisfying click and then never moved again in either direction. It is still on my shelf, permanently closed, holding the M3 bolts I no longer print.