Modeling threads

A thread is a helical groove wound around a cylinder. On paper it looks like just another detail. In FDM it isn't. A thread pushes three things your printer struggles with to the limit all at once: fine in-plane detail, overhangs that sag, and a fit that has to slide without seizing. That's why most threads people try to print come out useless: either they vanish because they're thinner than a single extruded bead, or they seize because they were modeled to exact size, or they come out hairy on the underside because they were printed lying down. You avoid all three in the modeling, before you slice anything. This article is about doing that.

Coarse pitches: M6 and up

The crest and root of a thread are tiny grooves, and when the cylinder stands upright what decides whether they're distinguishable is layer height, not bead width. With a vertical axis, the helix climbs layer by layer: every turn has to draw a crest, a flank, and a root by stacking 0.2 mm layers. At a 1 mm pitch, that's barely five layers per turn to form all three. Drop the pitch to 0.5 mm and you're left with two or three layers per turn — not enough to tell crest from root. Neighboring layers overlap, and the thread blurs into a cylinder with soft ripples that won't thread into anything.

The conservative starting point is a pitch ≥ 1 mm, which is where the coarse series starts at M6. It isn't an impassable physical boundary—with a rounded profile and a fine layer you can pull off lids at a 0.7–0.8 mm pitch—but below 1 mm you're leaning on a dialed-in layer height and good calibration, and the margin for error narrows fast. Watch out: a single nominal diameter can have both coarse and fine pitches. Don't go by the "M" alone: a coarse M8 at 1.25 mm runs loose, while a fine M8 at 1 mm sits right at the threshold. Always look at the pitch: that's the dimension that actually has to print.

Flank clearance to prevent seizing

A bolt thread and a nut thread modeled to the same nominal profile won't thread together in FDM. It's the problem of any printed fit, but worse: the male comes out oversized (the bead pushes the contour outward) and the female comes out tight (the bead eats into the gap), so the flanks collide before they ever touch. On top of that, the printed surface is rough: the layer ridges of one flank scrape against those of the other. A thread modeled to exact size doesn't just run tight—it won't go in, or it goes in one turn and jams.

The fix is to leave 0.1–0.2 mm of total flank clearance—the gap between the male flank and the female flank, measured perpendicular to the contact surface. That tenth or two of gap lets the male's oversize and the female's tightening coexist without interfering, and leaves a margin for the roughness. The honest way to split it is the same as for any printed fit: leave the male at its nominal size and open up the female, so you have a single dimension to tweak if it threads too tight. The full reasoning is in Captive nuts and clearance holes.

Don't confuse flank clearance with diameter clearance: they're the same idea seen from two sides, but you don't convert one into the other by a simple subtraction. On a 60° metric profile, the flank is angled, so the perpendicular gap you're after and the diameter shift you apply don't match: the diameter has to move a bit more than the flank clearance you're chasing. The clean way is to work on the pitch diameter (the mean diameter of the thread, where male and female touch) and leave your clearance there. But don't overdo it: with too much clearance, the thread wobbles, loses contact, and strips with almost no torque. The goal is to thread it by hand, with no hard spot, but without the play of a loose nut.

The profile: rounded, not a sharp V

The standard metric thread has a V profile with straight flanks and an almost-sharp crest. It's perfect for cutting into metal and terrible for depositing in plastic: the sharp crest is exactly the detail the nozzle can't resolve, and the peaked bottom of the root forces a change of direction the bead can't trace cleanly. Shrunk to printing size, a metric V is almost all gap: little material per turn and a lot of fragile crest.

Model a rounded or trapezoidal profile instead. A trapezoidal thread—gently ramped flanks, flat crest and root—is built from beads that lay down well and pack more material into each turn. That brings two advantages. The first is resolution: flat crests and roots come from clean stacked layers, without the tip the nozzle rounds off. The second is fault tolerance: with more cross-section per turn, an over-extruded crest here and there doesn't wreck the thread, because the neighboring turns keep gripping. A rounded thread forgives; a printed sharp V strips on the first turn.

The profile also decides how much the lower flank overhangs, so keep it gently ramped. On a profile with flanks at about 30° to the bolt axis—metric, trapezoidal—the lower flank of each turn stays well clear of a severe overhang and prints cleanly enough. By contrast, a profile with flanks nearly perpendicular to the axis—ACME or square—leaves the underside of each turn almost flat over the previous root: there you do have a hard overhang that comes out rough or sags. The generous, angled-flank profile is the one that both threads best and prints best.

Print with the axis vertical

Orientation decides whether the thread you dialed in on screen survives the print. Print the cylinder with its axis vertical, perpendicular to the bed. That way the helix climbs gently layer by layer: every turn rises one pitch over the full height, and the crest and root of each turn come from clean, stacked layers. It's the same self-supporting logic as any gentle slope: nothing hangs in the air because each bead rests on the one from the turn below.

There's one detail worth understanding, because it explains why even a well-oriented printed thread is never perfect. The helix rises very little per turn—on an M6×1, the helix angle is around 3°—so the problem isn't the incline, but that on each turn one flank faces up and the other faces down. The downward-facing flank is, technically, an overhang: it hangs over the root of the previous turn. With an angled-flank profile that overhang is gentle and comes out acceptable, if a bit rougher than the upper flank. That's why flank clearance isn't optional: part of it is eaten by that roughness on the lower flank. And that's why the generous profile pays off, giving margin for that flank to come out imperfect without ruining the fit.

Lay it down and you lose both: the helix stops climbing layer by layer and becomes a run of overhangs and bridges wrapping the cylinder. Half the thread ends up hanging in the air on its underside: it sags, loses its roundness, and the fit you measured vanishes. If a part has a thread you care about, orient it with that thread pointing up and deal with the rest afterward. Why the layer plane rules all of this is covered in How FDM shapes your design.

What a printed thread is actually good for

It's worth being honest about what you get. A thread printed in plastic, even well modeled and well oriented, is a light-load, occasional-use thread. It works beautifully for lids, jar caps, hand-tightened collars, big adjustment knobs, retaining rings, and tripod-style mounts: parts you thread by hand, without a tool, and that don't carry sustained structural load.

A printed thread is not a fastener. Don't use it for loaded joints, or anywhere you have to apply torque with a wrench, or in parts you assemble and disassemble daily. Plastic crests strip with little effort and, once stripped, don't come back: you're left with a smooth cylinder. If the thread has to take force or repeated tightening, don't model it in plastic; put metal in the hole—a heat-set insert or a tapped thread—and let a metal screw do the work.

The rule, in one sentence: if what fails when the thread strips is "I rethread the lid with a bit more care," print it; if it's "a part that was holding something comes loose," use metal. If so, continue with Designing for heat-set inserts, which explains how to leave the boss ready to receive the real thread.