Supports and bridging
A nozzle lays molten plastic on top of whatever is already underneath. That one simple sentence decides almost everything you can and can't print in a single piece. As long as each bead lands on top of the one in the layer below, all is well. The moment you ask it to deposit onto thin air, the plastic droops, cools crooked, and drags the next layer down with it. Support is the patch for that problem; bridging is the exception that escapes it; good design is the way to need neither. It helps to see all three for what they really are: consequences of how a part gets built, one layer on top of the last.
What a support really costs
Sometimes a surface hangs over empty space and, no matter how you rotate the part, there's no geometric fix. That's what support is for: a sacrificial scaffold the slicer builds under the overhang to give the layers something to land on, then you snap it off afterward. It works, but it isn't free.
It costs material, because that scaffold is plastic you print only to throw away. It costs time, because the nozzle has to trace it layer by layer just like it traces the real part. And you pay for it in finish: wherever the support touches the part you're left with a rough, pitted face — and the more horizontal the overhang it's holding up, the worse it gets. That interface is a permanent compromise between two competing demands. Bond it too tightly and you can't peel it off without tearing out material; leave it too loose and the overhang's first layers sag into the gap before the support catches them, and the face comes out just as ugly.
That's why a good designer's goal isn't to use supports well. It's to need them as little as possible. Most of the time, what looks like it's asking for support is really asking for a small change to the model.
The 45° rule
Everything keys off the overhang angle, measured from vertical. It's the same threshold that governs Orientation and overhangs, and here it sorts geometry into three bands.
Up to about 45° the part prints reasonably well: each bead overlaps the one below it enough to hold, though at a dead 45° the underside finish is already starting to degrade. Treat 45° as the limit, not as headroom. Past 55–60°, or facing a flat horizontal ceiling, the bead finds almost nothing to rest on and you need real support. Most real-world geometry lives in between. That's exactly the band beginners overlook: a moderately steep overhang almost never needs a scaffold — it needs you to stop treating it as an overhang.
Designing it out
Two tools handle most cases, and both come down to turning an impossible angle into a safe one.
Chamfer instead of overhang. Wherever a flat face would jut out horizontally — the shelf coming off a wall, the lip of an enclosure — replace the sharp underside with a 45° chamfer. A flat ceiling sags because its beads start out with nothing beneath them; a 45° slope prints as a self-supporting ramp, each bead resting on the last. Same bulk, same function, no scaffold.
Teardrop for horizontal holes. A round hole with its axis horizontal hides a treacherous overhang at the top. The last ceiling layers have to close the widest part of the circle almost horizontally, with no support, so they sag inward: the hole comes out warped, off-center, and out-of-round right at the top — exactly where a shaft or screw needs it round. The teardrop fixes this by capping the top in a point. That point isn't decorative: built at 45° or less, it turns the flat ceiling into two overhangs that never exceed 45°, so each layer rests comfortably on the one below and climbs to the apex without sagging. The hole stays round and accurate, and the internal support disappears. Use it as soon as the ceiling's roundness matters, not just on big holes: a horizontal M3–M4 shaft or screw already sags noticeably without it.
The third way out, when neither the chamfer nor the teardrop is enough, is to split the geometry: cut the part along a plane where each half prints fully supported, then join them afterward. An overhang that's impossible in one piece is often trivial in two. You'll see it in detail in How FDM shapes your design; here, just keep it in reserve for when reorienting and redesigning fall short.
Bridging: crossing the gap with nothing underneath
There's one case where the printer deposits onto thin air and it still comes out fine: the bridge. When the nozzle has a solid anchor on both sides of a gap, it doesn't let the bead drop — it stretches it from one support to the other, pulling a hot, straight strand that cools in flight before gravity wins. That's how you print the flat top of a hole, or the ceiling of a closed cavity, without support.
The mechanism is that strand pulled taut between two anchors, solidifying before it can sag. For it to work you need three things at once: tension between the two ends, a slow enough deposition speed, and enough cooling for the bead to hold its straight line while the supports keep it taut. In PLA with good cooling, FDM lays clean strands across about 10 mm with no special tuning, and reaches 20–30 mm if you drop the bridging speed and tune the flow so the strand doesn't thin out too much; beyond that the finish degrades even if the bridge doesn't collapse. PETG and ABS bridge considerably worse — less effective cooling, more sag — so scale these PLA figures down for them.
A bridge has two ways to fail. One is length: past a certain run, the strand cools before it crosses and sags under its own weight in the middle. The other is subtler: a bridge needs a nearby edge to pull from. If the gap is wide in both directions — a surface with no close edge in any sense — the beads in the middle have no nearby anchor to grab onto and they sag, even though each individual run looks short. A bridge lives on the tension between two close edges; take one away and it stops being a bridge.
Starting values, and when you can't avoid support
These are starting values in PLA with a 0.4 mm nozzle and decent cooling. As with everything in FDM, your machine will fine-tune the numbers; use them to design from, not as final truth.
| Situation | Safe limit |
|---|---|
| Overhang from vertical | up to ~45° without support |
| Chamfer on the underside | 45° prints clean enough |
| Clean bridge | up to ~10 mm with no special tuning |
| Rough but usable bridge | ~20–30 mm with tuned speed and flow |
| Horizontal hole | use a teardrop if the ceiling dimension matters (already from Ø3–4 mm) |
Even so, some parts genuinely need support: a wide flat ceiling, an island stranded in mid-air, an overhang no orientation can rescue. When you hit that case, two design decisions make the support bearable.
First, put it where the rough face doesn't matter. Support always leaves a mark, so orient the part — if the function lets you — so the supported surface is one nobody sees or touches: the base, the interior, a mating face that was already going to be hidden. The ugliness is inevitable; its location isn't.
Second, leave clearance for release. A support bonded to the part with no margin tears off material and leaves craters; one with a minimal gap pops off cleanly in a single pull. Part of that margin comes from the slicer's interface setting — a z-gap on the order of 0.1–0.2 mm in PLA is a good starting point — but the rest you set in the model: keep the surfaces that need support from ending in thin edges or in nooks the tool can't reach to pry at. The modeling job is to give the slicer the least surface to hold up — and to keep whatever's left accessible.
That closes out the geometric side of printability. The natural next step is the one that decides most of these angles before you even think about supports: how you seat the part on the bed. You'll see it in Orientation and overhangs.