Overhang and bridge failures

FDM builds by stacking beads, and every bead needs something underneath to land on. When a feature leans out past the layer below it, or spans a gap with nothing under it at all, the plastic has to hold its own shape in mid-air until it cools — and where it can't, it droops, curls or falls. The signatures are unmistakable: a rough, sagging underside on a leaning face; strands drooping across a gap that should have been a flat ceiling; or a show surface wrecked by the supports you added to prevent both. This is the one failure family that geometry causes and geometry mostly cures, which is why the design lever here is the whole game. The full model is in Orientation and overhangs and Supports and bridging; this article reads the failures.

Drooping undersides — overhang sag

An overhang is a wall that leans outward as it rises, so each layer sits partly off the edge of the one below. Up to a point that's fine — the new bead has enough of the previous layer under it to grip. Past roughly 45° from vertical, more than half of each bead hangs over open air, and with nothing to land on it sags before it sets. The signature is a rough, drooping, curled underside that gets worse the shallower the overhang, until it collapses entirely into strings.

The first fix is orientation, because the overhang is only an overhang in the direction you chose to print. Rotate the part so the offending face points up or runs vertical instead of leaning out over air, and the problem simply doesn't exist — no support, no scar, no sag. When geometry can't be reoriented out of trouble, cooling is the machine lever: more part-cooling fan freezes each overhanging bead faster so it holds its shape before it can droop (which is exactly why PLA overhangs so well and ABS, printed with little fan, so badly). Only when neither is enough do you add support — a sacrificial scaffold under the overhang — accepting that it costs plastic, time and a scarred surface where it touched.

Sagging across a gap — failed bridges

A bridge is the extreme case: a horizontal span printed across open space between two supports, with nothing beneath it at all — the ceiling over a window, the top of a horizontal hole, the roof of an enclosed cavity. The printer prints it as fast, taut, well-cooled strands stretched straight across, and a good bridge relies on that tension and quick cooling to hold the line flat until it sets. When the span is too long, the cooling too weak, or the plastic too runny, the strands sag in the middle and the ceiling comes out drooping or broken.

The machine levers are maximum cooling and the slicer's bridging settings — faster bridge speed and tuned flow to keep the strands taut. But the powerful move is again in the model: keep spans short where you can, and round the top of horizontal holes into a teardrop so the hole self-supports instead of trying to bridge a flat ceiling across its widest point. A horizontal hole printed round sags at the top and comes out oval; the same hole with a teardrop top prints clean without support. Reprofiling the geometry to avoid the bridge beats bridging it well.

The surface the supports ruined

Sometimes the overhang printed fine and the failure is the support scar left behind: a rough, pocked patch on a face that should be smooth, or supports fused so hard into a cavity you can't get them out. The fix is rarely a better support profile. It's to orient the part so the surface that matters never touches support — put the show face up or vertical, let the scars fall on a hidden underside, and choose the print direction around which face you're willing to sacrifice. This is the same decision that governs seams and elephant's foot: every part has faces that matter and faces that don't, and orientation is how you point the damage at the ones that don't.

Overhangs and bridges are the clearest case in all of FDM that the best troubleshooting happens before the print. A part reoriented so nothing leans past 45°, with horizontal holes teardropped and spans kept short, has no overhang problem to solve — the defect was designed out rather than tuned away. When you can't design it out completely, cooling and support catch what's left, in that order. But reach for the model first: it's the lever that costs nothing, leaves no scar, and travels with the part to every printer you own.