The 45° Overhang Rule, and When It Is Wrong

The 45 degree rule is not a physical limit. It is a rule of thumb about how much of each new layer the layer below is holding up, and at 45 degrees that fraction happens to be about half on a common set of printer settings. Change the layer height and the angle moves. Most printers will do 60 degrees at a fine layer height and fail at 40 degrees with a thick one.

A sloped wall is printed as a staircase. Each layer sits on the one below, shifted outward by however far the slope has moved in one layer of height. If the wall leans by an angle from vertical, and each layer is a certain height, then every layer steps out by

step = layerHeight * tan(angle from vertical)

and the extruded bead that has to bridge that step is one extrusion width across. So the quantity that decides whether the layer holds is a ratio, not an angle:

unsupported fraction = layerHeight * tan(angle) / extrusionWidth

At 45 degrees, with a 0.2 mm layer and a 0.4 mm wide bead, that is 0.2 divided by 0.4, or half the bead hanging over air. Half is roughly the point where surface tension and the bead's own stiffness stop coping and the underside turns rough.

That is the whole basis of the rule. It is not that 45 degrees is special. It is that 45 degrees produces a 50 percent overhang on the settings that were common when the rule became folklore.

Run the ratio at a few settings and the rule stops looking like a constant.

Table: Unsupported fraction of each bead, for a 0.4 mm extrusion width. Above 100 percent, a new bead lands entirely beyond the one below it.

Two things fall out of this table.

A fine layer at 60 degrees is better supported than a coarse layer at 45. A 0.1 mm layer at 60 degrees leaves 43 percent of each bead unsupported; a 0.2 mm layer at 45 degrees leaves 50 percent. The steeper overhang is the easier print. Anyone who has printed the same model at two layer heights has seen this and probably blamed something else.

There is a hard ceiling, and it is computable. The overhang reaches 100 percent when the step equals the bead width, which happens at an angle of atan(extrusionWidth / layerHeight). For a 0.4 mm bead that is 63.4 degrees at 0.2 mm layers and 76.0 degrees at 0.1 mm layers. Past that the printer is extruding into air with nothing beneath, and no amount of cooling saves it.

So the useful version of the rule is not a number. It is: decide the layer height first, then work out what angle that buys.

The ratio sets an upper bound. Several things pull the practical limit below it.

Cooling. A bead that is still molten sags regardless of how much of it is supported. Part cooling is the single biggest variable in overhang quality, which is why the same geometry succeeds on a machine with good ducting and fails on one without.

Material. PLA sets quickly and holds overhangs well. PETG stays soft longer and droops. ABS and ASA are printed in a warm chamber precisely so layers bond, which also means they stay soft longer. The material changes the achievable angle more than most design changes do.

Speed. Faster printing gives each bead less time to solidify before the next lands on it. Slowing overhang perimeters specifically is usually more effective than slowing the whole print.

Where the overhang is. An overhanging outer perimeter has nothing on its outboard side. The same angle in a region backed by infill is far more forgiving, because the bead is supported from behind even when it is not supported from below.

Curling. A drooping bead that then cools and contracts lifts its free edge upward into the path of the nozzle. This is the failure that turns a cosmetic problem into a crash, and it compounds over successive layers rather than staying local.

The practical consequence is that the angle a machine can hold is a property of that machine, that material and that profile. It is worth measuring once, with an overhang test print, and writing down — in the same way a printer profile records a flow rate.

A bridge is a horizontal span anchored at both ends. It is not an overhang at 90 degrees, and treating it as one gives the wrong answer.

An overhanging bead is stacked on a partial foundation. A bridging bead is stretched between two supports, and the slicer knows it: bridging typically prints in a single pass at a different speed and flow, with the fan at full, so the strand is pulled taut and cools before it can sag. A well-tuned machine bridges 20 mm or more cleanly and will still make a mess of a 70 degree overhang.

This is why a horizontal hole prints badly and a horizontal slot with a flat top prints fine. The hole's top is not a bridge — it is a continuously varying overhang that only reaches horizontal at the crown, passing through every bad angle on the way up. Two standard fixes follow directly: make the hole a teardrop, so the top becomes two steep walls meeting at a point, or make it a hexagon, so the top is a short bridge between two 60 degree walls.

The same logic covers unsupported floors over a cavity. A short span bridges. A long one sags in the middle no matter how good the cooling, because the strand is stretching under its own weight before it sets.

Supports work. They also cost material, time, and a surface finish, and they have to be removed from somewhere a tool can reach. Geometry that does not need them is usually cheaper than geometry that does.

Table: Overhang situations and the design move that removes them.

The first and last of these are the ones people forget. Chamfering an underside costs almost nothing in a model and removes a support region entirely. And reorientation is free: it changes no dimension, and it is the single highest-leverage decision available once the geometry exists.^[1]

The 45 degree rule has a rigorous form, and it appears somewhere most makers never look: topology optimization.

When an optimizer is asked to design a shape for stiffness, it will happily produce something no printer can build, full of unsupported overhangs. The fix is to write the overhang limit into the optimization as a constraint, so that every candidate shape is self-supporting by construction. That work is now a substantial literature. One approach adds an overhang constraint that eliminates sacrificial support material through the design itself.^[2] Another is a filter applied layer by layer, projecting each layer onto what the layer below can support, which has the appealing property of making any optimized result printable rather than checking it afterwards.^[3] A later constraint achieves the same end more cheaply.^[4] And an earlier, more pragmatic route post-processes the optimized topology into something printable instead of constraining the optimizer.^[5]

Two things are worth taking from this for ordinary design work. First, the rule is real enough to be formalised — the same angle threshold, treated as a hard constraint, produces parts that build without support. Second, the threshold in those papers is an input, not a discovery. The researchers pick it from process knowledge, the same way a maker picks it from experience. Nobody derived 45 degrees from first principles. It is the ratio in the second section, rounded to a number people can remember.

The geometry in this article is arithmetic and holds regardless of machine. The thresholds are a different matter.

The claim that 50 percent unsupported is roughly where quality falls apart is a summary of accumulated practice, not a measured constant. No widely cited study establishes it for desktop material extrusion, and the number would in any case depend on material, cooling and speed. The published overhang thresholds that do exist mostly come from the powder-bed fusion and topology optimization literature, where the constraint is chosen as an input rather than measured as an output.

What this means practically: the table in the second section tells you how much of a bead is hanging in air, which is a fact. Whether your printer tolerates 43 percent or 65 percent is something only your printer can tell you. An overhang test print takes twenty minutes and replaces every number here with one that is actually about your machine.

The other two design rules in this series end the same way — in a number that belongs to the machine rather than the model. How Thick Should 3D-Printed Walls Be? is about the nozzle setting the wall, not the design. Clearances for 3D-Printed Parts That Have to Move is about the gap two parts need between them before either will move. Both come with a coupon to measure it on your own machine.

For chamfering an underside rather than supporting it — and four other ways to break an edge when there is no fillet tool to click — see Fillets and Chamfers Without a Fillet Tool.

For geometry that sidesteps the problem entirely: sheet-based minimal surfaces never exceed a moderate overhang anywhere on the surface, which is why they print without support at densities where a strut lattice would not. Gyroid TPMS: When Geometry Becomes Structure covers why.