How Far Can a 3D Printer Bridge?

There is no bridging distance. One print service says keep a bridge under 5 mm, another says 10, the Prusa knowledge base gives no number, and a torture test on any forum shows clean strands across 100 mm. They are counting different things. What a bridge can span is set by how fast the strand cools and how hard it is pulled, not by the width of the gap, and the honest number is measured on your own machine.

An overhang is stacked on a partial foundation. A bridge has none. The strand is laid across air and has to hold itself up until it sets, which makes it a different mechanical problem from anything else in the print, and the slicer treats it as one: bridges get their own speed, their own flow, the fan at full, and a line direction chosen to run from one anchor to the other.

Follow one strand across the gap.

Leaving the near anchor. The strand is bonded to the anchor behind it and hanging from the nozzle in front. In between it is a molten cantilever, and it starts to droop the moment it leaves the nozzle.^[1] Nothing supports it yet. The only things holding it up are its own stiffness, which at extrusion temperature is close to nothing, and the pull of the nozzle moving away.

Landing. When the nozzle reaches the far anchor the strand has to bond to it, which means arriving still hot enough to fuse. Once it does, the mechanics change entirely. The strand is no longer a cantilever but a string with two ends, and a string sags by its own weight against the tension in it. The nozzle supplies that tension: with nothing beneath to press the bead onto, a bridging strand is drawn across the gap rather than laid, and if the flow is reduced it is pulled thinner than the line width the slicer asked for. That is what a bridge flow ratio below 1 is doing. A thinner strand weighs less per millimetre and cools faster, and both reduce the sag.^[7]

Setting. In free air the strand loses heat almost entirely by convection; there is no bed and no layer below to conduct it away.^[2] This is why the part fan matters more for bridges than for anything else in the print, why a bridge low over a heated bed is harder than the same bridge higher up, and why materials that need a warm chamber bridge badly. Once the strand is below its glass transition it is a solid beam, and whatever sag it has is frozen in.

Shrinking. It then keeps cooling, and contracts. A strand pinned at both ends and shrinking goes into tension, which pulls it straighter — a bridge that lands at all often ends up tauter than it was laid. On long spans the same contraction shows up as force on the anchors. In a 2026 study of long free-standing printed beams, one pier was dragged toward the other as the bridging layers stacked, and the authors concluded that thermal residual stress, not gravity, was the dominant cause of failure.^[2]

So a bridge is not one thing. It is a cantilever for a moment, a string once it lands, and a beam under tension once it cools. The span a printer can manage is the span over which the first phase stays short and the strand arrives at the second still hot enough to stick.

The standard advice for a bad bridge is to slow down. It is worth understanding what that does, because it is not what most people think.

A 2024 simulation of a single strand laid across a gap found the opposite of the folklore: within the range it tested, 50 to 250 mm/s, the lower the printing velocity, the greater the deflection of the unsupported portion.^[1] The reason is the cantilever phase. Slower travel means every millimetre of strand spends longer hanging from the nozzle before the far end is anchored, and the front of the strand has more time to bend downward before it lands. Speed is what gets the strand across the gap while it is still a cantilever. It is not what cools it.

The same model found temperature to be the sharper lever. Below about 210 °C for the PLA it modelled, sag was nearly insensitive to nozzle temperature; above it, sag rose steeply.^[1] A hot nozzle produces a strand with low viscosity and no skin, and that strand droops as it is drawn out. A cooler one has a stiffer surface and holds its shape.

Where slowing down does help is in the other direction: a slower pass gives the fan more time per millimetre, which matters if the strand is arriving at the far anchor too hot to hold its shape or too soft to stop stretching. The two effects fight. Slicers resolve the fight by giving bridges a moderate speed, full fan and adjusted flow, and by treating those as a set. Changing one of them alone — typically dropping the speed to a crawl — can make the bridge worse.

The practical conclusion is that bridge quality is a cooling and flow problem before it is a speed problem, and that the fan and the nozzle temperature are where to look first.

Table: Bridging figures from published guides, and what each is actually counting.

A print service is promising a surface finish on a part it has not seen, with a profile it will not tune for you, in a material you get to choose. Five millimetres is the span it can guarantee across all of that. A bridging test on a forum is asking a different question: did the strands stay up. Somewhere between "cosmetically perfect on any machine" and "did not fall" is the number a designer actually needs — the span at which the underside of this bridge, on this printer, is still flat enough for what that surface has to do — and there is no way to look it up, because it is not a property of the design.

It is a property of the machine, the material and the profile, the same as the wall thickness and clearance figures in the rest of this series, and it is measured the same way: print a run of bridges and look.

Everything that changes how far a bridge goes acts on one of the phases above.

Cooling. The biggest single lever, and the one most often at the wrong setting. Bridges want the part fan at full; slicers do this by default, and it is worth checking that a material profile has not turned it down. A bridge at low Z over a heated bed is in warm air, and a bridge inside an enclosure is in warmer air still, so geometry that bridges cleanly on an open printer in PLA may not in a chamber in ASA.

Material. PLA sets fast and bridges well. PETG stays soft longer, clings to the nozzle, and tends to sag and string; it bridges shorter and wants a cooler nozzle for the bridge. ABS and ASA are printed warm so that layers bond, which is the opposite of what a bridge wants. Flexible filaments barely bridge at all. Composite filaments can be worse than their base polymer; the Prusa knowledge base singles out a carbon-fibre PETG as one that needs its bridge flow adjusted.^[6]

Nozzle temperature. Run bridges at the cool end of the material's range. The 210 °C inflection in the simulation is for one PLA under one set of assumptions, so treat it as a direction rather than a setpoint: hotter strands sag more, and the effect steepens.^[1]

Flow. A bridge flow ratio below 1 draws a thinner strand that is lighter and cools faster; OrcaSlicer's documentation is explicit that reducing it helps minimise sagging on external bridges.^[7] The thick bridges option goes the other way, printing the bridge layer at a height equal to the nozzle diameter — a heavier strand that is stronger and more reliable over long spans but leaves a rougher underside.^[7] The two settings are a trade between finish and reach.

Anchors. A strand has to land on something. Slicers look for the direction in which both ends of the bridge region are supported and lay the bridge lines that way; Prusa's bridging angle setting exists for the cases where the automatic choice is wrong.^[6] If the far anchor is a single perimeter one line wide, the landing is a coin flip. Give both ends of a bridge a solid margin to fuse to.

The second layer. Everything above is about the first layer of the bridge. The layer after it is printed on top of whatever the first layer did, so a sagged first layer produces a sagged second layer with a rougher top; slicers offer extra bridge layers to recover the top surface.^[7] A bridge that has to have a flat top is a bridge whose first layer has to be right.

Table: Bridge situations and the design move that shortens or removes them.

The first row is the one that pays most often. A rib costs almost nothing in a model and halves a span, and halving the span does more than any slicer setting can.

Three of those moves, built as comparison prints. Each one puts the plain geometry next to the fix on a single short print, so the difference is something you can turn over in your hand rather than take on trust.

The physics in this article is well supported in direction. The numbers are another matter.

No study of desktop material extrusion establishes a bridging distance at which surface quality fails. The figures in published design guides are conservative guarantees, not measurements, chosen to hold across every material and profile a service accepts. The academic work that exists sits on either side of the span a designer usually cares about: a computational model of a single strand crossing a gap, which is good for the direction of each effect and silent on thresholds,^[1] and an experimental study of 60 to 140 mm free-standing beams, in which the failures were buckling, delamination and layer shift driven by thermal stress rather than sag.^[2] That second result is worth holding on to, because it means "how far can it bridge" has more than one answer even on a single machine. The span at which the underside turns rough, the span at which strands drop, and the span at which a stack of bridging layers starts to pull its own piers over are three different lengths.

Bridging torture tests fill the gap between them, and they are useful, but they are uncontrolled — different fans, different filaments, different profiles, and a pass criterion of "it stayed up." Their numbers do not transfer.

What this means practically: the direction of every effect above holds on any printer. The distance does not. The coupon is a short print, and it replaces every number here with one that is about your machine.

The overhang article in this series draws the line between an overhang and a bridge and covers the horizontal-hole problem in more depth: The 45° Overhang Rule, and When It Is Wrong. The other two design rules end the same way this one does, in a number that belongs to the machine: How Thick Should 3D-Printed Walls Be? and Clearances for 3D-Printed Parts That Have to Move.

For the ceiling of a hollowed part, which is the most common bridge people do not notice they have designed, see Shell, Hollowing, and Wall Thickness for Printing. For turning a flat roof into a pitched one when there is no fillet tool to click, see Fillets and Chamfers Without a Fillet Tool.