Design for 3D printing

Most failed prints are not printer problems. They are design problems that the printer faithfully reproduced. A wall too thin to bond, a hole that came out 0.3 mm undersize, an overhang the plastic had nothing to rest on. Each of these is decided before the file reaches the slicer, and each has a number attached to it.

This page is those numbers. They are starting points calibrated for fused deposition modelling with a 0.4 mm nozzle, which is what most desktop printers are. Your machine will differ slightly, and the last section explains how to find out by how much.

Wall thickness is a multiple of your nozzle

The printer lays down beads roughly as wide as the nozzle. A wall is made of a whole number of those beads, so a wall thickness that is a whole multiple of the nozzle width prints cleanly, and one that is not leaves a gap the slicer has to fill with a thin, poorly bonded bead.

With a 0.4 mm nozzle that gives you a simple ladder:

• 0.8 mm is two perimeters and the practical minimum for anything structural. A single 0.4 mm wall is one bead with nothing bonded to it and will snap in your hand.
• 1.2 mm is three perimeters and the sensible default for a part that carries load.
• 2 mm to 2.5 mm for enclosure walls that get handled, dropped, or screwed into.

The common mistake is picking a round number instead. A 1 mm wall is not a nicer version of 0.8 mm. It is two beads plus a 0.2 mm gap, and that gap is where it will fail.

Holes print undersize, so oversize them

Two things pull a printed hole inward. The printer approximates a circle with straight segments that sit inside the true curve, and the extruded plastic contracts slightly as it cools. The result is consistent: a hole modelled at 5 mm comes out somewhere near 4.7 mm to 4.8 mm.

Practical allowances, added to the diameter:

• Clearance hole for a bolt: about 0.2 mm. An M4 bolt wants a hole modelled around 4.5 mm, not 4 mm.
• Bearing or shaft seat: about 0.4 mm, then measure, because this is the fit that matters most and varies most.
• Self-tapping screw into a boss: undersize deliberately. A pilot hole around 2.5 mm for a 3 mm self-tapper gives the thread something to bite.

Orientation matters here too. A hole printed vertically, with its axis along the print direction, is far closer to nominal than one printed on its side, which sags at the top into a slight oval.

The 45 degree rule

Each layer needs the layer below to sit on. Lean a face more than about 45 degrees away from vertical and each new bead is hanging over air, so it droops. Past that angle you need support material, which costs time and leaves a rough surface where it was removed.

The fix is usually geometric rather than a slicer setting. Replace a horizontal ledge with a 45 degree chamfer and it prints unsupported. Bridge a short gap rather than filling it, because a bridge between two anchored points prints acceptably where an unsupported overhang does not. And where a hole must be horizontal, a teardrop profile prints where a circle sags.

Fits between two printed parts

When both halves are printed, both carry the same dimensional error, so the clearance you model is not the clearance you get. Reasonable starting points:

• Press fit, meant to stay put: about 0.1 mm to 0.2 mm total clearance. Tight enough to need a push, loose enough not to split the outer part.
• Slip fit, sliding or rotating: 0.4 mm to 0.5 mm.
• Loose fit, a lid that must go on easily: 0.6 mm or more.

These are per-pair, not per-face. If a 20 mm peg goes into a hole and you want a slip fit, model the hole at 20.4 mm rather than adding 0.4 mm to each side.

Elephant foot and the bottom edge

The first layer is pressed into the bed so it spreads wider than the layers above, leaving a small flare around the base. It is invisible until the bottom edge of your part has to fit into something, at which point it is the reason it will not.

A 0.5 mm chamfer on the bottom edge removes the problem entirely and costs nothing. Design it in rather than sanding it off.

Fillets where things meet

A tall feature meeting a flat base at a sharp internal corner is two problems at once. It is a stress riser, so the part cracks there, and it is where thermal contraction concentrates, so it lifts from the bed there. A 2 mm fillet at that junction addresses both.

Keep fillets off the very bottom face where possible. A fillet that meets the build plate tangentially creates a razor-thin first layer that will not stick.

Orientation is a design decision

A printed part is anisotropic. The plastic within a layer is solid; the bond between layers is a weld, and it is weaker. A part pulled apart along the print direction fails at a layer line well below the strength it would show in plane.

So decide early how the part will be loaded and design so it can be printed with the layers running along that load. A hook printed lying flat is strong. The same hook printed standing up will break at the first layer line under the same pull. This constraint often decides the geometry, which is why it belongs in the design stage and not in the slicer.

Putting the numbers into a prompt

Every rule above is a number, and numbers are exactly what a prompt should carry. State them and the model arrives print-ready rather than needing a round of corrections.

Calibrating to your own printer

Everything above is a starting point. Your printer has its own consistent bias, and finding it takes one print. Model a test plate with a row of holes at 4.2, 4.4, 4.6 and 4.8 mm and a row of pegs at 5.8, 6.0 and 6.2 mm. Print it, measure every feature with calipers, and record the difference between modelled and actual.

That difference is close to constant across parts, so once you have it you can apply it deliberately instead of guessing. It is the single most useful hour you can spend with a new printer, and it makes every subsequent part fit first time.

Where to go next

For the full prompt-to-slicer workflow see text to CAD for 3D printing, and for the prompt pattern itself see CAD prompt examples. The enclosures in the gallery show these rules applied to real parts, each with the prompt that generated it. If you are choosing a tool rather than a wall thickness, CAD for 3D printing covers that.