Preparing a file that prints.

An STL file is a mesh: a shell of triangles approximating your surfaces. It prints perfectly well, and if it is all you have, send it. But it has thrown away the design intent — there are no longer any circles, planes or dimensions, only triangles that happen to sit near where those things were.

A STEP file keeps the real geometry. A hole is still a hole with a diameter, a face is still a plane. That matters the moment something needs changing: opening a hole 0.3 mm, thickening a wall that will not survive, adding a fillet where a part will crack. In STEP those are small edits. In STL they are reconstruction work, and doing them well costs more than printing the part.

• STEP (.step, .stp) — preferred. True geometry, editable, measurable.
• STL (.stl) — accepted and prints fine. Export at a fine resolution so curves are not faceted.
• PDF drawing — send alongside either when specific dimensions are critical, with tolerances marked.
• Native CAD — usually fine too; tell us what produced it.
• No CAD at all — a dimensioned sketch, or the physical part, is a legitimate starting point. See product design engineering.

The most common avoidable failure is a wall too thin to be printed as a wall. As a working rule, keep any wall at least two to three times the nozzle diameter — around 1 mm on our 0.4 mm nozzle, and more if it carries load.

Below that, the wall becomes a single extruded line with nothing bonded either side of it. It looks fine and splits at the first stress. The features where this bites are predictable: embossed text, thin ribs, clip fingers, and the outer skin of a part someone hollowed out to save material.

This is the most under-appreciated factor in whether a printed part survives, and it costs nothing to get right.

An FDM part is built in layers, and it is substantially stronger along those layers than across the bond between them. So a load trying to peel layers apart will break a part that would have been comfortably fine printed at ninety degrees to that. A hook printed flat snaps; the same hook printed on edge holds.

We choose orientation, but we can only choose well if we know how the part is used. One sentence — "this bolts here and the load pulls this way" — changes the decision. Without it we optimise for surface finish and print time, which is the right default and the wrong one for a loaded part.

Printed holes come out slightly undersized, because the extrusion pulls inward on an inside curve. Small holes that must be accurate are best printed a little under and then drilled or reamed — which we can do, if the drawing says which holes matter.

• Fastener holes — design for a self-tapping screw or a heat-set brass insert rather than a printed thread. Printed fine threads strip.
• Clearance holes — add a little; a nominal clearance hole often prints as an interference fit.
• Shafts and bearings — print undersize and ream, or design for a pressed-in bush.
• Coarse threads — large-pitch threads do print usably. Anything M6 or finer, use an insert.

• Large flat unsupported spans — sag in the middle. Add a rib, a chamfer, or split the part.
• Tall thin towers — vibrate as the head moves and print poorly at the top. Brace them or reorient.
• Sharp internal corners — stress concentrators, and the first place a part cracks. A small fillet is nearly free.
• Knife edges — anything tapering below one extrusion width simply is not printed. Give it a flat of at least the nozzle width.
• Steep overhangs — beyond about 45° from vertical they need support, which costs time, material and surface finish. Chamfer instead of overhanging where you can.
• Enclosed voids — trap support material with no way to remove it. Add an access hole or split the part.

Send the file before you commit and the review comes back with these flagged. There is no charge for it — finding them after you have paid for parts serves nobody. See how printing is priced for which of these changes also make the part cheaper.