Design for the nozzle, not for the screen

Most failed prints we're asked to look at aren't settings problems. The part was drawn as if it would be milled from a block: holes at whole numbers, sharp square corners, a neat 1.5 mm wall, a 60° overhang because in CAD it looked harmless. The nozzle doesn't read drawings — it can only lay plastic onto plastic that's already there. This article is about where those two views disagree, and how to fix them in the drawing, where fixing is almost free.

Five thresholds to keep in mind while drawing

Overhangs: why the reference is 45°

45° is quoted so often it sounds like a law. It isn't. Protolabs Network's guide explains the mechanism: at 45°, the newly laid bead is supported by exactly 50 % of the width of the bead below — enough to bond and enough not to sag. Beyond that, more of the bead hangs out than is supported, and it starts to droop away from the nozzle.

There's a second overhang failure few people name correctly: curling. At the edge of an overhang, the new bead gets thinner and cools unevenly, and differential shrinkage pulls it upwards. That means overhangs fail not only from the angle but from cooling and material — the same 55° can be clean in PLA with full fan and ugly in ABS in an enclosure.

Bridges: 5 mm, then think again

A bridge happens when the nozzle has to travel from one support to another over empty space. The two commonest cases are horizontal holes through a wall and the top of a hollow pocket. Protolabs Network states it plainly: because of how FDM works, sag or support marks are almost always present to some degree unless the bridge is shorter than 5 mm. In other words, below 5 mm you can ignore it; above 5 mm you have to choose: accept the marks, split the part, or post-process.

The cheapest fix is often a fourth option few people think of first: reorient the part. A horizontal hole needs support; turn the part 90° and it becomes a vertical hole and the support disappears — bringing a different issue, because vertical holes print undersize.

Vertical holes are always smaller than you drew

This is where newcomers usually think the printer is wrong. The mechanism is purely mechanical: as the nozzle travels around the hole, it presses the fresh bead onto the layer below so the two bond. That pressure turns the bead's cross-section from rounded to flat and wider — better bonding, but the extra width eats into the hole. The hole comes out smaller than drawn, and the smaller the hole, the larger the error proportionally, because the ratio of hole to nozzle diameter gets worse.

Slicers compensate for this, but the amount varies by slicer, printer, hole size and material. Protolabs Network's recommendation when hole size matters is blunt: print the hole undersize, then drill it to the right diameter. For bolt clearance holes this is almost always quicker than three test prints.

Wall thickness is a multiple of the line width

This is the one threshold here that comes not from the printer's geometry but from how the slicer divides lines. A 0.4 mm nozzle usually lays lines about 0.42–0.45 mm wide. A 1.2 mm wall divides into exactly three lines; a 1.5 mm wall divides into three lines plus a narrow gap the slicer must decide about — fill it with a thin line, stretch the three lines to fit, or leave it open. All three options make the wall weaker and the surface worse than a wall drawn to a true multiple.

How each slicer handles that leftover is a setting of its own, and the options are described in the handbook. But settings only choose how to handle a problem the drawing created. Set wall thickness for the nozzle you'll actually use and the problem doesn't exist.

Corners, elephant's foot and bed edges

Because the nozzle is round, every corner and edge printed has a radius equal to the nozzle size — there are no perfectly square corners on an FDM part. On the first layer there's another effect: the same squashing that shrinks vertical holes makes the bottom edge flare out beyond its nominal size. That's elephant's foot, and it's a very common reason two printed parts won't fit together even though the drawings match.

Protolabs Network recommends a 45° chamfer or a radius on every edge touching the bed when overall size or fit matters. It does three things at once: elephant's foot no longer sits on a mating face, the part's edge stops being where cracks start under load, and a scraper slides underneath more easily when removing the part. Tackling elephant's foot from the settings and bed side is in first-layer adhesion.

Small pins are a predictable failure point

Locating pins and screw bosses look perfectly reasonable in CAD. The issue is how they print. Protolabs Network describes the difference: pins over 5 mm in diameter print with both walls and infill, so they join solidly to the rest of the part; pins under 5 mm may be walls only, with nothing inside, creating a discontinuity between pin and body — which is where it breaks. At worst, a pin too small doesn't print at all because there's not enough plastic for the next layer to land on.

• —Fillet the base of the pin. A small radius at the base removes the stress concentration and adds material exactly where bending load is highest.
• —For small, critical pins, don't print the pin. Print a hole in that position, drill it to size, and press in a bought metal pin.
• —Insert threads, don't print them. For screw holes that are assembled and disassembled repeatedly, heat-set inserts outlast printed threads — the process is in gluing and threaded inserts.

Orientation is a design decision

FDM parts are strong in XY and weak in Z, because the joint between two layers is where the continuous cross-section is smallest. Protolabs Network explains the cause geometrically: each line's cross-section is like a rectangle with rounded ends, so the interface between layers is actually a series of small grooves, and every groove is a stress concentration prone to cracking.

In practice, orientation has to follow the direction of the load, not whichever direction gives the nicest surface or the fastest print. A hook printed standing up can be less than half as strong as the same hook printed lying down. Read this alongside print settings for strength — and note the point from PLA, PETG or ABS: changing orientation often improves strength more than switching to a more expensive plastic.

Fixing an overhang in CAD takes three minutes. Fixing it with settings takes three prints, and every one has marks.

Splitting: when two parts beat one

Splitting a model into pieces sounds like defeat, but it's often the right call: it removes overhangs that need support, lets each piece print in its strongest orientation, and turns one 14-hour print that might fail into four short ones. Protolabs Network lists splitting alongside reorienting holes and choosing print orientation as three moves that together cut cost, cut time and add strength. How to glue the pieces — ABS and PLA being two different stories — is in gluing and threaded inserts.

1. 01

   Orient the part for the load first, then look at support

   The first question isn't how to print with least support, but which way load will run through the part. Layers should sit perpendicular to the force trying to pull them apart.

2. 02

   Mark every overhang beyond 45° and ask why it's there

   Many overhangs can be chamfered, filleted or turned into a 45° slope without changing function. What's left is where real support is needed.

3. 03

   Turn horizontal holes vertical where you can

   Support inside a horizontal hole is hard to remove and leaves a rough bore. When you must choose among several holes, prioritise: blind holes, then small holes before large, then the holes whose size matters most.

4. 04

   Chamfer every bed edge

   45° or a radius. One CAD command, and it saves an afternoon of sanding edges.

5. 05

   Set wall thickness for the nozzle you'll use

   A multiple of the line width. If you don't yet know which nozzle, draw for 0.4 mm — the most common size and the easiest to scale up from.

6. 06

   Print a cut-out of the difficult area, not the whole part

   If only one overhang and one pair of holes worry you, cut that region out as a small block and print it. Fifteen minutes instead of an overnight print.

Fit clearance is a number to measure, not to guess

Whether two parts fit doesn't depend on one design rule but on four things added together: first-layer elephant's foot, the slicer's hole compensation, the material's shrinkage, and the real line width. So starting clearances are a table, and that table is in the handbook with a note that it's a starting point, not a guaranteed tolerance. Once you've recalibrated flow ratio per calibrate your printer per spool, the number is stable enough to go into the drawing.