Gears are the most requested mechanical part in 3D printing, and the most reliably disappointing. A printed gear set usually fails in one of two ways: the teeth bind solid within a few turns, or a tooth shears off at the root the first time it sees real torque. Neither failure is the printer's fault. Standard CAD gear generators are written for machined metal, where the manufacturing process removes material and holds microns. FDM does the opposite — it deposits molten plastic that bulges outward, and it builds tooth flanks as a stack of horizontal steps.
This guide covers the geometry that decides whether a printed gear train runs for hours or strips in seconds: module selection, where backlash actually comes from, root fillets, orientation trade-offs, wall count, shaft attachment, and the material ranking backed by published wear testing.
Why Gears Fail Differently in Plastic
Before any numbers, it helps to understand what is different about a printed tooth. Three mechanisms drive almost every failure.
- Layer stepping adds material, it does not remove it. An involute tooth flank is a curve. Your printer approximates it with horizontal layers. The outer corner of each step protrudes beyond the theoretical curve, so the tooth comes out thicker than designed. At a 0.20 mm layer height this adds roughly 0.04–0.08 mm of effective tooth thickness per flank. A pair of gears designed to zero clearance therefore interferes by default.
- Bending stress on a tooth runs parallel to the layer lines. When you print a spur gear flat on the bed, the tooth is loaded in bending exactly along the layer interfaces — the weakest direction in any FDM part. This is why tooth roots shear cleanly and horizontally.
- Printed teeth reach only a fraction of bulk material strength. Load testing of printed gears commonly shows 25–35% of the theoretical bulk strength, because inter-layer bonding is weaker than the filament itself. Design loads must be derated accordingly.
None of this means printed gears are impractical. It means the design has to compensate deliberately instead of inheriting metal rules. The practical starting point is an understanding of how dimensional accuracy and tolerances behave on a desktop printer, because gear mesh is where those errors compound fastest.
Module: The One Number You Cannot Get Wrong
Module is tooth size, expressed as pitch diameter divided by tooth count. Two gears can only mesh if they share the same module and the same pressure angle. Get this wrong and no amount of backlash tuning will help.
The governing constraint for 3D printing is your nozzle. A tooth flank needs to be wide enough to be described by a full extrusion pass. Below module 1.0, the tooth becomes narrower than a standard 0.4 mm nozzle can cleanly resolve, and the involute profile rounds off into something that neither meshes smoothly nor carries load.
| Module | Nozzle fit | Use it for |
|---|---|---|
| Below 1.0 | Below the practical floor for a 0.4 mm nozzle | Avoid. Teeth round off, bind, or snap |
| 1.0 | Workable lower bound, demands good calibration | Light-duty mechanisms, display models, fine gear trains |
| 1.5 – 2.0 | The safe baseline | Most functional parts, motor drives, anything under load |
| 2.5 – 3.0 | Coarse and forgiving | Heavy torque at low speed, large slow mechanisms |
Two related numbers matter as much as module. Tooth count should stay at 17 or above to avoid undercut at a 20° pressure angle; 13 is the hard minimum before the tooth shape starts deforming. Face width (gear thickness along the axis) should reach 8–10× the module for printed gears — thinner gears tilt on their shaft and concentrate all the load on a tooth edge. A module 2.0 gear wants roughly 16–20 mm of face width.
Backlash: Build It Into the Tooth, Not the Frame
Backlash is the intentional clearance between meshing teeth. In metal gears it exists to absorb thermal expansion and lubricant film. In FDM it has to absorb much more: layer stepping, filament diameter variance, first-layer elephant's foot, and the general dimensional drift of a desktop machine.
The critical mistake is trying to create backlash by moving the shafts further apart in the frame. This does not work. Separating the centers changes the contact point on the involute and produces a gear pair that is simultaneously loose in one phase of rotation and tight in another. Backlash has to be cut into the tooth profile itself, by offsetting each tooth flank inward.
| Material | Backlash per side | Total meshing gap | Notes |
|---|---|---|---|
| Resin (MSLA) | 0.04 – 0.08 mm | 0.08 – 0.16 mm | Far more precise than FDM, but post-cure shrinkage varies by brand |
| Nylon (PA12) | 0.10 – 0.15 mm | 0.20 – 0.30 mm | Excellent layer adhesion; must be printed dry or it swells and binds |
| PETG | 0.12 – 0.15 mm | 0.24 – 0.30 mm | Low shrinkage and easiest to get right on the first attempt |
| PLA | 0.15 – 0.20 mm | 0.30 – 0.40 mm | Tends to run over-dimension; start at 0.20 mm if unsure |
| ABS / ASA | 0.18 – 0.25 mm | 0.36 – 0.50 mm | High shrinkage; use the higher value inside an enclosure |
| TPU / flexible | 0.25 – 0.35 mm | 0.50 – 0.70 mm | Compresses under load, eating its own clearance. Light duty only |
These figures assume a 0.4 mm nozzle, a 0.20 mm layer height, and a reasonably calibrated printer. Layer height shifts them. At 0.30 mm layers the staircase is coarser and you should add roughly 0.05 mm per side. At 0.15 mm layers the flanks are noticeably smoother and you can reduce the target by 0.02–0.03 mm per side. Below 0.10 mm the returns diminish.
A useful sanity check on material choice: PETG has a coefficient of thermal expansion around 68 µm/(m·°C). A 30 mm gear that warms from 20°C to 50°C during operation grows about 0.06 mm in diameter. A pair that only just clears when cold can bind after five minutes of running. Building the backlash in gives you that margin for free.
Root Fillets: The Line Between a Working Tooth and a Sheared One
This is the single highest-value change most people skip. Where a tooth flank meets the root circle, most CAD gear generators leave a sharp internal corner. In a printed part that corner is a stress concentration, and under bending load it cracks and propagates.
Adding a fillet radius equal to 0.3 to 0.5× the module distributes the bending load across a much larger area. For a module 2.0 gear that means a 0.6–1.0 mm radius. Many parametric gear generators expose this directly as a fillet factor; leave it in the region of 0.30 unless there is a reason to change it.
The same logic applies to hubs and bores. A thin ring of material around the shaft hole will split when torque is applied. A hub with an outer diameter at least 2× the shaft diameter, standing proud of the gear face, prevents the bore from cracking open.
Orientation: A Genuine Trade-Off, Not a Trick
There is a real conflict here, and it is worth naming honestly instead of repeating the usual advice.
| Orientation | Tooth profile accuracy | Tooth bending strength | When to choose it |
|---|---|---|---|
| Flat (shaft axis vertical) | Best — the involute is drawn entirely in XY, where motion systems are most precise | Weakest — bending load runs parallel to layer lines | Default for spur gears, and for any case where smooth meshing and a round bore matter |
| Vertical (shaft axis horizontal) | Poor — tooth flanks become stepped, bores need support | Better along the tooth root in specific load directions | Only when a specific load direction justifies the profile loss |
Printing flat is the right default. When you need more strength, the better lever is not reorienting the gear but increasing wall count and adding root fillets. Reorienting trades away the tooth accuracy that makes the gear mesh in the first place.
Walls Beat Infill for Gear Teeth
Infill percentage is the setting people tune first and it matters least. When a tooth is loaded, the stress runs through the outer perimeter — the shell — not through the sparse lattice in the middle. A tooth with 100% infill and two walls is weaker than the same tooth with 40% infill and five walls.
Set 4 to 6 solid perimeters and treat infill as secondary. This costs far less filament than a high infill percentage and puts the material exactly where the bending load travels. It also pairs well with correct layer bonding and orientation strategy for structural parts, since both are about respecting load direction rather than maximising mass.
Attaching the Gear to the Shaft
A gear that spins freely on its shaft has not transmitted anything. This is where printed gearboxes most often disappoint, and it is a creep problem as much as a geometry problem.
- Press fits relax over time. Plastic creeps under sustained stress, so a printed bore sized for a press fit on a smooth shaft will loosen. It may hold on day one and slip on day thirty.
- Use a form-locking drive instead. A D-shaped shaft, a square drive, a metal keyway, or a heat-set threaded insert with a set screw transmits torque through geometry rather than friction.
- Size the bore deliberately. A 0.1 mm bore tolerance relative to the shaft gives a tight fit that needs pushing on; 0.2 mm slides freely. Measure the actual shaft with calipers — nominal diameters are rarely exact.
The same principles govern functional threads and heat-set inserts, and they are worth reading together: both problems are about transferring load into plastic without relying on friction that will not survive.
Material Choice: What the Wear Testing Actually Shows
Material choice for gears is usually argued from generic datasheets. Comparative wear testing on FDM spur gears under identical load and speed gives a more useful answer.
- PETG outlasted PLA and ABS by 22.3% and 37% respectively in service life, with wear depths 30% and 40% lower. Its combination of rigidity and elongation absorbs the sliding contact at the tooth flank rather than abrading.
- PLA was about 12% more durable than ABS, but PLA's glass transition around 55–60°C makes it unsuitable anywhere the gearbox warms up under sustained load.
- Wear concentrates at the tooth root where the driving gear enters mesh — which is exactly where the root fillet needs to be.
- Carbon-fibre-filled nylon is a different tier. Process optimisation studies on CF-PA gears report wear lives in the tens of hours, with layer height, infill density, raster angle and nozzle temperature all statistically significant. The trade-off is abrasion: CF-filled filament will erode a brass nozzle, so it calls for a hardened nozzle.
For most indoor, lightly loaded gearboxes, PETG is the sensible first choice and the easiest to get meshing correctly. Where a gear runs warm, or outdoors, ASA holds dimensional stability to far higher temperatures and adds UV resistance. Where stiffness and wear life both matter, PETG-CF raises the modulus while PETG keeps the toughness underneath. Pure ABS is the weakest performer of the common three for this application, so it is worth choosing deliberately rather than by default.
One caution that applies to every hygroscopic material: wet filament foams as the moisture flashes to steam, giving an inconsistent extrusion and a tooth profile that varies part to part. Gears are dimensionally unforgiving, so dry the spool and print from a dryer cabinet rather than a shelf.
A Practical Build Sequence
- Fix module and pressure angle first. Match the mating gear or the motor pinion exactly. Default to 20°. Never mix.
- Keep tooth count at 17+. Below 13 the tooth form deforms regardless of other settings.
- Add root fillets at 0.3–0.5× module. Do this before anything else — it is the highest return per unit of effort.
- Cut backlash into the tooth flanks using the table above. Offset the flanks inward, do not move the shafts apart.
- Set face width to 8–10× module and add a hub at 2× shaft diameter.
- Model a form-locking shaft interface — D-shaft, keyway, or heat-set insert. Do not rely on a friction press fit.
- Orient flat, 4–6 walls, 0.15–0.20 mm layers. Layer height directly controls how much the staircase thickens the tooth.
- Print pins and a two-gear test rig at the true center distance before the real housing. Twenty minutes of testing beats a twelve-hour print of a gearbox that will not turn.
- Run it in under light load, then check backlash with a feeler gauge. The high points of the layer steps wear in slightly, and the mesh usually improves after a short run-in.
- Check operating temperature after ten minutes. If the housing is noticeably warm, thermal growth is eroding your clearance and the material choice needs revisiting.
When To Stop and Buy a Metal Gear
Printed gears are excellent in their range and poor outside it. It is worth being honest about the boundary. High continuous loads or shock loads, speeds where sliding friction heats the flank faster than it can dissipate, environments above the material's glass transition, and any safety-critical drive should use a metal gear. Additive manufacturing also stops being economical somewhere around a few hundred identical parts, above which injection moulding wins on unit cost.
Inside that range — prototyping, low-to-moderate torque, custom ratios, obsolete replacement parts, and geometries no machinist wants to quote — printed gears work well, and they work predictably once the mesh geometry is right. The gearbox you design is only as good as the fit between its parts, and the same discipline applies as for any functional assembly that has to fit the first time.
When a mechanism needs to move without any assembly at all, print-in-place design extends these ideas to joints that come off the plate already connected — and when a mating surface has to slide rather than roll, the same clearance logic governs printed sliding bearings and friction surfaces. Gears, hinges, and bushings are three faces of one problem: controlling fit in a process that does not naturally hold tolerance.