Planetary reducer: compact coaxial reduction

A planetary reducer packs a lot of reduction into very little volume, and it does it without pushing the output off to one side: power goes in at the center and comes out on the same axis. A sun sits in the middle, three or four planets surround it, and a ring gear toothed on the inside wraps them all. That's the whole machine. The trick is that the load isn't carried by a single pair of teeth but by several contacts working at once, and that the ratio you get doesn't come from running an extremely small pinion against a huge wheel, but from how the turns are shared among three rotating members. Yet that same elegance is what makes it hard to print well: if the planet centers don't land where they should, one of them carries the load for all the rest and a tooth snaps. Everything here is shared geometry, and the rest pay for an error in any one part.

The kinematics: fix one member and the ratio appears

A planetary train isn't a pair of gears with a fixed ratio; it's three coupled members—the sun, the carrier, and the ring—and the reduction depends on which one you hold. The planets don't count as a fourth degree of freedom: they're idler wheels that only carry motion between sun and ring; they don't set the ratio; they make it possible. That's why the whole assembly is described by a single equation, which ties the speeds of the three members together through the tooth ratio between sun and ring. Every possible configuration comes out of that, depending on which member you lock.

The usual case in a reducer is to fix the ring, feed torque in through the sun, and take it out through the carrier. The reduction is then one plus the ratio of ring teeth to sun teeth. A 12-tooth sun against a 36-tooth ring gives you a 4-to-1 reduction in a single stage, coaxial and inside the ring's diameter. Notice what doesn't show up in that arithmetic: the planet tooth count. The planets set the center distance, but they never enter the transmission ratio. They're the bridge, not the lever.

The fact that they don't enter the ratio doesn't mean any number will do. There are two geometric constraints you have to respect or the assembly won't mesh and won't assemble. The first sets the planet size: for the sun and ring to mesh with the same planet, the ring teeth must equal the sun teeth plus twice the planet teeth. In the example, the planet works out to 12 teeth: (36 − 12) / 2. The second is the assembly condition for the planets to come out equally spaced—which is exactly what gives you load sharing: the sum of sun and ring teeth must be divisible by the number of planets. With 12 and 36 and three planets, (12 + 36) / 3 = 16, exact, so they sit at 120°. Change the numbers without checking this and you'll find the third planet doesn't land where it should.

This has a design consequence worth being clear about before you draw anything: if you want more reduction, you grow the ring or shrink the sun—you don't touch the planets to change the ratio. And there's a physical floor: a sun that's too small runs out of healthy teeth and starts to suffer profile interference. So when one stage doesn't reach far enough, you chain two: the output of the first carrier drives the sun of the second, and the reductions multiply. Two stages of four-to-one give sixteen to one; with a bit more per stage you climb to forty or a hundred without growing much in diameter.

Why it shares the load: several contacts at once, all coaxial

The advantage that justifies the added complexity is load sharing. In an ordinary pair of gears, at any instant the torque is carried by one pair of teeth in contact; all the force passes through there. In a planetary with three planets, the sun's torque is split among three simultaneous contacts, and each planet again hands its share to the ring through another contact. The split between planets is one third: each one carries, in theory, a third of the total tangential force. That means a planetary transmits considerably more torque than a single pair of gears of the same diameter, or, for the same torque, it can be smaller. In FDM, where the tooth is plastic and the failure mode is shearing off at the root, that sharing is what makes viable a reducer that would otherwise lose its teeth the moment it saw a real load.

The second benefit is coaxiality. The input (sun) and the output (carrier) share an axis. There's no need to offset the output to one side as in an ordinary gear train, which saves volume and simplifies assembly: everything stacks around one line. It's exactly what you want when the reducer has to fit inside a tube, a wheel hub, or behind a robot's joint. Compact, coaxial, and with torque shared: those three things are why the planetary is the default reducer when space is tight.

Print it flat: clean profiles and teeth in the strong plane

Every wheel in a planetary—sun, planets, and ring—is printed flat, with the axis of rotation vertical, stacked layer on layer along the axis direction. There are two reasons, and both matter. The first is profile: a tooth printed standing up comes out stepped by the layers along its flank, and a stepped flank meshes unevenly, with backlash that varies along the contact. Printed flat, each layer reproduces the full tooth outline at the resolution of the XY plane, which is much finer than the Z axis, and the flank profile comes out clean and continuous.

The second is strength, and it ties in with Layer orientation for motion. The load on a tooth is bending at its root: the tooth works as a short cantilever beam that the mating tooth pushes on by the flank. If you print the wheel flat, that bending runs along the beads lengthwise, in the strong plane of the material, and the root holds. If you print it standing up, the line of maximum tension at the root falls right on a plane between layers—the weak bond of FDM—and the tooth delaminates, snapping off at the base as if it had been cut. A plastic tooth is already the fragile link in the assembly; making it work against interlayer adhesion on top of that is inviting the failure.

The ring deserves extra care. Its teeth point inward, but printed flat that's no problem: each layer is a closed, self-supporting toothed ring. What does matter is that its outer wall has enough body that the meshing hoop stress doesn't split it open. Treat it as a ring loaded from the inside, not as just another wheel, and give it a wall thickness in millimeters scaled to the diameter and torque, not just a perimeter count.

And don't forget the carrier itself as a structural part. If you print it as one piece—which is what you want—it usually carries cantilevered arms or columns between its two plates, and those arms work in bending and torsion when torque comes in. Printed flat, those arms can delaminate exactly like a tooth: the load line falls on the planes between layers. Orient the carrier thinking about where the stress passes, not just about whether it fits on the bed.

Clearances: every contact wants its own gap, and tolerances stack

Here's the part that decides whether the reducer turns or seizes. In a planetary there are two contacts in series per planet: sun against planet, and planet against ring. Each one needs its own backlash—the gap between unloaded flanks—just like any printed gear, and for the same reasons as any FDM fit: the tooth comes out thicker than you drew it because the bead width fattens the flank, and two flanks that barely touch on screen interfere in the part. Tolerances for moving parts works this out, and here it applies twice in a row.

The problem peculiar to the planetary is that these errors stack up in a chain. The distance from the sun's center to a planet's center, plus the distance from that planet to the ring tooth, plus the position error of the carrier hole, all add up. If each link drifts a tenth in the wrong direction, the mesh ends up tight and seizes; if they drift the other way, it ends up loose and the reducer has noticeable angular play at the output—you can turn the sun before the output responds. That's why a planetary isn't calibrated tooth by tooth, but as an assembly: you print, assemble, and check that it turns smoothly through the full revolution with no hard spots and no dead zones. If it seizes, you open the backlash on the two contacts a little and also check that the carrier center distance hasn't come up short.

The most critical dimension of all is the concentricity of the carrier. The carrier is the part that holds the three planet axles and keeps them equidistant from the center. If those three holes don't fall on a perfect circle, well centered on the sun's axis, the three planets don't mesh equally: one sits farther in and binds, another farther out and rattles. That's where load sharing is lost. That's why the carrier wants to be printed as one piece whenever possible, with its axle seats well dimensioned. Watch the planet axle clearance here: the planet has to spin freely, but if you give it too much diametral play, its center can shift more than the tolerance you demand of the carrier itself, and you unbalance the sharing again. It helps for the planet to self-position by the mesh—which is what it tends to do if the two contacts are well dimensioned—and for the axle to bear on a metal pin or bushing rather than on a printed plastic hole, which wears and goes off-center (covered in Embedded hardware: magnets, bearings, and inserts).

When to choose a planetary, and how it fails when you shouldn't

Choose a planetary when you need a medium or high reduction in a compact, coaxial volume: on the order of three to ten to one per stage, where a single pair of gears would come out too big to fit, or too heavily loaded for the plastic tooth to take the torque. If the ratio you're after is small—two to one—and you have room to spare, an ordinary pair of gears is simpler and has fewer things to calibrate; the planetary only earns its complication when space is tight or torque climbs. And if the reduction you need is very large, chain two planetary stages rather than forcing a single one with a tiny sun.

Before you assemble, make sure of two details that aren't kinematics but decide whether the reducer reduces. The fixed ring has to be truly anchored to the chassis, with an anti-rotation feature—keys, screws, a recess—not just held by friction: if the ring can turn, the torque carries it away and the reduction evaporates. And the input sun, if it's long and thin, needs a radial support to keep it aligned; a sun that flexes misaligns its mesh and unbalances the sharing from the very first moment.

It's worth naming the failure modes, because almost all of them come from the same place—shared geometry—and are recognized by their symptom:

• Uneven sharing from a center error. One planet meshes ahead of the others, carries most of the torque, and works as if the other two didn't exist. The symptom is a hard spot that shows up at the same angular position every revolution and, over time, the overloaded tooth coming off while the others look brand new. It's not bad luck: it's an off-center carrier or a planet with teeth a hair too thick.
• Tooth fracture from bending at the root, which in FDM is almost always delamination if the wheel was printed standing up, or pure overload if the sharing failed.
• Binding from stacked tolerances: the reducer is hard to turn or jams in certain positions because the sum of errors closed the backlash at some point in the travel.
• Carrier play: if it's not well guided on its axis, it goes off-center under load and drags the planets with it, turning a sharing that was good at rest into a bad one the moment torque comes in.

All four are prevented by the same things: precise centers, flat printing, backlash calibrated as an assembly, and a stiff, well-supported carrier.

If you're about to print your first planetary, start by settling the orientation of each wheel and the clearance of your gears before worrying about the ratio: Layer orientation for motion tells you why everything goes flat, and Tolerances for moving parts gives you the method for finding the backlash your printer actually needs.