Epicyclic Gearing

Why epicyclic is different from every other gear train

A pair of meshing spur gears does one job: it trades torque for speed in one fixed direction, at one fixed ratio, with one shaft in and one shaft out. Stack a few pairs and you get a gearbox — but the shafts still poke out the sides, the ratio is still fixed, and the volume is still proportional to torque.

Epicyclic gearing breaks all three of those assumptions at once. Three elements — the central sun gear, the orbiting planets mounted on a carrier, and the surrounding internal-tooth ring — share a single axis. Input and output sit on the same line. Pick any one of the three to hold stationary and the remaining two have a fixed speed ratio. So a single physical gearset has three possible gear ratios depending on which element is grounded. Stack two or three sets and apply a brake to a different element on each, and you have something that looks like one gearbox from the outside but behaves like eight.

That property — multiple ratios from one chunk of hardware, on one axis, with torque shared across three or more planets — is why epicyclic geometry is the dominant architecture wherever ratio flexibility, axial packaging, or torque-per-kilogram matters more than absolute simplicity. Automatic transmissions live and die on it. Geared turbofans, wind turbines, helicopter mains, harmonic-drive robot wrists, and the Toyota Prius power-split planetary all use it for different combinations of those reasons.

The kinematics — Willis equation in one line

Every operating mode of a single epicyclic set is captured by one identity. With ω_s, ω_c, ω_r the angular velocities of sun, carrier, and ring, and Z_s, Z_r the tooth counts of sun and ring:

(ω_s − ω_c) / (ω_r − ω_c) = − Z_r / Z_s

The minus sign is because sun and ring rotate in opposite directions relative to the carrier when one of them is held stationary. With twelve teeth on the sun and forty-eight on the ring (Z_r/Z_s = 4), the three canonical modes drop straight out:

Mode	Held	Input	Output	Ratio	Worked example
Reduction	Ring (ω_r = 0)	Sun	Carrier	1 + Z_r/Z_s	1 + 4 = 5:1, same direction
Overdrive	Ring (ω_r = 0)	Carrier	Sun	1 / (1 + Z_r/Z_s)	0.20:1 (×5 speed), same direction
Reversal	Carrier (ω_c = 0)	Sun	Ring	−Z_r/Z_s	−4:1 (opposite direction)
Modest reduction	Sun (ω_s = 0)	Ring	Carrier	1 + Z_s/Z_r	1 + 0.25 = 1.25:1
Coupled (1:1)	None — lock any two together	Any	Any	1:1	Direct drive — the whole set rotates as one
Neutral	Nothing engaged	—	—	—	Output freewheels regardless of input

That table is the entire gear-ratio chart of a torque-converter automatic transmission, expressed five different ways with the same three components. The control valve body's job is to apply hydraulic pressure to the right band or clutch pack so that exactly one of those modes is active at a time.

Worked example — a 5:1 stage in a GTF reduction gearbox

The Pratt & Whitney PW1000G geared turbofan was the first commercial high-bypass engine to put a single epicyclic stage between the low-pressure turbine and the fan. Let's size the geometry of a representative stage at the headline 3:1 ratio.

Target ratio:  N_in / N_out = 3.0
Hold:          ring
Input:         sun (LPT shaft)
Output:        carrier (fan shaft)
Willis:        1 + Z_r / Z_s = 3.0   →   Z_r = 2 × Z_s

Choose:        Z_s = 36,  Z_r = 72
Planet count:  N_p = 5  (load-sharing across 5 pinions)
Planet teeth:  Z_p = (Z_r − Z_s) / 2 = 18

Module (metric tooth pitch): m = 8 mm
Pitch diameters:
  D_s = m·Z_s = 288 mm
  D_p = m·Z_p = 144 mm
  D_r = m·Z_r = 576 mm

Assembly constraint:  (Z_s + Z_r) / N_p must be integer
  (36 + 72) / 5 = 21.6   ✗   — won't equally space 5 planets
  Re-tooth:  Z_s = 35, Z_r = 70 (ratio still 3.0:1)
  (35 + 70) / 5 = 21      ✓

That re-tooth is exactly the kind of constraint the engineer must solve before any cutting starts: it is geometrically impossible to assemble five equally spaced planets between a 36-tooth sun and 72-tooth ring without re-cutting at least one element. The Willis equation gives the ratio; the assembly constraint (Z_s + Z_r) ≡ 0 mod N_p tells you whether the design can physically be built.

At 10,000 rpm LPT input and 30 MW continuous, each of the five planets carries 6 MW of mesh power at the sun and another 6 MW at the ring. The lubrication system pumps ~200 l/min of synthetic oil through the planet bearings; thermal management is the dominant failure mode, not gear-tooth wear.

Compound layouts — Simpson, Ravigneaux, Lepelletier

Stacking sets in series multiplies the ratios available. The three named layouts behind almost every passenger-car automatic transmission are:

• Simpson. Two simple planetary sets sharing the sun gear; one input shaft, one output, two clutches, two brakes. Gives three forward ratios plus reverse. Dominant on 1960–1990 era three-speed automatics (Chrysler TorqueFlite 727, GM Turbo-Hydramatic 350). Cheap, robust, low-efficiency by modern standards.
• Ravigneaux. One compound set with two suns of different sizes, two sets of planets (short and long) on a shared carrier, and one ring. Gives four forward ratios from a single physical assembly — much more compact than Simpson. Standard on GM 4L60E, Ford AODE, Mercedes 722.4. Hugely common 1985–2005.
• Lepelletier. One simple set in series with one Ravigneaux set. Five rotating elements, five clutches and brakes, six to eight forward ratios. ZF's 6HP (six-speed) and 8HP (eight-speed) families, GM-Ford 10R80 / 10L80 (ten-speed). The dominant architecture from ~2005 onwards.

A modern 10-speed Lepelletier transmission has four planetary sets, six clutches, and a clutch-engagement chart with ten rows — but the gears themselves don't move, change diameter, or shift positions. All ten ratios come from the same fixed gear geometry; what changes is which two clutches are engaged at any moment.

Variants beyond simple spur-tooth planetaries

• Helical-tooth planetaries. Replace straight teeth with helical to spread load over multiple teeth simultaneously. Quieter (typical 3–5 dB lower noise floor), smoother torque, but introduces an axial thrust load that the bearings must absorb. Standard on premium passenger-car automatics and helicopter mains.
• Double-helical (herringbone) planetaries. Two helical sets of opposite hand on the same blank cancel the axial thrust internally. Used where bearing axial-load capacity is limited — marine propulsion gearboxes, large industrial drives.
• Compound planetary (Wolfrom drive). Two planet pinions on a shared shaft mesh with two different ring gears stacked along the axis. Ratios of 50:1 to 200:1 in a single stage become practical — used in robot wrist drives where ratio density is at a premium.
• Bevel-gear epicyclic (differential). Pinion shafts and side gears are bevel-cut to allow the carrier axis to be perpendicular to the output shafts rather than coaxial. Every automotive open differential is a bevel epicyclic.
• Harmonic drive (strain-wave gearing). Not strictly a planetary, but kinematically related: a flexible cup with external teeth meshes with a rigid ring two teeth larger, with a wave generator deforming the cup. Gives ratios of 30:1 to 320:1 in one stage with ~1 arc-minute backlash. Standard on robot joint drives.
• Cycloidal drive. A cycloidal disc with one fewer lobe than the housing pin count, driven by an eccentric input. Like the Wolfrom drive, achieves 50:1+ in one stage. RV Nabtesco units dominate industrial-robot wrist actuators.

Where epicyclic gearing actually shows up

• Automatic transmissions. Every torque-converter automatic from the 1940s ZF Hydramatic onward, every modern 6 to 10-speed planetary automatic. The CVT and DCT alternatives have not displaced it where pure refinement matters.
• Geared turbofans (GTF). Pratt & Whitney PW1000G family (PW1100G on A320neo, PW1500G on A220) — single epicyclic stage, 3:1 ratio, 30,000 kW. Rolls-Royce UltraFan (planned 2030+) takes the architecture to even larger fan diameters with the same kinematic principle.
• Wind-turbine main gearboxes. Three planetary stages in series step rotor speed (12–18 rpm at 4 MW) up to generator speed (1,000–1,800 rpm). Ratios per stage typically 4:1 to 6:1, total ratio 60:1 to 100:1. The Vestas V164 8 MW unit has a ~75 ton gearbox; the failure rate of these gearboxes is the single largest reliability problem in modern wind energy.
• Toyota Prius hybrid power-split. One simple planetary divides engine torque between a generator (MG1, on the sun) and the drive output (ring), with the engine on the carrier. By varying MG1 speed the gearbox behaves as a continuously variable transmission with no belts, pulleys, or hydraulic clutches.
• Helicopter main rotor gearboxes. Two or three planetary stages step the engine output (turboshaft at 20,000+ rpm) down to rotor speed (200–350 rpm) at multi-MW. The CH-53K Sikorsky uses a 7.5:1 epicyclic main reduction stage.
• Bicycle internally geared hubs. Shimano Nexus 7 and 8-speed, SRAM iMotion, Rohloff Speedhub 14 — all use stacked planetaries inside the rear-wheel hub instead of an external derailleur. The Rohloff achieves 14 ratios from four planetary sets, comparable to a 14-speed derailleur cassette but enclosed and weatherproof.
• Robot servo drives. Harmonic Drive HD-series, Nabtesco RV-series, Wittenstein cyber-planetaries — backlash under 1 arc-minute, torque density above 1 kNm/kg. The default architecture inside every industrial robot wrist axis.

Failure modes — where epicyclic drives actually break

• Planet bearing wear. Each planet pinion runs on a needle or journal bearing that sees both rotation and orbital acceleration. In wind-turbine main gearboxes the planet bearings are the single most common cause of premature failure, typically failing at 5–10 years versus a 20-year design life. Cure: oil-jet lubrication directly into the bearing race, larger bearing diameter, transition from needle to spherical roller.
• Carrier deflection. Under heavy load the planet carrier flexes asymmetrically and the planets no longer share load equally; one planet starts carrying most of the torque, runs hot, and pits. Cure: heavier carrier, integrated stiffening webs, FEM-optimised carrier geometry.
• Ring-gear bore fatigue. Internal teeth see fully reversed bending stress every revolution. Industrial-grade ring gears are case-hardened to ~60 HRC and shot-peened; consumer-grade transmissions get away with through-hardening at 45 HRC. Failure shows as pitting then tooth-root cracking.
• Sun-gear tooth scoring. Highest pitch-line velocity in the stage; first to suffer from oil-film breakdown if lubrication fails. Synthetic ester base stocks with antiwear additive packages are standard in automotive and aerospace applications.
• Planet phasing error. If the planets are not equally spaced (assembly error, or designed unequal spacing for noise reasons), torque distributes unevenly and the gearbox runs noisier. Modern designs frequently use deliberate unequal spacing — e.g. 0°, 120.5°, 239°, 358° — to spread mesh-frequency excitation across multiple peaks rather than one big spike.
• Lubricant contamination. Wear particles from any one bearing recirculate through the oil and accelerate wear in every other contact. Spectrographic oil analysis is standard practice on wind-turbine and helicopter gearboxes; trending elemental concentrations of iron, copper, and chromium predicts failures months in advance.

How epicyclic compares to alternative gear architectures

Property	Epicyclic	Parallel-shaft (spur)	Worm	Harmonic drive
Ratio per stage	3:1 to 8:1	2:1 to 5:1	10:1 to 100:1	30:1 to 320:1
Efficiency	96–98 %	97–99 %	40–90 %	70–85 %
Input / output axes	Coaxial	Offset	90° offset	Coaxial
Torque density (Nm/kg)	High	Medium	Low	Very high
Backlash	5–10 arc-min	10–30 arc-min	30–60 arc-min	0.5–3 arc-min
Backdriveable	Yes	Yes	No (self-locking)	Yes
Typical use	Transmissions, GTFs, wind	Industrial reducers, lathes	Conveyor drives, hoists	Robot joints

Common pitfalls when designing epicyclic systems

• Forgetting the assembly constraint. (Z_s + Z_r) must be divisible by the planet count, or the gears physically cannot be assembled equally spaced. Catch this in concept, not in the test cell.
• Designing for nominal load only. Planet-bearing load varies cyclically with carrier rotation. Peak bearing load is 30–50 percent higher than mean. Bearings sized to the mean fail prematurely.
• Ignoring planet-pin manufacturing tolerance. Pin-position error makes one planet carry the load while the others coast. Use floating pins, flexible carriers, or load-equalising mechanisms in high-torque applications.
• Skipping the dynamic analysis. Mesh frequency is the rotational speed times the number of teeth on the mesh — usually 2–10 kHz at full power. Excite a casing resonance there and the gearbox screams. NVH analysis is mandatory above ~50 kW.
• Treating the carrier as a rigid body. Real carriers flex. The deflection redistributes load between planets. Modern wind gearboxes use FEM-optimized hollow carriers that intentionally distribute deflection so that all planets share load even under peak torque.