Planetary Gearbox vs Harmonic Drive
vs Cycloidal — Which to Specify When

Rolling-contact involute gears

Sun gear drives multiple planet gears meshing simultaneously with a fixed ring gear. Load is distributed across N planet contacts (N=3–5), producing high torque density in a compact cylindrical envelope. Gear teeth roll and slide in contact — contact stress is Hertzian, proportional to applied load.

Elastic flexspline deformation

An elliptical wave generator cam elastically deforms a thin-wall flexible gear (flexspline) to mesh with a rigid circular spline at two diametrically opposite points. The tooth count difference between flexspline and circular spline produces the speed reduction. Backlash is near-zero by mechanism — no clearance required for deformation-driven mesh.

Eccentric cam + pin-gear mesh

An eccentric input crankshaft drives a cycloidal disc that rolls around the inside of a fixed ring of cylindrical pins. The eccentric motion of the disc minus one pin-pitch per revolution produces the speed reduction. Large contact area (half the pins engaged simultaneously) gives very high torque density and excellent shock tolerance. Output is taken through output pins in the disc.

The harmonic drive earns its cost premium in applications where backlash ≤0.5 arcmin is a genuine functional requirement — not simply a specification selected conservatively. Three Korean application categories justify harmonic drive selection despite the 3–8× cost premium over equivalent planetary units.

The cycloidal reducer’s distinguishing characteristic is its exceptional shock load tolerance — peak torque ratings of 4–6× the rated continuous torque are typical, compared to 2–3× for planetary and 1.5–2× for harmonic drives. This tolerance comes from the large contact area of the cycloidal mechanism: approximately half the output pins engage simultaneously, distributing any shock load across multiple contacts rather than concentrating it on the two mesh points active in a planetary gear.

In Korean industry, cycloidal reducers dominate three application categories where shock tolerance and stiffness are the overriding requirements:

Whether specified as EP-BPG energy-saving for conveyor replacement or EP-AB precision for servo axes, the planetary gearbox does not win on any single specification dimension against its specialised competitors. It wins on the combination of adequate performance across all dimensions simultaneously, combined with price, availability, and zero-maintenance sealed construction that no competing technology matches. In practice, approximately 80% of Korean servo drive applications require neither the sub-0.5 arcmin backlash of harmonic drives nor the 5× shock tolerance of cycloidal reducers — and for those 80%, the planetary is objectively the correct choice.

Efficiency advantage quantified: A Korean packaging line running 200 VFFS machines 21 hours per day, each with a cross-seal jaw servo at 750W nominal input power. At harmonic drive efficiency 80%, the system draws 937W per servo. At planetary efficiency 97%, the same servo draws 773W. Per machine: 164W difference × 21h × 330 days = 1,137 kWh/year. At Korean industrial electricity rates (₩150/kWh): ₩170,550 saved per machine per year. Over 200 machines: ₩34.1 million saved annually. Over a 10-year machine life: ₩341 million — for choosing planetary over harmonic on axes where backlash ≤0.5 arcmin is not actually required.

Zero-maintenance advantage: In Korean three-shift food packaging and logistics operations, maintenance windows are measured in minutes per machine per month. A harmonic drive requiring periodic grease inspection and a cycloidal reducer requiring oil level checks and annual oil changes both consume maintenance labour that a sealed planetary gearbox does not. The zero-maintenance sealed construction of Korea Ever-Power EP series is not a minor convenience feature — for facilities running 300+ machines, it is a significant operational cost advantage.

Harmonic drive catalogues typically specify peak efficiency at rated load and rated speed — conditions where the flexspline hysteresis losses are proportionally small relative to the transmitted power. However, Korean servo applications frequently operate at partial load (30–70% of rated torque) and variable speeds — conditions where harmonic drive efficiency drops significantly below its peak specification.

The efficiency-load characteristic of the three technologies diverges most sharply at partial load. Planetary gearbox efficiency is relatively flat across the load range — at 30% of rated torque, efficiency remains 94–96%. Harmonic drive efficiency at 30% of rated torque drops to 65–75% (the flexspline hysteresis loss is nearly constant in absolute watts regardless of load). Cycloidal efficiency at partial load is moderate — 80–88%.

This partial-load efficiency gap is particularly significant for Korean packaging and assembly machine servo drives that spend substantial time at partial load during acceleration ramps, dwell phases, and light-load product handling. A Korean cobot arm in pick-and-place operation may operate at full rated torque for only 10–20% of its cycle time — spending the remaining 80–90% at partial load. Under this duty cycle, the harmonic drive’s real-world average efficiency is closer to 70–75%, not the catalogue-stated 80–85%.

Otázka
Our Korean robot integrator recommends harmonic drives on all six joints. Is this necessary?

For joints J4–J6 (wrist) where the specification requires ≤0.5 arcmin backlash and the tool centre point accuracy target is ±0.05 mm at reach, harmonic drives are technically justified. For joints J1–J3 (base, shoulder, elbow), the TCP accuracy is dominated by structural compliance and joint stiffness at those large radii, not by gearbox backlash — P0 or P1 planetary gearboxes (≤1–3 arcmin) achieve the same robot TCP accuracy at J1–J3 at significantly lower cost and higher efficiency. A correctly specified 6-axis Korean cobot uses harmonic drives at J4–J6 and planetary gearboxes at J1–J3. If your integrator is specifying harmonic drives at all six joints, ask them to show the calculation connecting J1–J3 gearbox backlash to TCP accuracy — in most cases, the calculation will show that planetary P0 at J1–J3 achieves the same result.

Otázka
Can a Korea Ever-Power EP-AFH planetary gearbox replace a harmonic drive on a Korean cobot joint?

Ten/Tá/To EP-AFH ultra-precision series delivers ≤1 arcmin backlash as standard — approaching the harmonic drive’s ≤0.5 arcmin but not matching it. For J1–J3 joints on a cobot with 400–600 mm reach, EP-AFH is the correct replacement — TCP accuracy at those radii is not degraded by the 0.5 arcmin difference. For J4–J6 wrist joints where sub-0.5 arcmin is a genuine functional requirement, EP-AFH at ≤1 arcmin is marginally outside the specification. Korea Ever-Power recommends confirming the specific accuracy requirement at the joint — if the requirement is ≤1 arcmin at the joint (not at the TCP), EP-AFH is a direct cost-efficient replacement. If ≤0.5 arcmin at the joint is specified, the harmonic drive remains the correct choice.

Otázka
Why do Korean CNC machine tool builders use planetary gearboxes rather than harmonic drives for rotary tables?

Korean 5-axis machining centre rotary tables require ≤1 arcmin positioning accuracy and 2,000–5,000 N·m holding torque against machining forces. Harmonic drives at these torque levels become very expensive (3–8× the cost of equivalent planetary) and their peak torque multiple of 1.5–2× rated is insufficient for the interrupted cutting shock loads from milling titanium and high-hardness steels — peak cutting forces can exceed 2.5× the rated holding torque during toolpath reversals. The planetary gearbox (P0 ≤1 arcmin, peak torque 2–3× rated) meets the accuracy requirement and survives the shock loads at a cost that makes it the universal choice for Korean CNC rotary table drives.

Otázka
For Korean machine builders considering worm reducers as a cost alternative, how does the comparison extend?

Worm gear reducers occupy a different cost tier — typically 0.3–0.6× the cost of an equivalent planetary — but this cost advantage comes with meaningful capability trade-offs: efficiency of 40–70% (vs planetary ≥97%), backlash of 10–30 arcmin (vs planetary ≤1–5 arcmin), and torque output that degrades significantly above 40°C ambient as the worm lubricant viscosity drops. Worm reducers are appropriate for slow conveyor drives and agitator applications where efficiency and backlash are secondary to cost and self-locking capability. For any application requiring closed-loop servo position control, the worm’s 10–30 arcmin backlash effectively prohibits its use on precision axes — the planetary gearbox is the minimum precision standard for servo positioning.