Planetary Gearbox for Industrial Robots
and Collaborative Robot Joint Drives

A servo axis on a CNC machine tool reverses direction a few thousand times per production shift. A robot joint performing a welding or pick-and-place cycle reverses direction millions of times per year. At 60 cycles per minute for a Korean automotive welding robot running three shifts, the shoulder joint executes over 95 million direction reversals per year. Every one of those reversals is a tooth-flank stress event that accumulates toward the backlash growth that ends the gearbox’s useful life.

The four simultaneous requirements that define robot joint gearbox selection are unique in their combination: sub-arcminute backlash for TCP positioning accuracy, compact body diameter to fit within the arm link cross-section, minimum weight to maximise the arm’s usable payload, and a service life measured in tens of millions of reversals rather than the thousands typical of other servo applications. No industrial application imposes all four constraints simultaneously with the same severity as a robot joint drive.

The Korean collaborative robot market has added a fifth constraint: overall arm compactness. Korean electronics and automotive assembly cobots are designed for deployment in existing assembly stations originally dimensioned for human operators — the robot arm must fit in the same physical envelope. Every millimetre of arm cross-section savings at each joint reduces the total arm width by that amount, affecting whether the robot fits in its assigned station. This has made the compact body diameter of Korea Ever-Power’s EP-ADS series and the right-angle J1 configuration with EP-ABR directly relevant to Korean cobot OEM design decisions.

The Tool Centre Point (TCP) is the end-effector tip — the physical location in space where the robot delivers its work. TCP positioning accuracy is the robot’s primary performance specification, and it is the aggregate result of backlash and positioning errors at every joint in the kinematic chain. Understanding how individual joint backlash values combine at the TCP is essential for specifying gearbox grades correctly: over-specifying wastes cost; under-specifying produces a robot that fails its accuracy specification from day one.

The basic relationship: each joint’s backlash creates an angular uncertainty at that joint. This angular uncertainty propagates through all subsequent links to produce a linear position uncertainty at the TCP. The contribution of each joint to TCP error depends on the distance from that joint to the TCP (the effective lever arm) and the joint’s angular backlash.

Multi-joint combination (RSS method): When all six joints each contribute 1 arcmin P0 backlash, the worst-case linear combination would be 6 × 0.291 mm = 1.75 mm — but joints do not all act in the same direction simultaneously. The more accurate estimate uses Root Sum of Squares (RSS), assuming independent random joint errors:

Korean automotive welding robot target: ±0.1 mm TCP repeatability. This means the backlash contribution alone at RSS must stay well below ±0.1 mm — achievable only with P0 ≤1 arcmin on all joints J1 through J4, and with short lever-arm wrist joints J5–J6 contributing minimally even with slightly higher backlash.

Per-Joint TCP Contribution — 10 kg Cobot at 1 m Reach

L_eff = distance from joint to TCP. RSS = Root Sum of Squares combination. Worst-case linear sum = 1.24 mm. Actual repeatability better due to systematic error cancellation.

The elbow joint J3 is where Korean cobot OEMs most frequently encounter the body diameter constraint. The forearm link must accommodate the gearbox body, the motor, the motor encoder cable routing, and the structural shell — within a total envelope often set by the robot’s ISO 9283 arm diameter specification. Every millimetre of gearbox body diameter saved at J3 directly reduces the forearm cross-section, allowing the arm to reach deeper into jigs and fixtures.

The EP-ADS compact round flange series addresses this with a shorter body length than the standard EP-AD series at the same frame diameter — reducing the axial depth the gearbox occupies inside the forearm link. The round flange centres on the link bore, matching most Korean cobot J3 housing designs without a transition adapter. The non-standard ratios available in the ADS series — 16, 21, 31, 61, and 91 — solve a specific Korean cobot design problem that standard series ratios cannot.

The non-standard ratio problem: A Korean cobot J3 elbow servo motor running at 3,000 rpm must produce exactly 48.9 rpm at the output to achieve the designed joint speed profile without a VFD. The closest standard ratio is 60 (producing 50 rpm — close, but not exact) or 70 (42.9 rpm — too slow). The ADS series i=61 produces exactly 49.2 rpm — a 0.6% error, within the allowable variation for the motion profile. Without this non-standard ratio, the cobot OEM must either tolerate the speed error, add a VFD (cost and component count penalty), or redesign the joint geometry.

The waist joint J1 rotates the entire upper arm assembly in the horizontal plane. The conventional design places the servo motor vertically inside the robot base, with an inline gearbox driving the waist axis coaxially. This arrangement works mechanically but produces a tall, heavy base — the motor height plus gearbox length plus output bearing assembly stacks vertically, setting the minimum base height. For Korean cobots designed to mount on small-footprint pedestals or table-top stands in automotive assembly cells, this height is a competitive disadvantage.

The right-angle layout using EP-ABR060 P0 at i=80 repositions the motor horizontally within the base structure. The motor lies flat inside the base footprint rather than extending vertically above it. The right-angle gearbox changes direction 90° to drive the waist axis vertical output shaft. This configuration typically saves 40–50 mm of base height compared to the vertical motor inline layout — equivalent to a full standard motor frame length.

Korean cobot OEM case: 60 robots delivered with this configuration, zero joint rework incidents. P0 ≤1 arcmin confirmed at delivery on all 60 units. The base height reduction allowed the robot to fit on a standard 300 mm table-top pedestal rather than a custom 400 mm machined column — a procurement and production line flexibility benefit beyond the height saving alone.

The wrist joints J5 (wrist rotation) and J6 (tool flange) operate at low torques (5–30 N·m) and require the smallest possible body diameter to keep the wrist assembly within the arm’s distal envelope specification. For a Korean 10 kg cobot with a 70 mm wrist link diameter, the gearbox body can occupy no more than 40–45 mm of that 70 mm — leaving room for the motor, structural shell, and cable routing.

The Korea Ever-Power Economic Line PN II series covers J6 with body diameters starting at 17 mm (PN II 017) — the smallest planetary gearbox in the Korea Ever-Power catalogue. The 6–8 arcmin backlash of the PN II is acceptable for J6 tool flange rotation, because the TCP contribution from J6 backlash at a 60 mm lever arm is only 0.1 mm even at 6 arcmin — a negligible addition to the RSS error budget.

For J5 wrist bend where slightly higher precision is needed — Korean cobots performing screw-fastening or precision insertion at the wrist — the EP-ADS 047 (47 mm body) at P0 or P1 provides sub-3-arcmin backlash in a body small enough for the forearm wrist link. The ADS non-standard ratio availability (i=21) also helps match wrist speed profiles for precision assembly operations.

Backlash specification attracts most of the attention in robot gearbox selection — but inertia mismatch is responsible for more servo oscillation problems in Korean robot OEM commissioning than incorrect backlash grade. A robot axis with correct backlash but poor inertia ratio produces a servo loop that hunts, overshoots, and requires detuned gains — directly reducing the robot’s path accuracy and cycle speed.

The inertia ratio at a robot joint is: J_ratio = J_load_reflected / J_motor, where J_load_reflected = J_load / i². The gear ratio reduces the load inertia seen by the motor by the square of the ratio — which is why robot joint ratios are typically in the range i=20–100, even when the required speed reduction could be achieved with lower ratios. The high ratio is chosen primarily for inertia reduction, not speed.

This is why i=80 is the standard ratio for the J2 shoulder in a Korean 10 kg cobot: the inertia ratio drops to less than 0.002:1 — the motor is overwhelmingly dominant and the servo loop can be tuned aggressively for fast cycle times. Reducing the ratio to i=20 to use a single-stage unit would increase the inertia ratio 45-fold, requiring much softer servo gains and a slower cycle time.

Inertia Ratio Impact on Robot Servo Performance

Precision planetary gearboxes in robot joints do not fail suddenly. Backlash grows gradually over tens of millions of direction-reversal cycles as gear tooth flanks wear. The wear rate depends on the applied torque, the lubrication condition, and the impact severity at each reversal — aggressive deceleration followed immediately by acceleration in the opposite direction produces higher tooth flank contact stresses than smooth trapezoidal velocity profiles.

Korea Ever-Power P0 precision series are designed for ≤1 arcmin at delivery and a service target of ≤2 arcmin after 20,000 operating hours — approximately 8 years at standard three-shift Korean automotive production (2,500 operating hours/year). When backlash grows beyond 2× the original delivery specification, practical positioning accuracy is measurably degraded and joint replacement is warranted.

Q
Can a standard planetary gearbox be used for all six robot joints, or are some joints better served by harmonic or cycloidal drives?

Planetary gearboxes are viable for all six joints in a 6-axis robot, and Korean collaborative robot OEMs standardising on a single gearbox technology across the arm do so with planetary. Harmonic drives offer even lower backlash (sub-0.5 arcmin) and very high torque density in a thin profile — making them common in Japanese robot designs for J1–J3. Cycloidal drives offer extreme shock load resistance. Planetary gearboxes sit between these alternatives in cost and performance, and for Korean 10 kg cobots operating in assembly environments with low shock loads, planetary provides the backlash precision and cycle life required at a lower unit cost than harmonic. The choice between technologies depends on the robot’s payload class, shock load profile, and target unit cost — not on a universal technology preference.

Q
What is the difference between EP-AB and EP-ADS for robot arm joints?

Both series offer P0 ≤1 arcmin in the same frame diameter range. The key differences are: EP-ADS has a shorter body length (compact housing) than EP-AB at the same body diameter, reducing the axial depth inside the forearm link. EP-ADS uses a round flange for bore-centred mounting, while EP-AB uses a square flange for flat bolt-plate mounting. EP-ADS offers the non-standard ratios 16, 21, 31, 61, and 91 that EP-AB does not. Choose EP-ADS for J3 elbow (bore mount, compact, non-standard ratio) and EP-AB for J2 shoulder (square flange, standard mounting, widest ratio selection).

Q
How long do Korea Ever-Power robot joint gearboxes last in 3-shift automotive welding operation?

EP-AB P0 series is designed to maintain backlash ≤2 arcmin (2× the original ≤1 arcmin delivery spec) for 20,000 operating hours. At 2,500 operating hours per year for a three-shift Korean automotive welding robot, this is approximately 8 years of joint life. In practice, J1 and J2 shoulder joints — which carry the highest loads and the most reversals — tend to reach the replacement threshold first. J4–J6 wrist joints typically outlast the arm structure. Korean OEMs performing the annual backlash measurement protocol described in Module 8 have found that J1/J2 replacement at 15,000–18,000 hours is more common than reaching the theoretical 20,000-hour target, due to the combination of high reversal frequency and welding spatter contamination of the base environment accelerating seal wear.

Q
My robot tool-change system uses a vertical tool rack. Can I use planetary gearboxes for the rack-drive axis?

A vertical tool rack drive using an EP planetary gearbox will position tools correctly during powered operation — but when power is removed, the planetary gearbox is back-drivable and the vertical rack load will attempt to back-drive the gearbox, causing the tool rack to descend under gravity. For vertical tool racks where position must be maintained on motor power-off (during tool change, emergency stop, or maintenance), a worm reducer stage downstream of the planetary gearbox provides passive gravity load holding — the planetary delivers the speed reduction and positioning precision for the driven motion, and the worm stage contributes self-locking when power is off. Alternatively, an electromagnetic brake on the servo motor holds the axis on power-off, but requires electrical power to engage (fail-open, not fail-locked).