Why Robot Joint Testing Is Now Critical
The global collaborative robot (cobot) market is expanding at over 20% annually. Humanoid robots — the next frontier — are entering pilot production at companies across Asia, Europe, and North America. At the heart of every cobot arm and humanoid limb is the joint module: a compact actuator integrating a servo motor, harmonic drive reducer, torque sensor, encoder, and controller in a single unit.
Joint module performance directly determines robot accuracy, ripetibilità, payload capacity, e sicurezza. A joint with excessive backlash causes positioning error. A joint with high cogging torque produces jerky movement at low speeds. A joint with poor thermal management overheats under sustained load and triggers safety shutdowns on the factory floor.
Testing these properties requires a specialized test system — not a general motor test bench — designed for the unique mechanical and electrical characteristics of integrated joint modules.
Anatomy of a Robot Joint Module
Understanding what to test requires understanding what is inside a typical joint module:
- Servo motor: BLDC or PMSM, typically 100–500 W, high pole count for low cogging
- Harmonic drive reducer: Ratio 50:1 A 160:1, providing high torque density with near-zero backlash
- Output torque sensor: Built-in reaction torque sensor for compliance control (force/torque sensing)
- Dual encoder system: Motor-side encoder (for commutation) + output-side encoder (for position accuracy)
- Integrated controller: Motor drive, position/torque control loop, communication interface (EtherCAT, CAN)
- Gestione termica: Passive fins or active liquid cooling depending on power class
The test system must interface with all of these — mechanically, electrically, and via the communication protocol.
IL 8 Core Test Items for Robot Joint Modules
1. Output Torque Accuracy
The joint module is commanded to produce a series of torque setpoints via its controller interface. A reference torque sensor (external, independently calibrated) measures actual output torque at each commanded value. The deviation between commanded and actual torque defines torque accuracy — typically specified as ±1–3% of full-scale torque for production cobots, ±0.5% for precision models.
Key instrument: External reference torque sensor (precisione >10× better than the joint’s claimed accuracy — typically ±0.01–0.05% FS)
2. Torque Ripple and Cogging Torque
Cogging torque causes velocity ripple at low speeds — a significant problem for applications requiring smooth motion (assembly, surface finishing, medical robotics). The test system rotates the joint output at constant low speed while measuring torque fluctuation with a high-resolution dynamic torque sensor and angle encoder.
Key metrics:
- Cogging torque peak-to-peak amplitude (N·m)
- Cogging torque as % della coppia nominale (target: <0.5–2%)
- Cogging torque frequency (related to motor pole count × speed)
- Harmonic drive transmission error ripple (superimposed on motor cogging)
3. Torsional Stiffness
Torsional stiffness (N·m/rad) determines how much the joint deflects under load — directly affecting robot accuracy under payload. The test system applies a series of known torque loads to the joint output while measuring angular deflection with a high-resolution angle encoder. Stiffness = ΔTorque / ΔAngle.
Stiffness testing also reveals:
- Isteresi (difference in deflection curve between loading and unloading)
- Preload effects on harmonic drive stiffness
- Stiffness variation with torque magnitude (nonlinearity)
Stiffness test graph showing torque vs angular displacement with hysteresis loop highlighted — loading curve and unloading curve
4. Backlash Measurement
Harmonic drives are theoretically zero-backlash, but manufacturing variations, wear, and assembly tolerances introduce measurable backlash. The test system oscillates the joint output between positive and negative torque loads (±10% of rated torque) while measuring angular position with arc-second resolution. Backlash = total angular movement during torque reversal with no net displacement.
Acceptance criteria for cobot joints: <1–3 arc-minutes. For precision robots: <0.5 arc-minutes.
5. Joint Efficiency MAP
Joint efficiency = mechanical output power / electrical input power, measured across the full torque-speed operating space. The test system sweeps through a grid of output torque and output speed points, recording motor electrical input and joint mechanical output simultaneously.
Typical joint module efficiency at rated conditions: 65–80% (lower than standalone servo motors because harmonic drive losses of 15–25% are included). The efficiency MAP identifies the highest-efficiency operating zone — critical for battery-operated mobile robots where joint efficiency directly affects run time.
6. Thermal Performance Under Sustained Load
Robot joints in continuous-duty applications (welding robots, material handling cobots) must sustain rated output without thermal shutdown. The thermal test runs the joint at rated torque and speed for 60–120 minutes, logging:
- Motor winding temperature (via built-in thermistor)
- Harmonic drive housing temperature
- Controller board temperature
- Output torque degradation due to thermal derating
7. Holding Torque (Static Load)
When a cobot joint holds a position under gravity load — the most common static condition — it must do so without drifting or requiring continuous current. The holding torque test applies a static load to the joint output and measures angular drift over time (30–60 minutes). Any drift indicates insufficient position control gain, brake slippage (if a brake is present), or harmonic drive back-drivability issues.
8. Dynamic Response and Bandwidth
The test system applies sinusoidal torque commands at increasing frequency and measures the joint’s response (output torque amplitude and phase lag). The frequency at which output amplitude drops to 70.7% of input amplitude (−3 dB) is the joint’s torque bandwidth — a critical parameter for dynamic robot applications. Typical cobot torque bandwidth: 10–50 Hz.
Test System Configuration
Key Instruments and Why They Were Chosen
| Instrument | Specifiche | Why This Matters for Joint Testing |
|---|---|---|
| External reference torque sensor | Kistler 4503B, dinamico, 0–50 N·m, 0.01 N·m resolution | Independent of joint’s built-in sensor — validates claimed accuracy |
| High-precision angle encoder | ±1 arc-second, 23-bit absolute | Required for stiffness and backlash measurements |
| Load servo actuator | AC servo, 0–500 N·m, quattro quadranti | Applies controlled load in both directions for stiffness and hysteresis |
| Dynamic signal analyzer | 16-channel, 24-bit, 51.2 kS/s | Captures cogging torque harmonics and NVH |
| NI DAQ system | NI CompactDAQ, 16-channel AI/AO | High-speed synchronized data from all sensors |
| Motor performance analyzer | 4-channel, Precisione ±0,2%. | Electrical input power for efficiency calculation |
| EtherCAT / CAN master | Real-time fieldbus interface | Commands joint via native robot communication protocol |
| Vertical mounting fixture | Rigid, gravity-compensated | Eliminates rotor weight effect on torque measurements |
Vertical vs. Horizontal Mounting
Joint modules are typically tested in a vertical orientation (output shaft pointing up or down) to match their installed orientation in a robot arm and to minimize the effect of rotor weight on cogging torque measurements. The test fixture includes counterbalance weights or a gravity compensation mechanism to isolate the joint’s own torque generation from static gravitational effects.
Testing Humanoid Robot Joints: Emerging Requirements
Humanoid robot joint modules introduce additional test requirements beyond standard cobot joints:
- Multi-axis loading: Hip and shoulder joints experience combined axial, radiale, and torque loads simultaneously — requiring multi-axis load fixtures
- Impact resistance: Walking robots experience impact loads at heel strike — the joint must survive repeated impulsive loads 5–10× rated torque for <10 SM
- Back-driveability: Joints must be back-drivable for safe human interaction — the test system measures back-drive torque (torque required to rotate the joint output with motor unpowered)
- Fatigue life: 10 million cycle endurance testing at rated load for key structural joints
- Thermal cycling: Repeated power-on / power-off cycles simulating daily operation patterns
Domande frequenti
What is the difference between a robot joint test system and a servo motor test bench?
A servo motor test bench tests the motor alone — at the motor shaft. A robot joint module test system tests the complete integrated unit including motor, harmonic drive reducer, output torque sensor, encoders, and controller — at the joint output interface. The joint test system must handle very low output speeds (1–100 rpm after 100:1 reduction), very high output torques (50–500 N·m), and interface with robot communication protocols (EtherCAT, CAN). These requirements are very different from high-speed, low-torque servo motor testing.
How is harmonic drive efficiency measured separately from motor efficiency?
By installing torque sensors at both the motor shaft (input to harmonic drive) and the joint output shaft simultaneously. Motor shaft torque × motor shaft speed = mechanical input power to harmonic drive. Output torque × output speed = mechanical output power from harmonic drive. Harmonic drive efficiency = output/input. This dual-sensor measurement is a standard feature of the EconoTest joint test system.
What communication protocols does the test system support?
The test system includes interfaces for EtherCAT (dominant in industrial cobots — Universal Robots, FANUC, KUKA), CAN 2.0A/B (common in Chinese cobot brands and e-bike controllers), and RS-485 Modbus (legacy and low-cost actuators). Custom protocol adapters can be developed for proprietary interfaces.
Can the test system test joint modules designed for humanoid robots?
SÌ. The test fixture design accommodates joint modules from 0.5 N·m to 500 N·m output torque. For humanoid robot applications, we configure the system with impact load capability (pneumatic actuator for impulse testing), multi-axis load application, and back-drive torque measurement as additional modules.
What does a typical joint module validation program include?
A full design validation program for a new joint module typically includes: 3 days for static tests (stiffness, backlash, holding torque, coppia di cogging); 2 days for dynamic tests (efficiency MAP, larghezza di banda, precisione della coppia); 2 days for thermal characterization; 5–10 days for accelerated endurance testing. Production acceptance testing runs 15–30 minutes per joint.
Conclusione
Robot joint module testing is one of the most technically demanding applications for motor and actuator test benches — combining ultra-high torque accuracy, arc-second angle resolution, multi-protocol communication interfaces, and dynamic signal analysis in a single integrated system.
As cobot and humanoid robot production scales from hundreds to tens of thousands of units, the demand for reliable, automated joint test systems is growing rapidly. EconoTest’s joint module test system covers modules from 1 N·m to 500 N·m output torque, with full EtherCAT/CAN interface and automated multi-test sequence execution.
→ Contact our robotics testing team to discuss your joint module test requirements.
