By Theodore Li 
By Theodore Li, Technical Director, HONPINE. Adapted from the original published on honpine.com.
Harmonic joint modules are the muscles and joints of a humanoid, and they are also the cost centre: 50–60% of total robot cost. They set motion precision and power output, and therefore the robot’s flexibility, load capacity and what tasks it can be trusted with.
The mistake most programmes make is picking one module and repeating it. Different joints want different things — a wrist wants agility and accuracy, a hip wants torque, a neck wants nothing but small and light. This is a joint-by-joint decision.
Why these modules get specified at all
HONPINE’s harmonic joint modules are, in the company’s framing, compact integration champions: within a diameter of just 4 mm they integrate a harmonic reducer, a servo motor, a high-precision encoder and drive electronics.
Concretely, the four properties that matter:
Extreme integration
Traditional joints need a separate reducer, motor and drive, which means bulky wiring and a stack of interfaces. These modules weigh as little as 0.19 kg, which allows lighter, more compact joint designs — decisive in space-constrained areas like the neck and wrist.
Small size, high torque
Take the HPJM-RE30-40-PRO: compact, with reduction ratios of 51:1 or 101:1, peak torque up to 3.3–4.8 N·m, and average load capacity of 1.8–2.4 N·m. That is “big performance in a small package” territory — enough to support arm lifting and stable leg movement under the robot’s own weight.
Full-scenario variants
The same module family ships in three environment-specific versions:
- Brake version — auto-locks on power loss, for posture-holding tasks such as grasping and pausing mid-air
- Low-temperature version — reliable operation at −40 °C, for polar or cold-chain logistics
- Waterproof version — IP67, for dust, moisture and even deep-sea conditions on underwater robots
Integration-friendly design
- Hollow-shaft design — various bore sizes allow internal cable routing, which removes the wire clutter that ruins a slim limb design
- Versatile communication — single-encoder models support CAN/CANopen/EtherCAT (choose one); dual-encoder models support CAN/EtherCAT, except the small 30-40 and 40-52 sizes. One module fits multiple control schemes without swapping hardware
The two parameters that decide the choice
Positioning accuracy
| Configuration | Absolute | Repeatable | Fits |
|---|---|---|---|
| Single-encoder | ±0.05° | ±0.005° | most applications |
| Dual-encoder | better than ±0.03° | ±0.01° | high-precision tasks |
Note the exception: the 30-40 and 40-52 sizes do not support dual-encoder. If a joint needs that accuracy class, the size has to move up.
Environmental suitability
| Version | Rating | Temperature | Use |
|---|---|---|---|
| Standard | IP50 | −10 °C to +50 °C | indoor, derating at high temperature |
| Low-temperature | custom | −40 °C | polar, cold chain |
| Waterproof | IP67 | custom | deep sea, washdown, extreme cold |
Joint-by-joint selection logic
Different joints face vastly different demands. The practical mapping:
| Joint | Primary demand | Module choice |
|---|---|---|
| Neck | small, light, 2–3 DOF | smallest-bore module, single encoder |
| Shoulder | torque and accuracy | mid-size, dual encoder, brake |
| Elbow | torque, compact | mid-size, high ratio (101:1) |
| Wrist | agility, low mass | small module, hollow shaft for cabling |
| Hip | peak torque, shock | largest size, brake version |
| Knee | torque, impact | largest size, brake version |
| Ankle | fast response | small-mid, high bandwidth |
| Finger / hand base | minimum size | micro module or tendon path |
Two rules fall out of this table. First, size and accuracy trade against each other — the smallest modules cannot take dual encoders, so a joint that must be both tiny and ultra-accurate is a design problem, not a catalogue lookup. Second, hollow shaft is a cabling decision as much as a mechanical one: choosing a bore diameter early decides how the whole arm is wired.
The selection process
- Consultation — share load, operating scenario and precision requirement
- Technical discussion — engineers recommend the model based on the duty cycle, not the peak
- Order and delivery — factory direct
- After-sales — debugging and maintenance backed at factory level
The step that gets skipped and shouldn’t is the second one. The failure mode of a joint module is rarely “it does not fit”. It is “it fits, and it overheats at 80% of the duty cycle we actually run”. A duty-cycle conversation up front is what prevents that.
A short checklist before you order
- Reduction ratio and peak torque at your duty cycle, not the catalogue’s
- Encoder configuration the size actually supports
- Brake needed? Then check the power-loss behaviour, not just the holding torque
- Cable path — bore diameter and connector orientation
- Protocol — and whether your drive supports it without an adapter
- IP rating against the actual environment, including washdown
Source: adapted from “How to Select Harmonic Gear Joint Modules for Humanoid Robots?”, originally published by HONPINE at honpine.com. Specifications are HONPINE’s. Tables, joint mapping and checklist by Latest Robot News.
Sarah Xiao