The 1:43 cycloidal joint described here could be turned backward by hand, and a lever check with the input fixed felt rigid, but neither observation quantifies efficiency or proves zero backlash. The decisive measurements are output lost motion under a defined reversing torque and output deflection under a defined load. Size the joint from its actual torque and thermal duty, not from the arm’s payload figure or the stepper’s holding-torque label alone.
Separate backdrivability from backlash and stiffness
Backdrivability means output torque can drive the reducer and rotate its input; it does not mean the reducer has backlash. Backlash is the angular deadband seen when the input direction reverses before the output responds. Stiffness is the output’s elastic movement under load. A reducer may be backdrivable and still have low measured backlash, while a lever test may feel rigid without resolving small clearance or elastic movement.
The described arm uses the cycloidal drive on its fifth joint. The reported reduction is 1:43, with two cycloidal disks made from 4340 steel, a cross-roller bearing, and an 8 mm passage through the drive for wiring or an encoder. The disks are about 70 mm in diameter, 8 mm thick, and 160 g each. These dimensions describe the reported parts, not a generally suitable design envelope.
| Quantity | Reported value or observation | What it tells you |
|---|---|---|
| Reduction | 1:43 | Nominal speed and torque relationship; actual output depends on efficiency and losses. |
| Backlash | No numerical measurement reported | Must be measured over a defined reversal torque range. |
| Backdrivability | Output lever could turn the drive | Shows it can be driven from the output, but does not quantify efficiency. |
| Motor | NEMA 23 stepper, 3 Nm holding torque, for this joint | Holding torque is not a complete running-torque or thermal-duty rating. |
| Arm claim | About 1.1 m reach and at least 10 kg payload | Does not specify load position, joint pose, acceleration, or simultaneous rating. |
Read the load path before judging joint capacity
Payload mass alone does not determine the torque required at a joint. For a static load, the gravitational moment about a joint is the load force multiplied by its perpendicular distance from that joint. The relevant distance changes with arm pose; moving links, acceleration, friction, and external contact add to the joint demand. The stated 1.1 m reach and at least 10 kg payload therefore do not establish the fifth joint’s torque requirement or prove that this reducer can carry the rated payload.
For a first-order static estimate, calculate the moment at each joint from the mass and perpendicular lever arm in the pose being evaluated, then include the mass of every downstream link and tool. For motion, include acceleration and deceleration loads. Compare the resulting joint demand with the reducer, bearing, and motor ratings from their respective data sheets. A 3 Nm stepper holding-torque figure cannot be treated as available torque at speed, and the 1:43 ratio does not make output torque equal to exactly 43 times motor torque: losses reduce output, and the actual efficiency is not reported.
Thermal checks matter when a motor or reducer must sustain torque for long intervals. Use the motor’s torque-versus-speed and thermal data, the drive’s permitted duty, and the application’s motion profile. If the assembly warms during a repeated or sustained test, record motor current, speed, ambient and component temperatures, and duty cycle so the cause can be separated from mechanical binding.
Use reversal tests to quantify backlash and flex
The hand test—fixing the input and turning an aluminum extrusion attached to the output—reported no perceptible backlash or flex. It is a useful screening check for gross looseness, but it provides no angular resolution, applied torque, or repeatability. A long extrusion also magnifies small angular motion at its end, but without a measured force and displacement it cannot yield a backlash value.
Use a repeatable fixture and record input restraint, output angle, and applied torque. Test in both directions and distinguish the deadband near reversal from elastic deflection as torque increases. Report the measurement method and torque range with the result; “zero backlash” is not an engineering measurement unless the instrument resolution and test conditions are stated.
- Secure the reducer housing and restrain the input without adding compliance to the fixture.
- Attach an angular indicator or encoder to the output. Apply a small torque in one direction, reverse it, and record output travel before the load path engages in the opposite direction.
- Repeat at documented torque increments in both directions. Separate reversal deadband from the continuing, load-dependent angular deflection.
- Repeat the sequence after assembly and run-in, and compare results at the same fixture and torque conditions.
Measure output deflection separately when evaluating stiffness. Deflection that grows smoothly with torque indicates elastic compliance in the drive, bearing, housing, shaft, or fixture; a sudden jump or a dead zone around reversal points to clearance or looseness. Verify the fixture independently so it does not masquerade as gearbox flex.
Control pin fit and bearing fits as a tolerance chain
The reported machining order used the tightest tolerance option offered, stated as 0.02 mm. The drawing specified the center bore as +0.02 mm from nominal, with nominal bore size 0.01 mm larger than the bearing size. That description does not fully state the bearing’s actual size limits or the inspection method; verify the intended fit from the drawing’s complete tolerance notation and the bearing dimensional limits before ordering or assembly.
The designer had previously used a negative tolerance such as -0.02 mm for an interference fit, but found a slight clearance approach more suitable for larger bearings; the cycloidal disk bearings were 25 mm outside diameter. Bearings installed without major force, and retaining compound was used to secure them. Treat that as this assembly’s experience, not a universal fit recommendation. Excessive press force can damage a bearing, while too much clearance can allow movement and alter alignment.
Roller-pin housing inside diameter is especially sensitive: an oversized fit increases clearance and backlash, while an overly tight fit raises friction and can prevent proper operation. Six different printed pin housings with differing inside diameters were produced and tested to find the best working fit. Because printer accuracy varies, repeat the fit evaluation for the actual printer, material, process, and mating parts rather than assuming a prior printed dimension will transfer unchanged.
Diagnose friction, lost motion, and control faults separately
| Observation | Likely area to inspect | Discriminating check |
|---|---|---|
| High effort or binding through rotation | Pin-housing clearance, alignment, bearing seating, surface contact | Rotate the unloaded mechanism by hand through a full cycle and look for periodic drag or tight spots. |
| Output moves before the opposite flank engages | Clearance in pin fits, bearing fits, or other interfaces | Measure reversal deadband with controlled torque and an angular indicator. |
| Output motion grows under steadily increasing torque | Elastic deflection in reducer, bearing, mount, shaft, or test fixture | Plot angle against torque and isolate fixture compliance. |
| Motor holds but the joint misses commanded motion | Motor torque at operating speed, driver/control behavior, coupling, or mechanical load | Compare commanded motion with output encoder feedback and inspect motor current, speed, and fault indications. |
| Motor or reducer heats during repeated operation | Excessive sustained torque, friction, or inadequate duty margin | Record current, temperature, speed, and duty cycle; inspect for binding before retuning motion. |
The planned control arrangement uses a control board with an ESP32 to drive each stepper driver, with a Jetson Orin Nano Dev board handling trajectory planning and inverse kinematics. Closed-loop steppers and a second encoder at the joint output were planned. Output feedback can reveal that the motor turns while the joint fails to follow, but it does not by itself identify backlash, flex, or a slipping interface; compare output and motor positions during controlled reversals.
Verify the assembled joint before applying arm loads
- Check bearing seating, disk alignment, pin fit, and free rotation through a full input cycle. Resolve tight spots before motorized testing.
- Measure backlash and stiffness with the joint mounted as it will be used, then record the fixture, torque, angle resolution, and test direction.
- Run the motor across the intended speed and load range while monitoring commanded versus measured output movement, motor current, and temperature.
- Increase load in controlled steps and inspect for rising drag, loosened interfaces, abnormal noise, missed motion, or changing reversal error.
- Recheck backlash and fasteners after testing; compare with the initial measurements to identify settling or wear.
Keep the encoder route through the 8 mm hole clear of moving parts and verify cable movement through the full joint range. The opening is a packaging feature, not a substitute for checking cable bend, strain relief, and interference in the assembled arm.
Stop testing when the load or temperature leaves the rated envelope
Do not continue powered tests if the joint binds, the motor or reducer exceeds its published temperature or duty limits, a bearing shifts, or output motion becomes unstable; these conditions can damage the reducer, bearing, motor, or attached machine. Retrieve the actual component limits from the relevant data sheets rather than substituting the payload claim or holding-torque value. If the fit, bearing selection, motor sizing, or thermal result remains unresolved, stop loading the joint and consult the official support channel for the applicable component manufacturer or the machining provider.
FAQ
Can a backdrivable cycloidal drive still have low backlash?
Yes. Backdrivability describes whether output torque can turn the input; backlash is reversal deadband. Measure output angle under a documented reversing torque to quantify backlash.
Does the 1:43 reduction make the joint output torque 43 times motor torque?
No. The ratio gives the ideal relationship; real output is lower because of losses, and efficiency was not measured for this drive. Use motor torque at the operating speed and the reducer’s verified performance data.
Can I use the hand lever test to claim zero backlash?
No. It can reveal gross looseness, but the reported lever test had no specified torque or angular measurement. Use a restrained-input reversal test with an output angle measurement and report the torque range and resolution.