Humanoid Robot Ball Screw Nuts Machining

• 1
  Key takeaways
• 2
  Why the nut carries the accuracy budget
• 3
  Raceway geometry and lead tolerance
• 4
  Materials, hardening and the distortion trade
• 5
  Surface finish and lubrication retention
• 6
  When a ball screw nut is the wrong choice
• 7
  How to verify a nut before it goes into a robot
• 8
  Step by step: from blank to finished nut
• 9
  Machining routes for ball screw nuts

A ball screw converts motor rotation into linear motion at the joint. The screw and the nut share that job, but they do not share the error equally. The screw is usually ground on a dedicated thread grinder with a stable, known lead. The nut is the part that gets machined in small batches, heat treated, and sometimes split or flanged. Whatever lead error the nut raceway carries is added to the screw error in the assembled joint.

That matters in a humanoid robot because the actuator is not chasing a fixed path like a gantry. A hip or knee joint sees reversing loads, impact from foot strike, and torque spikes during balance recovery. A positioning error of a few microns is small in a milling machine, where a servo can correct it. In a walking robot it changes the phase of the whole gait.

The nut also has to be compact. Space inside a humanoid joint is tight, so the wall between the ball raceway and the outer diameter is often only a few millimeters. Thin walls distort during heat treatment and during clamping. That is where most of the real difficulty in ball screw nuts machining sits, not in the nominal tolerance value on the drawing.

• 1
  Nut raceway leadAdds directly to screw lead error in the assembled joint.
• 2
  Wall thicknessThin sections move during hardening and during clamping.
• 3
  Reversing loadPoints of contact change constantly, so wear must stay even.

The internal raceway is a helical groove with a Gothic arch or circular profile. Ball diameter, groove radius and contact angle are set by the design, and the machining has to reproduce them within a small band. For high-efficiency nuts, lead accuracy is commonly held to ±0.003 mm per 300 mm of travel, with raceway-to-screw matching inside ±0.005 mm.

Contact angle is the parameter that gets overlooked. A two-point contact Gothic profile nominally sits at 45°, but if the groove is milled slightly off-center, the ball rides on one flank. Axial stiffness drops and the nut develops backlash under reversing load even though the measured lead looks correct.

Multi-start threads and ball return channels add features that a conventional lathe cannot reach in one pass. Internal deflectors, crossover channels and flange mounting faces all need to be positioned relative to the same axis. Five-axis contouring lets the tool follow the helix, blend the return channel, and machine the flange face in a single setup.

The flange face perpendicularity to the screw axis is typically held within 0.01 mm. Miss it and the nut tilts in its housing, which shows up as a once-per-revolution ripple in joint position rather than a constant offset. That kind of error is hard to find after assembly.

• 1
  Lead accuracyTypically ±0.003 mm per 300 mm for high-efficiency nuts.
• 2
  Raceway matchHeld within ±0.005 mm to the mating screw.
• 3
  Flange perpendicularityAbout 0.01 mm to the screw axis.
• 4
  Contact angleOff-center grooves cause one-flank loading and backlash.

Most humanoid ball screw nuts are made from medium-carbon alloy steel or through-hardening bearing steel. 4130 and 4140 are common for carburized parts because the core stays tough while the raceway reaches high surface hardness. 440C stainless is used where corrosion resistance matters, though it is more expensive to machine and less forgiving in heat treatment.

Carburizing gives a hard case over a ductile core. Case depth is the number to negotiate, not just surface hardness. Too shallow and the raceway brinells under shock load. Too deep and the thin wall becomes brittle and may crack at the flange radius during assembly. Through-hardening is simpler and gives uniform hardness, but the whole nut becomes less tolerant of impact.

Heat treatment moves the part. A nut with a 4 mm wall can shift several microns in the raceway after quenching and tempering. The usual answer is to leave grinding stock, harden, then finish the raceway and bores in a second setup. That adds a step but keeps the final geometry under control.

For lower-load joints, some designs use nitrided steel or even hard-anodized aluminum nuts. These work in light-duty demonstrator robots. They are a poor fit for knee or ankle actuators where impact loads dominate.

• 1
  Carburized alloy steelHard case, tough core. Best for shock-loaded joints.
• 2
  Through-hardened steelUniform hardness, simpler process, less impact tolerance.
• 3
  440C stainlessCorrosion resistance, higher machining cost.
• 4
  Aluminum or nitridedLight-duty only, not for high-impact joints.

Raceway finish controls two things: friction torque and how long the lubricant film stays in place. A ground raceway in the Ra 0.2–0.8 μm range gives low, predictable torque. A milled raceway at Ra 0.8–1.6 μm is acceptable for many joints if the tool path is smooth and no cutter marks cross the ball path.

Deep cutter marks are worse than a uniformly rougher surface. A single circumferential scratch acts as a stress riser and slowly abrades the balls. Over a few million cycles, preload drops and the joint develops lost motion. That is why in-process inspection matters more on the raceway than on the outer diameter.

Lubrication retention depends on the surface having a slight texture, not a mirror polish. A perfectly polished raceway can shed grease and run dry in a vertical joint. Light bead blasting or a controlled grind pattern holds a thin film. The trade is that a rougher surface raises running torque slightly.

For sealed joints, the nut bore and seal counterbore have to be concentric. A seal that runs off-center leaks grease and lets dust into the recirculation path, which shortens life faster than any raceway finish issue.

• 1
  Ground racewayRa 0.2–0.8 μm, low and predictable torque.
• 2
  Milled racewayRa 0.8–1.6 μm, acceptable with a clean tool path.
• 3
  Isolated deep marksAct as stress risers and wear the balls.
• 4
  Seal counterboreMust stay concentric with the nut bore.

A ball screw nut is not always the right actuator element. It wins when you need high axial stiffness, low friction and long travel in a compact package. It loses when the joint needs very high torque at low speed, or when the motion is mostly rotary.

If the joint travel is under about 20 mm and the load is modest, a planetary roller screw or a high-ratio harmonic drive may package better. Roller screws carry more load in the same envelope but are harder to source in small quantities. Harmonic drives handle torque directly but add compliance that a walking gait may not want.

For very high cycle counts in a clean environment, a well-made ball screw nut is hard to beat. For dusty or poorly sealed joints, the recirculation path becomes a wear trap and life drops sharply. In that case, design the sealing first and pick the nut second.

One more boundary: if the actuator will be back-driven by gravity, low-friction ball screws can allow the joint to sag. Some humanoid designs add a brake or use a higher-lead screw to make back-driving predictable. That is a system decision, not a machining one, but it changes the nut specification.

• 1
  Good fitHigh stiffness, low friction, long travel in tight space.
• 2
  Poor fitVery high torque at low speed, or rotary-only motion.
• 3
  Dusty environmentRecirculation path becomes a wear trap.
• 4
  Back-drivingLow friction can let a gravity-loaded joint sag.

Start with the drawing, not the part. Lead, contact angle, raceway radius and flange perpendicularity should all be toleranced. If the drawing only says the nut is a ball screw nut, there is nothing to inspect against.

On the shop floor, the first check is dimensional: bore, outer diameter, flange face and raceway lead. A lead measuring instrument or a ball track gauge gives the number that matters most. Surface finish is measured on the raceway itself, not on a test coupon.

Then assemble the nut with its mating screw and measure torque over the full travel. Torque that rises in one region points to a local raceway defect or a slightly bent screw. Torque that is high everywhere usually means preload is too tight or the ball size is wrong.

Finally, run a short cycling test if the joint is safety-relevant. Even a few thousand cycles will expose a bad contact angle, a shallow case, or a recirculation channel that is too tight. Catching it before the robot walks saves a rebuild.

• 1
  Drawing firstLead, contact angle, raceway radius, perpendicularity.
• 2
  Measure the racewayLead gauge and finish check on the actual ball track.
• 3
  Torque sweepRising torque in one region points to a local defect.
• 4
  Short cycling testExposes contact angle and case depth problems early.