Mechanical
The Torque Limiter: How a Ball Detent Saves a Drivetrain
The Torque Limiter is a mechanical circuit breaker for rotating shafts. In a ball-detent (ball-and-spring) overload clutch, a handful of hardened steel balls sit in matching pockets and are clamped there by a stiff spring stack. Below a set torque they behave like solid shear keys and pass power rigidly, with no slip; the instant demand crosses the threshold — a machine jam, a servo crash, a seized bearing — the balls snap out of their pockets in a few milliseconds and the drive lets go, so the gearbox, shaft and ballscrew downstream never feel the overload. It is a threshold switch, not a governor, and the physics of that snap is what makes it fast, sharp and resettable.- Trip relationT ≈ r·F_a·tan(α+φ)
- Release time~1–5 ms
- Trip repeatability±5–10%
- Ball hardness58–64 HRC (100Cr6)
- Torque range~0.5–4000 N·m
- Preload sourceBelleville disc-spring stack (kN)
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
Balls held in pockets by a spring
Strip a ball-detent limiter and you find two coaxial members — a hub keyed to the shaft and a driven plate (often carrying the sprocket, pulley or gear). Between them sits a ring of n hardened balls, each resting in a conical or hemispherical detent pocket. A stack of Belleville (disc) springs presses an axial plate against the balls with a large, adjustable preload F_a, seating them like ball bearings squeezed into dimples.
Torque flows through the balls as if they were shear keys: the driving member tries to rotate, each ball is trapped between the wall of its pocket and the spring plate, and it transmits a tangential force to the driven member. Nothing slips, there is no wear, and the coupling is torsionally stiff — critical for positioning drives. The balls, seats and plate are through-hardened bearing steel (100Cr6/52100, 58–64 HRC) precisely because the contact patches run at Hertzian pressures of order 1–3 GPa. This is a standard machine element, refined by German coupling houses (Chr. Mayr's EAS line, R+W, Ringspann) since the mid-20th century.
The trip relation: T ≈ r·F_a·tan(α+φ)
The physics is a wedge on a ramp. Under torque T, each ball carries a tangential force F_t = T / (n·r), where r is the pitch radius of the ball circle. To escape, the ball must climb the pocket wall against the spring; the pocket geometry presents an effective ramp angle α, and lubricated steel-on-steel contributes a friction angle φ = arctan μ (μ ≈ 0.08–0.15). The classic incline-with-friction result gives the release condition per ball as F_t = (F_a/n)·tan(α+φ). Summed over all balls:
T_trip ≈ r · F_a · tan(α + φ)
Two things fall out of this. First, the trip torque is set almost entirely by the spring preload and the ramp geometry — turning the adjustment nut compresses the disc stack and slides the setpoint. Second, and counter-intuitively, the ball count n cancels: for a fixed total spring force, adding balls does not raise the trip torque. More balls simply share the same load, lowering per-ball contact stress and smoothing engagement.
Why release is a millisecond snap-through
The value of the device is not that it trips at a number — it is that it trips abruptly. Plot the holding torque against how far the ball has climbed. As torque rises the ball rides partway up its pocket wall and resistance grows, reaching a maximum right as the ball nears the pocket rim. Past that crest the geometry flips: the wall falls away, the resisting torque collapses, and the system has negative stiffness. There is no stable half-slipped equilibrium — it is all-or-nothing.
So the moment demand nudges past T_trip, the ball accelerates out of its seat, driving the spring plate axially back by a few millimetres. The two halves decouple in roughly 1–5 ms, faster than the shaft can wind up enough angle to overstress the gearteeth or ballscrew downstream. That axial lurch of the plate is also what usually throws a limit switch to e-stop the motor. Compare a sensor-plus-brake electronic system at tens of milliseconds: the ball detent wins because release is a mechanical instability, not a measurement.
Sizing one: a worked example
Say a servo drives a ballscrew: 20 N·m continuous, 40 N·m peak, and we want to protect the gearbox from a table crash. Set the limiter to trip at T_trip = 60 N·m — about 1.5× peak, so normal duty never trips it. Choose a ball circle of radius r = 25 mm, six balls, and an effective tan(α+φ) ≈ 0.6. Then:
F_a = T_trip / (r·tan(α+φ)) = 60 / (0.025 × 0.6) ≈ 4000 N
So the disc-spring stack must hold about 4 kN of axial preload — routine for a Belleville stack. Each of the six balls carries ≈ 667 N axially and ≈ 400 N tangentially at release. Now a crash spikes the demand toward 200 N·m: the balls climb out within a couple of milliseconds, and the gearbox and screw never see more than the 60 N·m setpoint. Reset by letting the balls drop back into their pockets and re-establishing preload — no parts consumed.
Ratchet, free-wheel, or synchronous
What happens after the trip is a design choice, and it matters:
- Ratcheting: the balls drop into the next pockets each revolution, giving a loud ratchet pulse per cycle. Fine if the jam clears quickly, but sustained ratcheting hammers the pockets.
- Full disengagement (free-wheel): an axial latch holds the balls fully out so residual torque is essentially zero until a manual reset. Best for crash protection, where you never want repeated impact loading.
- Synchronous re-engagement: a single set of pockets means the halves can only re-mate at one angular position, preserving phase. Essential for indexing tables, printing and packaging lines where timing is sacred.
The distinction is safety-relevant: a ratcheting unit that keeps slamming into an unresolved jam can fail worse than the drivetrain it was meant to save.
Where it lives and how it wears out
Ball-detent limiters guard machine-tool spindles and ballscrews, robot joints, conveyor and indexing drives, packaging and printing machinery, and PTO lines — anywhere a jam or crash could wreck an expensive gearbox. They are selected by service factor, not by code, but they are the mechanical backstop behind a lot of automation.
They are not immortal. Every trip is a Hertzian contact event, so repeated tripping spalls the balls and pockets by rolling-contact fatigue, exactly like a bearing raceway — a shifting, unreliable setpoint is the warning sign. The trip torque also drifts: grease breakdown changes μ, disc springs relax and lose preload, and temperature swings both. The governing relation assumes all balls share load equally and friction is stable — real repeatability lands at ±5–10%, and worse if pockets wear unevenly. Good practice is to trip-test and re-calibrate the preload periodically rather than trust the factory number forever.
| Device | Reset | Trip accuracy | Response | After-trip torque |
|---|---|---|---|---|
| Ball-detent torque limiter | Auto re-engage or manual | ±5–10% | ~1–5 ms snap | Ratchet pulses, or ~0 (disengage) |
| Shear pin / shear bolt | Replace the pin | ±15–25% | <1 ms (fracture) | 0 (link severed) |
| Friction slip clutch | Automatic | ±20–30%, drifts with wear | Continuous slip | ≈ constant slip torque |
| Magnetic hysteresis clutch | Automatic | ±5% | Continuous slip | ≈ constant torque, heats up |
Frequently asked questions
Does a torque limiter reduce or cap torque during normal running?
No. Below its setpoint a ball-detent limiter is rigid and slip-free — it transmits the full torque exactly, adding only a little compliance. It is a threshold switch that disengages once, not a governor that continuously holds torque at a value. That is why it protects positioning drives without introducing lost motion.
Do more balls give a higher trip torque?
No — a common misconception. For a fixed total spring preload F_a, the ball count cancels in T_trip ≈ r·F_a·tan(α+φ): the balls simply share the same axial force. Adding balls lowers per-ball Hertzian contact stress and smooths engagement and re-seating, but it does not raise the trip point. Change the setpoint with preload or radius, not ball count.
Ball-detent limiter versus a shear pin — when do you use each?
A shear pin is cheap and fractures in under a millisecond, but it is sacrificial (you replace it), less accurate (±15–25%), and offers no protection until swapped. A ball-detent unit is resettable, more accurate (±5–10%), throws a switch, and protects repeatedly — at higher cost and bulk. Use pins for rare, catastrophic events; use ball detents where jams are frequent or downtime is expensive.
How is the trip torque set and how repeatable is it?
You turn an adjustment nut that compresses the Belleville disc-spring stack, changing the axial preload F_a and thus T_trip ≈ r·F_a·tan(α+φ). Quality units repeat to about ±5–10%, but the setpoint drifts with lubricant condition, temperature and spring relaxation, so it should be trip-tested and re-calibrated against measured torque periodically.
Why disengage instead of just slipping continuously?
A friction slip clutch that slips continuously dumps the overload energy as heat and wears its faces, and its torque drifts with wear. The ball-detent snap-through gives a sharp, repeatable threshold and can drop residual torque to nearly zero, so the drivetrain sees a clean cut-off rather than a sustained hammering. The trade is that continuous ratcheting under an uncleared jam must be avoided — hence latching, full-disengagement designs for crash protection.