Whether a load can push the motor rather than the other way around isn't a side detail — it decides if you need a brake, whether the joint is safe to touch, and whether it holds position with the power off.
A backdrivable mechanism is one where force applied at the output (the load side) can drive the mechanism backward and turn the motor shaft. A non-backdrivable mechanism is one where output-side force, no matter how large, cannot turn the motor — the mechanism physically locks against back-driving. This single mechanical property, which is a function of gear geometry and friction rather than of the motor or controller at all, determines whether a joint feels alive and compliant in your hand or feels like it's bolted to the wall, and whether removing power leaves an axis free to fall or frozen exactly where it stopped.
Whether a mechanism is backdrivable is set by the geometry of the mesh, not by the motor, the drive, or the software. A worm gear's tooth contact happens at a shallow lead (helix) angle, so the friction force at the contact point has a large component directly opposing any attempt to slide the teeth — when that friction exceeds the tangential force the load can generate, the worm physically cannot turn, no matter how much torque is applied at the wheel. This is the same self-locking principle behind a lead screw with a shallow thread angle, or high-ratio planetary/harmonic drives with significant internal friction. A spur, helical, or bevel gear mesh (and belt/chain drives, and ballscrews with a steep thread angle) has a much higher lead angle and lower friction coefficient at the contact, so the tangential force from a load easily overcomes friction and the drive spins backward. As a rule of thumb, self-locking becomes likely once the coefficient of friction at the mesh exceeds the tangent of the lead angle — high gear ratios achieved through shallow-angle, high-friction mechanisms (worm, some harmonic drives) tend toward non-backdrivable; low-ratio, high-efficiency mechanisms (spur gears, ballscrews, belts, direct drive) tend toward backdrivable.
Gear ratio alone doesn't determine backdrivability — mesh geometry and friction do. A high-ratio planetary gearhead built with helical gears is still fully backdrivable; a low-friction ballscrew or a cycloidal drive can transmit large reductions while still passing torque back through efficiently. Conversely, even a modest-ratio worm set can be completely non-backdrivable because of its shallow lead angle, independent of the numeric ratio. This distinction matters enormously in application: a non-backdrivable mechanism is genuinely dangerous in a design that expects to fail safe by letting a load lower under its own weight if power is lost (e.g., an elevator or hoist relying on the motor to lower the load will simply stall, not descend), and equally, a backdrivable design used somewhere a fixed position must be held without power (a jack, a valve actuator, a parking brake) needs an explicit mechanical brake or detent added, because the drive train itself provides zero holding torque once current is removed.
Explains backdrivability — whether force applied at a mechanism's output can turn its motor — as a property of gear mesh geometry and friction rather than motor or control choice, using a worm gear (non-backdrivable) and a low-ratio spur gear joint (backdrivable) as contrasting examples.
Engineers often assume backdrivability is a function of gear ratio alone — "high ratio means it can't be backdriven." That is not reliable. Backdrivability is governed by the friction angle at the mesh relative to the mechanism's lead or helix angle, which depends on the specific gear type (worm, spur, helical, planetary, harmonic, ballscrew, lead screw) and its geometry, not simply on how large the overall reduction is. Two mechanisms with an identical numeric ratio can land on opposite sides of backdrivable depending on which gear technology produced that ratio.
Self-locking occurs when the friction force generated at the point of tooth or thread contact has a large enough component opposing sliding that it exceeds the tangential (driving) force component created by an applied output torque. This is most pronounced in worm gears, where the worm's thread has a shallow lead angle, and in some lead screws with a similarly shallow thread angle — friction there is large relative to the small lead angle, so it locks. Spur, helical, and bevel gears mesh at much steeper effective angles with lower sliding friction, so output torque passes back through the mesh with high efficiency and turns the input shaft freely — that is backdrivability. Ballscrews (which replace sliding friction with rolling ball bearings) and belt/chain drives are also generally backdrivable even at high efficiency and moderate ratios, for the same reason: low friction relative to the driving force component.
This property drives concrete design decisions. In robotics, backdrivable actuation (direct-drive motors, low-ratio transmissions, or series elastic actuators) is preferred for joints that need to be safe around humans or need force/torque sensing through motor current, because the joint yields naturally to unexpected contact. Non-backdrivable actuation (worm-geared, high-ratio harmonic drives with high internal friction) is preferred where holding position without continuous power draw matters — gate operators, jack lifts, some valve actuators — since the mechanism itself provides the holding torque and a power loss or controller fault does not let the load fall. The failure-mode implications are opposite and both need explicit engineering: a backdrivable joint needs an added mechanical brake wherever "hold position with no power" is a requirement, while a non-backdrivable joint used in a hoist or elevator lowering mechanism must never be relied upon to lower a load under gravity by simply cutting motor torque — it will just stall.
No. Self-locking is an inherent property of the drive train's geometry and friction — it locks passively, with zero added parts, purely because back-driving force cannot overcome mesh friction. A mechanical brake (electromagnetic, friction disc, or spring-applied) is a separate, dedicated component added specifically to hold position, typically used with backdrivable transmissions that have no inherent self-locking of their own.
Yes, but only actively — by continuously supplying holding current/torque through the motor and its control loop to counteract the load. This costs continuous power, generates heat, and fails immediately (the load moves) the moment power or control is lost — unlike a non-backdrivable mechanism, which holds passively with zero power draw.
It depends on the specific unit and ratio, but generally harmonic (strain wave) drives have notably lower backdrive efficiency than spur or planetary gearing of similar ratio, due to the flexspline's wave-generator friction — some higher-ratio harmonic drives are effectively non-backdrivable in practice, which is one reason many collaborative robots that want true torque-mode compliance use lower-ratio transmissions, series elastic actuators, or direct drive instead.
Yes, significantly. Estimating output torque from motor current works well only when the transmission has high, predictable, low-friction efficiency in both directions — exactly the property of a backdrivable mesh. A non-backdrivable, high-friction transmission (like a worm gear) has friction losses so large and load-dependent that motor current becomes a poor proxy for actual output torque, which is why torque-sensing robot joints favor backdrivable, low-friction transmissions or add a dedicated torque sensor instead.
They are closely related but not identical. Backdrivability is a binary-ish threshold condition (can output-side force turn the input at all), while forward efficiency is a continuous measure of power loss during normal motor-driven operation. A mechanism can have moderate forward efficiency and still be non-backdrivable if its self-locking friction is high enough in the reverse direction — worm gears are the classic example, often 50–90% efficient driving forward while being completely non-backdrivable in reverse.
Try our Mechatronics Studio
More calculators, simulators, and guides for this discipline.
Mechatronics Engineering Handbook (Full Access)
✨ Premium ContentA zoomable interactive reader — free preview, then unlock the full set.