A six-axis arm holding a 20 kg payload at full extension is a stored-energy system. Gravity constantly pulls the shoulder and elbow joints downward. During normal operation, only the servo loop's actively generated torque prevents that motion. When power disappears through an emergency stop, a tripped breaker, or a drive fault, that torque collapses within milliseconds.

The standard engineering answer is a fail-safe brake on each gravity-loaded axis, usually mounted on the motor shaft ahead of the gearbox. It looks like a simple component: a disc, some springs, a coil. In practice, brake selection and integration determine whether a robot holds its pose or lets a vertical axis sag onto a fixture, a part, or a person.

This article covers three parts of that problem. First, how electromagnetic brakes engage without power. Second, how to size them for both holding and stopping. Third, how to sequence them with the servo drive so that load torque passes cleanly between the motor and the brake.

The dominant design in robot joints is the spring-applied, electromagnetically released brake. Compression springs push an armature plate against a friction disc splined to the motor shaft, clamping it against a fixed face. Energizing the coil pulls the armature across a small air gap, typically a few tenths of a millimetre. This compresses the springs and frees the disc. When current stops, the springs win. The default state needs no energy, and that is what makes the brake fail-safe.

The permanent magnet brake works on the same power-off principle. A permanent magnet holds the armature against the friction face. To release the brake, the coil generates a field that cancels the magnet's pull. These brakes offer high torque density, zero backlash, and a compact form, which is why they are common in servo motors and collaborative robot joints. The trade-off is sensitivity. Release depends on the correct coil current, so overvoltage or reversed polarity can re-engage the brake. Magnet strength also falls as temperature rises.

Power-applied brakes reverse this logic: they need energy to clamp. They suit dynamic positioning tasks but are unsuitable as safety holding devices. A related distinction matters as well. Many brakes integrated into servo motors are holding brakes. They are rated to hold a stationary load and to survive a limited number of emergency stops, not to stop the axis routinely.

Friction surfaces wear, and wear changes behaviour. As a spring-applied brake's lining thins, the air gap grows. Eventually the coil can no longer pull the armature in, and the brake drags while supposedly released. Some designs allow gap adjustment or include release-monitoring switches. When choosing between types, weigh backlash requirements, operating temperature, maintenance access, and how the brake will be verified over its service life.

Takeaway

A fail-safe brake is defined by its unpowered state, so every engineering question about it comes down to what happens when the energy is gone and the springs or magnets take over.

Sizing starts with static holding torque. Calculate the worst-case gravity torque at the joint, with maximum payload and the arm at full horizontal extension. Then reflect that torque to the brake location. A motor-side brake sees the load torque divided by the gear ratio. Gearbox efficiency in back-driving can help, but conservative designs ignore it. Then apply a safety factor, commonly 1.5 to 2, because friction coefficients vary with contamination, temperature, and surface condition.

Dynamic braking capacity is a separate check. During an uncontrolled stop, the brake must absorb the axis's kinetic energy, ½Jω², where J is the total inertia reflected to the brake shaft, including the motor rotor. On a vertical axis, it must also absorb the potential energy released while the load falls during the stop. Dynamic torque ratings are often lower than static ratings because sliding friction is lower than static friction. Engagement delay matters too: before the brake grips, a gravity-loaded axis is already accelerating.

Thermal limits close the loop. Each stop converts energy into heat at the friction interface. Manufacturers specify maximum energy per engagement, allowable friction work per hour, and total friction work before adjustment or replacement. The coil also generates heat whenever the brake is released, and that heat adds to the motor's own temperature inside a sealed housing. For permanent magnet brakes, elevated temperature directly reduces available torque.

Size against the stop category that loads the brake hardest. In a Category 1 stop, the drive decelerates the axis before power is removed, so the brake does little dynamic work. In a Category 0 stop, the brake absorbs everything. Design for Category 0 at maximum speed and payload, even if it rarely occurs. Also account for burnishing: new friction surfaces may not reach rated torque until they are run in.

Takeaway

Holding a load and stopping a load are different jobs with different ratings, and a brake sized only for the first can fail at the moment the second is demanded.

The most common brake-related failure is not a weak brake but a poor handoff. When the robot is enabled, the correct sequence is: energize the drive, close the position loop at the current position so the motor produces holding torque, then command brake release. After release, wait the rated release time before permitting motion. If the brake opens before the servo carries the load, the axis drops by an amount that depends on loop response and backlash.

Disabling reverses the order. Decelerate to zero, hold position under servo control, command brake engagement, then wait the engagement time plus margin before removing motor torque. That engagement time depends on the coil circuit. A simple freewheeling diode slows the collapse of the magnetic field and can lengthen engagement time considerably. A Zener diode or varistor clamp allows faster field decay. Switching on the DC side of a rectified supply is also faster than switching on the AC side.

Coil drive strategy also matters. Overexcitation applies a higher voltage briefly to pull the armature in quickly, then drops to a lower holding voltage, often by PWM. This reduces coil heating and shortens release time. Voltage tolerance deserves attention. Long cable runs on a 24 V supply can drop enough voltage that a brake only partially releases. The result is drag, heat, and accelerated lining wear that is easy to misdiagnose as a mechanical problem.

In safety-critical applications, brake control becomes a safety function in its own right. Safe Brake Control, defined in IEC 61800-5-2, uses redundant switching channels and is evaluated against ISO 13849 performance levels. Periodic brake testing verifies that holding torque remains adequate: the drive applies a test torque against the engaged brake and confirms that the encoder registers no motion beyond a threshold. Robot safety standards such as ISO 10218 also expect a means to move axes without drive power, so that a trapped person can be freed.

Takeaway

Think of the brake and the servo as two hands passing a heavy object: at no moment should both let go, and the timing of the handoff deserves as much design attention as the grip itself.

Electromagnetic brakes are passive in principle but active in integration. The spring or magnet supplies the fail-safe behaviour. Sizing calculations and control sequencing determine whether that behaviour holds up under real loads, real temperatures, and real fault conditions.

Several practical recommendations follow. Size static torque against worst-case pose with margin. Check dynamic energy and thermal limits for Category 0 stops separately. Use fast field-decay circuits when engagement time matters. Never release a brake before the servo is carrying the load.

Finally, treat the brake as a component that degrades. Build in periodic brake testing and release monitoring, and track the friction work accumulated over the brake's life. A brake verified once at commissioning is a brake whose actual condition is unknown.