Troubleshooting
Servo Motor Rear Noise: Brake or Bearing, Not Gears
Short answer
Rear servo noise under the resolver cover is the spring-applied 24 V brake or the rear bearing, and the motor holds bearings, not gears. Verify brake release under electrical load first: brake drag shows up as current rise, heat and noise tied to release events, while bearing noise tracks shaft speed and survives a brake proven healthy.
On this page
- What sits under the resolver cover
- Why the noise location changes the diagnosis
- Check the brake release first
- Separating brake, bearing and reducer
- Worked example: reading 34,000 to 38,000 hours against a 40,000-hour expectation
- Inspection sequence and the commutation trap
- The decision for each case
- Questions asked
Noise from the rear of a servo motor, strongest under the resolver cover, is the brake or the rear bearing, and never small gears: the motor housing contains bearings and, on holding axes, a spring-applied brake, while the gear reduction lives in the reducer at the front flange. Work the brake first. A brake that never fully releases is the cheapest fault to find, and it announces itself as extra current, heat at the rear end and noise tied to release events. A bearing, by contrast, tracks shaft speed.
What sits under the resolver cover #
From the outside in you have the resolver reporting rotor angle, the brake coil and friction assembly on axes that must hold position, and the rear bearing at the end of the rotor. A brushless servo motor is an electromagnetic machine: stator, permanent-magnet rotor, bearings. When the cover is off and you hear what sounds like small gears, the real candidates are the brake dragging and the rear bearing. On these axes the brake is the usual spring-applied type fed from a 24 V circuit: coil powered means released, coil unpowered means clamped, so every supply-side fault shows up as drag rather than as a loose axis. If you are not sure which feedback device your motor carries, identifying geared resolver and encoder feedback on old servos tells them apart before you open anything.
Why the noise location changes the diagnosis #
Sound location and sound source are not the same thing. A worn reducer transmits vibration through the motor housing, and a dragging brake at the rear projects noise forward toward the reducer. Separate the two with a contact probe, a mechanic's stethoscope or similar, held on stationary metal while the axis repeats one slow move: rear motor housing, front flange, reducer housing. The loudest point wins the argument. Keep the probe off anything rotating. Do this before you condemn a reducer that may be perfectly healthy.
Check the brake release first #
The brake check is electrical and cheap, so it goes first. Measure, in this order:
- Voltage across the brake coil while it is connected and commanded released. A reading at the disconnected supply proves nothing: connector resistance, cable loss and a weak switching device only show up under load.
- Coil current against the figure in the service data for your exact motor. Take the release-voltage tolerance and the coil-current range from that document, because both vary with brake size and no generic number substitutes for them.
- Axis torque or current demand, logged at the same pose, speed, payload and direction on the noisy axis and on a quiet sister axis. Extra demand while the axis still follows its program is mechanical drag.
- Rear-end temperature after a fixed running interval, same comparison, same starting temperature.
- Release timing: whether the controller commands torque before the brake has physically let go, which hammers the friction faces at every start.
The physics behind the check is simple. The motor must produce torque to overcome whatever the brake still applies, and friction power grows with that residual torque multiplied by shaft speed. A partially released brake therefore delivers exactly the triad you are chasing: noise, current rise and localized heat.
Separating brake, bearing and reducer #
| What you observe | Points to | Confirming check |
|---|---|---|
| Noise, current rise and rear heat tied to release and apply events, changing with brake state | Brake not fully releasing | Voltage across the connected coil during a commanded release; coil current against service data |
| Noise tracks shaft speed up and down, unchanged by brake state | Rear bearing | Contact probe on rear housing versus front flange at fixed slow speed; shaft play checked in the safe state |
| Loudest at the reducer, varying with direction, load or arm position | Reducer | Backlash and lubricant inspection at the next planned access |
| Voltage collapses the moment the coil energizes | Supply path, not the motor | Measure at the drive brake output and at the motor connector; inspect terminals and cable |
Compare like with like: same program, speed, payload, starting temperature and direction. A quiet unit of the same configuration is your acoustic reference; a different machine family only tells you the noise is unusual, not what normal sounds like. Do not call the cause from the sound alone; each row closes only on its confirming check.
Worked example: reading 34,000 to 38,000 hours against a 40,000-hour expectation #
Inputs: 34,000 to 38,000 running hours on the used machines, taken from their hour counters, and 40,000 hours as the expected reducer life the fleet was bought against. That 40,000-hour figure is a reported expectation, not a verified rating for a specific reducer.
Rule in words: life consumed equals hours run divided by expected life, and the remaining margin is expected life minus hours run.
Arithmetic:
- 34,000 h ÷ 40,000 h = 0.85, so 85% of expected life consumed.
- 38,000 h ÷ 40,000 h = 0.95, so 95% consumed.
- Margin: 40,000 h − 34,000 h = 6,000 h; 40,000 h − 38,000 h = 2,000 h.
Result: this fleet has consumed 85 to 95% of the expected reducer life, with 2,000 to 6,000 nominal hours in hand.
Decision: past 80% consumed, book a reducer inspection, meaning lubricant condition and backlash measurement, into the next planned outage. Hours measure exposure, not the failed part, and the noise under the resolver cover is still worked brake-first regardless of what the hour meter says.
Inspection sequence and the commutation trap #
- Put the machine in its safe service state before anything else. Isolate electrical power under your energy-control procedure, which OSHA 1910.147 governs in the US, and mechanically support any axis gravity can move, because releasing the brake removes its holding torque and an unsupported arm or wrist falls.
- Record a baseline: axis, direction, commanded speed, payload, motor current, temperature, and the point in the move where the sound peaks. Repeat cold and warm.
- Run the brake test under load, from the release command through full physical release, on the noisy axis and a quiet one. If voltage sags when the coil energizes, fix connectors, cable or the drive brake output before touching any mechanical part.
- Localize with the contact probe at low speed. Stop and reassess if a housing heats quickly or the sound turns into impact noise.
- Remove the resolver only if you must reach the brake beneath it. The resolver sets commutation: it reports rotor angle so the drive keeps current in the correct electrical relationship to the rotor field. Scribe unambiguous alignment marks across the resolver body, its clamp and the end housing, photograph them, and follow the motor service procedure for reassembly. Rotated on reinstallation, it shifts the reported angle with the shaft standing still; torque per amp falls, current and heat rise, and only re-seating to the marks fixes it.
The decision for each case #
- Deficient brake supply, meaning low voltage or current under load: repair the wiring, connector or drive brake output before replacing any mechanical part. This alone closes a share of noisy rear ends.
- Supply healthy but drag signs persist: the brake assembly inside the motor is the suspect, so service or replace it per the motor procedure.
- Brake proven healthy and the noise still tracks shaft speed: rear bearing. Book the motor out for bearing service, and testing whether a servo motor is actually faulty covers the electrical checks worth running before you send it.
- Probe position, backlash and lubricant all point at the reducer: plan the reducer work, using the life-consumed figure from the example above to decide whether it is this outage or the next one.
Questions asked
What does a healthy holding brake sound like?
One distinct click when it releases and one when it re-applies. Continuous scraping, growling or squealing while the axis runs is not normal, and it points at dragging friction faces or a bearing rather than at the click of a working brake. Compare the noisy axis with a quiet sister axis and listen for that single event on each side.
Can I unplug the brake to see whether the noise stops?
Not as a running test. On a spring-applied brake, removing coil power keeps the brake clamped, so unplugging it adds drag and tells you nothing new. If you need a released-brake comparison, use the controller service-mode release with the axis mechanically supported against gravity, then compare slow rotation against a quiet axis.
Where do I find the brake coil current and release voltage tolerance?
In the service documentation for the exact motor, from the manufacturer or its distributors. Coil resistance and release voltage vary with brake size, so no generic tolerance is a safe substitute. Measure across the connected coil during a commanded release and compare that reading against the number in that document, not against another motor type.
Does 38,000 hours mean the gearboxes are finished?
No. At 38,000 of an expected 40,000 hours the reducers have consumed about 95% of their expected life, which means book an inspection with lubricant condition and backlash measurement at the next planned outage. Hours measure exposure, not the failed part, and noise located under the resolver cover is worked brake-first regardless of the hour meter.
What happens if the resolver goes back on slightly rotated?
Torque per amp drops because the drive now applies current at the wrong electrical angle to the rotor field, so the axis pulls more current for the same motion, runs hotter and can trip on overcurrent. No gain tuning fixes it. Re-seat the resolver to the alignment marks you scribed before removal, and the current draw should return to its baseline.