Troubleshooting
Why power cables hum during motor startup
Short answer
The hum is electromagnetic force between the phase conductors, not imbalance: attraction and repulsion reverse every half-cycle, loading the cables at 120 Hz, and force follows current squared. A 538 A start against 84 A running is 6.40 times the current and about 41 times the force. Restrain the conductors and cut the starting current; the measured numbers do not implicate the motor.
On this page
The hum is the feeder conductors being electromagnetically worked against each other and the raceway wall at twice line frequency, and the starting current is what works them: 538 A inrush against 84 A running is a 6.40 to 1 current ratio, and force between conductors follows the square of the current, so the mechanical load during that two-second acceleration runs about 41 times its running value. Balanced three-phase currents cancel the field at a distance — they do not cancel the forces between the conductors themselves, and that is the same double-frequency mechanism behind transformer hum. Nothing in the reported numbers implicates the motor. Prove where the movement is, meet the support rules, and cut the starting current with a soft starter, a partial-winding start, or a drive.
The force that balanced phases do not cancel #
When three conductors run close together, their external fields largely cancel a few conductor spacings away, which is why nearby circuits and clamp-on measurements stay quiet. Between the conductors themselves there is no such cancellation. Each phase current sits in the magnetic field of the other two, and for two parallel conductors the force on each metre of length is:
F/l = (μ₀ × I₁ × I₂) / (2π × r)
with μ₀ = 4π×10⁻⁷ H/m, currents in amperes and spacing r in metres giving newtons per metre. Same-direction currents attract, opposite currents repel, and two 60 Hz currents separated by 120 electrical degrees produce a product that swings through both signs every half-cycle: multiplying two sinusoids yields a double-frequency term, so a 60 Hz system loads its conductors at 120 Hz. At any instant a conductor is pulled toward one neighbour and pushed away from the other, and the resultant reverses 120 times a second.
Switchgear engineers run this same calculation against short-circuit currents, formalized in IEC 60865-1; for the rigorous treatment of that standard, see calculating busbar electromagnetic forces per IEC 60865-1. The part that matters here is the scaling: in a fixed geometry, force grows with current squared.
The force at the measured inrush, worked out #
The values measured on this installation:
| Item | Reported value |
|---|---|
| Motor | 40 hp, 200 V, three-phase, 60 Hz, 108 A nameplate, across-the-line start, partial-winding-start capable |
| Feeder | Three AWG 2 copper conductors in one metal raceway, about 250 ft (76 m), several 90° bends, vertical rise through the building |
| Starting current | 538 A ± 3 A, all three lines |
| Running current | 84 A ± 2 A, all three lines |
| Running voltage at panel | 210.2 / 210.3 / 212.0 V line-to-line |
| Running voltage at motor | 205.4 / 205.2 / 206.9 V line-to-line — a 5.0 V average drop, 2.4% |
Worked example, worst-case pair. Inputs: both conductor currents at the measured starting value, centre-to-centre spacing r = 0.05 m — about 2 in., realistic for conductors in a raceway of the reported roughly 2 in. outside diameter — and μ₀ = 4π×10⁻⁷ H/m.
F/l = (4π×10⁻⁷ H/m × (538 A)²) / (2π × 0.05 m) = (2×10⁻⁷ × 289 444) / 0.05 = 0.0579 / 0.05 ≈ 1.16 N/m, about 0.08 lb/ft per conductor pair.
Those use RMS currents; instantaneous peaks reach roughly twice that. The same arithmetic at the 84 A running current gives 0.028 N/m, about 0.002 lb/ft. The ratio is the square of the 6.40 to 1 current ratio, ≈ 41. Over the full 76 m run, 1.16 N/m would integrate to about 88 N (20 lbf) if one pair acted in a single direction the whole way — geometry and restraint cancel most of it, and what remains is still enough to slap an unrestrained conductor against the raceway wall 120 times a second for the two seconds the current takes to decay. That is the hum carried by the structure of a fifteen-storey riser.
Find where it moves first #
- Capture one complete start: three line currents and three line-to-line voltages at the panel and at the motor terminals, with a vibration channel on the raceway. Confirm the hum window matches the current-decay window — here, about two seconds.
- Walk the run with an accelerometer or a mechanic's stethoscope on a fixed mount, comparing amplitude and phase at the motor feet, the raceway at several points, each bend and pull point, and the panel. The dominant inrush-synchronous location is the problem; if the motor frame dominates and the raceway follows it, the noise is structure-borne from the machine instead.
- Keep hands and tools off the raceway and conductors during the start — the same 41× force acts on anything touching the conductors. Use instruments, not palms.
- Inspect supports and restraint only after lockout and after testing for absence of voltage: conductor supports in the vertical sections, wedges or spacers at pull points, and bright contact marks on the insulation or the raceway wall — marks are proof of past movement.
What each observation points to #
| Observation | Likely mechanism | Deciding check |
|---|---|---|
| Hum follows the inrush interval only | Electromagnetic conductor force during acceleration | Overlay current and vibration traces |
| Loudest at a bend, pull point, or long vertical section | Conductor slap: restraint gap or a roomy raceway | Compare vibration both sides of that point; look for contact marks |
| Motor frame dominant, raceway motion follows it | Structure-borne vibration from the motor or fan | Compare motor-foot, structure and raceway amplitude and phase |
| Starting currents differ noticeably between phases | Connection, contact or winding asymmetry | Capture all three current waveforms in one start |
| Panel voltage holds while motor-terminal voltage collapses during start | Feeder or connection impedance | Simultaneous two-end voltage capture during inrush |
| Balanced current, adequate voltage, one location still moves | Local looseness or resonance | Inspect supports, fittings and mounting at that spot |
The motor itself is not indicated. The measured starting current is 4.98 times the 108 A nameplate and balanced within ±3 A, which is ordinary across-the-line behaviour for a fan; the Code G letter on the nameplate exists to classify locked-rotor kVA, and a winding fault severe enough to shake its own feeder shows up as current imbalance or tripped overloads first. If running current ever climbs over nameplate, treat that as a separate investigation, as with a motor running over nameplate amps after soft starter bypass.
Cut the current and the force falls fast #
Because force follows current squared in the same geometry, every amp removed from the start counts twice:
- 30% less starting current leaves 0.7² = 0.49 — about half the force gone.
- Half the current leaves a quarter of the force.
- A soft-starter current limit of 400 A (illustrative setting) is 0.744 of the measured start, leaving 0.744² = 0.55 of today's force; a 350 A limit is 0.651 of it, leaving 0.42.
A soft starter on a fan is the usual first move: fan torque demand grows with speed squared, so the machine tolerates a long, current-limited acceleration, and the limit caps the conductor force for the whole start. The nameplate documents partial-winding-start capability, so a two-contactor sequence is also open — the first step draws the fraction of full inrush stated in the motor documentation, and the force falls with its square. A drive goes furthest: it accelerates the fan at close to running current, returning the conductor force to the 0.028 N/m running level and effectively ending the hum. The drive's output side then has cabling considerations of its own, covered in choosing a line reactor or motor reactor for a VFD.
Cable, code and starting-voltage decisions #
Two separate questions hide inside replace the cable — noise, and compliance.
Noise first: the 41× multiple is set by the current, not the conductor size. Heavier wire is stiffer and moves less for the same force, but this start in 3/0 copper still loads the conductors at the same 1.16 N/m scale. Fix the sound with restraint and reduced starting current.
Compliance: NFPA 70, the National Electrical Code (2023 edition), governs motor branch circuits — Article 430 sizes conductors at no less than 125% of the full-load current taken from its tables (430.6: nameplate current sets the overload, not the conductor), and 300.19 sets maximum support intervals for conductors in vertical raceways. On ampacity alone the existing feeder fails that method: AWG 2 copper in the 75 °C ampacity column carries 115 A, and 125% of the 108 A nameplate is already 135 A before applying the table value the code actually uses. Confirm the table figures, intervals and local amendments with the authority having jurisdiction and the equipment documentation; for the method end to end, see NEC motor branch circuit sizing with #14 AWG on a 20 A breaker.
Starting voltage is the third check. The reported 5.0 V line-to-line drop at 84 A scales, as a linear estimate, to 5.0 V × 6.40 = 32 V at full inrush — about 15% of 210 V, putting the motor terminals near 179 V at the worst instant. That is an estimate from running data; measure the real value during a start. A terminal voltage measuring at or below that level means acceleration is marginal and the drop is worth fixing. Note that a higher-impedance feeder trims peak current slightly and stretches the start — it does not silence anything, and the measured start current was taken on this very feeder.
The decision, case by case #
- Raceway peaks during inrush, motor frame quiet: conductor force confirmed. Verify or add conductor support to the 300.19 intervals, wedge or spacer the loud bends and pull points, and add reduced-voltage starting — expect the noise to fall with the square of the current reduction you achieve.
- Motor frame dominates: structure-borne noise from the machine. Balance the fan, check mounting and isolation, align the coupling; cable work will not touch it.
- The feeder: upsize it for the Article 430 minimum and for any measured starting-voltage shortfall, not for the hum — and still restrain the conductors and cut the starting current, because the squared force rides on any conductor size you pull in.
Questions asked
Why is the hum at 120 Hz on a 60 Hz system?
The force between two conductors is proportional to the product of their instantaneous currents, and the product of two 60 Hz sinusoids is a steady term plus a term at exactly twice the frequency. The steady term gives some pairs a bias toward attraction or repulsion, and the 120 Hz term is the shake you hear. The same math is why transformers and motors sing at twice line frequency.
The three conductors are already twisted. Why doesn't that stop the force?
Twisting reduces loop area and the external magnetic field, which helps with inductive pickup in nearby circuits, but the forces between the conductors themselves depend only on their currents and their spacing. A twisted trio that is loose in the raceway still slaps the wall during inrush. Support at the interval the code requires, wedging at pull points, and a lower starting current are what stop the movement.
How much current reduction does it take to quiet the run?
Force tracks the square of the current, so reductions compound quickly. Cutting starting current 30% removes about half the force, and a drive that accelerates the fan at close to the 84 A running level removes about 98% of it relative to an across-the-line start. With a soft starter, set a current limit, listen, and step the limit down until the building stops hearing the start.
If we upsize the feeder, will the hum stop?
Only partly. The pair force scale is set by the current, so the same start loads 3/0 copper the same way it loads AWG 2; heavier wire is stiffer and moves less, but a roomier raceway can also give the conductor more travel before impact, which makes each slap sharper. Upsize when the motor-circuit calculation or a measured starting-voltage shortfall calls for it, and plan on restraint plus reduced-current starting regardless.
Why can the whole building hear a two-second start?
The raceway wall is driven like a diaphragm 120 times per second wherever the conductors strike or load it, and metal conduit is clamped and bonded to building steel at every entry, so the structure carries the vibration to floors far from the motor. The two-second duration matches the current decaying from the starting value toward 84 A as the fan accelerates, which is the signature of inrush force rather than a running problem.