Electronics Guide

Magnetostriction as the Primary Noise Source

Magnetostriction is the dimensional change that occurs in ferromagnetic materials when magnetized. In transformer cores, the cyclic magnetization caused by AC current produces cyclic dimensional changes that vibrate the core at twice the power frequency (100 Hz for 50 Hz mains, 120 Hz for 60 Hz mains).

The magnetostriction of the core material depends on several factors:

• Core material composition: Silicon steel (used in most power transformers) has lower magnetostriction than plain iron. Higher silicon content further reduces magnetostriction but increases material cost and brittleness.
• Grain orientation: Grain-oriented silicon steel (GOSS) has very low magnetostriction when the flux is aligned with the rolling direction. Non-oriented steel has more uniform properties but higher overall magnetostriction.
• Flux density: Magnetostriction increases with flux density. Operating a core at lower flux density reduces noise but requires a larger, heavier, and more expensive core.
• Harmonic content: Harmonics in the supply voltage appear in the core flux. Because magnetostriction responds to the square of the flux, each odd flux harmonic adds vibration at further even multiples of the supply frequency. Harmonic load currents add to the winding forces described below.
• DC magnetization: A direct-current component in the magnetizing current, for example from half-wave rectified loads elsewhere on the supply or from geomagnetically induced currents in power networks, drives the core into saturation on one half-cycle. The vibration loses its symmetry, gains components at odd multiples of the supply frequency, and grows markedly louder. Toroidal transformers, which have no air gap, are especially sensitive: a DC offset of a few hundred millivolts on a 230 V supply can multiply a large toroid's no-load current and make it growl audibly.

The frequency doubling effect of magnetostriction is important: magnetostriction depends on the magnitude but not the polarity of the magnetic field, so both positive and negative flux peaks produce the same strain direction. A 50 Hz or 60 Hz magnetic field therefore produces mechanical vibration at 100 Hz or 120 Hz. The nonlinear shape of the magnetostriction curve adds harmonics at further multiples of that frequency: 200, 300, and 400 Hz on a 50 Hz supply, and 240, 360, and 480 Hz on a 60 Hz supply.

Maxwell Forces and Lamination Vibration

In addition to magnetostriction within the core material, electromagnetic forces between elements of the transformer structure contribute to noise:

• Lamination attraction: Adjacent laminations, and the core sections that meet at each joint, experience attractive forces due to the magnetic flux passing between them. Loose laminations can vibrate under these forces, producing a buzzing or rattling noise.
• Winding forces: Currents in transformer windings create a leakage flux that acts on the conductors themselves, producing forces between turns and between windings. Because the leakage flux is itself proportional to the current, these forces are proportional to the square of the current, producing vibration at twice the current frequency plus harmonic components.

The two families of noise behave differently. Core noise, often called no-load noise, depends on the applied voltage and flux density and is present whenever the transformer is energized; the nonlinear magnetostriction curve gives it a rich series of harmonics. Winding noise, or load noise, depends on the load current. With a nearly sinusoidal current it is concentrated at twice the supply frequency, and because the force, and with it the vibration amplitude and sound pressure, is proportional to I², the load-noise level changes by

ΔL = 40 log₁₀(I₂/I₁) dB

which is about 12 dB for each doubling of load current. Load noise is usually minor in small transformers but can become comparable to core noise in large power transformers built for low flux density. Loose laminations and hardware add rattle, which is irregular and richer in high frequencies than either.

Transformer Noise Reduction Techniques

Reducing transformer acoustic noise involves addressing both the sources of vibration and the transmission of that vibration to the environment:

• Core material selection: Magnetostriction varies widely between core materials, and the grade chosen for magnetic performance is not always the quietest. High-grade, domain-refined grain-oriented silicon steel has low magnetostriction along the rolling direction, and cobalt-based amorphous and nanocrystalline alloys, used mainly in smaller and higher-frequency parts, have lower still. Iron-based amorphous alloys are a notable exception: although valued for their very low core loss, their saturation magnetostriction (roughly 20 to 30 ppm) exceeds that of grain-oriented silicon steel (a few ppm), and amorphous-core distribution transformers tend to be noisier than their silicon-steel counterparts. Low-magnetostriction materials cost more but may be justified for noise-critical applications.
• Core construction: Tight lamination stacking with proper clamping prevents lamination vibration. Wound cores generally produce less noise than stacked cores because they have fewer joints. In stacked cores, step-lap joints, in which successive layers overlap in small steps, spread the transfer of flux across each joint and reduce both loss and noise compared with simple butt joints; all joints should be precisely cut and carefully assembled to minimize gaps.
• Operating flux density: Reducing the operating flux density reduces magnetostriction. This requires a larger core for the same power rating, increasing size, weight, and cost.
• Impregnation: Vacuum impregnation with varnish or resin fills gaps between laminations and windings, damping vibration and providing a more rigid structure.
• Mounting and isolation: Resilient mounting isolates transformer vibration from the chassis and enclosure. Proper isolation mounts have natural frequencies well below the fundamental vibration frequency (100 Hz or 120 Hz for line-frequency transformers), as the calculation below shows.
• Enclosure design: Sound-absorbing materials in transformer enclosures reduce radiated noise. The enclosure structure should avoid resonances at transformer vibration frequencies.

For an undamped mount with natural frequency f_n, the fraction of the vibration transmitted at frequency f is the transmissibility

T = 1 / |(f/f_n)² − 1|

Isolation begins only above f = √2 × f_n; below that the mount passes the vibration on or amplifies it. A mount tuned to 25 Hz under a transformer humming at 100 Hz gives f/f_n = 4 and T = 1/15, a reduction of about 23.5 dB. A stiffer mount tuned to 80 Hz gives f/f_n = 1.25 and T ≈ 1.8, nearly doubling the vibration it was meant to block. Damping in a real mount limits the peak at resonance but slightly reduces the isolation achieved well above it.

Special Considerations for Different Transformer Types

Different transformer applications have different noise characteristics and requirements:

• Distribution and power transformers: Pole- and pad-mounted distribution transformers, and the larger power transformers in substations, run continuously close to homes, and their low-frequency hum carries well and penetrates buildings. Their sound levels are measured by methods such as IEC 60076-10, purchasers commonly specify maximum levels, and low-noise designs may be required in sensitive areas.
• Power transformers in audio equipment: Mechanical hum from the mains transformer is audible in a quiet listening room, and its stray magnetic field can also induce electrical hum in nearby signal circuits. Toroidal transformers are popular in audio equipment for their low stray field, but their sensitivity to DC offset on the mains, described above, makes them prone to growling; commercial DC-blocking circuits exist for this reason.
• High-frequency transformers: Transformers in switching power supplies operate at frequencies from tens of kHz to MHz. While the fundamental frequencies are ultrasonic, burst-mode operation at light load, subharmonics, beat frequencies, and other modulation effects can produce audible noise.
• Reactors and gapped chokes: Iron-core line reactors and filter chokes contain air gaps, and the Maxwell forces pulling the gap faces together add strongly to their vibration. Gap spacers must be rigid and firmly bonded, and a gapped reactor can be louder than a transformer of similar rating.

Sources of Inductor Acoustic Noise

Inductors generate noise through the same magnetostriction mechanism as transformers, but with different characteristics due to their construction and operating conditions:

• Core magnetostriction: The core material expands and contracts with the magnetic field. For inductors in switching converters, this vibration occurs at the switching frequency and its harmonics. When the switching frequency is itself below about 20 kHz, the tone is heard directly; above that, audible noise usually comes from slower variations in the current, described next.
• Audio-rate current envelopes: The most common cause of coil whine in modern equipment is not the switching frequency but a variation of the inductor current at an audible rate. Light-load burst mode and pulse skipping deliver energy in packets whose repetition rate can fall into the kilohertz range, and loads that change in step with a periodic activity, such as a graphics processor rendering frames at a high rate or a processor entering and leaving sleep states, modulate the current at their own repetition rate. The magnetic forces follow the envelope, so the inductor sounds at the modulation frequency even though the converter switches at hundreds of kilohertz.
• Winding vibration: Forces between turns and between the winding and core can cause winding vibration. This is particularly problematic if windings are not tightly wound or securely bonded to the core.
• Air gap forces: The two faces of a discrete air gap attract each other with a force F = B²A / (2μ₀), where B is the flux density in the gap and A the area of the gap face. At 0.3 T this pressure is about 36 kPa, or 3.6 N on each square centimeter of gap, and a ripple of ±0.03 T riding on that bias swings it by about 20 percent in each cycle. Fringing fields near the gap also exert forces on nearby turns of the winding.
• Loose construction: Any loose elements in the inductor construction can vibrate independently, producing noise that may be worse than the inherent magnetostriction.

Factors Affecting Inductor Noise

Several factors influence the noise level produced by an inductor:

• Load current: Inductor noise often increases with load current because the magnetic field, the magnetostrictive strain, and the gap and winding forces all increase with current. This explains why coil whine often becomes noticeable only under high-load conditions, or under loads that pulse at an audible rate.
• Operating frequency: Frequencies above 20 kHz are inaudible to nearly all adults, so a whine from a converter switching at hundreds of kilohertz almost always comes from an audio-rate modulation of the current rather than from the switching frequency itself.
• Core material: Different core materials have different magnetostriction characteristics. Manganese-zinc power ferrites have low magnetostriction, and molypermalloy (MPP) and sendust (iron-silicon-aluminum) powders have compositions chosen for near-zero magnetostriction, whereas plain iron and 50 percent nickel-iron powders have more. A distributed-gap powder core also avoids the concentrated gap forces of a gapped ferrite, so material selection involves trade-offs on both counts.
• Core construction: The physical structure of the core affects vibration behavior. Potted or encapsulated cores tend to be quieter than open cores, and molded power inductors, in which a metal-powder composite is pressed around the winding, have neither a discrete gap nor a free winding.
• Temperature: Core properties and mechanical characteristics can change with temperature, affecting noise production.

Inductor Noise Reduction Strategies

Addressing inductor noise requires attention to component selection, circuit design, and mechanical construction:

• Component selection: Select inductors with low acoustic noise specifications for noise-sensitive applications. Some manufacturers offer "quiet" or "low-noise" versions of their inductor families, and molded and shielded constructions are generally quieter than open drum cores.
• Core material selection: Choose core materials with low magnetostriction for critical applications, such as manganese-zinc ferrite or MPP and sendust powders, and weigh the gap forces of a gapped ferrite against the distributed gap of a powder core.
• Potting and encapsulation: Potting inductors with a suitable compound damps vibration and constrains mechanical motion. The potting material should be selected for thermal conductivity, temperature range, and damping characteristics.
• Frequency and mode selection: Choose switching frequencies well above the audible range to avoid direct audibility, and keep the burst and pulse-skipping rates of light-load modes out of the audio band as well. Spread-spectrum or frequency-dithering techniques reduce tonal peaks when the offending tone is itself audible.
• Mounting: Secure mounting prevents the inductor from rattling against the PCB and enclosure. Resilient mounting materials can isolate vibration if direct coupling is problematic.
• PCB design: Locate inductors away from areas where acoustic noise is most objectionable (such as near ventilation openings in the enclosure). Use stiffened PCB areas to reduce vibration transmission.

Mechanism of Capacitor Acoustic Noise

Ceramic capacitor noise originates from the mechanical response of the dielectric to the applied electric field:

• Electrostriction: All dielectrics experience some dimensional change in response to an electric field. This electrostriction is proportional to the square of the field, producing vibration at twice the applied frequency. When a DC bias is superimposed on the AC signal (the usual case in a decoupling or filtering role), the squared term also generates a component at the fundamental frequency whose amplitude grows with the bias, which is why MLCC singing is typically worst on heavily biased rails.
• Piezoelectric effect: High-permittivity Class II ceramic formulations (X5R, X7R, Y5V, and similar types) are based on ferroelectric barium titanate. As fired, the ceramic's polar domains point in random directions and largely cancel, but a DC bias aligns them, and some of that alignment persists after the bias is removed. The biased or previously biased dielectric then strains in proportion to the applied AC voltage, at the fundamental frequency. This field-induced piezoelectric response and the bias term of electrostriction are closely related, and together they explain why a capacitor on a DC rail vibrates mainly at the frequency of the ripple across it.
• PCB coupling: The dielectric thickens along the field and shrinks across it, so the capacitor's length changes with every cycle, and the solder joints convert that change into bending of the board beneath it. The PCB then acts as a sounding board, radiating the sound.
• Resonance effects: The capacitor's own mechanical resonances lie in the megahertz range, but if the excitation frequency coincides with a resonance of the PCB, noise production can be greatly amplified.

Factors Affecting Capacitor Noise

Several factors determine the noise level from a ceramic capacitor:

• Dielectric type: Different ceramic formulations have different piezoelectric activity. Class I dielectrics such as C0G/NP0 use paraelectric (non-ferroelectric) materials and produce negligible noise, but offer limited volumetric capacitance. Class II dielectrics such as X7R, X5R, and Y5V use ferroelectric barium-titanate materials that offer far higher capacitance density but can produce significant noise.
• Capacitance and case size: Vibration grows with the number of active dielectric layers, and so with capacitance in a given case size, and a longer body changes length by more. High-value capacitors in large case sizes are therefore the usual offenders.
• Applied voltage: Noise increases with the AC voltage across the capacitor and, through the mechanisms above, with the DC bias on which it rides. The relationship is not linear, because the permittivity and piezoelectric activity of ferroelectric ceramics themselves change with bias.
• Frequency: Frequencies in the audible range (20 Hz to 20 kHz) produce directly audible noise. On a supply rail, such frequencies usually come from the ripple of a converter in burst mode or from a load that pulses at an audible rate, rather than from the switching frequency itself.
• Mounting orientation and position: Where the capacitor sits relative to the board's supports and vibration modes, and how it is oriented, change how strongly it drives the board.
• PCB characteristics: The PCB construction, thickness, and support conditions affect how capacitor vibration is transmitted and radiated as sound.

Capacitor Noise Reduction

Strategies for reducing capacitor acoustic noise include:

• Use low-noise dielectrics: C0G/NP0 capacitors produce minimal noise and should be used where noise is a concern, if their limited capacitance range is acceptable. Some manufacturers also offer Class II formulations with reduced piezoelectric activity.
• Substitute a non-piezoelectric type: Where a large capacitance is needed at a problem location, a polymer tantalum or polymer aluminum capacitor, whose dielectric is not piezoelectric, can replace the MLCC.
• Choose different package styles: Package construction determines how strongly a capacitor couples to the board. Metal-frame (lead-frame) MLCCs and interposer-mounted parts lift the ceramic body off the PCB so that less of its motion reaches the board, and manufacturers market both specifically to reduce singing.
• Mirror the layout: Placing identical capacitors directly opposite each other on the top and bottom of the board makes their bending moments oppose and largely cancel, a remedy that costs nothing in components.
• Constrain capacitor motion: Applying a damping compound or underfill to the capacitor can reduce vibration. The material must be compatible with the operating temperature range.
• Control excitation: Reducing the AC voltage across noise-producing capacitors reduces noise, and so does a lower DC bias or a part whose capacitance changes less with bias. This may involve circuit topology changes, adding damping components, or keeping a converter out of burst mode.
• PCB design: Locating capacitors near PCB supports or edges reduces PCB motion. Using thicker or stiffer PCB construction reduces sound radiation. Avoiding placement near enclosure openings reduces perceived noise.
• Mount on the side: Some studies report lower noise when a capacitor is mounted so that its internal electrodes stand perpendicular to the board rather than parallel to it; the benefit depends on the board and is not universal.

Electromagnetic Sources of Motor Noise

Several electromagnetic mechanisms contribute to motor acoustic noise:

• Radial forces: The electromagnetic interaction between stator and rotor creates radial forces that vary with angular position and time. These forces cause radial vibration of the stator, which radiates as acoustic noise.
• Torque ripple: Variations in electromagnetic torque during rotation create torsional vibrations that can couple to radial modes and produce sound.
• Magnetostriction: Magnetic materials in the motor experience magnetostrictive strain as the magnetic field varies. This is similar to transformer magnetostriction but with more complex field patterns.
• Slot harmonics: The discrete distribution of windings in stator slots creates spatial harmonics in the magnetic field. These harmonics interact with rotor position to produce forces at frequencies related to rotational speed and slot count.
• PWM-related noise: Motors driven by PWM inverters carry current ripple at the carrier frequency and in sidebands around it. Industrial drives commonly switch at a few kilohertz, which places the carrier tone and its sidebands squarely in the audible range; raising the carrier above the audible band removes the direct tone at the cost of higher switching losses, although modulation effects can still produce audible components.

Motor Types and Noise Characteristics

Different motor types have different noise signatures:

• DC motors: Brush-type DC motors produce noise from brush commutation in addition to electromagnetic sources. Brushless DC motors eliminate brush noise but may have significant electromagnetic noise components, particularly with six-step (trapezoidal) commutation, which produces torque ripple at each commutation event.
• Induction motors: The rotating magnetic field in induction motors creates characteristic noise patterns related to pole count and rotor slot count. Rotor eccentricity creates unbalanced magnetic pull, and it and mechanical unbalance both increase noise.
• Permanent magnet motors: Strong permanent magnets create high magnetic forces that can produce significant noise, including cogging torque as the magnets pass the stator teeth. Proper rotor magnetization and stator design are essential for quiet operation.
• Switched reluctance motors: SRMs have highly pulsating torque and radial forces that tend to make them noisier than other motor types unless carefully designed.
• Stepper motors: Stepper motors produce noise at stepping frequency and harmonics. Microstepping reduces noise by smoothing the current waveform.

Motor Noise Reduction

Reducing motor electromagnetic noise involves motor design, drive design, and system integration:

• Motor design: Proper selection of pole and slot combinations minimizes problematic harmonics. Skewing the rotor slots reduces slot harmonics. Careful magnetic design minimizes radial force variations.
• Structural design: Motor housings should be stiff to have high resonant frequencies and should avoid resonances at expected excitation frequencies. Damping treatments can reduce resonant amplification.
• Drive optimization: Sinusoidal current control produces less torque ripple than six-step commutation. Raising the PWM carrier frequency moves its tone out of the audible band, and random or spread-spectrum PWM spreads the carrier's energy over a band, replacing a discrete whine with a quieter, less tonal hiss when the carrier must stay audible.
• Mounting: Resilient mounting isolates motor vibration from the supporting structure. The mount natural frequency should be well below the lowest problematic motor vibration frequency.
• Enclosure: Motor enclosures can include sound-absorbing materials to reduce radiated noise. Ventilation openings should be designed as acoustic baffles where possible.

Sources of Switching-Related Acoustic Noise

Switching power conversion creates acoustic noise through multiple mechanisms:

• Transformer and inductor noise: As discussed, magnetic components vibrate at switching frequency and harmonics, and at any audio-rate modulation of their current.
• Capacitor noise: Ceramic capacitors in switching circuits experience AC voltage that produces piezoelectric and electrostrictive vibration.
• PCB vibration: The board itself generates little force, but it is the main radiator for the vibration of the capacitors and magnetic components mounted on it, and a large, lightly supported board radiates far more than a small, stiff one.
• Mechanical resonance excitation: The broadband spectral content of switching transients and load steps can excite mechanical resonances in any system component, potentially producing noise at resonant frequencies that differ from the switching frequency.

Frequency Considerations

The relationship between switching frequency and audible noise is complex:

• Sub-20 kHz switching: Switching frequencies in the audible range produce directly audible tones. Consumer products generally avoid them, but industrial motor drives and high-power converters, where switching losses make low frequencies attractive, often accept them.
• 20 to 30 kHz switching: Frequencies just above the nominal upper limit of hearing are inaudible to most people, though some young listeners can detect the low end of this band at high levels. The larger risk is that component tolerances, light-load frequency foldback, or subharmonics pull operation down into the audible range. These frequencies are common in older power supply designs.
• Above 30 kHz switching: Higher switching frequencies move the fundamental further from audibility but increase switching losses. Modern power supplies often use frequencies from 50 kHz to several MHz.
• Subharmonics and modulation: Even with high switching frequencies, load variations, pulse skipping, burst mode operation, and other modulation effects can produce audible frequency components. A peak-current-mode converter operating above 50 percent duty cycle without adequate slope compensation can also fall into subharmonic oscillation at half its switching frequency, which for a 30 kHz converter lands at an audible 15 kHz.

Switching Noise Reduction

Strategies for reducing acoustic noise from switching power conversion:

• Frequency selection: Choose switching frequencies well above the audible range, accounting for subharmonics and modulation effects.
• Light-load mode control: Burst mode and pulse skipping save energy at light load but create audio-rate current envelopes. Some controllers offer a forced continuous-conduction mode or a mode that holds the minimum switching or burst frequency above the audible band, trading a little light-load efficiency for silence.
• Spread-spectrum techniques: Modulating the switching frequency over a range spreads the energy across a band rather than concentrating it at discrete frequencies. When the tone is in or near the audible band, this reduces its tonal prominence and perceived annoyance.
• Component selection: Choose magnetic components and capacitors with low acoustic noise characteristics.
• PCB design: Keep singing capacitors and magnetic components away from the middle of large unsupported board spans, add supports or stiffeners where they must sit, and use the mirror-placement technique for ceramic capacitors.
• Mechanical design: Secure components to prevent rattling. Use damping materials to reduce resonant amplification. Design enclosures to avoid acoustic resonances at switching-related frequencies.

Corona Formation and Characteristics

Corona occurs when the electric field intensity exceeds approximately 30 kV/cm (peak) in air at atmospheric pressure. This typically happens at sharp points, edges, or small-radius conductors where the electric field is concentrated:

• Field concentration: Electric field strength is inversely related to the radius of curvature. Sharp points and edges create field concentrations that can exceed the ionization threshold even when the average field is well below it.
• Ionization cascade: When the field exceeds the ionization threshold, free electrons accelerate and collide with neutral molecules, creating more free electrons. This cascade creates a plasma region around the discharge point.
• Polarity effects: Corona characteristics differ between positive and negative polarity. Negative corona in air forms regular, highly localized current pulses (Trichel pulses), whereas positive corona develops as burst discharges and streamers. On AC systems the polarity reverses each half-cycle, and the sharp streamer activity of the positive half-cycle is typically a major contributor to the characteristic crackle of high-voltage lines.
• AC corona: In AC systems, corona polarity alternates with the applied voltage. The acoustic noise has components at twice the power frequency and higher harmonics.

The 30 kV/cm figure is the breakdown gradient of air in a uniform field. On a smooth round conductor the surface gradient at onset is higher, because the ionization zone must extend some distance from the surface, and Peek's empirical law gives the peak onset gradient as

E_c = 30 mδ(1 + 0.301/√(δr)) kV/cm

where r is the conductor radius in centimeters, δ is the relative air density (1 at 25 degrees Celsius and 101.3 kPa), and m, at most 1, is a surface irregularity factor (1 for a smooth, polished surface). For a smooth conductor of 1 cm radius in standard air the onset gradient is about 39 kV/cm; for a wire of 1 mm radius, about 59 kV/cm. Scratches, burrs, contamination, and water drops lower m and bring the onset down sharply, and the lower air density at altitude lowers it too.

Acoustic Characteristics of Corona

Corona produces acoustic noise through several mechanisms:

• Direct acoustic emission: The plasma in the corona region heats and expands the surrounding air, creating pressure waves that propagate as sound.
• Electrostatic speaker effect: The charged particles in the corona region experience forces from the alternating electric field, moving the air directly and creating sound.
• Streamers and partial arcs: Intense corona can develop streamers that propagate away from the discharge point, creating crackling sounds.

The acoustic spectrum of corona includes broadband noise (hiss) and tonal components related to the power frequency. At higher voltages or with contaminated surfaces, the noise becomes more impulsive (crackling).

Corona Prevention and Control

Preventing corona involves maintaining electric field stress below the ionization threshold:

• Geometry design: Use rounded shapes with large radii of curvature. Avoid sharp edges and points. Use corona rings or shields on high-voltage terminals.
• Voltage grading: Distribute voltage stress uniformly along insulators and through insulation systems to avoid local field concentrations.
• Insulation: Encapsulate high-voltage components in solid or liquid insulation with higher breakdown strength than air. Ensure no voids or inclusions that could create internal corona.
• Environment control: Higher gas pressure increases breakdown strength. Using SF₆ or other high-dielectric-strength gases in sealed enclosures allows higher operating voltages without corona, although SF₆ is a potent greenhouse gas whose use in new switchgear is being restricted.
• Surface treatment: Keep insulator surfaces clean and dry. Contamination and moisture reduce surface breakdown strength and promote corona.

Arc Acoustic Mechanisms

Arcs generate sound through thermal-acoustic coupling:

• Rapid heating: The arc plasma reaches temperatures of several thousand kelvins, and more than 10,000 K in the core of a high-current arc, rapidly heating the surrounding air and creating a pressure wave that propagates as sound.
• Arc instability: Arcs are inherently unstable, with the plasma column moving and fluctuating in intensity. These variations produce time-varying acoustic output.
• Contact separation dynamics: In switches and circuit breakers, the arc forms and evolves as contacts separate. The changing arc length and dynamics produce a characteristic acoustic signature.
• Magnetic forces: The arc current creates magnetic fields that interact with the arc plasma, causing motion and oscillation that contribute to acoustic noise.

Arc Noise in Electrical Equipment

Different equipment types produce characteristic arc noise:

• Relays and contactors: Contact arcing during opening and closing produces brief noise bursts. Severe arcing indicates worn or misaligned contacts.
• Circuit breakers: Interrupting high currents creates intense arcing that produces significant acoustic energy. Arc chutes and other arc-extinguishing features affect the noise characteristics.
• Arc welding: Welding arcs produce sustained acoustic noise with spectral content related to welding parameters and electrode motion.
• Arc faults: Unintentional arcing from damaged insulation or loose connections emits ultrasound as well as radio-frequency noise. Ultrasonic detectors use that emission to locate arcing and tracking during switchgear inspections, while arc-fault circuit interrupters recognize the arc from its electrical signature.

Arc Noise Considerations

Managing arc noise involves both preventing arcs where unwanted and controlling necessary arcs:

• Arc suppression: In switches and contactors, arc suppression networks (snubbers) reduce arc duration and intensity, reducing both electrical interference and acoustic noise.
• Solid-state switching: Replacing electromechanical switches with solid-state switches eliminates contact arcing and associated noise.
• Enclosure design: Equipment with unavoidable arcing (circuit breakers, arc welders) should have enclosures designed to contain both the arc and the resulting acoustic noise.

Motor Contributions to Fan Noise

The motors driving cooling fans produce electromagnetic noise as described in the motor noise section:

• Radial magnetic forces: Variations in the magnetic force between stator and rotor cause vibration that couples into the fan structure and radiates as sound.
• PWM drive effects: Fans driven by PWM speed controllers experience current ripple that can produce motor noise and excite structural resonances. Chopping a three-wire fan's supply at a low frequency is a common cause of ticking or clicking.
• Cogging torque: Variations in motor torque with rotor position create pulsating forces that can excite fan blade vibrations.

EMI Effects on Fan Control

Fan speed control systems can be affected by electromagnetic interference:

• Control loop disturbance: EMI affecting the fan speed control loop can cause speed variations that produce audible noise modulation.
• PWM timing errors: Interference with PWM generation can create irregular switching that produces acoustic artifacts.
• Sensor interference: Hall sensors or back-EMF sensing for motor control can be affected by external magnetic fields or conducted interference, affecting fan behavior.

Fan Noise Reduction

Reducing electromagnetic contributions to fan noise:

• Motor selection: Choose fan motors with low electromagnetic noise characteristics. Consider brushless DC motors with sinusoidal drive for minimum noise.
• PWM frequency: Use PWM frequencies well above the audible range. The four-wire fan interface specified by Intel for computers carries a separate 25 kHz control signal to the fan's own drive electronics for this reason. Avoid frequencies that coincide with fan blade resonances.
• EMI filtering: Ensure fan drive circuits have adequate filtering to prevent conducted interference from affecting control behavior.
• Mounting: Use resilient mounting to isolate motor vibration from the equipment structure.

Frequency Weighting and Perception

Human hearing sensitivity varies with frequency:

• A-weighting: The A-weighted sound level (dBA) is the most common single-number noise metric. The A curve is derived from the 40-phon equal-loudness contour, so it de-emphasizes low and very high frequencies roughly as the ear does for quiet sounds. It discounts a 100 Hz hum by about 19 dB, which is why a transformer can be clearly audible and annoying while its A-weighted reading looks modest.
• Loudness: Perceived loudness is measured in sones, computed from the spectrum by methods such as ISO 532-1 (the Zwicker method). One sone is the loudness of a 1 kHz tone at 40 dB SPL, and above about 40 phons each increase of 10 phons doubles the loudness in sones.
• Tonal content: Pure tones or narrow-band noise are perceived as more annoying than broadband noise of the same overall level. Standards for information technology equipment, such as ECMA-74 and ECMA-418-1, specify tone-to-noise ratio and prominence ratio procedures for deciding whether a tone is prominent, and environmental noise assessments add a tonal adjustment to the rated level.
• High-frequency sensitivity: Young people can hear frequencies above 15 kHz that older adults cannot. Products sold to diverse audiences must consider this variation in hearing ability.

Sound Quality Considerations

Beyond simple loudness, the character of sound affects its acceptability:

• Temporal variations: Steady sounds are generally less annoying than fluctuating sounds. Modulated noise (such as from variable-speed fans or PWM artifacts) can be more objectionable than steady noise of the same average level. Slow modulation, below about 20 Hz and most strongly near 4 Hz, is perceived as fluctuation strength.
• Impulsiveness: Impulsive sounds (clicks, pops, thumps) are perceived as more annoying than steady sounds. Relay actuation and contact arcing create impulsive noise that draws attention.
• Roughness: Amplitude modulation at frequencies from 15 Hz to 300 Hz creates a rough quality that increases perceived annoyance.
• Sharpness: High-frequency content increases the sharpness of sound, measured in acum, which is perceived as more annoying. Inductor whine often has high sharpness.

Application-Specific Requirements

Noise requirements vary greatly by application:

• Office equipment: Products used in quiet office environments face demanding noise targets. ISO 7779 and ECMA-74 specify the measurement methods for IT equipment, ISO 9296 and ECMA-109 specify how the results are declared, and ecolabels such as the Blue Angel set declared sound power limits.
• Home electronics: Consumer expectations for home equipment are high, particularly for devices used in bedrooms or living rooms.
• Medical equipment: Noise can interfere with patient communication, rest, and staff concentration. The general safety standard for medical electrical equipment, IEC 60601-1, limits the acoustic energy to which patients and operators are exposed.
• Industrial equipment: Higher noise levels may be acceptable in factory environments, but must not exceed occupational noise exposure limits, such as the 90 dBA eight-hour permissible exposure limit in the United States (29 CFR 1910.95) and the exposure action values of 80 and 85 dBA in EU Directive 2003/10/EC.
• Transportation: Vehicle noise standards and customer expectations drive noise requirements for automotive and aerospace electronics.

Sound Measurement Fundamentals

Key concepts in acoustic measurement:

• Sound pressure level: SPL in decibels is the logarithmic ratio of the RMS sound pressure p to a reference pressure p_ref = 20 µPa: SPL = 20 log₁₀(p/p_ref) dB. A pressure of 1 Pa is therefore 20 log₁₀(50,000) ≈ 94 dB, the level produced by many acoustic calibrators.
• Sound power level: The total acoustic power W radiated by a source, independent of measurement distance and room acoustics, expressed as L_W = 10 log₁₀(W/W₀) dB with W₀ = 1 pW (10⁻¹² W). Calculating sound power from pressure measurements requires knowledge of the measurement environment. Declared values for IT equipment are often quoted in bels, so 4.5 B is 45 dB.
• Adding levels: Levels from independent sources add on an energy basis, not arithmetically. Two equal, uncorrelated sources together are 10 log₁₀(2) ≈ 3 dB louder than either alone, and a source 10 dB below another adds only about 0.4 dB to the total.
• Frequency weighting: A-weighting (dBA) is most common for general noise assessment. C-weighting (dBC) is nearly flat across most of the audible range, falling by 3 dB only at about 31.5 Hz and 8 kHz, and is used for peak sound pressure measurements and to reveal low-frequency content. Z-weighting is flat.
• Time weighting: Fast (125 ms time constant), slow (1 s), and impulse (35 ms rise, 1.5 s decay) time weightings affect how transient sounds are captured. The equivalent continuous level, L_eq (L_Aeq when A-weighted), is the level of a steady sound carrying the same energy over the measurement period.

Measurement Environment

The acoustic measurement environment significantly affects results:

• Anechoic chambers: Fully anechoic rooms have negligible reflections and allow direct measurement of radiated sound. They are expensive and may not represent real operating conditions.
• Semi-anechoic chambers: These rooms have absorptive walls and ceiling but a reflective floor, providing a free field above a reflecting plane that matches equipment standing on a floor. They are commonly used for equipment noise testing under ISO 3744 and ISO 3745.
• Reverberation rooms: Highly reflective rooms are used for sound power measurement under ISO 3741, either directly, using the room's measured reverberation time, or by comparison with a calibrated reference sound source. The diffuse sound field allows total sound power to be determined from spatially averaged pressure measurements.
• Real environments: Measurements in real operating environments provide relevant data but are affected by background noise, reflections, and other sources that complicate interpretation. Sound intensity methods (ISO 9614) can determine sound power in place despite steady background noise.

Locating Electromagnetic Noise Sources

Finding which component on a crowded board is responsible is often the hardest part of solving an acoustic noise problem:

• Spectrum matching: A spectrum of the microphone signal gives the frequency of the offending tone, which can be compared with the switching frequency, the burst or pulse-skipping rate, the load's repetition rate, and twice the mains frequency. A tone that follows a change of load or operating mode points directly to its electrical cause.
• Close listening: A small probe microphone, or a length of nonconductive tubing used as a stethoscope, held near individual components localizes the source on the board. Hazardous voltages demand the usual precautions.
• Vibration measurement: Laser Doppler vibrometers measure board velocity without contact and map which areas of the board move; they are widely used to study capacitor singing.
• Sound intensity and arrays: Intensity probes and microphone arrays show where sound leaves the product, separating true sources from panels that merely radiate their vibration.
• Electrical correlation: Capturing the inductor current or rail ripple on an oscilloscope alongside the microphone signal shows whether the acoustic envelope follows the electrical one.

Standards and Specifications

Various standards govern acoustic noise measurement and limits:

• ISO 7779: Standard for measuring airborne noise from information technology and telecommunications equipment.
• ECMA-74: Measurement of airborne noise emitted by information technology and telecommunications equipment, corresponding to ISO 7779, with an annex on identifying prominent discrete tones.
• ISO 9296 and ECMA-109: Declaration of the noise emission values of information technology and telecommunications equipment.
• ECMA-418: Psychoacoustic metrics for such equipment: Part 1 covers prominent discrete tones, and Part 2 models perceived tonality and roughness.
• ISO 3741, ISO 3744, and ISO 3745: General methods for determining sound power in reverberation rooms, over a reflecting plane, and in anechoic and hemi-anechoic rooms.
• IEC 61672: Specification for sound level meters, defining performance classes and frequency and time weightings.
• IEC 60076-10: Determination of the sound levels of power transformers and reactors.
• Product-specific standards: Many product categories have specific noise limits and measurement procedures defined in relevant product standards.