Electronics Guide

Inductive Load Switching

Opening a circuit that carries current through an inductance generates one of the most severe switching transients. The inductance opposes any change in current according to v = L di/dt, so forcing the current toward zero in a short time drives the voltage across the opening switch upward. In an ideal circuit with a perfect switch, the voltage would rise without bound. In practice it rises only until the circuit finds another path: the switch arcs over, insulation breaks down, a semiconductor avalanches, or a protective device clamps the voltage.

The magnetic energy stored at the instant of interruption, E = ½LI², must go somewhere. A 100 millihenry relay coil carrying 200 milliamperes stores only 2 millijoules, yet that energy suffices to strike an arc and erode contact plating over many operations. Common contact metals sustain a metallic-phase arc only above a material-dependent threshold of roughly 10 to 15 volts across the gap and a few hundred milliamperes through it; silver and gold alloys are usually quoted near 12 volts and 0.4 ampere, and silver cadmium oxide near 10 volts. Circuits that operate below those thresholds erode their contacts far more slowly, which is why low-level signal relays specify gold-plated contacts and dry-circuit ratings.

As relay or contactor contacts separate, the widening gap withstands progressively more voltage, so a single opening operation usually produces not one arc but a rapid succession of them. Each restrike collapses the gap voltage in nanoseconds, and the resulting burst of steep pulses, known as a showering arc, carries spectral content into the hundreds of megahertz. This mechanism is the physical origin of the electrical fast transient phenomenon and of the burst waveform that laboratory immunity testing uses to reproduce it.

Mechanical relay and contactor switching of inductive loads is therefore a particularly demanding application. Alternating current assists interruption because the current passes through zero twice per cycle, giving the arc a natural opportunity to extinguish; contactors experience the worst stress when the contacts part near the current maximum, because the stored energy is then greatest. Direct current offers no such zero crossing, which makes direct current inductive switching considerably harder than the alternating current equivalent at the same voltage and current. Vacuum interrupters add a further effect called current chopping: they force the current to zero slightly ahead of the natural zero crossing, and the interrupted current then transfers to the stray capacitance of the load circuit, producing an overvoltage approximately equal to the chopped current multiplied by the surge impedance of that circuit, which is the square root of its inductance divided by its capacitance. Modern chromium-copper contacts chop at only a few amperes, but a high surge impedance can still turn that small current into a substantial voltage.

The practical remedy is to give the stored energy a deliberate path. A freewheeling diode across a direct current relay coil clamps the reverse voltage to roughly one diode drop above the supply, but it also lets the coil current decay slowly, lengthening armature release time and, in some designs, increasing erosion of the contacts that the relay itself operates. Adding a Zener diode or a resistor in series with the freewheeling diode raises the clamp voltage and shortens the decay in proportion, trading device stress against release time. Bidirectional transient voltage suppressors and metal oxide varistors serve alternating current coils, where a simple rectifier cannot be used, and a resistor-capacitor network placed directly across the contacts limits the rate of voltage rise that would otherwise restrike the arc.

Capacitive Load Switching

Closing a switch onto a discharged capacitor creates the mirror-image transient. The current into a capacitance follows i = C dv/dt, and at the instant of closure only the circuit resistance and the loop inductance limit it. Where resistance dominates, the current decays exponentially with the RC time constant as the capacitor voltage approaches the source voltage. Where resistance is small, the loop inductance dominates instead: the current rings at the resonant frequency of the loop, and its first peak approaches the applied voltage divided by the surge impedance of the loop rather than the value that resistance alone would allow.

Inrush current when energizing capacitive loads can trip overcurrent protection, weld switch contacts, and create voltage sags that affect other loads on the same supply. Energizing a shunt capacitor bank in a distribution system produces an oscillatory transient as the bank exchanges energy with the source inductance. With no trapped charge on the bank, the peak cannot exceed twice the crest of the system voltage, and losses, load, and other system capacitance typically hold it to roughly 1.2 to 1.8 times nominal, ringing at a few hundred hertz to about one kilohertz. Back-to-back switching, meaning energization of one bank while an adjacent bank remains in service, is far more severe: the inrush flows between the two banks through a small inductance, reaching much higher peaks at frequencies of several kilohertz. Circuit breaker application ratings therefore limit the permissible combination of inrush peak and inrush frequency, and utilities apply pre-insertion resistors or inductors, damping reactors, or synchronous closing controls that close each pole near a voltage zero.

A capacitor switching transient on the utility system can also be magnified inside a customer facility. When a lower-voltage capacitor bank and the transformer that feeds it resonate near the frequency of the incoming transient, the peak seen at the low-voltage bus is considerably larger than the event on the utility side. Nuisance tripping of the direct current bus overvoltage protection in adjustable-speed drives is the classic symptom, and it commonly appears at a fixed time of day when the utility switches its banks on a schedule.

At equipment scale, the bulk input capacitor of a switch-mode power supply presents the same problem in miniature: connecting a discharged capacitor across the rectified line draws a peak limited only by the line impedance and the loop inductance. An inrush limiter holds that first peak to a level the connector, fuse, and upstream breaker tolerate, usually a negative temperature coefficient thermistor in low-cost designs or a resistor bypassed by a relay or triac once the capacitor has charged.

Repetitive capacitive switching, as occurs in switched-mode power supplies and motor drives, creates conducted emissions that can violate electromagnetic compatibility limits. Each switching event injects a current pulse into the supply network, and the spectral content of those pulses follows from the switching frequency and the edge rate rather than from the average power drawn. Because commercial conducted emission limits cover 150 kilohertz to 30 megahertz, converters that switch in the tens to hundreds of kilohertz place their fundamental and many harmonics squarely inside the measured band, which makes input filtering a design requirement rather than an afterthought.

Power Electronics Switching

Modern power electronic converters rely on high-frequency switching to achieve efficient power conversion, and that switching generates transients that must be managed rather than merely tolerated. Silicon MOSFETs and IGBTs complete a transition in tens to hundreds of nanoseconds; silicon carbide and gallium nitride devices do so in a few nanoseconds to a few tens of nanoseconds, with voltage slew rates reaching tens of kilovolts per microsecond. The device never acts alone, however. The parasitic inductance of the commutation loop and the output capacitance of the switch together determine the voltage and current waveforms that actually appear.

Turn-off overshoot follows directly from that loop inductance. Interrupting 100 amperes in 50 nanoseconds through a commutation loop of 20 nanohenries adds 40 volts to the applied voltage, and the same inductance then rings against the output capacitance of the switch at a frequency of tens to hundreds of megahertz. Those resonances appear directly as peaks in the conducted and radiated emission spectrum, which is why an emission scan of a converter often identifies the offending loop before any probe touches the board. Minimizing loop area, choosing low-inductance package styles, and placing the direct current link capacitor immediately adjacent to the switch are the primary countermeasures. Slowing the switch is the fallback, and excessively fast switching can produce larger transients and worse emissions than a moderately slower transition, despite the lower switching loss.

Turn-on presents complementary challenges. The incoming switch must absorb the reverse recovery current of the freewheeling diode in addition to the load current, because the stored minority carriers in the diode must be extracted before it can block. Silicon carbide Schottky diodes store essentially no minority charge and remove that component entirely, which is one reason they entered hard-switched boost stages long before silicon carbide transistors became common. In synchronous rectifiers the body diode of the complementary MOSFET recovers instead, so the dead time between devices must be long enough to prevent shoot-through yet short enough to limit body diode conduction and its recovery penalty.

Fast voltage transitions also drive displacement current through every parasitic capacitance that couples the switching node to earth: transistor tab to heat sink, transformer primary to secondary, motor winding to frame. One picofarad of coupling at 20 kilovolts per microsecond passes 20 milliamperes, so the few tens of picofarads of a typical insulated heat sink mounting are enough to dominate common-mode emissions above a few megahertz. Isolated gate drivers face the same slew rate from the other side of the barrier, which is why their datasheets specify a common-mode transient immunity figure, commonly on the order of 100 kilovolts per microsecond for devices intended to serve wide-bandgap switches.

Digital Logic Switching

Every transition in a digital circuit is a switching event. Contemporary CMOS processes produce on-chip edges of a few hundred picoseconds, and even mainstream board-level logic families switch in one to two nanoseconds. The measure that matters for interference is the edge rate rather than the clock frequency: significant spectral content extends to approximately 1/(πt_r), so a 500 picosecond edge carries meaningful energy beyond 600 megahertz whether the signal repeats at 10 megahertz or at 500 megahertz.

During each transition, both the NMOS and PMOS transistors of a CMOS stage conduct briefly, producing a crowbar or shoot-through current from supply to ground. This adds to the generally larger current required to charge and discharge the load capacitance. Both components flow through the inductance of on-die interconnect, bond wires or bumps, package leads, vias, and board traces. A single via of roughly one nanohenry carrying a 50 milliampere transition in one nanosecond develops 50 millivolts, which is negligible on its own but multiplies by the number of drivers that switch together.

The voltage developed across these parasitic inductances appears as simultaneous switching noise, also called ground bounce, and it disturbs not only the switching gate but every circuit sharing the same supply and return. Wide parallel buses that change many bits at once represent the worst case, which is one reason high-speed interfaces migrated to differential serial signaling whose supply current stays nearly constant. Designers respond with a hierarchy of decoupling: on-die capacitance answers the fastest edges, package capacitance the next slower ones, and board capacitors the remainder, with each board capacitor useful only up to its series resonant frequency, above which its own inductance dominates.

Clock distribution in synchronous systems creates repetitive transients at the clock frequency and its harmonics. Large numbers of flip-flops clocking simultaneously draw substantial current pulses that must be supplied locally, and the resulting emissions often dominate the electromagnetic signature of a digital product, appearing as discrete spectral lines at multiples of the clock. Spread-spectrum clocking modulates the clock frequency by a fraction of a percent so that harmonic energy spreads across a band wider than the resolution bandwidth of the measuring receiver, lowering the measured amplitude by several decibels at low harmonic orders and by more at high orders. The technique redistributes energy rather than removing it, and the added jitter is unacceptable in some synchronous links, so designers apply it selectively alongside sound decoupling and layout.

Snubber Circuits

Snubber circuits provide controlled paths for transient energy, protecting switches and limiting electromagnetic emissions. The classic snubber places a resistor and capacitor in series, connected either across the switch or across the load. Across opening contacts, the capacitor accepts the inductive current while the contacts part and, more importantly, slows the rate of voltage rise across the gap so that it does not restrike; the series resistor limits the capacitor discharge current when the contacts close again. Across a semiconductor switch, the same network diverts current during the turn-off transition and damps the ringing between the commutation loop inductance and the device output capacitance.

Snubber design balances several objectives at once. Matching the resistance to the surge impedance of the circuit being damped, meaning the square root of the loop inductance divided by the total capacitance, gives close to critical damping and suppresses the ringing in roughly one cycle. The capacitance must be large enough to shape the transition, commonly two to four times the output capacitance of the device being protected, yet small enough to keep the discharge current and the dissipation within bounds.

Dissipation is the price. A snubber capacitor that charges and discharges once per switching cycle dissipates CV²f in the resistor, and that result is independent of the resistance value. One nanofarad at 400 volts switching at 100 kilohertz therefore dissipates 16 watts, a figure that explains why simple resistor-capacitor snubbers suit line-frequency contactors and low-frequency converters but become impractical as switching frequency rises. High-frequency converters instead use clamp networks, such as the resistor-capacitor-diode clamp that captures leakage-inductance energy in a flyback converter, or avoid the problem through soft switching.

Nondissipative, or energy-recovery, snubbers return the captured energy to the source or the load rather than converting it to heat. These circuits use auxiliary switches and magnetic components to transfer the transient energy back into the system. They eliminate the thermal burden of dissipative snubbers in high-power or high-frequency designs, but the added components, the auxiliary control, and the extra failure modes must be justified by the efficiency and performance gained.

Soft Switching Techniques

Soft switching modifies the switching trajectory so that voltage and current do not overlap during the transition. Zero-voltage switching turns a device on when the voltage across it has already fallen to zero, eliminating turn-on loss and the associated current spike; a resonant transition, often assisted by conduction of the device body diode just before turn-on, creates that condition. Zero-current switching turns a device off when its current has already reached zero, which suits IGBTs because it avoids dissipating the tail current. Both approaches reduce the dv/dt and di/dt that drive emissions, and they attack the transient at its source rather than damping it afterward.

Resonant and quasi-resonant converters use inductor-capacitor networks to establish these conditions. The LLC resonant converter, now widespread in high-density power supplies and offline adapters, maintains zero-voltage switching across a wide load range and permits switching frequencies of several hundred kilohertz with acceptable loss. Soft switching is not free, however: resonant operation circulates additional current that raises conduction loss, and it typically increases the peak voltage or current stress on the devices. The resonant network must be designed to preserve soft switching across the full operating range, because a converter that loses zero-voltage switching at light load can emit more than the hard-switched design it replaced.

Edge-rate control offers a simpler partial alternative. Gate resistors for MOSFETs and IGBTs set the rate at which gate charge is delivered and therefore the switching speed, and modern gate drivers frequently provide programmable slew rate or multi-level gate drive that shapes the turn-on and turn-off segments independently. Slower edges reduce dv/dt, di/dt, and high-frequency content, often by an amount that decides compliance, but they increase switching loss and shift the burden to the thermal design. The optimum is application specific and is best found by measuring emissions and efficiency together rather than by optimizing either alone.

Motor Starting Transients

Electric motor starting creates significant switching transients in both the motor supply and the connected power system. Induction motors started directly across the line draw locked-rotor currents of roughly six to eight times rated current until the rotor approaches operating speed, and the first half cycle peaks higher still because of the direct current offset present in an asymmetric transient. The resulting voltage dip may disturb other loads, trigger undervoltage protection, or produce visible light flicker; flicker limits in standards such as IEC 61000-3-3 and IEEE Std 1453 constrain how often and how abruptly a load may draw such current. Large motors therefore require starting methods that limit inrush while still developing enough accelerating torque.

Reduced-voltage starting limits inrush by applying less than full voltage during acceleration, and because torque falls with the square of voltage, every such method trades starting torque for current. A star-delta starter reduces the winding voltage by a factor of the square root of three and therefore reduces both starting current and starting torque to roughly one third of the direct-on-line values; its open-transition changeover reconnects the motor out of phase with its residual flux and can momentarily draw more current than the original inrush, which is why closed-transition designs are preferred where the supply is weak. Autotransformer starters provide selectable taps with a similar trade. Solid-state soft starters use thyristor phase control to ramp the voltage over several seconds, limiting both electrical and mechanical stress, but the phase-controlled waveform is rich in harmonics during the start interval and may require filtering in sensitive installations.

Variable frequency drives provide the most complete control by varying voltage and frequency together. Holding the volts-per-hertz ratio roughly constant while ramping frequency from near zero allows controlled acceleration at close to rated current. The inverter output, however, is a train of steep-edged pulses at a carrier frequency of a few kilohertz to about 16 kilohertz, and the motor cable behaves as a transmission line. When the one-way propagation delay of the cable exceeds roughly half the pulse rise time, the pulse reflects from the high-impedance motor terminals and nearly doubles; for the 0.1 microsecond rise times typical of IGBT inverters, drive manufacturers commonly cite runs beyond about 15 meters as the threshold. NEMA MG 1 Part 31 addresses the consequence by requiring the stator insulation of definite-purpose inverter-fed motors rated 600 volts or less to withstand line-to-line peaks of up to 3.1 times rated voltage with rise times of 0.1 microsecond or greater, which the standard tabulates as 1,431 volts for a 460 volt motor. Output reactors, dv/dt filters, sine-wave filters at the drive, and terminating networks at the motor all reduce the overshoot.

The same inverter also imposes a common-mode voltage that couples through the stray capacitance between stator winding and rotor, producing a shaft voltage that discharges through the bearing lubricant film once it exceeds the dielectric strength of that film. The resulting electrical discharge machining pits the bearing races and shortens bearing life, sometimes within months. Insulated bearings, conductive shaft grounding rings, symmetrical shielded motor cable bonded with a low-impedance connection at both ends, and common-mode chokes at the drive output are the established countermeasures.

Contact Bounce

Mechanical switch and relay contacts do not close or open cleanly in a single transition. The contacts strike, separate, and strike again several times before settling. Each bounce is a switching transient in its own right, generating a burst of high-frequency interference and potentially causing multiple false triggers in connected digital circuits. Bounce is generally more pronounced in light, fast-acting switches and less troublesome in heavy industrial contactors with high contact pressure.

The bounce interval depends strongly on the individual part, so a designer should measure it rather than assume it. Published bench measurements of a sample of common pushbuttons and toggle switches found an average bounce near 1.5 milliseconds and a typical worst case near 6 milliseconds, but one pushbutton in the sample bounced for roughly 157 milliseconds on release, and nominally identical parts differed from one another by a factor of two. Debounce intervals in the range of a few milliseconds to a few tens of milliseconds cover most parts, provided the interval is checked against the switch actually used.

Hardware debouncing typically combines a resistor-capacitor filter with a Schmitt trigger input, so that the filter slows the edge and the hysteresis restores a clean transition. The time constant must span the bounce interval of the specific switch without making the response feel sluggish. A set-reset latch offers an alternative that responds to the first valid edge and ignores the rest, although it requires a changeover contact that presents separate normally open and normally closed terminals.

Software debouncing samples the switch state periodically and accepts a change only after the input has held its new value for a defined number of samples, or after a fixed delay following the first detected edge. This approach eliminates external components and adapts easily, but it consumes processor attention and introduces latency between the physical action and the system response. Where the response must be immediate, such as in a safety interlock or a high-resolution timing measurement, hardware debouncing or a switch technology with inherently low bounce remains necessary.

Standardized Test Waveforms

Because switching transients vary so widely, standards bodies define a small set of representative waveforms so that equipment immunity can be specified and compared. Three of them dominate commercial practice, and military practice adds its own.

IEC 61000-4-4, in its third edition of 2012, defines the electrical fast transient and burst test that reproduces the showering arc of an inductive switching event. The generator delivers pulses with a 5 nanosecond rise time and a 50 nanosecond width into 50 ohms, repeated at either 5 kilohertz or 100 kilohertz. Those pulses arrive in bursts lasting 15 milliseconds at the 5 kilohertz rate or 0.75 milliseconds at 100 kilohertz, with a burst repeating every 300 milliseconds. Preferred test levels run from 0.5 to 4 kilovolts for power and earth ports and from 0.25 to 2 kilovolts for signal and control ports; the standard couples the pulses onto power lines through 33 nanofarad capacitors and onto signal lines through a capacitive coupling clamp.

IEC 61000-4-5, third edition of 2014 with its 2017 amendment, defines the surge test for the higher-energy, slower events produced by lightning and by heavy power system switching. Its combination wave generator produces a 1.2/50 microsecond voltage waveform into an open circuit and an 8/20 microsecond current waveform into a short circuit, giving an effective source impedance of 2 ohms, raised to 12 or 42 ohms when additional coupling resistors are used for particular port types. IEEE Std C62.41.2-2002 characterizes the same low-voltage environment in North American practice and adds the 0.5 microsecond, 100 kilohertz ring wave, which represents the oscillatory response of building wiring to a switching event; the standard assigns different source impedances to its location categories, for example roughly 30 ohms for a 6 kilovolt, 200 ampere Category A exposure and roughly 12 ohms for a 6 kilovolt, 500 ampere Category B exposure.

MIL-STD-461H, issued on April 17, 2026, superseding MIL-STD-461G, addresses switching transients through requirement CS115, which injects a pulse of 30 nanosecond width with 2 nanosecond rise and fall times onto cable bundles at an amplitude of 5 amperes, repeated at 30 hertz for one minute. The standard states in its appendix that the 2 nanosecond rise time reflects the transients that interrupting inductive devices can produce, and that CS115 was introduced to replace the earlier, poorly repeatable chattering-relay test of MIL-STD-461C. Requirement CS116 complements it by applying damped sinusoidal transients from 10 kilohertz to 100 megahertz, with compliance demonstrated at 0.01, 0.1, 1, 10, 30, and 100 megahertz.

Switching Noise Mitigation

Comprehensive control of switching transients requires attention at several levels. At the component level, snubbers and transient suppression devices protect sensitive circuits and limit emission amplitude at the source. Device selection matters as much as the suppression added around it, because a slower device generates a milder transient at the cost of higher switching loss. The balance among switching speed, efficiency, thermal design, and electromagnetic compatibility must be settled deliberately for each application rather than inherited from a reference design.

Circuit board layout determines how much of the transient escapes. Minimizing the area of loops that carry high di/dt reduces both radiated emission and the inductive voltage that appears across the switch. Decoupling capacitors placed immediately at the switching device supply local energy and keep transient current out of the wider power distribution. Deliberate return-path planning ensures that transient currents flow where the designer intends rather than through the reference of a sensitive analog or clock circuit, since the return current follows the path of least impedance, which above a few megahertz means the path directly beneath the outgoing conductor.

System-level measures include filtering at every port so that transients neither leave nor enter the enclosure, shielded cable with low-impedance terminations to limit radiated coupling, and isolation to break the loops that would otherwise carry switching current between units. Synchronizing multiple converters to a common clock, or deliberately interleaving them, prevents the worst-case alignment in which several transients coincide. In critical installations, isolation transformers, line filters, and uninterruptible power supplies add further barriers against transients arriving from the premises wiring.

No single measure controls switching transients, because the energy originates in the reactance of the circuit itself. The practical objective is management rather than elimination: give the stored energy a defined path, keep the loops that carry fast current small, slow the transition only as much as the loss budget permits, and then verify the outcome against the standardized waveforms that represent the environment in which the equipment must operate.

Related Topics

Switching transients connect to many aspects of electronic design and electromagnetic compatibility: