A switched-mode power supply (SMPS) is an electronic system that converts electrical power by repeatedly switching semiconductor devices between conducting and nonconducting states. Energy passes through inductors, transformers, capacitors, or combinations of these components, while a control circuit adjusts the switching pattern to obtain the required output voltage or current. Switched-mode supplies are a major application of power electronics and appear in equipment ranging from miniature sensors and mobile-phone chargers to computers, telecommunications installations, and industrial machinery.
Their principal advantages are high efficiency and the possibility of using relatively small magnetic components. Unlike a linear regulator, which controls voltage by continuously dissipating excess electrical power in a regulating device, an SMPS ideally transfers energy through switches that dissipate little power in either steady state. Real switches nevertheless incur conduction and switching losses, and practical supplies must manage electrical noise, heat, insulation, transient behavior, and fault conditions.
The term usually denotes a supply providing regulated direct-current power, whether its input is alternating current or direct current. The underlying switching methods also appear in motor drives, battery chargers, and bidirectional converters, although those systems are commonly classified by their particular function rather than simply called power supplies.
Operating principles and energy conversion
An SMPS does not obtain an intermediate output voltage by holding its main regulating device continuously in a partially conducting state. Instead, it applies a sequence of circuit configurations, each lasting part of a switching cycle. An inductor limits the rate of current change, while a capacitor limits the rate of voltage change. Together with the load, these components convert pulsed electrical quantities into an output with a comparatively small alternating component, known as ripple.
For a switching frequency , the switching period is . If a switch conducts for a time during each period, its duty ratio is
Changing , the switching frequency, or the relative timing of several switches changes the energy transferred per cycle. In many supplies the switching frequency lies between tens of kilohertz and several megahertz, although the appropriate range depends strongly on power level, topology, semiconductor technology, and magnetic-component design. The output is not necessarily generated by duty-ratio control: resonant converters, for example, often regulate primarily by changing frequency.
The governing relationships include for an ideal inductor and for an ideal capacitor. In periodic steady state, an inductor’s current must return to its starting value after each complete cycle, so its average voltage is zero. Similarly, a capacitor’s net charge change over a cycle is zero. These volt-second and charge-balance conditions provide a direct way to derive many converter voltage relationships.
Efficiency is defined by
The difference becomes heat or other unwanted losses. A supply delivering 100 W at 90% efficiency draws approximately 111 W and dissipates about 11 W. At 95% efficiency, it dissipates approximately 5.3 W. Thus, a seemingly small percentage-point improvement can substantially reduce cooling requirements.
A switching supply is not inherently regulated, isolated, or unidirectional. Those properties result from its circuit and control system. An unregulated converter may merely provide a fixed approximate conversion ratio; a regulated supply adjusts its operation through feedback. Likewise, high efficiency does not imply that every SMPS is more efficient than every linear supply. At very light loads, control-circuit consumption and switching overhead can dominate, while a linear regulator operating with a very small input–output voltage difference may incur little loss.
Non-isolated converter topologies
The simplest switching regulators are forms of DC–DC converter in which input and output share an electrical reference. Their basic circuits illustrate how switching and energy storage produce step-down, step-up, or polarity-inverting conversion. The familiar ideal voltage-ratio equations assume steady-state operation, negligible losses, and continuous inductor current; practical limits and other conduction modes modify these relationships.
A buck converter produces an output below its input. During the main switch’s on-time, the source supplies the load and raises the inductor current. During the off-time, current continues through a diode or a second controlled switch. For an ideal buck converter,
A 12 V input can therefore produce 3 V at an ideal duty ratio of 0.25. The output inductor carries relatively continuous current, but the input current is strongly pulsed, making local input decoupling important. Buck converters are widely used for intermediate supply rails and low-voltage electronic loads.
A boost converter produces an output above its input. When its main switch conducts, the inductor stores energy and the output capacitor supports the load. When the switch opens, the inductor drives current through a rectifier into the output at a voltage above the source voltage. Its ideal relationship is
The equation suggests an unlimited voltage increase as approaches one, but real conversion is constrained by resistance, switching losses, current limits, and the diminishing interval available to deliver energy to the output. A conventional boost circuit also has an input-to-output conduction path through its inductor and diode, so stopping switching does not necessarily disconnect its output.
The basic inverting buck–boost converter can produce an output magnitude above or below the input, but with reversed polarity:
Non-inverting buck–boost circuits, including four-switch arrangements, serve applications in which a battery voltage moves above and below the required output. Other configurations, such as SEPIC and Ćuk converters, provide different combinations of voltage polarity, current continuity, and component stress. Flying-capacitor circuits form another family, transferring charge among capacitors rather than relying primarily on magnetic energy storage.
In continuous-conduction mode, inductor current remains above zero throughout the cycle. In discontinuous-conduction mode, it reaches zero for part of the cycle, and the conversion ratio generally depends on load as well as duty ratio. Many converters deliberately change operating mode at light load. Interleaving several identical phases spreads their switching events across the period, reducing combined ripple and distributing current among components; this approach is particularly important for high-current processor supplies.
Isolated converters and magnetic energy transfer
An isolated supply uses an electrical transformer to transfer energy across an insulation barrier. Isolation can separate hazardous input circuits from accessible outputs, interrupt direct ground-current paths, or allow an output to float relative to the input. A transformer also changes voltage according to its turns ratio. Unlike a mains-frequency transformer, an SMPS transformer operates at the converter’s switching frequency and can therefore be much smaller for a given application, although winding insulation and thermal requirements still impose physical limits.
The flyback converter is common in lower-power adapters and auxiliary supplies. Its magnetic component is more precisely understood as a coupled energy-storage inductor. While the primary switch conducts, magnetizing current rises and energy accumulates in the magnetic circuit; the secondary rectifier is normally reverse-biased. When the switch turns off, winding voltages reverse and energy flows to the secondary. For ideal continuous-conduction operation,
where is the secondary-to-primary turns ratio and the output polarity is taken as positive for the usual winding arrangement. Leakage inductance causes voltage spikes that require suitable clamping or energy-recovery arrangements.
A forward converter transfers energy to the secondary while its primary switch is on. An output inductor maintains load current between transfer intervals. Unlike the flyback’s coupled inductor, the forward transformer is not intended to store most of the energy delivered each cycle. Its magnetic flux must nevertheless be reset so that successive pulses do not drive the core into saturation. Single-switch forward circuits use a reset winding or another reset mechanism; two-switch versions provide a different path for returning magnetizing energy.
Push–pull, half-bridge, and full-bridge circuits apply alternating excitation to a transformer. Their different switch counts and voltage stresses suit different input voltages and power levels. A full bridge can apply both polarities of the input voltage across the primary and is widely used where substantial power must cross an isolation barrier. In a phase-shifted full bridge, relative timing between bridge legs controls effective energy transfer and can enable reduced-loss switching under suitable conditions.
Resonant converters use inductance and capacitance to shape current and voltage waveforms. The LLC converter employs a resonant inductance, a resonant capacitance, and transformer magnetizing inductance; it often regulates by varying frequency around its characteristic resonances. Its gain and switching conditions depend on load, component ratios, and operating frequency rather than on a single fixed conversion equation. Resonant operation can reduce switching loss, but introduces circulating current and more complicated behavior across wide input and output ranges.
Semiconductor devices and passive components
The main power switch determines much of a supply’s conduction loss, switching speed, voltage capability, and fault behavior. Early transistorized supplies commonly used bipolar junction transistors, which require base current and can exhibit charge-storage delays at turn-off. Modern low- and medium-power supplies predominantly use the metal–oxide–semiconductor field-effect transistor, or MOSFET. Its insulated gate requires little steady-state drive current, but charging and discharging gate capacitances at every transition consumes energy and requires a suitable driver.
Silicon MOSFETs serve a broad range of voltages and currents. Gallium nitride power transistors can offer low switching charge and fast transitions, supporting compact high-frequency converters. Silicon carbide devices are important in higher-voltage and high-temperature power conversion. No semiconductor material removes the need to evaluate conduction resistance, capacitance, drive requirements, thermal resistance, and transient ratings together. Faster switching can reduce transition losses while increasing overshoot and electromagnetic emissions.
Rectifiers provide a path for current when the main switch is off or when transformer secondary voltage has the appropriate polarity. Diode forward voltage can represent a large fraction of the output voltage in a low-voltage supply. Synchronous rectification replaces the diode’s principal conducting function with a controlled transistor, reducing conduction loss when its channel resistance is sufficiently low. Timing is critical: overlapping conduction can create a short circuit, while poorly controlled reverse current can send energy in an unintended direction.
Inductors and transformers are constrained by both electrical and magnetic effects. Core saturation sharply reduces effective inductance and can cause destructive current rise. Ferrite cores are common at high switching frequencies because of their relatively low eddy-current loss. Air gaps allow energy-storage magnetic components to support larger magnetizing currents without reaching saturation. At elevated frequency, skin and proximity effects redistribute winding current, increasing alternating-current resistance; foil, litz wire, and carefully arranged windings address different aspects of this problem.
Capacitors must withstand ripple current, applied voltage, temperature, and mechanical or environmental stress. Aluminum electrolytic capacitors offer substantial bulk capacitance, while ceramic and film capacitors serve high-frequency and other specialized roles. Effective ceramic capacitance can decrease significantly under direct-current bias, depending on dielectric and construction. Equivalent series resistance and inductance affect ripple, heating, and transient response. Consequently, nominal capacitance alone does not describe a component’s usefulness in a switching supply, and replacing one capacitor technology with another can alter both filtering and control-loop stability.
Regulation, feedback, and dynamic behavior
A regulated SMPS uses negative feedback to compare a measured electrical quantity with a reference and adjust switching accordingly. The regulated quantity may be output voltage, output current, input current, or a combination of these. The switching power stage, output filter, sensing network, controller, and load form a dynamic system studied using control theory. Good steady-state accuracy alone does not establish that this system will remain stable during changing input voltage or load.
Pulse-width modulation controls the duration of switching pulses, commonly at an approximately fixed repetition frequency. Voltage-mode control derives the switching command primarily from output-voltage error. Peak current-mode control adds a fast inner mechanism that terminates a pulse when switch or inductor current reaches a commanded level. This can simplify some aspects of compensation and provide cycle-by-cycle current limiting. In common fixed-frequency peak current-mode implementations, operation above a duty ratio of one half can require slope compensation to prevent subharmonic oscillation.
A controller cannot respond instantaneously to a load step. Initially, the output capacitor supplies or absorbs the difference between load current and converter current. Neglecting other effects, a current deficit lasting for produces an approximate capacitive voltage change
Capacitor resistance adds an immediate voltage step, and connection inductance contributes faster transients. These effects explain why a supply can meet its average voltage specification yet still produce unacceptable short-duration excursions at a sensitive load.
Control design must account for the power stage’s poles, zeros, delays, and changing operating conditions. A continuous-conduction boost converter, for example, has a right-half-plane zero in its control-to-output response: increasing duty ratio initially reduces the interval during which energy reaches the output before the larger inductor current produces the intended voltage rise. This limits achievable control bandwidth. Interactions with input filters and downstream regulated loads can create further instability, especially when a downstream converter behaves approximately as a constant-power load.
In an isolated supply, feedback may cross the insulation barrier through an optocoupler, a magnetic or capacitive isolation channel, or a separate winding-based sensing arrangement. Primary-side regulation estimates output conditions from information available on the primary side, avoiding a conventional secondary feedback channel but introducing sensitivity to winding and rectifier characteristics. Digital controllers replace or supplement analog functions with sampled measurements and programmed control laws. Sampling delay, quantization, computational timing, and protection independence then become additional design considerations rather than eliminating the underlying power-stage dynamics.
Losses, thermal design, and power density
Switching operation improves efficiency by avoiding prolonged simultaneous high voltage and high current in the regulating device, but actual transitions take finite time. A rough estimate for hard-switching overlap loss is
where and represent switching voltage and current, and and characterize transition durations. This simplified expression does not fully capture nonlinear device capacitances, diode reverse recovery, resonant transitions, or circuit parasitics, but it shows why increasing frequency can raise semiconductor loss.
Conduction loss in a MOSFET is approximately , with resistance dependent on temperature and operating conditions. Diode losses include forward conduction and, for many diode types, reverse recovery. Gate-drive loss scales approximately with gate charge, drive voltage, and switching frequency. Transformers and inductors dissipate energy through winding resistance and magnetic-core loss. Capacitor ripple currents generate heat in effective series resistance, and sensing circuits, controllers, and auxiliary supplies consume additional power.
Raising switching frequency can reduce the energy that must be stored and transferred during each cycle, enabling smaller magnetic and filtering components. It does not automatically reduce total supply size. Greater switching loss may demand larger heat-removal structures; tighter winding arrangements may increase capacitance; and safety spacing cannot simply shrink in proportion to electrical component size. Power density, commonly expressed as output power per unit volume, therefore reflects a balance among conversion frequency, component utilization, cooling, insulation, and packaging.
Soft switching reduces some transition losses by arranging for a device to turn on or off when its voltage or current is near zero. Zero-voltage switching can avoid much of the loss associated with discharging switch-node capacitance through the transistor at turn-on. Zero-current switching reduces transition stress associated with current interruption. These conditions often depend on load and stored reactive energy. Additional circulating current or auxiliary circuitry may offset part of the gain, particularly outside the intended operating range.
Heat must travel from semiconductor junctions, windings, and other hot components to the surroundings. Thermal resistance, interface materials, airflow, enclosure temperature, and neighboring heat sources all influence component temperature. Efficiency is therefore best represented as a function of input voltage, load, and temperature rather than as a single maximum figure. At light load, burst operation, pulse skipping, or reduced switching frequency can cut overhead consumption, but may increase low-frequency ripple, audible noise, or the time required to respond to a sudden load increase.
Mains input stages and power-factor correction
An AC-powered SMPS commonly begins with surge protection, an interference filter, a rectifier, and energy-storage capacitance. A diode bridge converts alternating input voltage to a unidirectional waveform, while a bulk capacitor stores energy between portions of the line cycle. With a capacitor-input rectifier, the unloaded bus voltage approaches the peak of the sinusoidal supply: nominal 230 V RMS corresponds to approximately 325 V peak, before allowing for diode drops and actual line conditions.
A large capacitor connected after a bridge draws current mainly when the instantaneous mains voltage exceeds the capacitor voltage. The resulting narrow current pulses contain harmonics and produce an RMS current greater than that of a sinusoidal current delivering the same real power at the same voltage. Power factor is defined as
For distorted waveforms, it is not adequately described by the cosine of a single voltage–current phase angle. Harmonic distortion and displacement both contribute to the result.
Active power-factor correction controls input current so that it approximately follows the shape of the line voltage. A boost converter is a common implementation because it can draw controlled current while producing a DC bus above the instantaneous rectified input. Universal-input supplies often use a regulated bus around 380–400 V, although actual values depend on the intended line range and architecture. A downstream isolated converter then transforms this bus into the required output voltage.
Single-phase input power still varies at twice line frequency even with ideal unity power factor. Some energy-storage element must absorb this variation if the output power is nearly constant. Bulk capacitors commonly perform this task and also support brief line interruptions. Their stored energy is
The usable hold-up energy depends on both the initial bus voltage and the minimum bus voltage at which the downstream converter can maintain its output. Hold-up time therefore cannot be inferred from capacitance alone.
Large input capacitors create inrush current when first connected to the mains. Thermistors, resistors with bypass switches, and actively controlled circuits limit this current. More advanced input stages include interleaved boost converters and bridgeless arrangements such as totem-pole PFC, which can reduce conduction loss but place additional demands on switching and control. Three-phase and bidirectional systems use other active-rectifier structures. Harmonic-emission requirements, including relevant parts of the IEC 61000 series, influence input-stage design according to equipment class and applicable jurisdiction.
Electromagnetic interference and physical layout
Rapid switching produces time-varying electric and magnetic fields and drives unwanted currents through parasitic capacitances and inductances. Electromagnetic compatibility concerns both limiting these emissions and maintaining correct operation in the presence of external disturbances. A supply can be efficient and well regulated while still interfering with radio reception, communication interfaces, or nearby analog circuitry. Conversely, electrostatic discharge or an input surge can upset a poorly protected controller without causing an obvious steady-state electrical fault.
Conducted interference is commonly separated into differential-mode and common-mode components. Differential-mode noise flows outward on one power conductor and returns on another. Common-mode noise involves conductors moving together relative to chassis, protective earth, or another reference, with return paths through intended or stray capacitance. A transformer’s interwinding capacitance can conduct common-mode displacement current even though it blocks direct-current conduction and provides an insulation barrier.
Filter elements address these modes differently. Differential inductance and capacitors across the supply conductors attenuate differential noise. Common-mode chokes present substantial impedance to currents flowing in the same direction through their windings while largely cancelling flux from the intended opposing load currents. Safety-rated X capacitors are used in appropriate line-to-line positions, while safety-rated Y capacitors serve specified line-to-earth or barrier-related functions. Their permissible application depends on insulation and leakage-current requirements, not merely capacitance and voltage rating.
Circuit-board layout is part of the power circuit rather than an independent finishing step. The loop containing a switching transistor, commutation path, and local bypass capacitor can carry abrupt current changes. Its stray inductance produces voltage according to , so even small inductances can create substantial overshoot. High-voltage switch nodes also couple noise capacitively into nearby conductors. Compact commutation loops, controlled return paths, and separation between noisy and sensitive circuitry affect both emissions and device stress.
Snubbers and clamps damp ringing or limit voltage excursions; shields redirect electric-field coupling; and controlled gate drive trades transition speed against loss and noise. Spread-spectrum modulation distributes switching energy across a frequency range, potentially reducing narrowband peaks, but does not remove the underlying noise energy. Verification typically combines conducted and radiated emission measurements with immunity testing. A line impedance stabilization network provides a defined measurement environment for many mains-conducted emissions tests, while near-field probes help locate local sources during development. Mechanical vibration of magnetic or ceramic components can also produce audible sound, particularly during burst-mode operation.
Electrical safety, protection, and reliability
A mains-powered switching supply may contain hazardous voltage whenever connected to power and can retain energy after disconnection. Electrical isolation is therefore a system property involving the transformer, feedback path, circuit-board spacing, connectors, enclosure, and any components crossing the barrier. A transformer symbol in a schematic does not establish adequate insulation. The required protective measures depend on working voltage, expected transients, environment, equipment category, and the applicable product standard.
Two important geometric quantities are clearance, the shortest distance through air, and creepage distance, the shortest relevant distance along an insulating surface. These address different failure mechanisms and are affected by factors such as altitude, contamination, and material properties. Insulation may be classified as functional, basic, supplementary, reinforced, or double according to its role and the governing standard. Requirements consequently differ between an industrial control supply, an accessible consumer adapter, and equipment intended for a medical environment.
Protection circuits address both external faults and internal abnormal operation. Overcurrent protection may limit current continuously, shut down until reset, or enter a repeated start-and-stop pattern known as hiccup mode. Overvoltage protection can use an independent sensing path to disable power transfer or trigger another protective action. Undervoltage lockout prevents operation when the controller or gate drive cannot function correctly. Overtemperature protection limits operation when a monitored location becomes too hot, although it cannot guarantee that every component remains within its rating.
Start-up and shutdown deserve separate attention because a converter is not in periodic steady state during these intervals. Soft start limits the rate at which the output target or current command rises. Transformer flux imbalance, capacitor charging, prebiased outputs, and synchronous-rectifier timing can otherwise produce unexpected stresses. Short circuits and open feedback paths may expose devices to conditions absent during normal regulation. Protection effectiveness therefore depends on fault response time, stored energy, and the behavior of the complete circuit.
Reliability is strongly influenced by temperature, electrical stress, environmental exposure, and component construction. Electrolytic capacitors can lose electrolyte and develop increased resistance; solder joints and semiconductor interconnections experience thermomechanical fatigue; insulation can deteriorate under electrical and environmental stress. Product qualification uses combinations of thermal cycling, operating-life tests, surge tests, and fault evaluation as appropriate. Statistical reliability figures such as mean time between failures describe a model or population under stated assumptions; they are not the expected service life of every individual unit. Safety standards such as IEC 62368-1 and IEC 60601-1 address different equipment domains and do not replace application-specific reliability assessment.
Historical development
The conceptual ancestry of switched-mode conversion extends beyond semiconductor electronics. Electromechanical interrupters and vibrator supplies converted battery direct current into alternating or pulsating waveforms that could drive a transformer. Such equipment was used, for example, to obtain higher voltages in vehicle-mounted vacuum-tube radios. These arrangements shared the principle of periodic switching and magnetic conversion, but lacked the speed, controllability, and durability of later semiconductor supplies.
The development of power transistors made compact electronic switching practical. During the 1950s and 1960s, switching converters became increasingly attractive in aerospace, communications, and other systems where weight, heat, and battery utilization mattered. Their adoption depended not only on switches but also on magnetic materials, rectifiers, capacitors, and feedback techniques. High-frequency conversion was therefore not a single invention that appeared fully formed in one product; it developed through multiple circuit families and manufacturing advances.
Dedicated controller integrated circuits reduced the number of discrete components needed to implement oscillation, reference generation, error amplification, and pulse control. Silicon General’s SG1524, introduced in 1976, was an influential early integrated switching-regulator controller. Such devices helped standardize converter architectures and made switching supplies more accessible to equipment manufacturers. Later controllers incorporated current sensing, protection, high-voltage start-up functions, and, in some cases, the main power switch itself.
Personal computers contributed to widespread commercial adoption. The Apple II, introduced in 1977, used a switching supply designed by Rod Holt; it was a notable application rather than the origin of switching power conversion. The IBM PC and its successors further established switching supplies in office computing. As digital systems required more power at lower voltages, supply architecture shifted from a small number of centrally regulated outputs toward multiple local converters close to their loads.
Power MOSFETs, improved ferrites, low-resistance synchronous rectifiers, and better packaging progressively raised efficiency and switching frequency. Energy-efficiency requirements also placed greater emphasis on losses over an entire load range, standby consumption, and input-current quality rather than full-load performance alone. More recently, gallium nitride and silicon carbide devices have extended useful operating ranges and supported compact converters. Their adoption has continued the longstanding pattern in which semiconductor progress creates opportunities that must be matched by advances in magnetics, control, packaging, and electromagnetic compatibility.
Applications and system architectures
Computing equipment commonly uses a hierarchy of switching stages. A mains supply produces one or more intermediate DC rails, and local point-of-load converters generate the low voltages required by processors, memory, and peripheral circuits. Processor supplies often use interleaved synchronous buck stages because the required current can be high while the permissible voltage deviation is small. Locating conversion near the load reduces distribution loss and connection inductance, although it places heat-generating components close to temperature-sensitive electronics.
Telecommunications equipment frequently uses a nominal 48 V DC distribution system, commonly with a negative polarity convention in traditional installations. AC rectifier modules, backup batteries, and isolated board-level converters form a coordinated power system. The distribution voltage reduces current compared with directly distributing low logic voltages, while local converters provide the required isolation and regulation. Similar intermediate-bus approaches appear in data centers, where changes in distribution voltage and conversion staging seek to reduce total electrical and cooling losses.
Battery-powered products use buck, boost, and buck–boost converters to accommodate changing battery voltage and several internal supply rails. A battery management system monitors and protects an energy-storage pack, while switching converters perform the associated voltage or current conversion; the two functions are related but not interchangeable. In an electric vehicle, an auxiliary DC–DC converter supplies low-voltage equipment from the traction battery, and an onboard charger converts external AC power into controlled charging power. Some designs support bidirectional energy flow, but this requires suitable power-stage and control provisions.
Industrial supplies must accommodate electrical noise, broad load variation, and demanding environmental conditions. They power programmable controllers, sensors, actuators, and communication equipment, often through standardized DC distribution voltages. LED drivers are another major application: their principal regulated quantity is usually current rather than voltage, and their control must accommodate the LED string’s electrical characteristics and any dimming requirements. Laboratory supplies may combine a switching preregulator with a linear output stage to obtain both manageable dissipation and low output noise.
Switching conversion also links supplies to renewable-energy and backup systems. A photovoltaic DC–DC stage can adjust its operating point to extract available power from a solar array, while a power inverter converts DC into AC for a load or grid connection. An uninterruptible power supply may contain rectification, battery charging, DC–DC conversion, and inversion within one product. These examples illustrate the boundary of the term: the same switching devices and energy-transfer principles can serve several functions, but the equipment’s name usually follows the complete system’s purpose, output form, and role in the electrical installation.