Electronics Guide

Power Supply Topologies

Power supply architectures fall into two broad categories: linear regulators and switching regulators. Each approach offers distinct advantages that make it suitable for specific applications within an embedded system.

Linear regulators operate by dissipating excess voltage as heat, functioning essentially as variable resistors that maintain constant output voltage. Their simplicity provides inherently low noise output and fast transient response, making them ideal for powering noise-sensitive analog circuits, precision references, and RF stages. Ignoring quiescent current, their efficiency cannot exceed the ratio of output voltage to input voltage. A 3.3 V regulator fed from 5 V therefore wastes at least a third of the input power, and the same regulator fed from 12 V wastes more than 70%.

Switching regulators achieve high efficiency by rapidly switching power transistors on and off, storing energy in inductors and capacitors during switching cycles. Buck converters step voltage down, boost converters step voltage up, and buck-boost topologies can do either. While more complex and inherently noisier than linear regulators, well-designed synchronous switching converters routinely exceed 90% efficiency and can exceed 95% at favorable conversion ratios, dramatically reducing heat generation and extending battery life.

A third family, switched-capacitor or charge-pump converters, moves energy between capacitors rather than through an inductor. Charge pumps suit low-current rails and fixed conversion ratios such as doubling or halving, and they eliminate the magnetics entirely. Their efficiency falls sharply when the output must deviate from the ratio the topology naturally produces, so they are best reserved for modest loads such as bias generation, LCD supplies, and gate drive.

Isolation and Off-Line Conversion

Embedded systems that connect to AC mains or to hazardous voltages require galvanic isolation between the input and the output. Isolation is provided by a transformer, and feedback crosses the barrier through an optocoupler or a digital isolator. The flyback converter dominates low-power off-line supplies because it needs only one magnetic component and supports multiple outputs; forward, half-bridge, and LLC resonant converters take over as power levels rise and efficiency requirements tighten.

Isolation is also used within low-voltage systems, where it breaks ground loops rather than protecting people. Isolated bias supplies for gate drivers, isolated rails for 4–20 mA current loops, and isolated sensor front ends all rely on small transformer-based converters. Because the safety requirements differ, designers must distinguish functional isolation, which serves signal integrity, from reinforced isolation, which serves user safety and is governed by product safety standards and creepage and clearance rules.

Power Architecture

Above a handful of rails, individual regulator choices matter less than the architecture that connects them. A distributed architecture converts the input once to an intermediate bus, commonly 12 V or 5 V, then places small point-of-load converters next to the circuits they feed. Short, low-impedance connections between the converter and its load preserve regulation accuracy and transient response, which matters most for the low-voltage, high-current rails feeding processors and FPGAs.

Portable and cost-sensitive designs more often use a power management integrated circuit, a PMIC that packages several bucks, boosts, and LDOs with sequencing, supervisory, and often battery-charging functions behind a single digital interface. PMICs matched to a specific processor family remove much of the sequencing risk, at the cost of flexibility if the design later needs a rail the device does not provide.

Input sources shape the architecture as much as the loads do. USB Type-C with Power Delivery negotiates 5 V, 9 V, 15 V, or 20 V within the Standard Power Range up to 100 W, and USB PD 3.1 adds Extended Power Range fixed voltages of 28 V, 36 V, and 48 V for up to 240 W. A design that accepts USB PD must therefore tolerate a wide and dynamically changing input, which usually argues for a buck-boost front end.

Efficiency Considerations

Power conversion efficiency directly impacts system thermal management, battery runtime, and operating costs. Understanding efficiency characteristics across the full operating range is essential for optimal design.

Switching converter efficiency varies with load current, input voltage, and switching frequency. At heavy loads, conduction losses in switches and inductors dominate. At light loads, switching losses and quiescent current become more significant. Many modern converters implement pulse-skipping or burst modes that improve light-load efficiency by reducing switching frequency when demand is low.

For battery-powered systems that spend most of their time in sleep modes, light-load efficiency often matters more than peak efficiency. A converter that achieves 90% efficiency at full load but draws 100 microamperes quiescent current may be inferior to one with 85% peak efficiency but only 10 microamperes quiescent current for applications with low duty cycles.

Voltage Regulation Accuracy

Modern digital circuits operate at increasingly low voltages with tight tolerance requirements. A processor specified for 1.0V core voltage with plus or minus 3% tolerance leaves only 30 millivolts of margin for all error sources combined. This budget must accommodate initial accuracy, line regulation, load regulation, temperature drift, and transient deviations.

Load regulation describes how output voltage changes with load current variations. Tight load regulation requires fast control loops and low impedance output stages. Line regulation describes output variation with input voltage changes, particularly important when operating from batteries whose voltage declines as they discharge.

Copper resistance consumes part of the budget before the rail reaches the load. A 10-ampere rail crossing five milliohms of plane and via resistance loses 50 millivolts, which alone exceeds a 3% budget on a 1.0 V rail. Two remedies apply. Remote sensing routes the feedback connection to the load rather than to the converter output, so the loop regulates the voltage that the load actually sees. Where remote sensing is impractical, the converter is placed close enough to the load that the drop becomes negligible, which is the central argument for point-of-load architectures.

A complete rail budget also accounts for output ripple and for transient deviation during load steps. Because these add to the static error terms, a specification quoted only as initial accuracy understates the design problem considerably.

Buck Converter Fundamentals

The buck converter is the workhorse of embedded power systems, efficiently stepping down higher input voltages to the lower rails required by modern integrated circuits. Understanding its operation provides the foundation for effective power supply design.

During the on-time, the high-side switch connects the input to the inductor, causing current to ramp up as energy is stored in the magnetic field. During the off-time, the switch opens and inductor current continues flowing through the low-side switch or diode, transferring stored energy to the output. The output capacitor filters the resulting current ripple into a stable DC voltage.

In continuous conduction mode and neglecting losses, the duty cycle of an ideal buck converter equals the ratio of output voltage to input voltage. Converting 12 V to 1.2 V therefore calls for roughly a 10% duty cycle, which at a 500 kHz switching frequency leaves an on-time near 200 nanoseconds. Very small duty cycles stress the controller's minimum on-time specification and often force a lower switching frequency or a two-stage conversion. The corresponding relationship for an ideal boost converter is that the output voltage equals the input voltage divided by one minus the duty cycle, which is why extreme step-up ratios become impractical as the duty cycle approaches unity.

Key design parameters include switching frequency, inductor value, and output capacitance. Higher switching frequencies allow smaller inductors and capacitors but increase switching losses and electromagnetic emissions. The inductor is commonly sized so that peak-to-peak ripple current is roughly 30% of the maximum load current: less ripple demands a bulky inductor and slows the response to load steps, while more ripple raises core and capacitor losses and pushes peak current toward saturation. The inductor must carry that peak current without saturating, and output capacitors must provide both ripple filtering and the charge that holds the rail up during a load step.

Synchronous versus Non-Synchronous Rectification

Traditional buck converters use a diode for the low-side switch, conducting current during the off-time. While simple and reliable, the diode forward voltage drop causes significant power loss, particularly at high currents and low output voltages. A Schottky diode dropping 0.4 V in a 1.2 V output converter wastes power roughly in proportion to that drop relative to the output voltage. Synchronous converters replace the diode with a controlled MOSFET whose on-state drop is a small fraction of that, reducing conduction losses and improving efficiency.

Synchronous rectification requires more complex control to prevent shoot-through, a condition where both high-side and low-side switches conduct simultaneously. Dead-time intervals between switching transitions prevent this condition but introduce additional losses, because the body diode conducts during the dead time. Modern integrated controllers handle these timing requirements automatically, often with adaptive dead-time control, making synchronous designs practical for most applications.

Synchronous stages also conduct current in reverse, which allows true continuous conduction at light load but wastes energy circulating between the output capacitor and the input. Most converters therefore offer a diode-emulation or discontinuous mode that blocks reverse current at light load, trading a small amount of output ripple for a large gain in light-load efficiency.

Control Modes and Loop Compensation

Every switching regulator closes a feedback loop around its power stage, and the choice of control mode shapes both stability and transient behavior. Voltage-mode control compares the output to a reference and adjusts duty cycle directly; it tolerates noise well and behaves predictably at very small duty cycles, but the output filter contributes a two-pole response that requires more elaborate compensation. Peak current-mode control adds an inner loop that regulates inductor current cycle by cycle, which simplifies compensation, provides inherent current limiting, and improves line transient rejection. Its drawback is subharmonic oscillation above roughly 50% duty cycle, which is corrected by adding slope compensation.

Hysteretic and constant-on-time controllers dispense with a conventional error amplifier and respond to load steps within a switching cycle, at the cost of a switching frequency that varies with operating conditions. That variation complicates EMI planning and makes synchronization to a system clock difficult, so these controllers are most common where transient response outweighs spectral predictability.

Compensation design targets a loop crossover frequency well below the switching frequency, commonly between one-tenth and one-fifth of it, with at least 45 degrees of phase margin and typically 45 to 60 degrees in production designs. Higher crossover improves transient response but leaves less margin against component tolerance, ceramic capacitor derating, and layout parasitics. Loop gain is measured on hardware by injecting a small signal across a series resistor in the feedback divider and sweeping frequency with a network analyzer, which reveals margins that simulation alone may miss.

Boost and Buck-Boost Converters

When the required output voltage exceeds the available input voltage, boost converters step voltage up by storing energy in an inductor during switch on-time and releasing it at higher voltage during off-time. This topology finds use in battery-powered systems where the voltage rail must remain stable as battery voltage declines below the required output level.

Buck-boost topologies provide both step-up and step-down capability, essential when input voltage may be either above or below the required output. Single-inductor designs include inverting buck-boost, SEPIC, and Cuk converters. Four-switch buck-boost converters provide non-inverting operation with seamless transitions between buck and boost modes as input voltage varies.

Multiphase Converters

High-current applications benefit from multiphase converter architectures that parallel multiple power stages with interleaved switching. This approach distributes current among multiple inductors and switches, reducing component stress and allowing use of smaller, less expensive parts. Interleaved switching phases partially cancel input and output ripple currents, reducing filtering requirements and electromagnetic emissions.

Modern processors and FPGAs may require over 100 amperes at sub-volt voltages. Multiphase voltage regulator modules with six or more phases handle these extreme requirements while maintaining efficiency and transient response specifications.

Low-Dropout Regulators

Low-dropout regulators, commonly known as LDOs, provide linear regulation with minimal voltage differential between input and output. Where traditional linear regulators require several volts of headroom, modern LDOs can regulate with dropouts below 100 millivolts, enabling efficient post-regulation of switching converter outputs.

LDO dropout voltage depends on load current and pass element design. PMOS pass transistors offer the lowest dropout but may have stability challenges. NPN-based designs provide better transient response but require more headroom. Understanding these trade-offs guides selection for specific applications.

Noise Performance

Linear regulators inherently provide cleaner output than switching converters because they lack the high-frequency switching that generates noise. LDOs additionally filter noise present on their input, providing power supply rejection that attenuates disturbances from upstream converters.

Power supply rejection ratio, or PSRR, quantifies this filtering capability as a function of frequency. At low frequencies, PSRR may exceed 70 decibels, meaning input noise is attenuated by more than 3000 times. However, PSRR degrades at higher frequencies, typically falling to minimal values above 1 MHz. Understanding this frequency-dependent behavior is crucial when using LDOs to filter switching converter ripple.

Ultra-low-noise LDOs designed for sensitive RF and precision analog applications achieve output noise below 10 microvolts RMS, enabling direct powering of voltage-controlled oscillators, phase-locked loops, and high-resolution data converters.

Transient Response

Load transients occur when circuits rapidly change their current consumption, such as when processors wake from sleep or begin intensive computation. The power supply must respond quickly to maintain voltage within specification during these events.

LDO transient response depends on control loop bandwidth, output capacitance, and capacitor equivalent series resistance. Higher bandwidth enables faster response but may compromise stability. Output capacitors provide energy storage that limits initial voltage deviation while the regulator responds. Capacitor ESR affects both the magnitude of transient deviation and regulator stability.

Thermal Considerations

Linear regulators dissipate power equal to the voltage differential multiplied by load current. This power becomes heat that must be removed to maintain junction temperature within safe limits. Thermal design involves selecting appropriate packages, providing adequate copper area for heat spreading, and potentially adding heatsinks for high-dissipation applications.

Modern LDOs include thermal shutdown protection that reduces output current or disables the regulator entirely when junction temperature exceeds safe thresholds. While this protection prevents damage, designs should ensure normal operation remains well below shutdown thresholds to maintain reliability.

Sequencing Requirements

Complex integrated circuits often require multiple supply voltages that must power up and down in specific sequences. Core voltages typically must be established before I/O voltages to prevent latch-up conditions where parasitic thyristors conduct destructively high currents. Memory interfaces may require termination voltages to track data voltages within specified ratios.

Device datasheets specify sequencing requirements that must be followed for reliable operation. Violating these requirements may cause immediate damage, gradual reliability degradation, or unpredictable behavior depending on the specific violation and device design.

Sequencing Implementation

Power sequencing can be implemented through several approaches depending on complexity requirements. Simple systems may use the enable pins of individual regulators connected to voltage supervisors or RC delay networks that establish timing relationships. More complex systems employ dedicated sequencing controllers that manage multiple rails with programmable timing, monitoring, and fault handling.

Power management integrated circuits, or PMICs, integrate multiple regulators with built-in sequencing logic optimized for specific processor families. These devices simplify design by handling sequencing requirements automatically while providing monitoring and protection features.

Soft Start and Inrush Control

Switching converters and large output capacitors can draw substantial current during startup if voltage rises too quickly. This inrush current may cause input voltage droop, trigger upstream protection, or stress components. Soft start circuits limit the rate of output voltage rise, spreading inrush current over a controlled interval.

Soft start time must balance inrush control against system startup time requirements. Overly slow soft start extends the period when supplies are outside regulation, potentially causing issues with circuits that begin operating before all rails are valid.

Power-Down Sequencing

Controlled power-down sequencing is often as important as power-up sequencing. Supplies may need to discharge in reverse order from power-up, with active discharge circuits ensuring rapid voltage decay. Some systems require specific voltage relationships during power-down to prevent data corruption in nonvolatile memory or damage to sensitive circuits.

Emergency shutdown scenarios, such as input power failure, may not allow time for controlled sequencing. Designs must ensure that even abrupt power loss does not cause damage, potentially using bulk capacitance to maintain power long enough for critical operations to complete.

Switching Noise Sources

Switching converters generate noise through multiple mechanisms. The fundamental switching frequency and its harmonics appear on both input and output. High-frequency ringing occurs during switching transitions as parasitic inductances and capacitances resonate. Gate drive circuits generate noise as they charge and discharge switch capacitances.

Noise couples through conducted and radiated paths. Conducted noise travels through power and ground connections to affect other circuits sharing those paths. Radiated noise propagates through space, potentially interfering with nearby sensitive circuits or violating electromagnetic compatibility regulations.

Regulatory limits define the frequencies that matter. For multimedia and information technology equipment, CISPR 32 and its European counterpart EN 55032 specify conducted emissions from 150 kHz to 30 MHz and radiated emissions from 30 MHz upward, extending above 1 GHz for equipment with high internal clock rates. Automotive electronics are held to CISPR 25, which is considerably stricter and begins at 150 kHz. Because the AM broadcast band spans roughly 530 kHz to 1710 kHz, automotive converters are frequently designed to switch above 2 MHz so that the fundamental and its first harmonics fall clear of that band.

Layout Best Practices

PCB layout critically affects switching converter noise performance. The switched current loop, comprising the input capacitor, switches, and inductor, should be as compact as possible to minimize loop inductance and radiated emissions. Ground planes provide low-impedance return paths that contain high-frequency currents.

Component placement should minimize the area of high-current loops while keeping sensitive circuits physically separated from noise sources. Switching converter sections should be isolated from analog circuits, with dedicated power and ground connections preventing noise coupling through shared impedances.

Several specific practices repay the effort. Place the input ceramic capacitor directly across the high-side switch and the low-side return, with the shortest possible path, and keep an uninterrupted ground plane on the layer immediately beneath the hot loop so return current can mirror the forward path. Keep the switch node copper only as large as thermal requirements demand, because it swings the full input voltage at high slew rate and behaves as a radiating plate. Route the feedback trace as a quiet, narrow path away from the switch node and the inductor, referenced to the ground plane, and connect it to the output at the point that regulation must hold. Where a sense resistor is used, route its connections as a differential pair back to the controller.

These practices are far cheaper to apply during layout than to retrofit. Emissions problems traced to a large hot loop or a poorly referenced feedback trace usually require a board respin, whereas the same geometry costs nothing if it is planned from the start.

Filtering Techniques

Input and output filtering attenuates conducted noise to acceptable levels. Input filters prevent switching noise from propagating to the power source where it might affect other circuits or violate EMI specifications. Output filters reduce ripple voltage delivered to the load.

LC filters provide effective attenuation above their resonant frequency but can amplify noise near resonance. Damping networks or carefully chosen component values prevent resonant peaking. Ferrite beads provide high-frequency filtering in a compact form factor, particularly effective against high-frequency ringing.

Spread Spectrum Modulation

Fixed-frequency switching concentrates noise energy at specific frequencies, potentially causing interference with sensitive circuits or failing EMI compliance at those frequencies. Spread spectrum modulation varies the switching frequency over a defined range, spreading noise energy across a wider bandwidth and reducing peak emissions.

The benefit depends on the measurement. Spreading the frequency lowers the reading of a receiver whose resolution bandwidth is narrower than the spread, which is the usual case for the quasi-peak and peak detectors used in compliance testing. It does not reduce total emitted energy; it redistributes it. Applications sensitive to broadband noise, or those that must synchronize converters to a system clock to keep beat frequencies out of a signal band, may be better served by fixed-frequency operation with better filtering.

Post-Regulation for Noise-Sensitive Loads

Combining switching converters with linear post-regulators provides both efficiency and low noise. The switching converter handles bulk power conversion efficiently, while the LDO filters switching noise and provides clean, well-regulated voltage to sensitive circuits. This hybrid approach is standard practice for powering precision analog circuits, RF stages, and high-speed clocks.

The dropout voltage required by the LDO represents unavoidable efficiency loss, but this loss is often acceptable given the substantial efficiency improvement of using a switching converter for the majority of voltage conversion rather than a pure linear approach.

Inductor Selection

The inductor is often the largest and most critical component in switching converter design. Key parameters include inductance value, saturation current, DC resistance, and core losses. The inductor must handle peak current without saturating, which would cause inductance to collapse and current to spike uncontrollably.

Shielded inductors contain magnetic fields more effectively, reducing coupling to nearby circuits. Ferrite cores offer low losses at high frequencies but may saturate abruptly. Powdered iron cores handle higher currents with gentler saturation but exhibit higher core losses. Understanding these trade-offs enables selection of inductors optimized for specific applications.

Capacitor Selection

Input and output capacitors store energy, filter ripple, and supply current during transients. Key parameters include capacitance value, voltage rating, equivalent series resistance, and equivalent series inductance. Ceramic capacitors offer low ESR and ESL in compact packages, but Class II dielectrics such as X5R and X7R lose capacitance under DC bias, often more than half of the nominal value at a significant fraction of the rated voltage, with the loss growing as case size shrinks. Designers therefore derate deliberately, choosing a voltage rating well above the applied rail and verifying the effective capacitance from the manufacturer's bias curves rather than the marked value. The same dielectrics also drift with temperature and age, and they are piezoelectric, which turns ripple voltage into audible singing on some boards.

Electrolytic and polymer capacitors provide high capacitance density for bulk energy storage but have higher ESR than ceramics. Aluminum electrolytics are inexpensive and lose capacitance and gain ESR at low temperature; polymer and hybrid types hold their characteristics better and last longer, at higher cost. Tantalum parts offer stable capacitance in small volumes but require careful voltage derating. Many designs combine types deliberately, using bulk electrolytic or polymer capacitance for low-frequency energy storage and small ceramics placed close to the load for high-frequency decoupling.

Input capacitor selection deserves particular attention on buck converters, where the input current is discontinuous and the capacitor must supply the full switched current pulse. Its RMS current rating, not its capacitance, is usually the limiting specification, and it must sit as close as possible to the switch node's current loop.

Controller and Integrated Solutions

Power supply controllers range from simple fixed-frequency PWM controllers to sophisticated digital power management devices. Integrated converters combine the controller with power switches in a single package, simplifying design at the cost of flexibility and power handling capability.

Selection criteria include switching frequency range, protection features, feedback options, and packaging. For high-current applications, external switches provide better thermal management and allow optimization of switch characteristics. For low-power applications, highly integrated solutions minimize board space and component count. Modules that place the controller, switches, and inductor in one package trade cost and flexibility for a qualified layout and a short design cycle, which is often the right bargain when board area is scarce or the schedule is tight.

Switch technology increasingly influences these choices. Silicon MOSFETs remain the default below about 100 V. Gallium nitride transistors switch faster with lower gate charge and no reverse-recovery charge, allowing higher frequencies and smaller magnetics, but their fast edges demand disciplined layout and gate drive to control ringing and emissions. Silicon carbide devices serve higher-voltage rails, well above what most embedded systems require.

Overcurrent Protection

Power supplies must limit current under overload or short-circuit conditions to prevent damage to the supply itself and downstream circuits. Current-mode control inherently provides cycle-by-cycle current limiting, while voltage-mode controllers require separate current sensing and limiting circuits.

Foldback current limiting reduces output current as output voltage drops below normal, limiting power dissipation during sustained shorts. Hiccup mode periodically attempts restart after fault conditions, preventing continuous power dissipation while enabling automatic recovery when faults clear.

Overvoltage and Undervoltage Protection

Overvoltage conditions can damage sensitive load circuits, making overvoltage protection essential. Supervisory circuits monitor output voltage and trigger protective action when thresholds are exceeded, typically disabling the supply or activating crowbar circuits that clamp voltage by forcing a short.

Undervoltage lockout prevents operation when input voltage is insufficient for proper regulation. This protection ensures the supply either operates correctly or remains off, preventing unpredictable behavior during brownout conditions.

Thermal Protection

Integrated thermal protection monitors die temperature and reduces power or shuts down when safe limits are approached. This protection prevents damage from excessive ambient temperature, inadequate cooling, or overload conditions that generate abnormal heat.

Designs should ensure thermal protection activates only under abnormal conditions. Normal operation should maintain temperatures well below protection thresholds to ensure long-term reliability.

Reverse Polarity and Transient Protection

Systems that connect to external power sources must handle connection errors and transient events. Reverse polarity protection prevents damage if power connections are accidentally reversed, using series diodes, MOSFETs, or specialized protection ICs. Transient voltage suppressors clamp voltage spikes from electrostatic discharge, load dump, or lightning-induced surges.

Requirements Definition

Successful power supply design begins with clear requirements definition. Input voltage range must account for source variations, cable drops, and transient conditions. Output voltage and tolerance specifications come from load device requirements. Load current profiles should capture average, peak, and transient characteristics including startup and sleep mode currents.

Efficiency targets may be specified at multiple operating points. Thermal constraints establish maximum allowable temperature rise. Size and cost targets influence topology selection and component choices. EMI requirements define acceptable emission levels.

Topology Selection

Matching topology to requirements is fundamental to achieving design goals. Linear regulators suit low-power applications where simplicity and low noise outweigh efficiency concerns. Buck converters serve step-down applications with wide input voltage ranges. Boost and buck-boost topologies handle applications where output voltage may exceed input.

Hybrid approaches combining switching and linear regulation often provide optimal solutions for systems with both efficiency and noise requirements. Multiple conversion stages may be appropriate when voltage differentials are large or when isolation is required.

Simulation and Verification

SPICE simulation enables verification of power supply behavior before committing to hardware. Transient simulations verify startup sequencing, load step response, and stability. AC analysis confirms loop gain and phase margin for stable operation. Parametric sweeps explore sensitivity to component variations.

Many controller manufacturers provide simulation models and reference designs that accelerate development. These resources should be starting points that are adapted to specific requirements rather than used without modification.

Prototype Testing

Hardware testing verifies that designs meet requirements under real-world conditions. Efficiency measurements across the operating range confirm thermal design adequacy. Load transient testing validates dynamic response. EMI testing ensures compliance with applicable standards. Temperature testing verifies operation over the specified range.

Oscilloscope measurements should examine switching waveforms for excessive ringing or other anomalies that might indicate layout issues or inadequate snubbing. Measurement technique matters as much as the instrument: a probe with a long ground lead forms a loop that picks up switching fields and manufactures ringing that is not present on the board, so ripple and switch-node measurements should use a short ground spring or a tip-and-barrel connection across the capacitor terminals. Thermal imaging identifies hot spots that might limit reliability, and it is most informative when the board is run at worst-case ambient temperature rather than on the bench.

Two tests are commonly deferred and should not be. Loop gain measurement across the full range of load, input voltage, and temperature catches marginal phase margin that only appears at a corner. Repeated power cycling, including brownouts and slow input ramps, exposes sequencing and undervoltage lockout faults that a clean single power-up will never reveal.