A switching regulator (sometimes loosely called an SMPS, though that term more precisely refers to the broader power-supply system in which a switching regulator is the regulation stage) is a power conversion circuit that regulates an output voltage by rapidly switching one or more transistors on and off, storing and releasing energy primarily in inductors or transformers (with capacitors mainly serving to filter and smooth the output), rather than dissipating excess energy as heat. Conversion efficiencies of 80–95% are common, making switching regulators the dominant choice when power budget or thermal constraints rule out linear regulators.
In practice
Switching regulators appear in a wide range of embedded systems, particularly those running from a battery or an unregulated supply rail where efficiency is a priority. Common topologies include buck (step-down), boost (step-up), buck-boost (inverting or non-inverting), and flyback (for isolated supplies). A buck regulator is the most common choice for powering an MCU or SoC from a higher-voltage rail — for example, stepping a 12 V supply down to 3.3 V or 1.8 V for a Cortex-A SoC core. Boost converters are typical in battery-powered designs where the cell voltage can fall below the required rail, such as driving a 3.3 V system from a single-cell Li-ion that discharges down to 3.0 V.
The switching frequency has a direct impact on the size of the external passive components. Controllers switching at 400 kHz–2 MHz allow smaller inductors and capacitors than older designs at 50–100 kHz, which matters for board-area-constrained designs. Many modern regulators integrate the power FETs into the controller IC alongside synchronous rectification (which replaces the recirculation diode with a second controlled FET), reducing external component count and improving efficiency. Devices such as the Texas Instruments TPS62xxx family or the Analog Devices (Maxim) MAX17xxx family are common examples used in embedded hardware.
Output noise is the principal drawback compared to linear regulators. The switching action injects ripple onto the output rail at the switching frequency and its harmonics, typically in the range of a few millivolts to tens of millivolts. Sensitive analog front ends, PLLs, RF circuits, and precision ADC references often require additional LC or RC post-filtering, careful PCB layout to minimize switching-node radiation, or a separate linear regulator fed from the switcher output. The blog series "Byte and Switch (Part 1)" and "Byte and Switch (Part 2)" on EmbeddedRelated cover practical switching circuit considerations worth reviewing.
Feedback loop stability is another common pitfall. The error amplifier and compensation network must be designed for the selected output capacitor's ESR and capacitance; swapping capacitor types (e.g., replacing an electrolytic with a low-ESR ceramic) can shift the loop's phase margin and cause oscillation or poor transient response. Always verify stability with a loop analyzer or at minimum check transient load-step response when changing BOM components.
Frequently asked
When should I choose a switching regulator over a linear regulator?
Use a switching regulator when the input-to-output voltage difference is large, when load current is significant, or when battery life matters. A
linear regulator drops (Vin - Vout) * Iload as heat; at 12 V in and 3.3 V out at 500 mA, that is 4.35 W of dissipation, requiring a heatsink and wasting most of the input power. A switching regulator in the same scenario will produce substantially lower losses, though the exact figure depends on topology, switching frequency, load, and component choices. Linear regulators remain preferable for low-noise requirements, very low dropout situations, or when simplicity and cost outweigh efficiency.
What is the difference between a synchronous and an asynchronous (non-synchronous) switching regulator?
An asynchronous regulator uses a
diode (often a
Schottky) to provide the recirculation path for
inductor current when the main switch is off. A synchronous regulator replaces that diode with a second controlled FET, which has a lower forward voltage drop than the diode and therefore lower conduction losses. Synchronous designs are more efficient, especially at higher currents, but the controller must ensure the two switches are never on simultaneously (shoot-through), which adds complexity.
How do I reduce switching noise affecting my analog circuitry?
Several techniques are used in combination: keep the switching node copper area small to reduce radiated
EMI; use a separate analog ground plane and connect it to the power ground at a single point; add a small LC or ferrite-bead RC filter on the output feeding analog rails; increase output capacitance to reduce ripple amplitude; and consider a low-dropout linear post-regulator for the most sensitive circuits such as
ADC references or
RF VCOs. Switching frequency selection also matters — a higher frequency reduces
inductor size but may move noise into more sensitive bands.
What is discontinuous conduction mode (DCM) and does it matter for my design?
In continuous conduction mode (CCM),
inductor current never reaches zero during a switching cycle. In discontinuous conduction mode (DCM), the inductor fully discharges before the next cycle begins. DCM occurs at light loads and affects control loop dynamics — the small-signal transfer function changes, which can destabilize a compensator optimized for CCM. Many modern regulators handle this automatically (some enter a pulse-skipping or burst mode at light load for efficiency), but it is important to verify stability and output ripple across the full load range, not just at full load.
Can I power an MCU directly from a switching regulator output, or do I need a linear regulator in between?
Most MCUs tolerate switching regulator ripple without issues, provided the ripple is well within the supply voltage tolerance, though the acceptable level depends on the specific
MCU, supply rail, and which analog functions are in use. For microcontrollers with on-chip ADCs or analog peripherals, excessive ripple on the analog supply pin (AVDD/VDDA) can degrade measurement accuracy. A common practice is to power the digital supply directly from the switcher and add a small LC or ferrite-bead filter plus a
bypass capacitor on the analog supply pin, or use a small
LDO just for the analog rail.
Differentiators vs similar concepts
Switching regulators are frequently contrasted with
linear regulators (such as the LM317 or
LDO types like the AMS1117). A linear regulator controls output voltage by operating a pass
transistor in its linear (partially-on) region, dissipating the excess voltage as heat; efficiency is directly proportional to Vout/Vin, which becomes poor at large step-down ratios. A switching regulator avoids this by using the transistor only as a switch (fully on or fully off), storing energy in reactive components instead. The trade-off is that switching regulators require external
inductors and
capacitors, generate switching noise, and involve more complex feedback compensation, while linear regulators are simpler, quieter, and lower-cost for small voltage drops and low currents.
