123 Main Street, New York, NY 10001

A synchronous buck replaces the rectifier diode with a low-side MOSFET, cutting conduction loss and boosting efficiency—especially for low output voltages and higher currents. This child page focuses on single-phase synchronous bucks; Multiphase VRs, Boost/Buck-Boost, and PMBus/PSM telemetry live on sibling pages.

Introduction & Use Cases

Why synchronous rectification matters—quick math: for 12 V → 1.2 V @ 5 A, duty ≈ 0.1. With a Schottky diode (Vf ≈ 0.5 V), loss is P ≈ I * Vf * (1 − D) ≈ 5 * 0.5 * 0.9 ≈ 2.25 W. Using a low-side MOSFET (Rds_on ≈ 10 mΩ), conduction loss becomes P ≈ I² * R * (1 − D) ≈ 25 * 0.01 * 0.9 ≈ 0.225 W — about 2 W saved.

12 V bus → 1.0–1.8 V / 3.3 V / 5 V

MCU/SoC/FPGA cores and I/O; radios and mixed-signal auxiliaries.

24 V industrial → 5 V / 3.3 V

PLC/servo controllers, motor boards, sensors, HMIs.

Automotive 9–16 V (crank) → logic rails

Consider AEC-Q100 parts, cold-crank and EMI constraints.

Choose a synchronous buck when VOUT is low, IOUT is moderate-to-high, and the design is constrained by efficiency, thermal headroom, and EMI predictability.

Need 3 pin-to-pin options or AEC-Q100 replacements? Submit your BOM (48h).

Synchronous buck topology with HS/LS MOSFETs, inductor, and key current paths.
Fig. 1 — Topology and current paths in a single-phase synchronous buck.

Looking for Boost/Buck-Boost or Multiphase VRs? See Buck-Boost, Boost, Multiphase Controller, or PMBus/PSM.

How a Synchronous Buck Works

HS/LS MOSFET roles & the SW node

The high-side MOSFET connects VIN to the inductor; the low-side MOSFET (replacing the diode) provides the return path when HS is off. The SW node swings between VIN and ground as the inductor current remains continuous (CCM) or reaches zero (DCM).

Dead time, body diode, and reverse recovery

Controller dead-time avoids shoot-through. If too long, the LS body diode conducts longer, raising loss and EMI via reverse-recovery charge (Qrr); if too short, overlap current causes heat. Modern controllers use adaptive dead-time to minimize diode conduction without overlap.

CCM vs DCM

In CCM, inductor current never hits zero—ripple is lower and control is simpler, but light-load efficiency suffers. In DCM, current reaches zero—light-load efficiency improves at the expense of ripple and potential acoustic noise. Minimum on-time and valley detection set practical limits for very small duty cycles.

Where efficiency goes

  • Conduction loss: I²R in HS/LS FETs, inductor DCR, PCB copper.
  • Switching loss: gate charge (Qg), Coss, Qrr events, and driver loss.
  • Frequency trade-off: higher FSW shrinks magnetics and improves transient response but raises switching loss and EMI.
Current-mode vs voltage-mode control in a synchronous buck, loop blocks and signals.
Fig. 2 — Control abstraction: voltage-mode vs current-mode (detailed design in Control & Compensation).
Next: Control & Compensation — why Type-III networks are common with ceramic-cap bucks and how slope compensation prevents sub-harmonic oscillation at high duty.

Control Modes & Compensation

Control overview

Voltage-mode offers low noise and predictable duty but can interact with input filters. Peak current-mode embeds a current loop that improves line transient and eases compensation; at duty > 50% you need slope compensation. Valley current-mode / COT can achieve short minimum on-time and fast step response with simple sensing.

Stability targets (pragmatic)

  • Crossover frequency f_c ≈ Fsw / 5 … Fsw / 10
  • Phase margin ≥ 50–60°, gain margin ≥ 8–10 dB
  • Ceramic outputs (low ESR) often require a Type-III network

Type-II / Type-III placement

LC double pole: ω0 = 1 / √(L·C); ESR zero: ωz,ESR = 1 / (RESR·Cout). Type-II suits higher ESR or added RC-zero outputs. Type-III places two zeros around the LC double pole, one pole above crossover, and a high-frequency pole for roll-off.

Slope compensation (peak current-mode)

To avoid sub-harmonic oscillation at duty > 0.5, inject a down-slope at the PWM ramp. A practical rule is m_slope ≥ (m_down − m_up) / 2, or at least half the inductor current down-slope.

Internal vs external compensation

  • Internal: faster bring-up and layout simplicity, but limited tuning range across L/C spread.
  • External: maximum bandwidth/stability control, ideal for wide VIN/IOUT or tight transient specs.

Feed-forward capacitor (Cff)

Cff in parallel with the upper divider resistor (Vout→FB) improves high-frequency gain and reduces overshoot. Start small (e.g., 10–100 pF) and sweep upward; verify noise coupling and ensure stability at high VIN.

Tip: with ceramic outputs, a Type-III network plus modest Cff often yields clean load steps without sacrificing phase margin.
Continue to Inductor & Switching
Type-III compensation with ESR zero and crossover placement for a synchronous buck regulator.
Fig. 3 — Type-III compensation: zeros near LC double pole; pole above crossover.
Current-mode vs voltage-mode control in a synchronous buck, loop blocks and signals.
Fig. 4 — Control abstractions: voltage-mode, peak/valley current-mode.

Inductor & Switching Network

Ripple target and initial sizing

Start with ΔIL ≈ 0.2–0.4 · IOUT. For a buck, D = VOUT / VIN and an initial estimate L ≈ (VIN − VOUT) · D / (ΔIL · Fsw). Smaller ΔIL lowers AC loss and ripple but increases inductor size and can limit bandwidth.

Saturation and thermal current

Ensure Isat ≥ IOUT,max + ΔIL/2. Thermal RMS current may limit earlier than Isat; DCR and winding geometry strongly affect efficiency and temperature rise.

FSW / loss / EMI trade-offs

  • Higher Fsw shrinks magnetics and improves transient response.
  • But switching loss (Qg, Coss, reverse recovery) and EMI rise; check minimum on-time at high VIN/low VOUT.
  • Typical starting points: 400–800 kHz for mid-current rails; verify with your EMI and thermal budgets.

Synchronous vs diode rectification

Synchronous offers much lower conduction loss at low VOUT/high IOUT but requires careful dead-time control; excessive dead-time leads to body-diode conduction and higher Qrr loss. Schottky is simple and robust but loses efficiency quickly at low VOUT—sometimes useful at ultra-light load to suppress audible noise.

MOSFET selection quick notes

Balance Rds_on vs Qg/Coss. Tune gate resistors and driver strength; manage Miller effect, dv/dt, and ringing. Check package thermal (θJA) in your board stack-up.

Pro tip: co-design magnetics and compensation. ΔIL and Fsw choices change LC poles/zeros and therefore the compensation you need.
Download worksheet & calculator
Inductor selection by ripple current, saturation current, and switching frequency trade-offs in a synchronous buck.
Fig. 5 — ΔIL target vs Fsw: ripple, size, and bandwidth move together.

Layout matters next—see the critical switching loop and return paths in Layout & EMI Notes.

Light-Load Strategy

PFM / skip vs forced-PWM

PFM / skip reduces switching events at light load to cut switching loss—excellent efficiency but higher ripple and potential audible noise. Forced-PWM keeps a fixed switching frequency for low ripple and RF/audio-friendly noise, at the expense of light-load efficiency.

Audible/EMI noise in skip modes

  • Mitigations: raise PFM threshold, limit minimum on-time, or enable forced-PWM on sensitive rails.
  • Add small post-filtering (RC/LC) or a downstream LDO for audio/RF paths.
  • Verify burst frequency minima, ripple (mVpp), and demodulated EMI “birdies”.

Low quiescent current (Iq) & standby

Iq is the regulator’s self-consumption when lightly loaded or idle. µA-class Iq greatly extends battery life in always-on systems. Typical Iq depends on mode and VIN/VOUT conditions—check typical and max ratings.

Mode thresholds & validation

  • Set automatic CCM/DCM/skip thresholds after magnetics and Fsw are chosen.
  • Record wake-up/exit hysteresis, minimum on-time limits, and valley/zero-current detection behavior.
  • Confirm no accidental hiccup during normal light load; hiccup is a fault-mode behavior.
Tip: For rails feeding audio codecs or radios, prefer forced-PWM below a defined load current, then automatically switch to PFM when noise constraints relax.
Continue to Protections & Soft-Start
Light-load operation modes in a synchronous buck: PFM/skip versus forced PWM and their ripple/noise impact.
Fig. 6 — PFM/skip vs forced-PWM at light load.

Protections & Soft-Start

Over-current protection (OCP)

  • Peak OCP: clamps the inductor peak current—fast response, more sensitive to ripple/transients.
  • Average/Valley OCP: closer to real thermal load current—robust with large ΔIL.
  • Foldback current limit: reduces VOUT under heavy overload/short, minimizing thermal stress.

UVLO / OVP thresholds

  • UVLO: defined turn-on/turn-off with hysteresis to prevent chatter during brown-out.
  • OVP: protects against feedback faults and load-dump-like events; verify trip/reset points.
  • Automotive/industrial: consider cold-crank and load-dump; reverse battery and surge handled by front-end protection.

Over-temperature protection (OTP)

OTP trip/clear points should align with your board’s θJA, copper area, and airflow. Avoid thermal oscillation by ensuring margin between steady-state temperature and the OTP threshold.

Soft-start & pre-biased startup

  • Soft-start ramp controls inrush and overshoot; coordinate with Cff and loop bandwidth to prevent ringing.
  • Pre-bias: keep the low-side MOSFET off during ramp if VOUT is already charged; validate power-cycle, brown-out, and hot-reset.
Validation checklist: OCP trigger (room/high temp), UVLO/OVP points, short-circuit hiccup period, soft-start time, and pre-bias success rate.
Next: Layout & EMI Notes Download worksheets
Design trade-offs context for protections: balancing thermal, EMI, and time-to-market.
Fig. 7 — Protections sit within broader thermal/EMI trade-offs.

Controller vs Power Module

When to pick a power module (SiP/integrated): you need compressed BOM, predictable EMI/thermal results, compact layout, and faster time-to-market—ideal for small teams or space-constrained boards.

When to pick a controller (PWM + external FET/inductor): you prioritize cost optimization, tunability, and efficiency/thermal limits at extremes (high VIN, high IOUT, very low VOUT). Freedom to choose magnetics and FETs enables the best W/$ at volume.

Comparison matrix

Dimension Module (with inductor) Module (power stage only, no inductor) Controller (external FET & inductor)
Cost ($/W) Highest, pay for integration Medium Lowest at volume
BOM & sourcing Minimal lines, single vendor Reduced but still source L/C Full BOM, multi-vendor
Area / height Small footprint, height depends on inductor Small footprint, inductor height external Variable; can be larger but flexible
EMI predictability Best—pre-validated layout/magnetics Good—hot loop internalized Depends on your layout & parts
Thermal path Known θJA; easy thermal modeling Good; external inductor spreads heat Fully tunable with packages & copper
Peak efficiency Very good within spec window Good; some FET freedom Highest with optimized FET/inductor
Loop bandwidth / tuning Limited (often internal comp) Moderate (some tuning) Maximum (external comp)
Validation effort Lowest Low–medium Highest (but most flexible)
Lifecycle / AEC options Check series longevity Check stage & passives options Wide choices incl. AEC-Q100

Inductor-included vs power-stage-only

  • Inductor-included module: simplest path to success; pay premium and accept limited tunability/height.
  • Power stage only (no inductor): hot loop is integrated, but you still choose the inductor/Cout—middle ground for cost/control.

3-step decision

  1. Time & resources: tight schedule or small team → start with a module.
  2. EMI/compliance risk: high risk → module; normal → go to step 3.
  3. Cost/efficiency extreme: extreme VIN/IOUT or cost-sensitive → controller; otherwise prototype both and decide by thermal/EMI.
Practical path: prototype with a module to de-risk EMI and bring-up, then cost-down to a controller after requirements stabilize.
Submit your BOM (48h)
Design trade-offs context for protections: balancing thermal, EMI, and time-to-market.
Fig. 8 — Module vs Controller at a glance.

Design Flow & Quick Worksheet

Six steps from targets to validation

  1. Inputs/outputs & ripple: define VIN(min/typ/max), VOUT, IOUT(max), ripple target and transient spec.
  2. Pick Fsw: balance EMI/efficiency/size (often 400–800 kHz to start); mind minimum on-time at high VIN/low VOUT.
  3. Estimate L / ΔIL: set ΔIL ≈ 0.2–0.4 · IOUT, D = VOUT/VIN, then
    L ≈ (VIN − VOUT) · D / (ΔIL · Fsw) as an initial value.
  4. Power stage & drive: trade Rds_on vs Qg/Coss; size gate resistors/driver; check thermal path and θJA.
  5. Compensation draft: target f_c ≈ Fsw/5…Fsw/10, phase margin ≥ 50–60°; with ceramic output caps prefer Type-III; add modest Cff if needed.
  6. Test & calibrate: load/line steps, pre-biased start, short-circuit/hiccup, thermal sweep/derating, EMI and acoustic noise.
Co-design matters: ΔIL and Fsw choices shift LC poles/zeros—revisit compensation after magnetics are finalized.
Download Buck Quick Worksheet (PDF) Thermal Budget Calculator (XLSX) See all resources
Design-flow from targets to compensation tuning with validation steps for a synchronous buck converter.
Fig. 9 — Targets → Fsw/ΔIL → Loss/θJA → Compensation → Validation.

Brand & Part Examples

Below are series-level examples for single-phase synchronous buck controllers, integrated regulators (with FETs), and power modules. We can provide 3 pin-to-pin alternatives within 48h—submit your BOM at the bottom.

Brand Controller Families Integrated Buck Families (with FETs) Power Module / SiP Families
TI TPS40xxx, LM51xx TPS62xxx, TPS56xx TPSM82xx, TPSM53xx
ST L67xx, L59xx L698x, L798x STPMIC / selected µSiP series
NXP PF / PF8xxx (PMIC buck control), VR55xx (auto PMIC) PF50xx / PF81xx PMIC buck rails Module availability varies by platform
Renesas ISL81xx, RAA48xxxx ISL85xx, RAA21xxxx ISL82xxM, RAA22xxM
onsemi NCP15xx / NCP81xx (incl. NCV variants) NCP31xxx / NCV series NCPM series (per VIN/IOUT)
Microchip MIC21xx, MIC28xxx MIC23xxx, MCP1650x, PMIC buck rails MIC2830x, MIC43xx
Melexis Focused on sensors/actuators; dedicated synchronous buck families are limited—ask us for cross-brand replacements.

* Series names are illustrative and commonly encountered. For lifecycle (NRND/EOL) and pin-compatibility, use the RFQ form. See also: Controller vs Power Module.

Submit your BOM (48h)

Layout & EMI Notes

Core principles

  • Minimize the hot loop: keep VIN → HS FET → SW → CIN extremely tight; place input MLCCs next to the power stage.
  • Return path: provide a low-impedance path for high di/dt currents; single-point tie AGND to PGND.
  • Split grounds: route FB/COMP/RT on clean AGND, then join to PGND at the IC’s recommended star point.
  • Sensing: take FB from load or Cout pads; keep it away from the SW copper; use differential routing if available.
  • Hot vs cold nets & shielding: keep SW copper contained; fence with ground pour/vias where needed.

Placement priorities

  1. Input MLCCs (CIN) hugging VIN/GND pins and HS/LS devices.
  2. Inductor close to the power stage for a short SW→L→Cout path.
  3. Output capacitors placed to tightly close the loop.
  4. COMP/FB RC near the IC and far from SW.

Vias, copper & coatings

  • Use multiple stitching vias for grounds and high current paths; expose thermal pads with copper pours.
  • Avoid unnecessary splits near SW; prevent slot antennas.
  • Ensure solder mask coverage on high dv/dt regions; keep BOOT/gate traces short and direct.

EMI & damping

  • Add RC snubbers to tame SW ringing if needed; verify their thermal impact.
  • Check input π-filter interaction with control loop—avoid creating an under-damped peak.
  • Reserve pads for A/B tests (snubbers, gate resistors, small RC across inductor).
Measurement checklist: short ground spring or coax at SW, load/line step captures, LISN pre-scan (conducted EMI), near-field scan, SW overshoot/ringing, FB noise coupling, Cout ripple, thermal images and θJA.
Download EMI & Layout Checklist (PDF) Back to Design Flow
Critical switching loop and ground return layout for synchronous buck converters.
Fig. 10 — Critical hot loop and clean return paths.

FAQs

Current-mode vs voltage-mode — which should I choose?
Current-mode embeds a current loop, simplifying compensation and improving line/transient response. Voltage-mode offers low noise and predictable duty but can interact with input filters. For wide L/C spread or fast steps, start with current-mode; for fixed-frequency, low-noise rails, voltage-mode remains strong. Learn more in Control & Compensation.
Why is Type-III compensation common with ceramic outputs?
Ceramic capacitors exhibit very low ESR, so the ESR zero shifts high and the LC double pole dominates. A Type-III network adds two zeros to tame the double pole and maintains phase margin at a practical crossover. It’s the usual choice for fast, low-ripple designs. See Control & Compensation.
How much slope compensation do I need in peak current-mode?
Above 50% duty, add ramp to prevent sub-harmonic oscillation. A practical rule is injecting at least half the inductor down-slope; many controllers expose a pin or internal setting. Verify stability across VIN/VOUT corners and temperature. Start conservative, then optimize bandwidth. Details in Control & Compensation.
Should I add a feed-forward capacitor (Cff)?
Cff across the upper divider increases high-frequency gain, improving load-step overshoot. Begin small (10–100 pF), sweep upward, and re-verify phase margin and noise pickup at high VIN. Cff is optional but effective with ceramic outputs when properly tuned. Learn how to place it in Control & Compensation.
How do I select ΔIL and the inductor value?
Start with ΔIL around 20–40% of IOUT. Use D = VOUT/VIN and estimate L ≈ (VIN − VOUT)·D/(ΔIL·Fsw), then refine for thermal and transient goals. Smaller ΔIL reduces ripple but enlarges L; larger ΔIL improves bandwidth yet raises EMI. See Inductor & Switching.
Which limits first — inductor Isat or thermal Irms?
In practice, RMS heating often limits earlier than Isat. Ensure Isat ≥ IOUT,max + ΔIL/2, but also model DCR and core loss at your Fsw and ripple. Validate with temperature rise on your board, not only datasheet curves. More guidance in Inductor & Switching.
PFM/skip produces audible noise — how to mitigate?
Raise the PFM threshold, limit minimum on-time, or switch to forced-PWM on sensitive rails. Add small RC/LC post-filters, or a quiet LDO downstream for audio/RF paths. Confirm the lowest burst frequency and ripple under worst-case VIN and temperature. See Light-Load Strategy.
What does “low Iq” mean, and how do I verify standby?
Low Iq is the regulator’s self-consumption at light load or idle. µA-class parts dramatically extend battery life. Measure Iq across modes and VIN/VOUT corners; note typical versus maximum ratings. Beware measurement artifacts from instruments. More tips in Light-Load Strategy.
Peak vs average/valley OCP — when, and why foldback?
Peak OCP reacts fastest but is ripple-sensitive. Average/valley reflects thermal reality and tolerates large ΔIL. Foldback limits power during overload/short, protecting silicon and copper. Ensure startup exceptions and recovery behavior are well characterized. Learn the options in Protections & Soft-Start.
Soft-start and pre-biased startup — what should I check?
Soft-start shapes inrush and overshoot; coordinate with compensation and Cff to avoid ringing. For pre-bias, the low-side MOSFET must remain off until regulation, preventing sink current. Validate power-cycle, brown-out, hot-reset, and brief interruptions. See Protections & Soft-Start.
Module or controller — how do I decide quickly?
Choose a module for compressed BOM, predictable EMI/thermal, and faster time-to-market; pick a controller for ultimate efficiency, tunability, and cost at volume. Prototype with a module, then cost-down to a controller once requirements stabilize. Compare options in Module vs Controller.
What crossover frequency and margins should I target?
A practical starting point is fc ≈ Fsw/5 … Fsw/10 with phase margin ≥ 50–60° and gain margin ≥ 8–10 dB. Ceramic outputs often benefit from Type-III networks. Verify across VIN/IOUT and temperature. More in Control & Compensation.
How does Fsw affect efficiency and EMI?
Higher Fsw shrinks magnetics and improves transient response but increases switching loss (Qg, Coss, Qrr) and EMI. Check minimum on-time limits at high VIN/low VOUT. Many designs start around 400–800 kHz and iterate with thermal/EMI data. See Inductor & Switching.
How do I minimize the hot loop and manage return paths?
Keep VIN→HS→SW→CIN extremely tight; place MLCCs next to the power stage. Route feedback on clean AGND and join at a single star point. Fence SW copper with ground vias as needed, and reserve pads for snubbers. Practical tips in Layout & EMI Notes.

Resources & RFQ

Downloads

We return 3 pin-to-pin alternatives within 48 hours, including lifecycle notes.

Submit Your BOM & Contact

Prototype with a module, or cost-down with a controller—your call. We’ll suggest both paths.