A 5 V, 20 A output with a Schottky rectifier burns more power in the diode than in the transformer. That single fact drove synchronous rectification (SR) from a premium feature to a standard requirement in server, telecom, and charger designs. This article explains how SR MOSFETs replace the output diode, walks through a loss comparison with real numbers, and lists the parameters that separate a good SR MOSFET from a mediocre one.
The Efficiency Problem in Low-Voltage Outputs
Consider a power supply delivering 5 V at 20 A—100 W of output power. A Schottky rectifier with a forward drop of 0.4 V at that current dissipates roughly 8 W of conduction loss, about 8% of the output. Push the output to 1.2 V or 1.8 V for modern CPUs and the diode drop becomes a third of the output voltage; the rectification stage can waste more energy than the switching stage, transformer, and magnetics combined.
The physics is unforgiving: a diode’s forward voltage is tied to the semiconductor junction built-in potential and current density, so no silicon Schottky process can shrink it to millivolts. A MOSFET, however, conducts through a resistive channel; its “forward drop” is I × RDS(on), which can be tuned down with a bigger die, better technology, and the right package.
How Synchronous Rectification Works
In a synchronous rectifier, the output diode is replaced by a MOSFET (or a pair of MOSFETs) driven by an SR controller. The controller monitors the transformer winding voltage or the MOSFET drain-source voltage and turns the SR switch on exactly when the body diode would conduct, then turns it off before the voltage reverses.
The control challenge is timing:
- Turn-on timing. The SR MOSFET must turn on when its body diode starts conducting—early enough to capture most of the conduction interval, late enough to avoid shoot-through with the primary switch.
- Turn-off timing. It must turn off before the current reverses, otherwise the output capacitor is discharged back into the transformer.
- Dead time. During the transition, current flows through the body diode for a few tens of nanoseconds. That is acceptable loss, but it is why SR controllers monitor the drain voltage with fast comparators.
The result: the conduction loss drops from “diode forward drop × current” to “I² × RDS(on),” which at low voltage and high current is dramatically smaller.
A Loss Comparison With Real Numbers
To see the difference, compare a 40 V Schottky diode and a 40 V SR MOSFET in a 5 V, 20 A output, both operating at a 50% rectifier duty cycle and a junction temperature of 100 °C.
| Item | Schottky diode | SR MOSFET |
|---|---|---|
| Conduction parameter | VF ≈ 0.45 V at 20 A (hot) | RDS(on) ≈ 5 mΩ at 100 °C |
| Conduction loss (50% duty) | 0.45 × 20 × 0.5 = 4.5 W | 20² × 0.005 × 0.5 = 1.0 W |
| Switching/recovery loss | Small at line frequency, larger in high-frequency SR | Body-diode dead-time loss ≈ 0.2–0.4 W typical |
| Total rectification loss | ~4.5–5 W | ~1.2–1.4 W |
| Efficiency impact at 100 W out | ~4.5–5 points | ~1.2–1.4 points |
The numbers are illustrative, not a guarantee for any specific part—real values depend on the exact diode, MOSFET, temperature, and dead time—but the order of magnitude is the point: synchronous rectification recovers several efficiency points at low output voltages. At 1.8 V and higher currents, the gap widens further because the diode drop becomes a larger fraction of the output voltage.
Diagnosing SR Dead-Time Loss
The SR stage’s most common efficiency leak is dead time: the interval when neither the SR MOSFET nor the rectifier path is conducting properly, so the current flows through the body diode at a fixed forward drop. The table below ties the symptom you would measure to its cause and the fix:
| Symptom on the bench | Likely cause | Fix |
|---|---|---|
| Drain-source waveform shows a long flat conduction interval at turn-on | Controller turning on late; sensing window misaligned | Shorten the turn-on delay or adjust the drain-voltage threshold |
| Efficiency drops sharply at light load | SR still active below the load where gate-drive loss pays back | Enable diode-emulation mode below the threshold |
| Ringing at the drain node after turn-off | Body-diode recovery interacting with loop inductance | Add a small parallel Schottky or tune the blanking time |
| One of several paralleled SR parts runs hotter | Asymmetric gate drive or layout | Symmetrize gate traces and current paths; verify sharing on the bench |
| Shoot-through current at mode transitions | SR on-time extends past current reversal | Tighten the minimum on-time and off-time limits |
The parameters behind the diagnosis remain the same four: hot RDS(on) for conduction loss, gate charge and threshold/plateau against the controller’s output, body-diode forward drop and recovery for the dead-time interval, and the package thermal path that decides the practical current limit. The structure choice behind the low RDS(on) at these voltages is covered in the companion article on trench vs planar MOSFETs in this series.
Layout and Thermal Design for SR Stages
The SR MOSFET sits in the output current loop, which is also the most loss-sensitive loop on the board:
- Keep the gate loop short. The SR controller’s gate output to the MOSFET gate and back to the controller source must be a small, tight loop to avoid ringing and false triggering.
- Minimize the drain-source loop inductance. The commutation loop between the transformer winding, SR MOSFET, and output capacitor determines the voltage spike at turn-off. A low-inductance layout lets the design use a lower-voltage MOSFET, which has better RDS(on) per cost.
- Put copper where the heat is. The SR MOSFET’s thermal pad should connect to a solid copper plane with vias down to inner layers. The datasheet’s thermal resistance assumes a defined copper area; a smaller pad raises the junction temperature quickly.
- Measure the dead time. An oscilloscope across the SR MOSFET drain-source during full load shows the body-diode conduction interval. If it is long, the controller timing or the gate drive is not optimal.
A Worked Selection Example
Design target: 12 V input, 3.3 V / 30 A output buck converter with synchronous rectification, 500 kHz.
- Voltage class. The low-side SR MOSFET sees about the input voltage plus ringing. A 30–40 V part is typical; 40 V gives margin for layout spikes without sacrificing RDS(on).
- RDS(on) target. With 30 A and a target SR conduction loss below 1.5 W at 100 °C, RDS(on) hot should stay below about 1.7 mΩ at the rectifier duty cycle. That points to a low-voltage trench MOSFET in a DFN or DPAK-class package with generous copper.
- Gate charge. At 500 kHz, gate drive loss is Qg × Vgs × fsw. A part with 40 nC at 4.5 V costs about 0.09 W at 500 kHz—acceptable, but a lower-Qg part helps if the controller is small.
- Package check. Verify the thermal resistance and footprint against the board’s copper budget before finalizing.
- Prototype verification. Measure dead time, drain ringing, and efficiency at full load and at light load; the diagnosis table above is the checklist for interpreting what you see.
For the discrete MOSFET families and packages available for this kind of stage, the SMPS application section shows how rectifier, MOSFET, and bridge choices combine in a complete supply design, and the Documents section carries the datasheets and package outlines needed for footprint design.
SR Controller Selection and Light-Load Operation
The MOSFET is only half of a synchronous rectifier; the controller’s sensing and timing determine how much of the theoretical gain is realized. SR controllers fall into two sensing families:
- Voltage-sensing controllers watch the SR MOSFET’s drain-source voltage to decide when to switch. They are simple and work well in flyback and forward topologies, but the sensing window must be tuned to avoid false turn-on from ringing.
- Current-sensing controllers use a sense resistor or the transformer’s secondary current information. They handle wider load ranges but add a sensing element and its loss.
Light-load operation is where SR designs often fail to deliver. At low output current, the benefit of synchronous rectification shrinks because the fixed gate-drive loss and the controller’s quiescent current dominate. Many designs disable the SR MOSFET below a load threshold and let the body diode or a parallel Schottky conduct instead—the “diode emulation” mode. The transition between modes must be smooth, because repeated switching between diode and synchronous modes at the boundary can produce audible noise and efficiency dips.
When selecting an SR controller, check the minimum on-time, the blanking time after turn-on (to ignore the initial ringing), and the maximum frequency against the converter’s operating range. The controller’s propagation delay adds directly to the dead time, so a fast, well-blanked controller is worth more than a slightly lower-Qg MOSFET in many designs.
Frequently Asked Questions
Why not just use a lower-drop Schottky diode? Silicon Schottky forward voltage is bounded by the junction physics; at low output voltages the drop still costs several points of efficiency. SR MOSFETs reduce conduction loss to I²R and are standard for outputs below 12 V.
Does synchronous rectification work at all output voltages? It pays off most below ~12 V and at high current. At higher voltages, the diode drop is a smaller fraction of the output, and the added controller and driver complexity may not be justified.
What happens during SR MOSFET dead time? The body diode conducts during the transition. The loss is small if the dead time is short, but a long dead time or poor timing adds loss and can cause reverse current. Fast drain-voltage sensing in the SR controller keeps this interval tight.
Can one SR MOSFET replace a dual-diode center-tap rectifier? Yes—a center-tap secondary typically uses two SR MOSFETs, one per half-winding, driven by the same controller with proper timing.
Why does efficiency drop at light load even with synchronous rectification? The gate-drive energy and the controller’s quiescent current are fixed costs, so the SR benefit shrinks as load current falls. Diode-emulation mode, which switches the SR off below a threshold, recovers most of the light-load efficiency.
Making the SR Decision
Synchronous rectification turns the output diode’s fixed forward drop into a resistive loss you can engineer down. The gain is largest at low output voltage and high current, where a few milliohms of RDS(on) replace several watts of diode loss. The decision sequence is short: confirm the output voltage makes SR worthwhile, pair the controller with a MOSFET whose threshold and gate charge fit its drive, then use the dead-time diagnosis table on the bench. Measured dead time and light-load behavior—not the datasheet summary—are the acceptance criteria.