A flyback converter is one of the most widely used isolated switch-mode power supply topologies for output powers roughly below 100–150 W, and its design revolves around a handful of tightly coupled decisions: the transformer (really a coupled inductor), the operating mode, the control loop, and the snubber or clamp that handles leakage energy. Getting any one of those wrong can mean excessive voltage spikes, poor efficiency, or a power supply that oscillates instead of regulating. The topology looks simple on a schematic, but the interactions between magnetics, switching behavior, and feedback make it one of the more nuanced converters to design well.
Why the Flyback Is So Popular for Low-to-Medium Power
The flyback converter stores energy in the magnetizing inductance of a transformer during the switch-on time, then delivers that energy to the output during the switch-off time. This store-and-forward behavior is what distinguishes it from a forward converter, which transfers energy directly while the switch is on. The practical payoff is simplicity: a flyback needs only one magnetic component and one main power switch, with no output inductor on the secondary side. That makes it compact and cheap, which is why it dominates in phone chargers, laptop adapters, standby supplies, and auxiliary rails inside larger systems.
The topology also makes it straightforward to generate multiple output voltages from a single transformer by adding extra secondary windings. Each winding produces a voltage proportional to its turns ratio, and only one output needs tight regulation through the feedback loop. The other outputs are “semi-regulated,” tracking the regulated output loosely. A design study for a universal AC adapter, for instance, demonstrated a two-output flyback producing 15 V and 32 V outputs from a single gapped-core transformer operating at 85 kHz, with a methodology for selecting duty cycle and magnetizing inductance to keep both outputs in specification.1IET Power Electronics. Magnetising inductance of multiple‐output flyback dc–dc convertor for discontinuous‐conduction mode
The cost of all this simplicity is that the single switch sees high voltage stress (the input voltage plus the reflected output voltage plus the leakage spike), and the pulsating current on both sides of the transformer makes EMI filtering more demanding than in topologies with continuous current flow. These trade-offs are manageable at lower power levels but become increasingly painful above 150 W or so, which is where half-bridge and full-bridge topologies tend to take over.
Continuous Versus Discontinuous Conduction Mode
The single most consequential early decision in flyback converter design is whether to operate in continuous conduction mode (CCM) or discontinuous conduction mode (DCM). The difference is about what happens to the current in the transformer’s magnetizing inductance during each switching cycle. In CCM, the magnetizing current never reaches zero before the next switch-on event; in DCM, it falls to zero and stays there for a dead time before the switch turns on again.
DCM is the more common choice at lower power levels, and for good reason. The transformer core fully resets each cycle, so there is no residual energy to worry about. The secondary-side diode turns off naturally at zero current, which eliminates reverse-recovery losses and simplifies the design. Peak currents are higher than in CCM for the same average power, which means the switch and transformer must handle larger current spikes. But the control loop is easier to stabilize, which matters when you want a simple, cheap controller IC.
CCM keeps peak currents lower and generally improves efficiency at higher power, but it introduces a control challenge that has earned its own reputation: the right-half-plane zero. Designing the magnetizing inductance is the lever that determines which mode you land in. A smaller magnetizing inductance pushes the converter toward DCM; a larger one keeps current flowing continuously. The boundary between the two modes shifts with load, so a converter designed for CCM at full load may slip into DCM at light load, and the control loop needs to handle both gracefully.
For multiple-output designs operating in DCM, the magnetizing inductance has a maximum value that must not be exceeded if discontinuous operation is to be maintained across the full load range. Methods exist to derive this boundary rigorously for multi-winding transformers and to select the duty cycle based on the rated output voltages of each winding.2IET Power Electronics. Magnetising inductance of multiple‐output flyback dc–dc convertor for discontinuous‐conduction mode
Designing the Transformer
Calling the flyback’s magnetic component a “transformer” is a useful convention, but it can mislead. A conventional transformer transfers energy instantaneously from primary to secondary; a flyback transformer stores energy in its magnetizing inductance during the primary-side conduction interval and releases it during the secondary-side conduction interval. The air gap in the core is not a manufacturing defect. It is the energy storage element. Without it, the core would saturate almost immediately because the magnetizing inductance would be far too high for the flux swing involved.
The core selection starts with the energy storage requirement. You need to store enough energy each cycle to deliver the required output power, accounting for losses. That energy is proportional to the magnetizing inductance and the square of the peak magnetizing current. A ferrite core with an appropriate air gap is the standard choice. The gap length sets the magnetizing inductance for a given number of primary turns, and it also determines how much DC flux the core can handle before saturating. Larger gaps store more energy but require more turns to achieve a given inductance, which increases copper losses and may not fit in the winding window.
Winding layout matters more than many designers initially expect. The primary and secondary windings are magnetically coupled, but not perfectly. Leakage inductance, the portion of flux that links one winding but not the other, is the enemy. It stores energy that cannot transfer to the output when the switch turns off, producing a voltage spike on the drain of the MOSFET. Interleaving primary and secondary layers, using wider bobbins, and minimizing the number of winding layers all reduce leakage. Sandwich windings, where the secondary is placed between two halves of the primary, are a common and effective technique.
For designs with multiple outputs, the transformer winding design becomes more involved. A three-winding transformer (one primary, two secondaries) for a gapped core requires careful allocation of the winding window area among the windings, and the turns ratios must reflect the desired output voltages while accounting for diode drops and resistive losses in each path.3IET Power Electronics. Magnetising inductance of multiple‐output flyback dc–dc convertor for discontinuous‐conduction mode
The Right-Half-Plane Zero and Loop Stability
If you have ever tried to push a flyback converter’s bandwidth in CCM and found the output ringing or oscillating, the right-half-plane (RHP) zero is almost certainly the culprit. This is a feature of the flyback’s transfer function in CCM that limits how fast the control loop can respond. When the duty cycle increases to deliver more current, there is an initial period where the output voltage actually dips before rising, because the switch is on longer and the secondary diode conducts for a shorter interval. The control loop sees this as the output moving in the wrong direction, which is exactly what a right-half-plane zero does mathematically.
The practical consequence is a hard ceiling on crossover frequency. You cannot compensate your way past the RHP zero the way you can with a left-half-plane zero. The loop gain must roll off well below the RHP zero frequency, which means the converter responds slowly to load transients. For applications that need fast dynamic response, this is a real limitation. Research into modified flyback topologies has explored ways to alleviate the RHP zero’s impact on small-signal loop stability and dynamic response, including circuit modifications that alter the energy transfer path during the switching cycle.4Engineering Science and Technology, an International Journal. Modeling, design and control of a modified flyback converter with ability of right-half-plane zero alleviation in continuous conduction mode
In DCM, the RHP zero moves to a much higher frequency or disappears entirely from the frequency range of interest, which is one of the main reasons designers choose DCM despite its higher peak currents. If your application tolerates the higher peak currents and slightly lower efficiency, DCM sidesteps the whole RHP zero headache. If CCM is necessary for efficiency at higher power, the standard approach is to accept a relatively low bandwidth (often in the range of a few kilohertz for a converter switching at 65–150 kHz) and use output capacitors with enough bulk to ride through load steps while the loop catches up.
Snubbers, Clamps, and Leakage Energy
Every flyback converter needs a way to deal with the energy stored in the transformer’s leakage inductance. When the primary-side switch turns off, the magnetizing current commutates to the secondary winding and starts delivering energy to the output. But the leakage inductance current has nowhere to go. It cannot transfer to the secondary because, by definition, it is not coupled. That current, trapped in the leakage inductance, causes the voltage on the switch drain to spike upward until something absorbs the energy or the switch breaks down.
The simplest solution is an RCD (resistor-capacitor-diode) snubber across the primary winding. The capacitor absorbs the leakage energy, the resistor dissipates it as heat between switching cycles, and the diode steers the current path. This is cheap and works well enough for many designs, but the energy lost in the snubber resistor directly reduces efficiency. At higher power levels or in efficiency-sensitive applications, an active clamp circuit can recover the leakage energy and return it to the input or output, improving efficiency by a few percentage points.
Sizing the snubber or clamp is not optional. Undersizing it lets the drain voltage exceed the switch’s breakdown rating. Oversizing wastes power and may change the converter’s operating behavior. The clamp voltage should be chosen to limit the peak drain voltage to a safe margin below the MOSFET’s rated breakdown, while not clamping so aggressively that excessive current flows through the snubber during normal operation. A common design target is to keep the peak drain voltage at roughly 80–90% of the MOSFET’s voltage rating to leave margin for transients and component variation.
Startup and Auxiliary Bias Circuits
A flyback converter needs power for its own controller IC before it can start switching. This creates a chicken-and-egg problem: the controller needs a bias supply to operate, but the bias supply is usually derived from the converter’s own transformer. The standard solution is a startup circuit that trickles current from the high-voltage input rail through a high-value resistor (or a depletion-mode MOSFET acting as a current source) into the controller’s supply capacitor. Once the converter starts switching, an auxiliary winding on the transformer takes over and provides the bias voltage, and the startup resistor is effectively disconnected or its power dissipation becomes negligible relative to the converter’s output.
This auxiliary winding approach has an important benefit beyond just startup. A well-designed self-biasing auxiliary winding can improve overall efficiency at no load and light load, where the ratio of housekeeping power to output power is at its worst. In one investigation of a modified double-switch flyback-forward topology for low-power applications, the auxiliary winding was designed specifically for self-biasing to improve efficiency in the no-load condition, alongside a startup voltage regulator circuit and a modified bootstrap gate driver.5IET Power Electronics. Design, modelling, and implementation of a modified double‐switch flyback‐forward converter for low power applications
Light-load and no-load efficiency has become increasingly important as energy standards (such as the U.S. Department of Energy’s Level VI and the EU’s ErP Lot 6 regulations) now impose strict limits on standby power consumption. A flyback converter in a phone charger or laptop adapter spends most of its life at no load or very light load, so the startup and bias circuits’ power consumption can dominate the efficiency picture during those periods. Many modern controller ICs address this by entering a burst-mode or pulse-skipping mode at light load, where the converter switches only intermittently to maintain the output voltage, dramatically reducing switching losses and bias current draw.
Output Rectification and Synchronous Switching
On the secondary side, a diode rectifies the pulsating current delivered from the transformer. For low output voltages, the forward voltage drop of this diode becomes a significant fraction of the output, eating directly into efficiency. A standard silicon diode might drop 0.7–1.0 V, which on a 5 V output represents a 15–20% penalty before you even account for other losses. Schottky diodes reduce this to roughly 0.3–0.5 V but still leave a noticeable loss.
Synchronous rectification replaces the output diode with a MOSFET that is turned on during the secondary conduction interval. A well-chosen MOSFET can have an effective voltage drop of 50 mV or less at the load currents typical in flyback designs, which represents a substantial efficiency gain at low output voltages. The control of the synchronous rectifier MOSFET adds complexity, requiring either a dedicated SR controller IC or careful timing derived from the secondary voltage waveform. In DCM, turning off the synchronous MOSFET at the right instant is critical: if it stays on too long after the current reaches zero, current flows backward from the output capacitor through the MOSFET and back into the transformer, wasting energy and potentially disrupting the converter’s operation.
For output voltages above about 12–15 V, the diode drop is a smaller fraction of the total, and the added cost and complexity of synchronous rectification may not be justified. Most designers use Schottky diodes at those voltage levels unless the application has very tight efficiency requirements.
Selecting the Switching Frequency
Switching frequency is one of the first parameters chosen, and it affects almost every other design decision. Higher frequencies shrink the transformer because less energy needs to be stored per cycle (the converter delivers energy to the output more frequently, in smaller packets). This allows smaller cores and fewer turns, which reduces size and cost. The trade-off is that switching losses increase with frequency: the MOSFET dissipates more power during each turn-on and turn-off transition, and core losses in the transformer scale with frequency as well.
Common flyback switching frequencies range from about 50 kHz to 150 kHz for offline (AC mains input) designs. Some newer designs push higher, into the 200–300 kHz range, using advanced MOSFET technologies or gallium nitride (GaN) switches that have much lower switching losses than conventional silicon MOSFETs. A design for a universal power supply adapter, for example, used a switching frequency of 85 kHz, which sits comfortably in the middle of the typical range and balances transformer size against switching losses for a moderate-power application.6IET Power Electronics. Magnetising inductance of multiple‐output flyback dc–dc convertor for discontinuous‐conduction mode
Quasi-resonant (QR) flyback converters vary their switching frequency to turn on the MOSFET at the valley of the drain voltage ringing that occurs after the secondary current reaches zero. This valley switching reduces turn-on losses and conducted EMI. The downside is that the switching frequency changes with input voltage and load, which can make EMI filter design more complex since the noise energy spreads across a wider frequency band instead of concentrating at a fixed fundamental and its harmonics.
EMI and Layout Considerations
The flyback topology is inherently noisy. The pulsating currents on both primary and secondary sides produce both conducted and radiated emissions. The high-frequency ringing at the drain node after turn-off, driven by the resonance between the leakage inductance and the parasitic capacitance of the switch and transformer, is a particularly strong source of radiated EMI in the tens of megahertz range.
Good PCB layout is not just helpful, it is a prerequisite for a working design. The primary-side high-current loop (from the input capacitor through the transformer primary and the MOSFET back to the input capacitor) and the secondary-side high-current loop (from the transformer secondary through the rectifier and output capacitor back to the transformer) should each be as small and tight as possible. Every square centimeter of loop area acts as a transmitting antenna. The gate drive trace should be short and routed away from sensitive signal nodes. The feedback optocoupler (used to cross the isolation barrier in most flyback designs) should be placed close to the controller IC, and its ground reference should be carefully managed.
An input EMI filter, typically a combination of common-mode and differential-mode inductors with X and Y capacitors, is required for any flyback converter connected to AC mains. The filter design interacts with the converter’s switching frequency and the snubber design: a converter with a well-damped snubber and a clean drain waveform needs less filtering than one with large high-frequency ringing. Starting the layout and EMI filter design early rather than treating them as afterthoughts can save considerable debugging time later.
Quasi-Resonant and Active-Clamp Variants
The basic hard-switching flyback has spawned several variants aimed at improving efficiency or reducing EMI. Quasi-resonant operation, mentioned earlier in the context of switching frequency, is one of the most common. By detecting the drain voltage valleys and timing the turn-on to coincide with a valley, the energy stored in the MOSFET’s output capacitance at turn-on is minimized. Many modern flyback controller ICs include valley detection circuitry as a standard feature.
Active-clamp flyback (ACF) designs go further. Instead of an RCD snubber that wastes the leakage energy as heat, an active-clamp circuit uses an auxiliary switch and capacitor to capture the leakage energy and recycle it. The clamp capacitor voltage resonates with the leakage and magnetizing inductances in a way that allows the main switch to turn on at or near zero voltage, eliminating turn-on switching losses almost entirely. This approach, sometimes called zero-voltage switching (ZVS), can push flyback efficiency above 93–94% in well-optimized designs. USB Power Delivery chargers rated at 45–65 W have driven widespread adoption of ACF topology, since the combination of small size, high efficiency, and the availability of integrated ACF controller ICs has made the design accessible to more engineers than the older discrete approaches.
GaN switches have accelerated this trend. Their lower output capacitance and zero reverse-recovery charge compared to silicon MOSFETs make them natural partners for active-clamp and quasi-resonant flyback designs. The combination of GaN switches and ACF control has enabled USB-C chargers that deliver 65 W from packages roughly the size of a standard 5 W phone charger from a decade ago, a compression of size that would not have been practical with silicon switches and hard-switching topologies alone.