Flyback Converter

A transformer that isn't really a transformer

The headline component of a flyback converter is its transformer. Open one up and you see a primary and a secondary winding wrapped on a shared ferrite core, the same way a 50 Hz mains transformer is wound. But it does not work like a mains transformer at all.

In a mains transformer, the primary current flows whenever the secondary current flows. Energy passes through the core in real time, with the primary winding sucking it from the source and the secondary winding pushing it into the load at the same instant. The energy stored momentarily in the core is small — just whatever leakage and magnetising inductance demands.

A flyback transformer is the opposite. The primary conducts when the secondary doesn't, and vice versa. While the primary is energised, energy accumulates in the core's magnetic field. When the primary opens, the field collapses, the polarity across the secondary reverses, the output diode forward-biases, and that stored energy flushes out through the secondary into the load capacitor. Energy passes through the transformer in two phases, not one. Strictly speaking, the flyback transformer is a coupled inductor with a turns ratio.

This is why every flyback transformer has a small air gap milled into its centre limb. The air gap is what stores the energy — ferrite saturates at about 0.3 T and cannot hold much by itself. By breaking the magnetic circuit with a few hundred microns of air, the effective permeability drops by 1000× or more, and the field strength H can rise enough to store useful energy density without saturating the ferrite. About 95 % of the cycle's energy lives in that 0.2 mm slot of air.

The simplest isolated topology

                            air gap
       ┌───── N_p ─────┐│┌───── N_s ─────┐
   V_in ●               │││               ● ── D ──┬── V_out
       │               │││               │        │
       │    primary    │││   secondary   │       ═ C  ⏚ load
       │               │││               │        │
       └─[SW]───────────┘│└──────────────────┘────┴── GND_sec
       (primary GND)        (isolated GND)

    SW = primary-side MOSFET
    D  = output rectifier diode (Schottky or PN ultrafast)
    C  = output capacitor

The "dots" on the windings — by convention the points where instantaneous voltage shares the same sign — are arranged in anti-phase. While the switch is on, the primary's dotted end pulls positive (V_in across primary), so the secondary's dotted end goes positive too, and the diode (with its anode tied to the un-dotted end) reverse-biases. While the switch is off, the magnetising inductance flips polarity, the dotted ends reverse, the secondary's dotted end now sits at the secondary GND through the inductor, and the un-dotted end (the diode anode) flies positive — the diode forward-biases.

The two phases of every switching cycle

Switch closed (on-time, fraction D). The primary winding sees V_in. Primary current ramps up at dI_p/dt = V_in / L_pri. Magnetic field in the air gap grows; energy ½ L_pri I_p² accumulates. On the secondary side the diode is reverse-biased; the output capacitor alone delivers load current. Output voltage sags slightly under load.

Switch open (off-time, fraction 1 − D). The FET turns off. Primary current cannot flow — but the magnetic field cannot collapse instantaneously without dumping infinite voltage somewhere. The collapsing field re-appears across the secondary winding (where there is a current path: through the now-forward-biased diode into the load). Secondary current starts at I_p · (N_p/N_s) and ramps down at dI_s/dt = V_out · (N_s/N_p) / L_sec. The capacitor recharges, the load draws steadily, and the cycle ends when the FET closes again.

In CCM (continuous-conduction mode), the secondary current never quite reaches zero before the next switch turn-on. In DCM (discontinuous-conduction mode) — common at light load — the secondary current rings to zero and sits idle until the next turn-on. CCM gives lower peak currents; DCM is easier to compensate and is the default for low-power off-line designs.

The voltage transfer relation

The volt-second balance is applied to the primary inductance:

primary volt-seconds:  V_in · D · T  =  (V_out · (N_p/N_s)) · (1 − D) · T

solving:               V_out  =  V_in · (N_s/N_p) · D / (1 − D)

So the ratio depends on two design knobs: the turns ratio (sets a coarse step) and the duty cycle (fine regulates). A typical 230 V mains flyback with a 6:1 step-down (N_p = 30, N_s = 5) at D = 0.4 gives V_out ≈ 230 × (1/6) × (0.4/0.6) ≈ 25 V — appropriate for a laptop charger pre-regulator. The same transformer at D = 0.2 gives ~10 V; at D = 0.55 gives ~50 V.

Worked example: 230 V mains → 5 V / 2 A USB charger

Design a 10 W USB charger flyback at f_sw = 100 kHz, V_in = 325 V DC (after rectification + PFC of European mains), V_out = 5 V at 2 A, target DCM operation, η = 80 %, max duty D_max = 0.45.

Turns ratio. Pick N_p/N_s to allow D_max under the lowest input voltage. From the steady-state relation: N_p/N_s = (V_in / V_out) · D / (1 − D). At D = 0.45, V_in = 325, V_out = 5 (plus diode V_F ≈ 0.4 V, so V_out + V_F ≈ 5.4):

N_p/N_s = (325 / 5.4) · 0.45 / 0.55 ≈ 49

Round to a manufacturable ratio: N_p = 50 turns, N_s = 1 turn (impractical),
or N_p = 100, N_s = 2 — round-trip more turns gives better coupling.
Production charger typically uses N_p = 80, N_s = 4 (20:1), bias winding N_aux = 6.

Primary inductance. For DCM operation: L_pri = (V_in · D)² / (2 · P_in · f_sw):

P_in = P_out / η = 10 / 0.8 = 12.5 W
L_pri = (325 · 0.45)² / (2 · 12.5 · 100 000)
      = (146.25)² / 2.5e6
      = 8.6 mH

Pick a commercial 8.5 mH–10 mH flyback transformer with 200 µH leakage and ~0.2 mm gap on a EFD20 or EE19 ferrite. Peak primary current: I_pk = V_in · D / (L_pri · f_sw) ≈ 0.17 A. Peak secondary current: I_pk · (N_p/N_s) ≈ 3.4 A.

Output capacitor. For 100 mV ripple at 2 A load, with secondary current peaking at 3.4 A in pulses: C_out ≥ I_out · D / (ΔV · f_sw) ≈ 2 · 0.45 / (0.1 · 100 000) = 90 µF. Use two 47 µF low-ESR aluminium-electrolytic caps in parallel.

Snubber. Leakage energy = ½ · L_leak · I_pk² ≈ ½ · 200 µH · 0.17² ≈ 2.9 µJ per cycle. At 100 kHz that's 290 mW dissipated in the RCD snubber resistor. Pick R_snub ≈ 100 kΩ, C_snub ≈ 1 nF, diode UF4007.

Efficiency. Combined switch + snubber + magnetic + diode + capacitor-ESR losses total around 2.5 W on 12.5 W input — η ≈ 80 % at full load. Add a synchronous rectifier on the secondary and η climbs to 87 %.

Flyback variants you'll meet in the wild

Variant	Switch count	Snubber	Efficiency	Typical use
Single-switch flyback (asynchronous)	1 + diode	RCD or zener-TVS	75–82 %	Phone chargers, standby rails, white goods
Synchronous-rectifier flyback	1 + SR FET	RCD	85–88 %	Mid-range USB-C, monitor PSUs
Active-clamp flyback (ACF)	2 + SR FET	Active clamp recycles leakage	92–94 %	High-density 65 W and 100 W USB-PD chargers, GaN-based
Two-switch flyback	2 + diode	Two clamp diodes (no snubber)	83–88 %	Industrial PSUs above 100 W, three-phase auxiliary supplies
Quasi-resonant flyback (QR)	1 + diode	RCD, valley switching	82–87 %	LED drivers, mid-power standby
Interleaved flyback	2 + 2 diodes	Per-phase RCD	83–87 %	Rare; higher power with reduced ripple

The single-switch flyback dominates below 30 W. Above that, active-clamp variants take over to recycle leakage-inductance energy and shrink the magnetics. GaN-FET active-clamp flybacks are the technology behind 65 W chargers the size of a matchbox; modern Anker, Apple, and Samsung USB-PD chargers all use this architecture.

Leakage inductance — the flyback's worst enemy

The primary and secondary windings are never perfectly coupled. A fraction of the primary inductance — the leakage inductance L_leak, typically 1–3 % of L_pri — represents flux lines that don't reach the secondary. When the FET turns off, the magnetising inductance dumps its energy through the secondary in the normal way, but the leakage inductance has no such path. Its stored energy must go somewhere.

Absent a clamp, the leakage rings up the FET drain to potentially destructive voltages — easily 2 × V_in plus 200–500 V of ringing. Two solutions dominate:

• RCD snubber. A diode + capacitor clamps the drain voltage at a chosen ceiling; a resistor in parallel discharges the cap each cycle, burning the leakage energy as heat. Cheap, robust, lossy. Typical loss: 0.5–2 W in a 30 W flyback. The resistor wattage is sized to the leakage energy × f_sw.
• Active clamp. A second FET + capacitor catches the leakage energy and releases it back into the primary on the next cycle, recovering most of it. Adds parts and complexity but pushes flyback efficiency above 92 % and enables compact 100 W chargers. The active-clamp FET is usually driven complementarily to the main FET with adaptive dead-time.

Tight winding design — bifilar wound, sandwich-wound, or split-bobbin construction — minimises leakage in the first place. Magnetics vendors compete on coupling coefficient (k = 0.98+) because every 1 % of leakage = 1 % of efficiency loss.

Where flyback converters show up

• USB and USB-C chargers below ~150 W. Every wall wart, laptop charger, and phone brick under 150 W contains a flyback. From the $2 white-label phone charger to Apple's 30 W USB-C brick, the basic topology is unchanged in 40 years — only the controller, GaN switches, and packaging have evolved.
• Standby supplies in TVs, monitors, and appliances. The 5 V "always-on" rail in your TV that wakes it on remote keypress is generated by a tiny standby flyback running at 100 mW. Power Integrations' TinySwitch family dominates this segment with 3-pin integrated FET-plus-controller chips.
• LED drivers under 50 W. Constant-current flybacks drive LED strings in retrofit bulbs, MR16 lamps, and downlights. The output diode + capacitor handles ripple; current feedback closes the regulation loop.
• Auxiliary bias supplies in larger PSUs. A server PSU has an 800 W main rail (LLC) plus a 5–15 W flyback auxiliary that powers the controller, fans, and standby logic. The flyback runs whenever AC is present; the main rail is gated by the wakeup logic.
• Industrial 24 V DIN-rail supplies under 100 W. Phoenix Contact, Mean Well, and TDK-Lambda's small DIN rail PSUs are typically flybacks with PFC front-ends. Reliability targets 100k+ hours MTBF; convection-cooled.
• Smart-home, IoT, and white-goods PSUs. A $30 smart plug has a 1 W flyback inside running its Wi-Fi module from mains. Refrigerator and washing-machine control boards use 5–10 W flybacks for their microcontrollers.

Common design pitfalls

• Underestimating leakage inductance. A datasheet specifying L_pri = 10 mH and L_leak = 100 µH might actually deliver 250 µH after a batch swap. The FET drain stress doubles. Always measure leakage on production samples and verify snubber margin.
• Core saturation under transient. A flyback transformer must not saturate even at the worst-case duty cycle and lowest input voltage. A nominal 8 mH gapped transformer at 0.3 T might saturate at 0.35 T; running it at D = 0.5 instead of 0.4 (during low-line transient) pushes peak primary current 25 % higher and tips the core into saturation. Magnetics specs include I_sat current rating for exactly this reason.
• Right-half-plane zero in CCM. Flybacks in CCM share the buck-boost family's RHP zero. Loop bandwidth must stay below the RHP-zero frequency (typically a few kHz), making transient response slow. DCM operation eliminates the RHP zero — a major reason mid-power flybacks default to DCM.
• Output diode reverse-recovery. A standard silicon diode at the secondary can take 50–100 ns to recover when the FET turns on again. During recovery, current flows backward through the diode, dumping energy into the FET. Schottky diodes eliminate this for V_out below ~60 V; SiC Schottkys handle higher rails.
• Bias winding miscount. The auxiliary (third) winding that powers the controller IC must have the right turns ratio to V_out — too few turns and the IC browns out at light load; too many and it over-volts at heavy load. A 1-turn error on a 6-turn aux winding is a 17 % regulation error.
• Y-cap leakage current. The primary-to-secondary safety capacitor must keep leakage current to ground below 250 µA (IEC 60950) or 100 µA (60601 medical). A 2.2 nF Y2 cap typically gives ~150 µA at 230 VAC — fine for most applications, fail for medical without further reduction.
• EMI from switching transitions. The drain of the primary FET swings hundreds of volts at di/dt = 100 A/µs+. Without careful layout (short snubber loops, RC damping on the gate, primary winding nearest the bobbin centre), the converter fails CISPR 22 / FCC Part 15 conducted-emission limits by 10–20 dB. Layout, not BOM, is what passes EMI.

Historical context

The flyback's lineage traces to flyback transformers in vacuum-tube television sets of the 1950s and 1960s. Those transformers produced the high-voltage anode supply (10–25 kV for the CRT) by accumulating energy during each horizontal-line sweep and releasing it through a high-impedance secondary on the "flyback" (return) phase of the sweep — the origin of the name. Engineers in the 1970s realised the same topology, scaled down and run at higher frequency by a transistor switch, made an excellent isolated power supply.

The 1980s commercialised the design. Power Integrations (founded 1988) shipped the first integrated-controller-plus-FET flyback IC (the TOPSwitch) in 1994, cutting the part count from ~30 to ~5 and standardising the architecture. By 2000 essentially all consumer-electronics wall warts were flybacks. The 2010s pushed GaN switches into the topology, enabling the modern 100 W USB-PD charger the size of a US power plug. The 2020s saw integrated active-clamp flybacks (Power Integrations' InnoSwitch3-Pro, Navitas's NV6125 GaNFast) bring 95 % efficiency and 30 W/in³ density to mass-market chargers.