Push-Pull Converter

The dual-switch isolated buck — two switches take turns driving a centre-tapped transformer, so the core is used in both directions and the output is fed twice per cycle. Higher power than a flyback or forward, with a simple ground-referenced gate drive.

Table of Contents:

Introduction — two forwards taking turns

The forward converter passes energy through its transformer while the switch is on, then spends the rest of the cycle idle — the core just resetting. That is a bit wasteful: the transformer is used in only one magnetic direction, and half the time it does no work. The push-pull converter fixes both problems by using two switches that take turns. While one switch drives one half of the transformer, the other rests; then they swap. The result is a transformer driven back and forth like an AC square wave, and an output that is fed twice per cycle.

The cleanest way to picture it: a push-pull is essentially two forward converters sharing one centre-tapped transformer and one output filter, firing on alternate half-cycles. Because the two halves push the core in opposite directions, the core swings through both magnetic polarities on its own — so, unlike the forward, it needs no reset winding. And because both switches sit with their sources at ground, they are easy to drive. That combination makes the push-pull a natural step up in power, into the 150 W–1 kW range.

This page builds the push-pull from that idea: the parts, what happens in each of its three switching states, why the switches must never overlap, the waveforms you would see on a scope, the flux-walking problem that comes with two switches, the formulas that size it, and a full worked example with real numbers.

What is a Push-Pull Converter?

A push-pull converter is an isolated switch-mode DC-DC converter that behaves like a buck converter fed by two alternating switches through a centre-tapped transformer. It is built from these parts:

  • A centre-tapped primary transformer. The primary is split into two equal halves that meet at a centre tap; the input Vin connects to that centre tap. The secondary is centre-tapped too.
  • Two switches (MOSFETs), Q1 and Q2. Each connects one end of the primary to ground. They are ground-referenced, so their gate drives are simple, and they conduct alternately, 180° apart.
  • Two rectifier diodes, D1 and D2. On the centre-tapped secondary, they form a full-wave rectifier — one conducts for each switch, and both share the current during the dead time.
  • An output L-C filter. A series inductor L and shunt capacitor C, exactly as in a buck or forward converter, that smooth the rectified pulses into clean DC.

The trick is the alternating drive. When Q1 turns on, current flows from the centre tap through the upper primary half; the induced secondary voltage forward-biases D1 and power reaches the filter. When Q2 turns on half a period later, current flows through the lower primary half in the opposite direction, D2 conducts, and power reaches the filter again. Because the two halves magnetise the core in opposite senses, the core is driven symmetrically both ways — and that is what removes the need for a reset winding.

The one-sentence version. A push-pull converter is two forward converters sharing one centre-tapped transformer and one output filter, fired alternately: each switch drives half the core one way, so the core resets itself, the output is fed twice per period, and both switches are easy to drive from ground.

Block Diagram

Block diagram of a push-pull converter: a DC input feeds two alternately-driven primary switches through a PWM controller, then a centre-tapped isolation transformer, then a two-diode full-wave rectifier, then an L-C output filter, then the load, with an isolated feedback path returning to the controller.
Figure 1: Push-pull converter block diagram — two alternating switches drive a centre-tapped transformer into a full-wave rectifier and L-C filter

Read it left to right as a power path: the DC input is chopped by the two switches, driven through the centre-tapped transformer, full-wave rectified, smoothed by the L-C filter, and delivered to the load. Two things mark it as a push-pull. First, the PWM controller drives two gates 180° apart instead of one. Second, the transformer is driven with a symmetric AC square wave, so — unlike the forward — there is no reset winding hanging off it; the core resets itself. As always, feedback crosses the isolation barrier through an opto-coupler.

Circuit Diagram & Construction

Circuit diagram of a push-pull converter: the DC source Vin connects to the centre tap of a centre-tapped primary; the two ends of the primary go to switches Q1 and Q2, both referenced to ground; the centre-tapped secondary feeds rectifier diodes D1 and D2 into a common node P, followed by an output inductor L, capacitor C and load R producing Vo.
Figure 2: Push-pull converter power circuit — centre-tapped primary + two switches, centre-tapped secondary full-wave rectifier + L-C filter
Centre-tapped transformerPrimary split into two equal halves (each Np) meeting at the centre tap, which connects to Vin. Secondary split into two halves (each Ns). Isolates and scales; because it is driven both ways it needs no air gap and no reset winding.
Switches Q1, Q2Each grounds one end of the primary. Ground-referenced, so gate drive is simple. They conduct alternately (180° apart) and must never overlap. Each must block 2·Vin when the other conducts.
Rectifier diodes D1, D2Form a full-wave rectifier on the centre-tapped secondary. D1 conducts with Q1, D2 with Q2, and both conduct together (sharing the inductor current) during the dead time. Each blocks 2·n·Vin.
Output filter L, CA buck-type L-C filter. Because the output is fed twice per period, the inductor current ripples at twice the switching frequency, so the filter is smaller than a single forward’s for the same ripple.
Centre-tap driveThe defining feature. Grounding one end pulls Vin across one primary half; grounding the other end pulls Vin across the other half in the opposite direction — a symmetric AC drive that fully uses the core.
The two switches must never be on at the same time. If Q1 and Q2 ever overlap, both primary halves are grounded at once and the transformer presents a near short across the supply — a destructive shoot-through. That is why each switch is limited to less than 50 % duty with a guaranteed dead time between them, and it is the reason the maximum output is bounded.

Principle of Operation

The two switches are driven by the same PWM controller but 180° out of phase, each on for a fraction D of the full period (D < 0.5). Over one period there are therefore four intervals: Q1 on, dead time, Q2 on, dead time. The transformer and the output inductor split the work:

  • When Q1 is on, Vin appears across the upper primary half; a scaled voltage n·Vin appears on the secondary, D1 conducts, and energy passes straight through to the L-C filter. The inductor current ramps up.
  • When Q2 is on, Vin appears across the lower primary half in the opposite polarity; the same n·Vin reaches the filter, this time through D2. The inductor charges again.
  • In the dead time (both off), the transformer is quiet, but the output inductor keeps its current flowing — and with no switch driving, both diodes conduct together, each carrying half the inductor current back to the load. This is the freewheel.

So the load sees an energy pulse twice per switching period, and the transformer sees a symmetric AC square wave. The output voltage is fixed, as in a buck, by volt-second balance on the output inductor; the symmetric drive automatically keeps the core’s flux balanced, which is the job the forward converter needed a reset winding to do.

Modes of Operation

A push-pull has two driven states — one per switch — separated by a short dead time in which both switches are off and the output inductor freewheels. Here are the two driven states on the circuit, with the conducting path highlighted.

Mode 1 — Q1 ON (upper half drives)

Push-pull converter with switch Q1 on: current flows from Vin through the centre tap and upper primary half through Q1 to ground, while on the secondary the induced voltage forward-biases D1 and drives current through the output inductor L into the load. Q2, D2 are off.
Figure 3: Mode 1 — Q1 on. The upper primary half drives the core one way; D1 feeds the filter
  • Q1 grounds the top end of the primary, so Vin sits across the upper half. The core magnetises in the “positive” direction.
  • The secondary produces n·Vin, forward-biasing D1. The inductor sees (n·Vin − Vo) and its current ramps up, delivering power to the load.
  • Q2 and D2 are off. The idle switch Q2 has to block 2·Vin.

Mode 2 — Q2 ON (lower half drives)

Push-pull converter with switch Q2 on: current flows from Vin through the centre tap and lower primary half through Q2 to ground, driving the core the opposite way; on the secondary D2 conducts and drives the output inductor. Q1, D1 are off.
Figure 4: Mode 2 — Q2 on. The lower primary half drives the core the other way; D2 feeds the filter
  • Half a period later Q2 grounds the bottom end of the primary, putting Vin across the lower half in the opposite polarity. The core now magnetises the other way — this is what resets it.
  • The secondary again produces n·Vin, now forward-biasing D2, so the inductor charges just as it did in Mode 1. From the output’s point of view Modes 1 and 2 are identical.
  • Q1 and D1 are off, and Q1 now blocks 2·Vin. A short dead time then completes the period before Mode 1 begins again.

Waveforms Explained in Detail

Push-pull converter waveforms over two switching periods: the two gate signals firing alternately; the bipolar primary voltage; the bipolar magnetising current; the smooth output-inductor current rippling at twice the switching frequency; the two diode currents which always sum to the inductor current; and the switch-1 voltage that reaches twice the input voltage while switch 2 conducts.
Figure 5: Push-pull waveforms — gates, primary voltage, magnetising current, inductor current, diode currents and switch voltage

Read the traces together and the whole converter is visible in one picture:

  • Gates Q1, Q2. Two pulses, each on for D of the full period, 180° apart, with a dead gap between them. They never overlap — the gap is the shoot-through guard.
  • Primary voltage vpri. A symmetric AC square wave: +Vin under Q1, −Vin under Q2, zero in the dead time. This bidirectional drive is the essence of the push-pull.
  • Magnetising current im. Ramps up under Q1 and down under Q2, swinging symmetrically about zero. Equal up and down means the flux returns to where it started every period — the core resets itself, with no reset winding.
  • Inductor current iL. A smooth triangle that never reaches zero, rippling at twice the switching frequency because the filter is topped up twice per period. Its average is the load current. The doubled ripple frequency is why the filter can be small.
  • Diode currents iD1, iD2. Each carries the full inductor current while its switch drives, and half during the dead-time freewheel. At every instant iD1 + iD2 = iL, so the output current never breaks.
  • Switch voltage vQ1. Zero while Q1 conducts, Vin in the dead time, and 2·Vin = 200 V while Q2 conducts — because the driven half reflects onto the idle switch. That 2·Vin rating is the topology’s main price.

Bidirectional Core & Flux Walking

Two stacked plots of transformer core flux in a push-pull converter over four periods. The top plot, a balanced drive, shows the flux as a symmetric triangle swinging equally positive and negative about zero and repeating unchanged, safely inside the saturation limits. The bottom plot, an unbalanced drive, shows the flux staircasing upward period by period and drifting toward positive saturation, the flux-walking failure.
Figure 6: Bidirectional flux — balanced drive resets the core (top); a small imbalance makes the flux “walk” into saturation (bottom)

Driving the core both ways is the push-pull’s great strength — it uses the full B-H loop, so the transformer is smaller than a same-power forward’s and needs no reset winding. But it comes with a catch. The two half-cycles must be perfectly matched. If Q1 conducts even slightly longer than Q2, or has a slightly lower on-state drop, each period adds a tiny DC step to the flux. Over many cycles that step accumulates and the flux “walks” steadily toward one polarity until the core saturates — at which point the magnetising current explodes and the switch conducting at that moment can be destroyed.

The fix: current-mode control. Rather than commanding a fixed duty, the controller senses each switch’s peak current and terminates its on-time when the current reaches a limit. Any imbalance shows up as a current difference, which the loop corrects cycle by cycle, automatically re-centring the flux. This is why almost every practical push-pull uses peak-current-mode control rather than plain voltage-mode.

Key Formulas

All of these come from volt-second balance on the output inductor, with n = Ns/Np (per half winding) and D the duty of each switch (D ≤ 0.5).

Output voltage (CCM). The filter is fed n·Vin for a total of 2D of the period (two pulses) and 0 for the rest; the inductor’s average voltage is zero:

(n·Vin − Vo)·2D·T = Vo·(1 − 2D)·T  →  Vo = 2 · n · D · Vin

At the D = 0.5 ceiling this becomes Vo = n·Vin — the idealised square-wave result. A straight line in D, like a buck, but with twice the slope of one forward.

Duty ratio for a target output. Rearranging:

D = Vo / (2 · n · Vin)   (must stay < 0.5)

Inductor-current ripple and average. Standard buck filter, but the ripple repeats every half period, so it is small:

ΔiL = (n·Vin − Vo)·D·T / L   |   IL,avg = Io

Device stress. The idle switch and the off diode see the reflected opposite half:

Switch:  VQ,off = 2 · Vin   |   Diode:  VD,rev = 2 · n · Vin

Switch current. Each switch carries the reflected inductor current plus magnetising during its on-time; on average the input current splits between the two:

IQ,avg ≈ D · n · Io   |   Iin,avg = 2 · D · n · Io

(Switch and diode drops, winding resistance and losses are neglected here; a real push-pull is typically 85–92 % efficient. The source article’s IA = Is/2 and 2·Vs switch stress are the same results.)

Voltage Gain & Turns Ratio

Voltage gain of a push-pull converter versus per-switch duty ratio for a turns ratio of 0.25. The gain M equals two times n times D is a straight line from the origin, twice the slope of a single forward converter shown dashed. A vertical dashed line marks the hard duty limit at 0.5 per switch. The operating point at duty 0.4 gives a gain of 0.2.
Figure 7: Push-pull gain M = 2n·D — twice a single forward’s, capped at the D = 0.5 shoot-through limit

The push-pull’s gain is a straight line, M = 2n·D — easy to control, and twice the gain of a single forward converter at the same turns ratio and duty, because two pulses reach the filter each period. Like the forward it is fundamentally a step-down, with the turns ratio doing the coarse scaling and the duty trimming around it. Key points:

  • The turns ratio sets the operating point. An offline supply dropping a high rectified voltage to a low output picks n so the duty lands comfortably below the 0.5 limit.
  • The duty is hard-capped at 0.5 per switch. Above it the switches would overlap and short the supply, so the maximum output is n·Vin. Real designs leave margin for dead time, so the practical ceiling is a little below 0.5.
  • Continuous inductor current keeps it linear. As in any buck-derived converter, very light load can push the inductor into discontinuous conduction, where the output rises above 2n·D·Vin and depends on load. The inductor is sized to stay continuous over the load range.

Worked Example

One consistent design, used by every figure above. Follow the numbers and the formulas check out.

GivenVin = 100 V (centre tap to each end), Ns:Np = 1:4 (n = 0.25), D = 0.4 per switch, f = 50 kHz (T = 20 µs), L = 100 µH, Lm = 1 mH
Output voltage   Vo = 2·n·D·Vin2 × 0.25 × 0.4 × 100 = 20 V
Load (for Io = 4 A)   R = Vo/Io, Po = VoIoR = 5 Ω, Po = 80 W
Secondary pulse voltage   n·Vin0.25 × 100 = 25 V (applied 2D = 0.8 of the period)
Inductor ripple   ΔiL = (n·Vin−Vo)·D·T/L(25−20) × 0.4 × 20µ/100µ = 0.4 A at 2f (Imax 4.2, Imin 3.8 → CCM)
Magnetising swing   Δim = Vin·D·T/Lm100 × 0.4 × 20µ/1m = 0.8 A bipolar (±0.4 A about zero)
Switch voltage stress   2·Vin2 × 100 = 200 V
Diode reverse voltage   2·n·Vin2 × 0.25 × 100 = 50 V
Input current (avg)   2·D·n·Io2 × 0.4 × 0.25 × 4 = 0.8 A → Pin = 100 × 0.8 = 80 W = Po ✓

The last row is the sanity check: with no losses, input power equals output power, 80 W in and 80 W out. It is worth comparing with the forward converter doing the same 100 V → 20 V / 4 A job: the push-pull needs only half the turns ratio (n = 0.25 vs 0.5) for the same output because it delivers two pulses per period, its inductor ripple is smaller (0.4 A vs 1.2 A) thanks to the doubled frequency, and it carries the load with two switches sharing the work — the reasons it scales to higher power.

Advantages & Disadvantages

Advantages

  • Full core utilisation. The bidirectional drive uses both halves of the B-H loop, so the transformer is smaller than a same-power forward’s and needs no reset winding.
  • Simple gate drive. Both switches are ground-referenced, so no floating high-side driver is needed — a real advantage over half-bridge and full-bridge topologies.
  • Small output filter. The output is fed twice per period, so the inductor current ripples at twice the switching frequency and the L-C filter can be smaller for a given ripple.
  • Good efficiency and higher power than flyback or forward — typically 150 W to about 1 kW — with the two switches sharing the current.
  • Isolation and multiple outputs, like any transformer-isolated converter, with a clean, low-ripple output.

Disadvantages

  • High switch voltage stress — each switch must block 2·Vin, which limits the topology at high input voltages (where half-bridge and full-bridge, at only Vin, take over).
  • Flux walking / staircase saturation. Any mismatch between the two half-cycles drives the core toward saturation; it practically demands current-mode control.
  • Duty capped below 50 % per switch, with a mandatory dead time to prevent shoot-through — limiting the achievable gain and input range.
  • Centre-tapped transformer needs two primary and two secondary windings, using copper less efficiently than a bridge; the diodes also see 2·n·Vin.
  • Two switches instead of one, adding cost and a second gate drive compared with a flyback or forward.

Applications

  • Medium-power isolated supplies (150 W–1 kW) — the classic role, above the forward converter and below the bridge topologies.
  • Low-input-voltage, high-current supplies — battery, 12 V, 24 V and 48 V inputs, where the 2·Vin stress is easily handled and the ground-referenced drive is a big plus.
  • Automotive and telecom DC-DC converters running from a battery bus.
  • DC-AC inverters and solar / photovoltaic converters, where a push-pull front end boosts and isolates a low DC input before the inverter stage.
  • Industrial and instrumentation supplies needing isolation, several hundred watts and a clean output.
  • Uninterruptible power supplies (UPS) and similar battery-fed power stages.

The rule of thumb: with a low or moderate input voltage and a few hundred watts to deliver, the push-pull is hard to beat — simple ground-referenced drive, a well-used core, and a small filter. When the input voltage climbs (making 2·Vin switches expensive), a half-bridge or full-bridge converter takes over.

Frequently Asked Questions – FAQs

A push-pull converter is an isolated DC-DC converter that uses two switches to drive a centre-tapped transformer on alternate half-cycles. While one switch drives one half of the primary, the other rests; then they swap, so the transformer is driven back and forth like an AC square wave. The centre-tapped secondary full-wave rectifies this into an L-C filter to give a smooth DC output. It behaves like a buck converter fed through a transformer, delivers power twice per cycle, and is used for medium-power supplies from about 150 W to 1 kW.

In continuous conduction the output voltage is Vo = 2 · n · D · Vin, where n = Ns/Np is the turns ratio per half winding and D is the duty of each switch (D less than 0.5). The factor of two is because the filter is fed twice per switching period. This is the buck relation scaled by the turns ratio, with twice the slope of a single forward converter. At the maximum duty of 0.5 it reduces to Vo = n · Vin, the idealised square-wave result. The duty for a target output is D = Vo / (2 · n · Vin).

When one switch conducts, it applies Vin across its half of the primary. Because the two primary halves are magnetically coupled on the same core, that voltage is reflected onto the other half, so the idle switch sees the input voltage plus the reflected voltage — a total of 2·Vin. In the worked example on this page, with Vin = 100 V, each switch must block 200 V while the other conducts. This 2·Vin stress is the push-pull's main drawback and the reason half-bridge and full-bridge converters, whose switches see only Vin, are preferred at high input voltages.

The two switches must never conduct at the same time. If they overlapped, both ends of the primary would be grounded together, the transformer would present a near short across the supply, and a destructive shoot-through current would flow. So each switch is limited to less than half the period, with a guaranteed dead time between them. This is why the maximum output is bounded at n · Vin, and why real designs leave extra margin below 50 % for the dead time.

Flux walking is a drift of the transformer core flux toward saturation caused by an imbalance between the two half-cycles. If one switch conducts slightly longer, or has a slightly lower voltage drop, than the other, each period leaves a small net DC flux increment. Over many cycles these add up and the flux "walks" steadily toward one polarity until the core saturates, at which point the magnetising current spikes and can destroy a switch. The standard cure is current-mode control, which limits each switch's peak current and automatically corrects the imbalance cycle by cycle.

A single forward converter drives its core in only one direction, so it needs a reset winding to remove the magnetising flux each cycle. A push-pull drives the core in both directions: one switch magnetises it one way, the other switch magnetises it the opposite way. As long as the two half-cycles are balanced, the flux swings symmetrically about zero and resets itself automatically — so no separate reset winding is required. This bidirectional use of the core also means the transformer can be smaller than a same-power forward converter's.

A forward converter uses one switch and drives its transformer in one direction, so it needs a reset winding and delivers one energy pulse per cycle, giving Vo = n·D·Vin. A push-pull uses two switches that drive a centre-tapped transformer in both directions, so it needs no reset winding, delivers two pulses per cycle, and gives twice the output, Vo = 2·n·D·Vin, with the output ripple at twice the switching frequency. The push-pull handles more power but its switches must block 2·Vin, whereas a forward's block Vin(1 + Np/Nt). Roughly, a push-pull is two forward converters sharing one transformer.

Because energy reaches the output filter twice in every switching period — once when Q1 conducts and once when Q2 conducts. Each of those pulses charges the output inductor, so the inductor current rises and falls twice per period. The ripple frequency the filter has to handle is therefore twice the switch frequency. A higher ripple frequency is easier to filter, so a push-pull can use a smaller output inductor and capacitor than a single-ended converter running at the same switch frequency for the same output ripple.

During the dead time both switches are off, so the transformer delivers no power — but the output inductor keeps its current flowing. With neither half driven, the inductor current divides equally between the two secondary halves, so both rectifier diodes D1 and D2 conduct together, each carrying half the inductor current back to the load. The secondary is effectively shorted, so the inductor sees minus the output voltage and its current ramps gently down until the next switch turns on. This freewheeling keeps the output current continuous and is the same action a buck converter's freewheel diode provides.

A push-pull converter is typically used from about 150 W to 1 kW. It is especially well suited to low or moderate input voltages — battery, 12 V, 24 V or 48 V rails — where its 2·Vin switch stress is easily handled and its ground-referenced gate drive is a real advantage. Common applications include automotive and telecom DC-DC converters, solar and photovoltaic front ends, DC-AC inverter input stages, UPS systems, and industrial isolated supplies. Below this range a flyback or forward is cheaper; above it, or at high input voltage, a half-bridge or full-bridge converter is preferred.

Related Topics