PWM / Active-Front-End Rectifier

A fully controlled three-phase rectifier that shapes its own grid current into a sine at unity power factor, holds the DC bus wherever you ask — and, unlike every rectifier before it, can send the power straight back to the grid.

Table of Contents:

Introduction

An ordinary three-phase diode bridge is cheap and rugged, but it is also dumb. It gulps current in ugly spikes, throws harmonics back into the grid, gives you whatever DC voltage the mains happens to provide, and it is a one-way street — energy can only flow from the grid into the load. The moment your load tries to give energy back (a motor braking, a lift coming down, a battery discharging), a diode bridge simply cannot accept it.

The Active Front End — also called a PWM rectifier or active rectifier — replaces those six diodes with six controllable switches. Because the converter can now decide, thousands of times a second, exactly what voltage to put on each phase, it can shape the grid current into a clean sine, hold it in phase with the voltage for unity power factor, regulate the DC bus to a chosen value, and — the big one — run the power backwards into the grid when the load wants to regenerate. That is why AFEs sit in front of variable-speed drives and fast EV chargers. This page covers the topology, the control that makes it work, and the waveforms.

What is an Active Front End?

An Active Front End (AFE) belongs to the family of controllable rectifiers: AC-to-DC converters whose switches are actively driven rather than left to conduct on their own. Physically it is the very same three-phase, two-level bridge used as an inverter — only here it is run in reverse, pulling power from the grid into a DC bus. Its headline abilities are:

  • Bidirectional power flow — it rectifies, and it can inject regenerative power back into the grid.
  • Current shaping — the grid current is forced to be sinusoidal, giving low THD.
  • Power factor control — usually run at unity power factor, but it can deliberately supply reactive power if asked.
  • A regulated DC bus — the output voltage is held at a setpoint you choose (it is a boost-type converter, so that setpoint must sit above the grid’s line-to-line peak).
Active Front End block diagram: the three-phase grid connects through an L or LCL filter to a two-level six-switch bridge that feeds a regulated DC bus and load, with a controller performing PLL synchronisation and dq current control, and power able to flow in both directions.
Figure 1: Active Front End block diagram — a six-switch bridge behind an L/LCL filter, feeding a regulated DC bus, with power free to flow either way.

Circuit Diagram & Construction (Typical AFE Topology)

The typical AFE is a three-phase, two-level voltage-source bridge. Its parts are few and familiar:

  • Three half-bridge legs — six switches (S1…S6): each leg is an upper and a lower switch (MOSFETs or IGBTs), each with an anti-parallel diode. The midpoint of each leg is that phase’s output. Upper and lower are driven complementary (with a small dead-time).
  • Grid filter (Lg, Rg): an inductor per phase between the grid and each leg midpoint. It carries the grid current ig and smooths the switching ripple. For tighter harmonic attenuation an LCL filter is often used instead of a plain L.
  • DC-link capacitor (Cdc): stiffens the DC bus Vdc and absorbs the switching and transient currents.
  • DC load: the drive inverter, charger stage or whatever the bus feeds.
Typical Active Front End topology: a three-phase grid feeds through per-phase filter inductors into a two-level bridge of six switches with anti-parallel diodes, with a DC-link capacitor and load across the DC bus
Figure 2: Typical topology of an Active Front End
The DC bus must be boosted. For the converter to keep control of its currents, the DC voltage has to stay above the grid’s line-to-line peak: Vdc ≥ VLL,peak. Below that the anti-parallel diodes simply take over and conduct on their own, and the current is no longer controllable. That is also why the bus is pre-charged before the switching starts.

Why an AFE Instead of a Diode Rectifier?

Swapping six diodes for six switches buys four things a diode bridge can never give you:

  • Regeneration. Power can flow both ways. A braking motor or a discharging battery pushes energy back to the grid instead of burning it in a braking resistor.
  • Clean current. The current is actively shaped into a sine, so THD is low and the grid stays happy.
  • Power factor you choose. Normally unity; but the same control can inject reactive power on purpose.
  • A DC bus you choose. The output is regulated to a setpoint and stays there through load steps — a diode bridge just follows the mains.

Control Diagram of the Active Front End

All of the AFE’s cleverness lives in its control. The standard scheme is voltage-oriented control: line up a rotating reference frame with the grid voltage, then the messy three-phase AC quantities become two steady DC-like numbers that a plain PI controller can handle. It is built as two nested loops:

Control diagram of the Active Front End: an SRF-PLL provides the grid angle, an outer DC-link voltage PI sets the d-axis current reference, inner d and q axis PI current controllers with decoupling produce the voltage references, which are transformed back to abc and modulated into PWM gate signals
Figure 3: Control diagram of the Active Front End example
  • Grid synchronisation (SRF-PLL): a phase-locked loop tracks the grid’s phase and frequency and gives a clean angle θ. Everything else is timed from θ. Switching is only allowed once the PLL is locked (and the pre-charge relays are closed).
  • abc → dq: using θ, the three measured currents become just two: ig,d (the active, power-carrying part) and ig,q (the reactive part). In steady state both are flat DC values, which is why simple PI controllers work so well here.
  • Outer loop — DC-link voltage PI: compares Vdc with its setpoint and outputs the d-axis current reference i*g,d — in plain words, “draw this much active current to keep the bus up”. It is deliberately slow and uses saturation with anti-windup.
  • Inner loop — dq current PIs: two fast PI controllers drive ig,d and ig,q to their references. Setting i*g,q = 0 is what gives unity power factor; ask for a non-zero q and the AFE supplies reactive power instead.
  • Decoupling & feed-forward: the d and q axes naturally pull on each other through the filter inductance (the ±ωLgi terms). Those terms are calculated and added back so each axis behaves independently, and the measured grid voltage is fed forward so the PIs only have to correct the small remainder.
  • dq → abc and PWM: the two voltage references are rotated back into three-phase quantities using θ and handed to the modulator, which produces the gate pulses for the six switches.

Line-Cycle Waveforms (Deep Dive)

Over one grid cycle the AFE looks beautifully boring on the grid side — which is exactly the point — while the converter terminals are chopping hard.

Active Front End line-cycle waveforms: three sinusoidal grid voltages 120 degrees apart, three sinusoidal grid currents in phase with them, the two-level converter pole voltage switching between plus and minus half the DC voltage with a sinusoidal average, and the regulated DC bus voltage.
Figure 4: One grid cycle. The three currents are clean sines locked in phase with their voltages. The converter’s pole voltage jumps between only two levels — +Vdc/2 and −Vdc/2 — but its pulse widths follow a sine, and the filter inductor averages that into the smooth current above. The DC bus sits at its setpoint.
  • Grid currents: sinusoidal and in phase with the voltages — near-unity power factor and low THD, which is the whole reason to use an AFE.
  • Two-level pole voltage: notice the contrast with the Vienna rectifier, which has three levels. Here every switching step swings the full bus, so the devices must block all of Vdc and the filter has more ripple to mop up — the price of full bidirectional control.
  • Pulse widths tell the story: the pulses are widest at the positive peak and narrowest at the negative peak. That sinusoidal modulation is what the inductor turns into a sinusoidal current.
  • DC bus: flat, because the outer PI keeps it there regardless of what the grid or the load does.

Switching-Cycle Waveforms (Deep Dive)

Zoom in on a few switching periods of one leg and the mechanism is obvious:

Active Front End switching-cycle waveforms: the upper switch gate, the complementary lower switch gate, the two-level pole voltage that follows the upper switch, and the resulting triangular ripple on the grid current.
Figure 5: Inside a few switching periods. The upper and lower switches are complementary, so the pole voltage is a clean two-level square whose width the controller sets. The inductor sees the difference between that square and the (slowly moving) grid voltage, so its current ramps up and down — a small triangular ripple riding on the sine.
  • Complementary switching: when S1 is on the phase is tied to the top rail; when S2 is on it is tied to the bottom rail. A short dead-time between them stops the leg from shorting the bus.
  • Current ripple: the slope of ig is set by (vpole − vgrid)/Lg. A bigger inductor or a faster switching frequency means less ripple — and an LCL filter cleans up what is left.
  • The anti-parallel diodes carry the current whenever it flows against the switch it is in — which is what lets the same bridge work in both power directions.

Rectifying vs Regenerating (Deep Dive)

This is the AFE’s party trick, and it is easiest to see in one picture. Nothing changes in the hardware — the controller simply asks for a negative d-axis current, and the current flips over. The voltage keeps doing exactly what it was doing:

Rectifying versus regenerating in an Active Front End. In the top trace the grid current is in phase with the grid voltage so power flows from the grid into the DC bus. In the bottom trace the same grid voltage is drawn but the current is inverted, 180 degrees out of phase, so the power flows from the DC bus back into the grid.
Figure 6: Same hardware, same grid voltage — only the sign of the d-axis current reference changes. Top: current in phase with voltage, so the AFE takes power (rectifying). Bottom: current flipped 180°, so the AFE returns power to the grid (regenerating). A diode bridge can only ever do the top trace.
  • Power is voltage × current. When v and i rise and fall together, their product is positive on average — power flows in. Flip the current over and the product goes negative — power flows out. That is all “regeneration” is.
  • The power factor is still unity in both cases: the current is still a clean sine perfectly aligned (or anti-aligned) with the voltage. Only the direction changed, not the quality.
  • In practice this is a motor braking, a lift descending, or a battery discharging — energy that a diode bridge would have to burn in a resistor is instead sold back to the grid.

DC-Bus Step Response & dq Currents

Because the outer loop is a real controller, you can ask the bus for a new voltage and watch it get there. Here the reference is stepped up and the bus follows in a few milliseconds with only a small overshoot, while the current controllers do the work underneath:

DC bus step response of an Active Front End. The DC voltage reference is stepped up and the measured DC voltage follows within about nine milliseconds with under ten percent overshoot, settling in about fifty milliseconds. Below, the d-axis current surges to charge the bus and then settles to a new steady value, while the q-axis current stays at zero throughout.
Figure 7: A +50 V step on the DC-bus reference. The bus rises in roughly 9 ms with under 10% overshoot and settles in about 50 ms. Underneath, ig,d surges to pour extra active current in and then relaxes to its new steady level, while ig,q never moves off zero — proof that the two axes really are decoupled.
  • The outer loop is deliberately slow (a bandwidth of roughly 50 Hz here). It only has to keep the bus steady, and going faster would just fight the natural ripple and the inner loop.
  • ig,d is the throttle: to lift the bus you must temporarily draw extra active current from the grid, so d surges, then falls back once the bus is at its new level.
  • ig,q staying flat at zero through the whole transient is the visible proof that the decoupling terms are doing their job — the power factor never wobbles even while the bus is moving.
  • Load steps behave the same way: a sudden extra DC load pulls the bus down a few volts, and the loop recovers it in a few tens of milliseconds.

Key Parameters & Formulas

QuantityFormula & Value
Boost condition (must hold)Vdc ≥ VLL,peak — below this the diodes conduct and the current is uncontrollable
Pre-charge threshold before switchingVdc,min = 0.9 · √3 · Vphase,peak
Unity power factor conditioni*g,q = 0 (all current on the d-axis)
Current-loop PI (Magnitude Optimum)Kp,I = Lg / (2Td) , Ki,I = Rg / (2Td)
Voltage-loop PI integral gainKi,V = Kp,V · 2πfBW / tan(φPM)
Rise-time estimatetr ≈ 0.35 / fBW (≈ 7 ms at fBW = 50 Hz)
Levels per phase2 (+Vdc/2, −Vdc/2)
Example bench values230 Vrms/50 Hz grid, Vdc 650–800 V, Lg = 950 µH, Rg = 54 mΩ

AFE vs Other Rectifiers

Feature6-Pulse Diode BridgeVienna RectifierPWM / AFE Rectifier
Active switches036
Levels—32
Power factor / THDPoorExcellentExcellent
DC bus regulationNo (follows mains)YesYes
Power directionUnidirectionalUnidirectionalBidirectional
Reactive power controlNoLimitedYes
Device voltage stressFull busHalf busFull bus
Best forCheap, tolerant loadsEfficient one-way chargersRegeneration, drives, V2G

Advantages & Disadvantages

Advantages

  • Bidirectional — regenerative energy goes back to the grid instead of into a brake resistor.
  • Sinusoidal grid current, low THD and a power factor you control (unity, or leading/lagging on demand).
  • Regulated, boosted DC bus that holds through load and grid disturbances.
  • Same hardware as an inverter, so it reuses standard half-bridge modules and pairs naturally into back-to-back converters.

Disadvantages

  • Six active switches plus gate drives — more cost and more to go wrong than a diode bridge.
  • Complex control — PLL, dq transforms, two tuned loops, decoupling, plus pre-charge and protection logic.
  • Two-level switching means full-bus voltage steps: higher switching loss, more EMI and a bigger filter than a three-level design like the Vienna rectifier.
  • Boost only — Vdc can never go below the line-to-line peak.

Applications

  • Variable-speed drives — the classic AFE job, letting motors brake regeneratively (cranes, lifts, centrifuges, test benches).
  • Fast EV chargers and bidirectional / vehicle-to-grid charging.
  • Back-to-back converters — an AFE plus an inverter, as used in wind turbines and motor drives.
  • Grid-tied inverters, battery storage and microgrids, where the same bridge runs as a grid-following (or grid-forming) converter.

Frequently Asked Questions – FAQs

An Active Front End is a controllable AC-to-DC rectifier built from a three-phase two-level bridge of six actively switched devices, fed through an L or LCL filter, with a DC-link capacitor on the output. Because the switches are controlled with PWM, it shapes the grid current into a sine, runs at unity power factor with low THD, regulates the DC bus to a chosen voltage, and can pass power in both directions. It is also called a PWM rectifier or active rectifier.

A diode bridge is passive: it draws distorted current, cannot control the power factor, gives whatever DC voltage the mains provides, and only passes power one way. An AFE replaces the diodes with controlled switches, so it can shape the current into a sine, choose the power factor, regulate the DC bus to a setpoint, and send regenerative energy back to the grid. The cost is six switches, gate drives and a much more complex controller.

It uses voltage-oriented control with two nested loops. A PLL locks onto the grid and provides the angle theta, which is used to transform the measured currents from abc into the dq frame. An outer PI controller regulates the DC-link voltage and outputs the d-axis current reference. Two inner PI controllers force the d and q currents to their references, with cross-coupling and grid feed-forward terms added. The resulting dq voltage references are transformed back to abc and modulated into PWM gate signals.

When the reference frame is aligned with the grid voltage, the d-axis current carries the active power and the q-axis current carries the reactive power. Setting the q-axis reference to zero means no reactive current is drawn, so the grid current stays exactly in phase with the grid voltage and the power factor is unity. If you deliberately command a non-zero q-axis current, the AFE will supply or absorb reactive power instead.

Nothing changes in the hardware. The controller simply commands a negative d-axis current reference, which flips the grid current 180 degrees so it is now in anti-phase with the grid voltage. The average of voltage times current then becomes negative, meaning power flows from the DC bus back into the grid. This happens naturally when a motor brakes or a battery discharges, and the current stays a clean sine throughout.

The AFE is a boost-type converter. If the DC bus falls below the line-to-line peak of the grid, the anti-parallel diodes across the switches start conducting on their own like an ordinary diode bridge, and the controller loses command of the current. Keeping Vdc at or above the line-to-line peak keeps the switches in charge. This is also why the DC bus is pre-charged through a resistor and relay before the PWM is enabled.

Related Topics