Why Class D Amplifiers Feel Like Magic
The switched output stage trades thermal mass for 80 to 95 percent efficiency, shrinking the heat sink and enabling battery-powered audio. The cost lands as switching ripple on the tweeter and weaker damping at the woofer. Compact and portable systems gain; critical-listening setups with sensitive speakers lose.
- Efficiency is the prize. The switched topology sits fully on or fully off, cutting thermal loss from tens of watts to just a few. That trade makes small heat sinks and battery-powered portable audio workable at high output levels.
- The cost shows up in the highs and the grip. EMI and switching ripple excite the tweeter cone. The PWM stream also makes back-EMF absorption from the woofer voice coil less clean, reducing damping control.
- The other road still wins in specific cases. A low-power Class A stage feeding a sensitive speaker in a small room gets a smoother rail and tighter damping, at the price of heat and size that the room can absorb.
- The “black technology” label is a perception gap. The PWM modulation itself is straightforward. What makes the result look like magic to a listener is the output-filter design and the EMI-management work hidden inside the enclosure.
The Goal On The Drawing Board
The problem was simple to state and hard to solve: deliver enough power to move a speaker cone without turning the amplifier into a heater. A linear Class A or Class AB output stage wastes power as heat whenever voltage across the transistor and current through it are both non-zero at the same moment. Push that stage to high output, and the heat sink has to shed tens of watts of waste heat. The enclosure grows. The weight climbs. A battery, if one is in the loop, drains fast.
The engineering goal became: make the output transistor spend its time either fully on (zero voltage across it, so no power is lost) or fully off (zero current through it, so no power is lost). That is the core idea behind a switching, or Class-D, output stage. Think of it like a bicycle light that blinks a tiny LED on and off thousands of times per second instead of running a dim bulb. The LED is either lit or not. The average brightness looks smooth to your eye. The battery lasts far longer.
The specific target was 80 to 95 percent efficiency at audio power levels, so that the thermal design could shrink to a fingertip-sized heat sink or disappear entirely for battery-powered devices.
What Physics Would Not Allow
Even with the cleanest switching stage, a handful of physical limits set the outer boundary for every design in this category.
Thermal loss still exists at the edges. No real transistor switches instantaneously. During the brief overlap when the device is neither fully on nor fully off, both voltage and current coexist, and that overlap is where the remaining power loss lives. You can make the overlap shorter with faster gate drivers, but you cannot make it vanish. The result is a small but real wattage of heat, typically a few watts even in a well-tuned design.
Fast switching radiates. Usually a conductor snaps from zero current to full supply voltage in a fraction of a microsecond, it launches a burst of electromagnetic energy across a wide band of frequencies. EMI (electromagnetic interference) is that radiated energy. The faster the edge, the wider the band, and the harder it is to keep that energy out of the audio path. No layout, no filter, and no enclosure can hide that radiation entirely. It couples into nearby traces, into the crossover network, and into the tweeter voice coil as a high-frequency vibration that the cone translates into a faint, harsh edge on the sound.
Back-EMF absorption wants a steady rail. When the woofer cone moves, the voice coil cutting through the magnet field generates a voltage called back-EMF (back-electromotive force). That voltage flows back into the amplifier. If the output presents a smooth, low-impedance DC rail (as a Class A stage does), the back-EMF is absorbed cleanly and the cone stops when the signal says stop. If the output is a PWM stream (a rapid on-off pulse train whose average value follows the audio signal), the impedance the back-EMF sees is no longer a flat, low value. It varies with the pulse duty cycle. The absorption is less clean, and the woofer cone can ring a little longer after a note ends.
Filter size and switching frequency trade off. To strip the switching carrier from the output, you place an LC filter (an inductor and a capacitor) after the output stage. The higher the switching frequency (how many times per second the output transistor toggles), the smaller the inductor and capacitor can be. But raising that frequency pushes more EMI energy into higher bands and increases the loss at each switching transition. The designer picks a middle ground, and that middle ground sets a floor on how small the filter can be and how quiet the EMI can be.
What Was Given Up To Get It
| Design direction | What it gains | What it costs | Where you meet it |
|---|---|---|---|
| Switched (Class-D) output stage | 80 to 95 percent efficiency; heat sink shrinks to a small block or vanishes; battery-powered designs become practical at useful volume | Switching ripple excites the tweeter; EMI couples into the crossover; damping factor drops; back-EMF absorption is less clean | Portable Bluetooth speakers, compact powered bookshelves, laptop sound cards, car audio |
| Linear (Class-A) output stage | Smooth, low-impedance DC rail; high damping factor; no switching artifacts; back-EMF absorbed cleanly | Tens of watts of quiescent heat; large thermal mass; battery life suffers; enclosure must vent heat | High-end hi-fi in small to medium rooms; critical-listening setups with sensitive, high-efficiency speakers |
The table is not a scorecard. Each row is a physical reality. A Class-D design that ignores the cost column will sound thin in the top end and loose in the bass. A Class-A design that ignores its own cost column will run hot, need a bigger enclosure, and drain a battery fast enough that portable use is out of the question.
How You Hear The Decision
In the bass, the difference shows up in how a note ends. A well-designed Class-D stage still delivers strong fundamental energy to the woofer. But when the cymbal crashes and the kick drum drops out, the woofer cone on a switched amp can keep vibrating for a few extra milliseconds. The sound has a slightly looser, rounder tail. A Class-A stage, with its high damping factor and smooth rail, snaps the cone to a stop more crisply. At low to moderate listening levels in a small room, that crispness is what a critical listener notices first.
In the midrange and vocals, both topologies usually sound close. The voice sits in the band where the speaker cone is most controlled by the crossover, and the EMI energy is well above the range where the midrange driver is sensitive. You will not hear a major difference in a clean vocal line unless the system is very loud or the EMI shielding is poor.
In the highs, the cost column of the table becomes audible. The tweeter cone is light, fast, and directly in the path of switching ripple. A Class-D output that leaks even a small amount of ripple in the 200 to 500 kilohertz range into the tweeter driver makes cymbals and sibilance sound slightly edged or metallic. The effect is subtle. A listener comparing the two side by side will describe the switched version as “a touch harder up top.” The Class-A version, free of that ripple, reads as smoother and more open in the air and cymbal detail.
The losing side of the fork, the linear stage, would have sounded more controlled at every register. But it would have been a brick of aluminum heat sink on top of the amplifier, running warm to the touch, and it would have drained a battery fast enough that portable use was out of the question.
When The Other Choice Wins
A small room with a sensitive, high-efficiency speaker at low to moderate volume. The room is maybe 12 by 14 feet. The speaker moves a lot of air for the power it draws. The listener plays at a level where a low-power Class A stage stays thermally comfortable. In that context, the damping advantage of the smooth rail and the clean back-EMF absorption matter more than the watts a switched stage would save. The bass is tighter. The tweeter is free of ripple. The cost is a warm enclosure, which the room absorbs without issue.
A system where the woofer excursion is large and the crossover is set low. A single-driver full-range speaker or a two-way design with a crossover around 800 hertz means the woofer does most of the work from 20 hertz up. The back-EMF signal is large because the cone travels far. A switched output stage, with its time-varying impedance, absorbs that energy less cleanly. The bass loses focus and detail in the upper register where the woofer is still working. A linear stage holds the cone more firmly through the full excursion range.
A battery-powered setup where the music level is low. A phone, a laptop, or a small Bluetooth speaker playing at a volume where only a few watts reach the ear. The efficiency advantage of 80 to 95 percent over 50 to 60 percent is real, but at very low power the absolute number of watts saved is small. A compact Class A stage, drawing a few watts of quiescent current, is thermally fine in a small enclosure and gives the listener the smoothest possible rail at that power level. The battery still lasts hours.
How Designs Got Here
In the early 1990s, the first consumer Class-D amplifiers switched at around 20 to 40 kilohertz. The output filters were large: inductors the size of a thumb and capacitors that needed real board space. EMI was a practical headache. Enclosures had to be metal, and the filter inductors needed careful placement to keep the switching noise out of the tweeter. The sound was clean but not as tight as a well-run Class A stage, and reviewers described the bass as “a little loose.”
By the mid-2000s, gate-driver technology and faster MOSFETs pushed switching frequencies into the 300 to 500 kilohertz range. The filters shrank to tiny surface-mount inductors and capacitors. But the faster edges launched more EMI energy into higher bands, so the layout work and shielding got more demanding. The “black technology” label started to stick: to a listener comparing a compact Class-D receiver to a heatsink-laden Class AB unit from the 1980s, the smaller device sounded as good or better, and the internal trickery was opaque.
The real turning point was not a single breakthrough but a market shift. Battery-powered portable audio, from phones to Bluetooth speakers to in-car systems, created a demand where even an 80 percent efficiency figure was a major gain compared with the 50 to 60 percent of a linear stage. The thermal design that had to shed 30 watts of waste heat simply could not fit in a phone. The switched topology was not the only answer, but it was the only answer that worked at that scale. The cost column (ripple, EMI, reduced damping) was acceptable because the power level was low and the listener was not sitting three feet from a tweeter in a quiet room.
Today, multi-phase topologies, advanced gate drivers, and careful EMI shielding have pushed the efficiency figures toward the upper 90s while keeping the ripple well below the level a typical listener would flag in casual listening. The engineering is mature, but the perception gap remains: people still call it “black technology” because the filter design and the EMI management are not visible, and the PWM idea, once you see it, looks almost too simple to be the whole story.
What Engineers Get Asked
Does Switching Ripple Actually Reach The Tweeter In Most Well-Built Designs?
In most well-shielded designs, the ripple that reaches the tweeter driver is very small, well below the level where a casual listener would call it out. But “most” is doing a lot of work in that sentence. A compact Bluetooth speaker with a thin board and no metal enclosure will let more switching energy couple into the tweeter crossover than a metal-chassis, multi-layer-board design. The effect scales with how far the output stage sits from the tweeter and how much physical separation and shielding the layout provides.
How Does The Damping Factor Change When The Output Is A PWM Stream Instead Of A Smooth Rail?
A Class A stage presents a very low output impedance, often described by a damping factor in the hundreds or thousands. The back-EMF from the woofer is absorbed almost as if the speaker were connected to a stiff mechanical damper. A Class-D stage, even after the LC filter, has a slightly higher and time-varying output impedance because the filter inductor and the PWM duty cycle both play a role. The practical effect is a lower damping factor, usually by a meaningful margin. The woofer cone gets a little more freedom to ring after a note ends.
Why Does The “Black Technology” Label Stick Even Though The PWM Concept Is Simple?
The PWM idea is easy to state: modulate the width of a pulse to encode a continuous value. Anyone who has run a dimmer on a DC motor gets the principle. What is not simple is the filter design that removes the carrier without smearing the audio, the gate-driver timing that keeps the overlap loss low, and the layout and shielding that keeps EMI out of the audio path. That last group of problems is where the engineering depth lives, and it is invisible to the person holding the speaker.
At What Power Level Does The Efficiency Advantage Of A Switched Stage Stop Mattering?
For a stationary hi-fi amplifier in a medium-sized room, the efficiency advantage stops mattering once the thermal design can shed the heat without the enclosure running uncomfortably warm or the battery lasting all night. In practice, that threshold sits somewhere around 20 to 30 watts per channel into a typical 8-ohm load. Below that, a well-ventilated Class A or Class AB stage is thermally fine, and the listener gets the smooth rail and high damping factor for free. Above that, or in a battery-powered context, the switched topology becomes the practical choice.
Can A Class-D Stage Ever Match The Damping Control Of A Well-Run Class A Stage?
At the power levels where a Class A stage is thermally comfortable, a Class-D stage can come close but typically does not match the damping factor. The LC filter and the time-varying impedance of the PWM stream set a floor on how much back-EMF the stage can absorb. At higher power levels, where a Class A stage would be running so hot that the thermal design becomes the limiting factor, the Class-D stage is not just better on damping. It is the only topology that keeps the enclosure at a safe temperature while delivering the wattage. The two topologies are not in direct competition at those levels.
(This piece is an original hifireport.com publication. Reuse without consent is prohibited.)