Why Class D Amplifiers Got A Worse Rep Than They Deserve
Answer-First Snippet
Switching an amplifier on and off thousands of times per second buys you efficiency and lower heat, but it drags in a large output filter and radio-frequency noise. Early designs paid for that in the treble. Modern driver ICs and smaller inductors shrank that penalty so the efficiency gain now outpaces the old filtering cost.
Key Takeaways
- Efficiency over smoothness: Switching transistors on and off keeps the supply rail steady under load, but the raw waveform must be smoothed by an LC filter before it reaches the speaker wire.
- The filter is the price tag: Every Class D design carries an inductor and capacitor between the switching node and the speaker. The bigger the inductor, the more heat and board space you spend, and the looser the bass grip.
- The treble used to pay: Early boards with large toroidal inductors and discrete gate drivers left a faint switching grain in the high band. Modern integrated ICs with on-chip gate drivers and ceramic-capacitor filters pushed that grain below the noise floor in most listening rooms.
- A small, quiet system still favors linear: When the speaker is efficient and the room is small, the extra headroom a switching topology offers sits unused, and the filter’s residual ripple becomes just barely audible in the cymbals.
The Job The Design Had To Fill
The problem was heat. A traditional amplifier that runs its output transistors as variable resistors dumps a large share of the supply voltage as waste heat into the chassis. The louder you turn it, the hotter the case gets, and the more ventilation or fans you need. In a home theater rack with five or six amplifier channels, that heat adds up fast.
The engineering goal was simple: deliver the same wattage into the same speaker while cutting the wasted heat by a wide margin. Think of a thermostat. A dimmer switch wastes energy by running the heater at half power continuously. A relay that snaps the heater fully on or off wastes far less. Class D does the relay trick at audio speed.
The concrete listening goal behind that trick: the amplifier must hold the supply rail steady even when the speaker is drawing heavy current during a loud bass passage. If the rail sags, the top end thins out and the bass loses its weight. A switching topology keeps the rail firm by drawing only the power the signal actually needs, not a constant amount.
The Constraints Nobody Could Argue With
Four physical limits sit under every Class D design, and no amount of clever engineering removes them.
The switching frequency sets the filter floor. The raw output of a switching amplifier is a square wave riding at the carrier frequency. That square wave is not music. Before it can drive a speaker, an LC output filter (a combination of an inductor and a capacitor wired in series and parallel) must smooth the square wave into a clean, sine-like signal. The filter works by attenuating energy at and above the carrier frequency. If you set the carrier low, the inductor and capacitor must be physically large to keep the attenuation deep. If you set it high, the components shrink. The hard floor is that the carrier must sit well above the top of the audible range, which tops out around 20,000 hertz. In practice, designers place it further up to leave room for the filter’s transition band.
EMI radiates from the switching node. The same square wave that the filter tames also radiates electromagnetic noise. The faster the edges, the broader the spectrum. Early boards with discrete transistors and long gate-drive traces had slow, uneven switching edges. Those edges radiated into the mid and high frequencies, and the filter could not catch them all. You heard it as a faint, gritty texture behind the treble, especially on quiet passages with a high-sensitivity tweeter.
The heat budget still exists, just smaller. Switching does reduce the wasted power, but the gate drive, the output filter inductors, and the rectifier front-end all generate their own heat. In a compact enclosure with little airflow, that heat still needs somewhere to go. Designers must balance how much thermal mass the chassis can hold before the top case temperature becomes uncomfortable to touch.
Damping depends on the filter. The damping factor (the ratio of the speaker’s nominal impedance to the amplifier’s output impedance at low frequency) tells you how tightly the amp grips the woofer cone on the downstroke. In a Class D design, the output filter inductance sits in series with the speaker. That inductance is a dead weight at low frequencies. It raises the output impedance exactly where you want it lowest. The result: the amp grips the cone a bit less tightly than a linear design with near-zero output impedance. You feel it as a slightly softer, less controlled bass note on very low frequencies.
What Was Given Up To Get It
| Design direction | What it gains | What it costs | Where you meet it |
|---|---|---|---|
| High carrier frequency with a small LC filter | Compact chassis, lower EMI, cleaner treble | Higher gate-drive losses, faster thermal cycling of the output stage | The board layout and filter component values; you hear it as a slightly noisier floor on very quiet passages |
| Low carrier frequency with a large toroidal inductor | Lower switching loss per cycle, simpler gate-drive circuitry | Bigger physical inductor, more EMI leakage, higher output impedance at bass | The size of the chassis; you feel it as looser sub-bass grip and a gritty high band on a sensitive tweeter |
| Integrated driver IC with on-chip gate driver | Faster, more even switching edges, smaller filter, less board-level EMI | Higher bill-of-material cost, tighter thermal design around a single hot chip | The heat-sink path on the PCB; you notice it as a cleaner treble and a quieter noise floor |
| Discrete output transistors with external gate-drive network | Lower part cost, easier to repair, more design freedom | Slower and less symmetric switching, larger filter, more radiated EMI | The gate-drive traces on the PCB; you hear it as a slight hardness in the upper register on quiet material |
| Ceramic-capacitor filter instead of electrolytic | Smaller filter footprint, lower ESR, tighter damping | Lower energy storage, more sensitive to ripple and DC offset | The output stage layout; you feel it as a snappier low end but a slightly harder top end on very fast transients |
| Wide carrier with feed-forward error amplifier | Lower in-band distortion, more linear phase through the audible band | More complex control loop, higher design iteration cost | The PCB routing and the amplifier firmware; you hear it as smoother vocal transitions and less ringing on plucked strings |
The Trade-Off In Your Ears
Start with the bass. The winning side of the trade-off gives you a supply rail that stays firm while the woofer is being slammed. Because the amp draws only the power the bass note demands, the rail does not sag the way it would in a linear design under the same load. The low end stays weighted and even. The damping grip is a shade looser than a linear amp because of that filter inductor in series, but in a typical listening room the difference sits right at the edge of what you can tell.
Move to the vocals. A clean, well-tuned filter keeps the switching harmonics out of the 500-hertz to 4-kilohertz band where a singer’s consonants live. The result is a voice that sits in the middle of the stage with natural sibilance, not a veiled or harsh one. The old designs with large toroidal inductors and slow gate-drive edges let a smudge of switching ripple bleed into that range. You would hear it as a faint metallic edge on the “s” and “t” sounds, more obvious when the music drops to a whisper.
Now the high frequencies. This is where the early trade-off hurt most. The filter could not fully suppress the carrier and its first few harmonics. A small, fast-responding tweeter picked up that residual ripple as a faint, granular texture above 10 kilohertz. It was not loud. It was more like a background hiss that changed character when the music got quiet. On a large, less sensitive speaker the same ripple got buried under the cone’s own mechanical noise. On a small, efficient tweeter in a quiet room, it stood up.
The losing side, a well-driven linear amplifier, would have given you a smoother treble floor and a slightly tighter low-end grip. But it would have run hotter, needed more ventilation, and drawn a constant current even when the music was silent. The trade was: cleaner highs and tighter bass at the cost of a warm chassis and a bigger power bill.
Cases For Going The OtherWay
A small, high-sensitivity loudspeaker in a quiet listening room. This is the scenario where the linear road wins. The speaker only needs a few watts to fill a small space. The extra headroom the switching topology offers is taxed rarely, if at all. Meanwhile, the filter’s residual ripple sits just above the noise floor, and a quiet room means you can hear that last bit of treble grain. The linear amp’s near-zero output impedance also gives the tweeter a firmer grip on fast transients, and its smoother top end has no switching residue to worry about.
A system where the amplifier chassis shares a shelf with a turntable or a passive preamplifier. The EMI from a Class D switching node, even well-filtered, can couple into an unshielded phono stage or a thin-film preamp on the same shelf. A linear amplifier radiates far less broadband noise. The audible result is a cleaner low-level detail on a well-mastered classical recording, without the faint digital-grain texture that a nearby switching supply can add.
A large, low-sensitivity speaker that demands real current and voltage swing. Here the switching amp’s headroom advantage is the whole point, and you should take the trade-off. But if you pair that topology with a very high-sensitivity small speaker, you are carrying the extra headroom for no reason, and the filter’s small imperfections become the most visible part of the sound. A well-matched linear amp at the sensitivity level the small speaker actually needs will sound smoother in the upper band and costs you nothing in the bass grip, because the speaker does not need the extra current.
How The Decision Evolved
In the early days of switching amplifiers for audio, the output stage was built from discrete power transistors. Each transistor had its own gate-drive resistor and its own layout trace. The edges of the switching waveform were rounded by parasitic capacitance and inductance in those traces. To get a deep filter notch, the designers reached for large toroidal inductors, sometimes the size of a coffee mug. Those inductors sat in series with the speaker and raised the low-frequency output impedance. The EMI from the wide, uneven switching edges bled into the treble. The heat, though lower than a linear design, still packed the enclosure, and the board needed generous airflow.
Then integrated driver ICs appeared. The gate driver moved onto the same silicon die as the output transistors. The switching edges got faster and more symmetric because the gate-drive path was no longer a millimeter-long PCB trace; it was a micrometer-wide metal layer inside the chip. The filter shrank. A large toroidal inductor gave way to a smaller, lower-inductance coil and a bank of ceramic capacitors. The ceramic caps have a much lower equivalent series resistance than electrolytic capacitors, so they grip the woofer a bit more tightly and add less coloration to the fast transients.
The thermal picture shifted too. The old design spread its heat across several discrete transistors on a large heatsink. The new design concentrates it in one IC and a small number of external pass devices. The board is smaller, but the thermal spot is hotter and more local. Designers had to rethink the enclosure: a small, sealed box with a vented back panel replaced the large, open chassis.
The net effect: the filter got smaller, the EMI got quieter, the damping got a little better, and the efficiency gain that motivated the whole topology held. The trade-off did not disappear. It moved. The penalty that used to live in the treble as an audible grain now lives mostly in the last decibel of the noise floor, where most listening environments mask it.
What Engineers Get Asked
Why Does The Output Filter Have To Sit Between The Switching Node And The Speaker?
The speaker is a mechanical load. It cannot see a square wave. If you fed raw PWM into a voice coil, the cone would oscillate at the carrier frequency and produce a loud buzz. The LC filter is the only thing that turns the on-off pulses into a smooth voltage the coil can follow. It has to live in that path because the switching node is the source of the square wave, and the speaker is the destination. No other placement does the smoothing.
How Does A Bigger Inductor Make The Bass Feel Looser?
The inductor sits in series with the woofer. At the low frequencies where the bass lives, the inductor’s opposition to current is at its lowest, but it still adds a small amount of impedance that a linear amp does not have. The woofer cone is trying to stop on the downstroke, and the amp has to push current back into the speaker to brake it. That extra series impedance means the braking force is a shade weaker. The note rings a tiny bit longer instead of stopping crisply. You hear it as a slightly less controlled low end.
Why Did Early Class D Amps Sound Cold Or Harsh On A Tweeter?
Two things stacked up. The slow, uneven switching edges from discrete gate-drive circuits radiated EMI into the upper band. The large filter, tuned for those slow edges, left a residual ripple that a small, fast tweeter converted into a faint grain. The result was a high band that lost its smoothness. It was not a loud artifact. It was a texture. Quiet passages with a solo cymbal or a whispered vocal made it most obvious.
What Changed When The Gate Driver Moved Onto The Chip?
The gate-drive trace went from a millimeter of copper on a PCB to a micrometer of metal on a die. The switching edges got steeper and more symmetric. A steeper edge means the filter sees a cleaner pulse shape, so the filter components can be smaller and the notch deeper. The EMI spectrum also tightened because the rise and fall times matched more closely. The filter shrank. The treble cleaned up. The efficiency held.
Does A Class D Amp Still Waste Heat At All?
Yes, but far less. The gate drive, the output transistors during the brief overlap when both are partially on, the inductor’s coil resistance, and the rectifier front-end all generate heat. The total is well below a linear design running at the same output power. In a sealed, compact enclosure that heat still matters. The chassis can still get warm to the touch after an hour of loud playback. The topology reduced the heat, but it did not remove it, and the enclosure design must still plan for it.
(This piece is an original hifireport.com publication. Reuse without consent is prohibited.)