Class D Amplifier
A Class D amplifier abandons linear amplification entirely and instead switches its output devices fully on or fully off at a carrier frequency far above the audio band, encoding the input signal as a pulse-width-modulated square wave that a passive LC filter then smooths back into an analog waveform, reaching 90%+ efficiency because an ideal switch dissipates almost nothing in either state — only during the brief transitions between them.
The whole efficiency gain comes from never letting a device sit in its lossy middle region. A linear stage dissipates power proportional to current times the voltage dropped across the device, and that voltage is largest exactly when the device is partly on; Class D removes that condition by keeping each output transistor either fully saturated (near-zero voltage drop) or fully off (near-zero current), so the only real loss mechanisms left are switching transitions, gate-drive overhead, and the output filter's parasitic resistance — which is why Class D chips like Texas Instruments' TPA3255 or Infineon's MERUS family can deliver 100+ watts while barely warming up, something no linear class can approach.
The carrier frequency is a tradeoff the designer has to choose deliberately. Pushing the switching frequency higher — commonly 300 kHz to over 1.5 MHz in modern audio Class D ICs — shrinks the LC filter's inductor and capacitor and reduces audible switching artifacts, but it also increases switching losses in the output devices roughly in proportion to frequency, so GaN-based Class D designs that can switch faster with lower loss (companies like Infineon and Navitas have pushed this with gallium-nitride FETs) directly unlock smaller, more efficient amplifiers rather than just a marginal improvement.
The output filter is doing exactly the same job the Class C tank does, for a different reason. Both circuits need to strip a non-sinusoidal switching artifact and recover a clean signal, but Class D's LC low-pass filter has to pass the entire audio band uniformly rather than a single narrow carrier, which makes filter design in Class D fundamentally a wideband problem — get the corner frequency wrong and you either leave audible switching noise on the output or roll off real audio content, which is the main engineering tension that separates a cheap Class D module from a well-engineered one.
It displaced Class AB in almost every power-sensitive application once the control loops matured. Early Class D amplifiers from the 1980s and 90s suffered audible distortion and poor PSRR, but modern feedback topologies and self-oscillating or digital PWM control (used across the industry by companies like Hypex, ICEpower, and Purifi) have closed the measured-performance gap enough that Class D now dominates battery-powered portable speakers, automotive subwoofer amplifiers, and even high-end home audio, leaving Class AB mainly to applications where its slightly simpler analog signal path is still preferred.
Read a Class D amplifier through a *digital-at-the-power-stage* lens rather than an *analog amplifier* lens: the output devices are not amplifying a waveform at all, they are acting as a binary switch driven by a PWM encoder, and the entire art of Class D design is choosing a carrier frequency and filter that make that switch train indistinguishable from a linear amplifier once it reaches the speaker.