Short answer: the two classes
- Class A keeps the output transistor conducting for the whole signal cycle. It is the most linear, with low distortion, but it wastes a lot of power as heat.
- Class D switches the output transistors fully on and off at a high frequency, and uses a filter to recover the audio. It is much more efficient, but it needs careful switching and filter design.
The numbers below are design calculations based on stated assumptions. They are not measured results. Check them against your chosen parts before building anything.
Class A: how it works
In a single-ended class A stage, a power transistor sits between the supply and ground. A bias network sets a quiescent current, Iq, that flows even when there is no signal. The audio signal rides on top of that current, so the current never falls to zero.
For the output to swing both ways without clipping, Iq must be at least as large as the peak load current. A speaker is usually connected through a large coupling capacitor, or through a choke for a DC path, so the transistor sees the speaker as its load.
Typical roles in the stage:
- Power transistor: carries the bias and signal current, and must be rated for the supply voltage and the heat it produces.
- Emitter resistor (Re): stabilizes the bias. It sets the quiescent voltage across the emitter, and so the quiescent current.
- Base bias divider: sets the base voltage. It should carry much more current than the base needs, often about ten times as much.
- Coupling capacitor: passes the audio to the speaker and blocks DC.
Class A example: design numbers
Assume a 12 V supply, an 8 Ω speaker and a target output of 1 W.
- Peak voltage across the speaker: Vpeak = √(2 × P × R) = √(2 × 1 W × 8 Ω) = 4 V
- Peak current: Ipeak = Vpeak / R = 4 V / 8 Ω = 0.5 A
- Quiescent current: Iq must be at least 0.5 A. Choose about 0.6 A for margin.
- Power drawn from the supply at idle: about 12 V × 0.5 A = 6 W, almost all of it dissipated as heat in the transistor and resistors.
- Efficiency at 1 W output: 1 W / 6 W ≈ 17%.
For a single-ended class A stage with a resistive load, the theoretical maximum efficiency is 25%. A transformer-coupled stage can reach about 50%. Either way, the transistor needs a heatsink sized for the idle dissipation.
Emitter resistor example. With Iq = 0.5 A and a 1 V drop across Re, Re = 1 V / 0.5 A = 2 Ω. The emitter then dissipates about 0.5 W. The base must sit about 0.6 V above the emitter for a silicon transistor, so the base divider should hold the base near 1.6 V. Use the transistor’s gain to size the base current, then choose divider resistors that carry about ten times that current.
Check the transistor’s safe operating area (SOA) in its datasheet. The combination of about 12 V and 0.5 A is 6 W, which many small transistors cannot handle without a heatsink.
Class D: how it works
A class D amplifier turns the audio into pulses whose width follows the signal, then uses a filter to recover the audio.
audio in ──► [ PWM comparator ] ──► [ gate driver ] ──► [ half-bridge MOSFETs ] ──► [ LC low-pass filter ] ──► speaker
▲
│
triangle wave oscillator (typically hundreds of kHz)
- A triangle-wave oscillator runs at a high frequency, typically several hundred kilohertz.
- A comparator compares the audio with the triangle wave. Its output is a pulse train whose width follows the audio amplitude (pulse-width modulation, or PWM). The op amp comparator article explains the comparator.
- The gate driver turns the MOSFETs on and off. The half-bridge switches the output between the positive supply and ground.
- The LC low-pass filter removes the switching frequency and passes the audio to the speaker.
A small dead time is inserted so both MOSFETs are never on at the same moment. Without it, the supply is shorted through the bridge.
Because the MOSFETs are either fully on or fully off, little power is dissipated in them. Many commercial class D amplifiers reach efficiencies above 80%.
Class D example: filter values
For a second-order Butterworth-style LC filter with a load R, the cutoff and the component values are linked by:
- fc = 1 / (2π × √(LC))
- R = √(L / C)
Solving these for L and C:
- L = R / (2π × fc)
- C = 1 / (2π × fc × R)
Worked example. For an 8 Ω speaker and a cutoff of 40 kHz:
- L = 8 Ω / (2π × 40,000 Hz) ≈ 32 µH
- C = 1 / (2π × 40,000 Hz × 8 Ω) ≈ 0.5 µF
At 20 kHz, the response of this filter is about −0.3 dB, so the audio band is only slightly affected. The cutoff should be well below the switching frequency. Real class D chips often specify their own filter, so use the chip’s recommended values if it has them.
Comparing the two
| Class A | Class D | |
|---|---|---|
| Linearity | Very good | Good, with careful design |
| Efficiency | Low (25% or less, single-ended) | High (often above 80%) |
| Heat | High, needs a heatsink | Low |
| Circuit complexity | Simple | Needs switching, dead time and filter design |
| Main risk | Heat and thermal runaway | Switching noise, shoot-through and EMI |
Build safety notes
- Mains power. Any amplifier that runs from mains voltage carries a shock risk. Use a commercial module or a tested kit, and do not build a mains-powered amplifier from an informal diagram.
- Heat. Class A stages get hot. Fit a heatsink sized for the idle dissipation, and check the temperature with the amplifier running.
- Shoot-through. A class D bridge without dead time can short the supply. Use the gate driver recommended in the chip datasheet.
- Check the parts. Confirm the transistor’s voltage, current and power ratings, and the MOSFET’s voltage rating, against the design values here before you build.
- Start on the bench. Use a current-limited supply and a dummy load, and increase the voltage step by step.