Audio Power Amplifiers
The power amplifier is the final stage in any sound system — the stage that takes a line-level signal of a volt or so and delivers watts, or hundreds of watts, into a loudspeaker. It is found in every hi-fi amplifier, powered speaker, guitar and bass amplifier, car stereo, PA system, television and Bluetooth speaker. Whether it is a pair of big transistors on a heatsink or a chip the size of a fingernail, the same principles decide how much power it can deliver, how much heat it makes, and how clean it sounds.
Audio power amplifiers present engineering challenges that do not exist in small-signal design. The load is a loudspeaker — a low, reactive, frequency-dependent impedance — rather than a high-impedance input. The currents are measured in amperes, so supply design, wiring and heat become central. And the amplifier must survive the things users do to it: shorted speaker leads, overdriven inputs, and enclosures with no ventilation.
Inside a Class AB audio power amplifier: power supply, output transistors on the heatsink, and the protection relay feeding the speaker terminals. Inset: a Class D board delivering similar power with a far smaller heatsink.
View LargerLoudspeaker Loads and Power Ratings
A loudspeaker is specified by its nominal impedance — usually 4, 6 or 8 Ω for home and car speakers, 8 or 16 Ω for guitar cabinets. The real impedance varies with frequency: it rises to a peak at the speaker's bass resonance, dips to roughly 80% of nominal (the DC resistance of the voice coil plus a little) somewhere above it, and rises again at high frequencies because of the voice coil's inductance. An “8 Ω” speaker may measure 6 Ω at its minimum and 40 Ω at resonance. The amplifier must be designed for the minimum, because that is where it delivers the most current.
Amplifier power should be quoted as continuous (RMS) power into a stated load, over a stated frequency range, at a stated distortion — for example “50 W RMS per channel into 8 Ω, 20 Hz–20 kHz, at 0.1% THD.” Figures labelled “peak” or “PMPO” can be many times higher and are not comparable. Speaker power ratings describe how much heat the voice coil can survive, not how much it needs.
How loud the result is depends heavily on the speaker's sensitivity, given in dB SPL at 1 m for 1 W input. Every doubling of power adds only 3 dB of sound level. A speaker with 90 dB/W sensitivity reaches 100 dB SPL at 1 m with 10 W; a 84 dB/W speaker needs 40 W for the same level. Choosing an efficient speaker is often cheaper than buying a bigger amplifier.
Power, Voltage, Current and Supply Rails
For a sine wave into a resistive load, the output power, peak voltage and peak current are linked by:
The supply must provide at least Vpeak plus the output stage's “headroom” — the voltage lost across the output transistors and emitter resistors at full current, typically 2–5 V for discrete Class AB stages and less than 1 V for MOSFET Class D stages.
Vpeak = √(2 × 50 × 8) = √800 = 28.3 V
Ipeak = 28.3 / 8 = 3.54 A (and 4.4 A if the speaker dips to 6.4 Ω)
A Class AB stage with about 3.5 V of headroom needs split supply rails of about ±32 V. Under music the rails sag, so a real design uses a transformer and filter capacitors that hold ±32 V or more at full load — or accepts a little less than 50 W.
Into 4 Ω the same ±32 V rails would deliver up to about 100 W — but at twice the current, which the output transistors, heatsink and power supply must be able to handle.
Class AB Output Stages
Almost every traditional (linear) audio power amplifier uses the same basic structure: an input differential pair, a voltage-amplifier stage that provides most of the gain, and a push-pull output stage of emitter followers — one NPN and one PNP, or Darlington pairs for more current gain. Overall negative feedback from the output back to the input sets the gain (typically 20–30 times) and reduces distortion.
- Bias: a VBE multiplier transistor (one transistor and two resistors, one of them a trimmer) sits between the bases of the output pair and sets the idle current, typically 20–100 mA. It is mounted on the heatsink so that its voltage falls as the output transistors warm up, preventing thermal runaway.
- Emitter resistors: small resistors (0.1–0.47 Ω) in each output emitter share current between paralleled transistors and stabilize the bias. The idle current is usually set by measuring the voltage across them.
- Zobel network: a series resistor and capacitor (typically 10 Ω and 100 nF) from the output to ground keeps the load from looking inductive at very high frequencies, which would otherwise make the amplifier unstable.
- Output inductor: a small air-cored coil (a few microhenries, with a resistor across it) isolates the amplifier from the capacitance of long speaker cables, again for stability.
Integrated Class AB amplifiers such as the LM386 (under 1 W), TDA2030 (about 15 W) and LM3886 (about 60 W) contain the whole circuit, including protection, and need only a few external components and a heatsink.
Class D Amplifiers
A Class D amplifier converts the audio into a high-frequency pulse-width-modulated (PWM) square wave, typically switching at 300 kHz to 1 MHz. The output MOSFETs are always either fully on or fully off, so they dissipate very little power. An LC low-pass filter between the MOSFETs and the speaker (typically 10–22 µH and 0.47–1 µF per side) removes the switching frequency and leaves the audio. Practical Class D amplifiers achieve 85–95% efficiency.
That efficiency is why Class D has taken over: a 2 × 50 W Class D chip such as the TPA3116 or TDA7498 runs from a single 12–24 V supply with a heatsink the size of a postage stamp, where an equivalent Class AB amplifier would need a large heatsink and a heavy transformer. The trade-offs are that the switching edges can cause radio interference if the layout and filter are poor, that the output filter's response depends on the speaker impedance, and that very cheap modules may have higher distortion than a good Class AB design.
Bridge-Tied Load (BTL)
In a bridge-tied load connection, the speaker is connected between the outputs of two amplifiers driven in opposite phase, instead of between one amplifier and ground. When one side swings up, the other swings down, so the speaker sees twice the voltage swing — and four times the power from the same supply. BTL also needs no output coupling capacitor and no split supply, because both ends of the speaker sit at the same DC voltage. Almost all Class D chips and car audio amplifiers use it.
From a 12 V car supply, a single-ended amplifier can swing about ±6 V around half-supply. Into 4 Ω: P = 6² / (2 × 4) = 4.5 W.
Bridged, the speaker sees about ±12 V: P = 12² / (2 × 4) = 18 W — four times as much, before losses. Real BTL head-unit chips deliver 15–20 W per channel this way, and anything more powerful uses a DC-DC converter to raise the supply voltage.
Note that a BTL output has no ground side: neither speaker terminal may be connected to ground or to the other channel's negative terminal, or the output stage will be short-circuited.
Efficiency, Heat and Heatsinking
A Class AB amplifier's efficiency rises with output level, reaching about 60–70% at full power and much less at the low levels where music spends most of its time. Its worst-case dissipation does not occur at full power but at about 40% of maximum output voltage. For a Class B/AB push-pull stage with supply rails ±VCC and load R:
Maximum dissipation = 2 × 32² / (9.87 × 8) = 2,048 / 79 = 25.9 W, or about 13 W per transistor.
For comparison, a Class D version at 90% efficiency dissipates only 50/0.9 − 50 = 5.6 W at full power.
Output transistors: θJC = 1.0 °C/W; insulating pad θCS = 0.3 °C/W
Heatsink: θSA = 1.5 °C/W; ambient inside the enclosure: 40°C
Heatsink temperature = 40 + 25.9 × 1.5 = 79°C
Junction temperature = 79 + 13 × (1.0 + 0.3) = 96°C — comfortably below the 150°C maximum.
The heatsink at 79°C is too hot to hold, which is normal; a smaller heatsink, or a closed cabinet raising the ambient temperature to 60°C, would erode the margin quickly.
Mount output transistors with thermal compound (and an insulating pad where the transistor tab is not at ground), keep the heatsink fins vertical, and leave space above and below for air to flow. Most amplifier failures that are not caused by shorted speaker wires are caused by heat.
Protection Circuits
Short-circuit and overload protection
A shorted speaker cable presents almost zero ohms, and the output transistors try to deliver enormous current. Protection circuits sense the current (through the emitter resistors) and, in more sophisticated designs, the voltage across the transistors as well, and limit the drive before the transistors leave their safe operating area. Integrated amplifier chips include this protection internally.
DC offset and speaker protection relay
If an output transistor fails short, the full supply rail appears across the speaker as DC, which will burn out the voice coil within seconds. Most amplifiers above a few watts drive the speakers through a relay that opens if DC is detected on the output. The same relay is delayed for a few seconds at switch-on and opens instantly at switch-off, which removes the thumps as the supply rails settle.
Thermal protection
A thermistor or temperature-sensor IC on the heatsink reduces the output, starts a fan or shuts the amplifier down if the heatsink temperature exceeds a threshold, typically 70–90°C. Class D chips shut down internally and restart when they have cooled.
Audio Amplifier Calculator
Audio Power Amplifier Calculator
Enter the continuous (RMS) output power you need, the speaker impedance and the amplifier type. The calculator gives the peak output voltage and current, the minimum supply rails, the DC input power and the heat to be removed at full power.
Frequently Asked Questions
Can I connect 4 Ω speakers to an amplifier rated for 8 Ω?
Only if the manufacturer says so. Halving the load impedance roughly doubles the output current and the power, and it roughly doubles the heat the output stage must dissipate. An amplifier with generous output transistors, power supply and heatsinks will cope, and many are rated for both 8 and 4 Ω. One designed only for 8 Ω may overheat, trip its protection on loud passages, or fail. Connecting two 8 Ω speakers in parallel to one output creates a 4 Ω load, so the same caution applies.
Why does my amplifier sound harsh when I turn it up?
It is almost certainly clipping. When the output tries to swing beyond the supply rails, the tops of the waveform are cut off flat, generating strong odd harmonics that sound harsh and can damage tweeters (clipped waveforms carry far more high-frequency energy than the music itself). The supply rails also sag under heavy load, so clipping starts sooner with sustained bass than a sine-wave test suggests. The cure is a more powerful amplifier, more efficient speakers, or less volume — not a bigger speaker power rating.
Do Class D amplifiers sound worse than Class AB?
Not inherently. A well-designed Class D amplifier can measure as cleanly as a good Class AB design, and most powered studio monitors and high-end active speakers now use Class D. Early and very cheap Class D modules earned a poor reputation through high distortion at high frequencies, a response that changed with the speaker load because of the output filter, and switching noise. Judge an individual amplifier by its measurements and by listening, not by its class.
Why does my amplifier click a few seconds after I switch it on?
That is the speaker protection relay closing. At switch-on the supply rails and the amplifier's DC operating points take a moment to settle, and connecting the speakers immediately would produce a loud thump. The protection circuit waits a few seconds, checks there is no DC on the output, and then connects the speakers. If the relay never closes, or clicks open while playing, the amplifier has detected a fault — often DC on an output caused by a failed transistor — and is protecting your speakers.
Check Your Understanding
Answer each question and click Check to see the explanation.