A power amplifier is an electronic circuit that takes a low-voltage input signal and scales up both its current and voltage to drive a low-impedance, high-power load like a loudspeaker. While a preamplifier or op-amp circuit might give you 10 volts of swing, it will collapse if you ask it to source 3 amps into a 4-ohm speaker; the power amplifier's job is to provide that muscular current delivery without the voltage sagging. In a real circuit, power amplifier design changes a fragile, high-impedance voltage signal into a robust, low-impedance current source capable of doing physical work—specifically, moving a speaker cone to displace air. Think of a voltage amplifier as a high-pressure garden hose with a tiny nozzle: it has plenty of force but delivers almost no volume. A power amplifier is the municipal water main, providing the massive volume required to fill a pool.
Beginners frequently confuse voltage gain with power gain, assuming an op-amp driving a transistor base is sufficient, or they mistakenly assume Class D switching amplifiers inherently sound worse than Class AB linear amplifiers, ignoring modern implementations that achieve >110dB SINAD (Signal-to-Noise and Distortion ratio).
The Math of Moving Air: A Worked Thermal and Power Example
The hardest part of linear power amplifier design isn't the audio path; it's the thermal management. Let's look at a classic chip-amp build using the LM3886 Class AB amplifier IC, targeting 50W RMS into an 8-ohm load.
To get 50W into 8 ohms, you need roughly 20V RMS at the output, which means a peak voltage of about 28V. We will power the chip with a dual ±35V DC supply (70V total rail-to-rail) to allow for headroom and internal voltage drops. The maximum power dissipation ($P_{D(max)}$) in a Class B/AB output stage occurs at roughly 63% of maximum output voltage, calculated as:
P_D(max) = (2 × V_CC^2) / (π^2 × R_L)
Plugging in our numbers: (2 × 35^2) / (9.87 × 8) = 2450 / 78.96 = 31 Watts. This means the chip will dissipate 31W of pure heat under worst-case continuous sine-wave conditions.
Now we calculate the required heatsink. The LM3886 has a maximum junction temperature ($T_J$) of 150°C, but we derate to 125°C for reliability. Assume an internal chassis ambient temperature ($T_A$) of 40°C. The thermal resistance junction-to-case ($θ_{JC}$) is 1°C/W, and case-to-sink ($θ_{CS}$) using a mica insulator and thermal paste is roughly 0.5°C/W.
θ_{SA} = [(T_J - T_A) / P_D] - (θ_{JC} + θ_{CS})
θ_{SA} = [(125 - 40) / 31] - (1 + 0.5) = 2.74 - 1.5 = 1.24°C/W
You need a heatsink rated at 1.24°C/W or lower. A standard 2-inch extruded aluminum sink is usually around 4.0°C/W. If you use that, your chip will hit 150°C and trigger its internal thermal shutdown (SPiKe protection) within seconds. You must use a massive finned sink, bolt multiple chips to a shared chassis, or add forced-air cooling. For deeper reading on thermal runaway and sink sizing, Rod Elliott's Audio Power Amplifier Design archives remain the gold standard for bench builders.
Where You Meet Power Amplifier Design in Practice
You don't just encounter power amplifiers in hi-fi listening rooms. The principles of low-impedance, high-current delivery show up across several disciplines:
- Active Studio Monitors: Internal Class D modules (like Hypex Ncore or Purifi Eigentakt) drive woofers directly, requiring tight integration between the DSP crossover and the amp's PWM switching frequency.
- Ham Radio RF Transmitters: RF power amplifiers use LDMOS transistors to drive 50-ohm antenna loads at hundreds of watts, where impedance matching networks replace audio inductors.
- Subwoofer Plate Amplifiers: High-efficiency Class D designs that must handle massive reactive current swings at 20Hz without the power supply capacitors browning out.
- Piezo and Ultrasonic Drivers: Industrial power amps driving highly capacitive loads, requiring specialized output networks to prevent high-frequency oscillation and MOSFET destruction.
Scenario Walkthrough: The Melted Class-AB Output Stage
Theory is clean; the workbench is unforgiving. Here is a real-world failure that illustrates why bias stability is the most critical aspect of discrete power amplifier design.
The Setup: A hobbyist builds a 100W Class AB amplifier using discrete ON Semi MJL21193 (NPN) and MJL21194 (PNP) output transistors. The bias is set to 50mA quiescent current per pair, powered by an unregulated toroidal transformer yielding ±50V at idle.
The Numbers: With 4 output devices (2 NPN, 2 PNP), the total quiescent dissipation at 50mA per device across a 100V total rail is 0.05A × 100V = 5W per device. That's 20W of heat generated just sitting idle. The heatsinks are rated at 3.0°C/W.
The Outcome: The hobbyist plays a 4-ohm test tone. The unregulated supply sags to ±42V under heavy load, but the thermal mass of the small heatsinks cannot absorb the dynamic switching losses plus the 20W idle heat. The transistors heat up rapidly. Within 45 seconds, a loud pop destroys the test speaker, and the emitter resistors are visibly charred.
What Went Wrong: The designer forgot to thermally couple the $V_{BE}$ multiplier (the bias transistor) directly to the main heatsink. As the output transistors heated up, their internal $V_{BE}$ dropped. Because the bias circuit was tracking the ambient PCB temperature instead of the heatsink temperature, it failed to reduce the base drive. This caused thermal runaway: the hotter the transistors got, the more bias current they drew, which generated more heat. The bias current spiked from 50mA to over 2 amps, burning the 0.22-ohm emitter resistors open and shorting the output devices, sending 42V DC straight into the tweeter.
Choosing Your Topology: Class AB vs. Class D in 2026
When starting a new power amplifier design, your topology choice dictates your PCB layout, power supply requirements, and thermal strategy. While Class AB was the undisputed king of fidelity for decades, modern Class D architectures have largely closed the gap.
| Criteria | Class AB (e.g., LM3886, Discrete) | Class D (e.g., TPA3255, Hypex) |
|---|---|---|
| Efficiency | 50% - 65% | 90% - 95% |
| Heat Dissipation | Massive heatsinks required | Minimal; often bare PCB copper |
| Audio Fidelity (THD+N) | 0.005% (Excellent) | 0.001% (State-of-the-art) |
| PCB Footprint | Large (due to heatsinks/caps) | Compact |
| EMI Shielding | Not required | Strict layout and LC filtering needed |
| Power Supply | Linear, heavy toroidal | SMPS, high-frequency capable |
If you are building a desktop headphone amp or a low-power vintage restoration, Class AB remains forgiving and easy to debug with a basic oscilloscope. If you are designing a 500W subwoofer amplifier or a multi-channel active speaker, modern Class D chips like the Texas Instruments PurePath console series are the only logical choice to avoid building a space heater.
Setting Quiescent Bias: A Step-by-Step Guide
If you are designing or repairing a Class AB discrete amplifier, setting the idle current correctly is mandatory before connecting a speaker.
- Insert a Dummy Load: Never test bias with a speaker connected. Connect a 4-ohm or 8-ohm high-wattage power resistor across the output terminals.
- Power via Dim-Bulb Tester: Wire a 60W incandescent bulb in series with the AC mains. If the bulb glows brightly during bias adjustment, you have a short or severe thermal runaway; cut power immediately.
- Measure Across Emitter Resistors: Set your multimeter to DC millivolts. Measure the voltage drop across one of the 0.22-ohm emitter resistors on the NPN output devices.
- Adjust the Trimpot: Slowly turn the bias trimpot. For standard thermal stability, target a voltage drop of 11mV to 15mV across the 0.22-ohm resistor (which equates to 50mA - 68mA of quiescent current).
- Thermal Soak: Let the amplifier run for 20 minutes. Re-measure. If the voltage has crept up to 25mV, your thermal tracking is inadequate, and you must readjust downward once the heatsink reaches equilibrium.
Frequently Asked Questions
Do I need a dual (split) power supply for a power amplifier?
Not strictly, but it is highly preferred for audio. A dual supply (e.g., ±35V) allows the output to sit at exactly 0V DC at idle, meaning you don't need a massive output coupling capacitor to block DC from reaching your speaker. Single-supply designs require an output capacitor that acts as a high-pass filter, which can muddy the low-frequency bass response unless the capacitor is impractically large.
Why do output transistors need emitter resistors?
Emitter resistors (typically 0.1Ω to 0.33Ω) provide local negative feedback and force current sharing among parallel output devices. Because bipolar transistors have slight manufacturing variations in their $V_{BE}$ turn-on thresholds, one transistor will naturally try to hog all the current and overheat. The emitter resistor drops a small voltage proportional to the current, effectively balancing the load across all parallel devices.
