If you want high-fidelity audio without the high-frequency switching noise of modern Class D chips, the discrete Class AB complementary emitter-follower remains the gold standard for bench builds. A properly biased audio power amplifier circuit in this topology delivers low total harmonic distortion (THD), excellent thermal stability, and a transparent soundstage. This guide walks through a 50W into 8-ohm design, mapping the exact node topology, calculating the VBE multiplier bias network, and detailing the failure modes that will destroy your output stage if you ignore them.
Anatomy and Node Topology of the Class AB Stage
A standard three-stage discrete amplifier consists of a differential input pair, a Voltage Amplification Stage (VAS), and a complementary output stage. To understand how signal and bias flow, we map the circuit to four critical nodes:
- Node A (Input & Diff Pair): The non-inverting input receives the AC audio signal. The inverting input receives the global negative feedback from Node D. The diff pair converts the input voltage error into a differential current.
- Node B (VAS Collector): The Voltage Amplification Stage takes the differential current and converts it back to a high-gain voltage signal. A Miller compensation capacitor (typically 47pF to 100pF) connects from Node B to the base of the VAS transistor to enforce dominant-pole stability.
- Node C (VBE Multiplier / Bias): Driven by the constant current source feeding the VAS, Node C sets the quiescent bias voltage for the output transistor bases. It must track the temperature of the output devices to prevent thermal runaway.
- Node D (Output & Feedback Tap): The emitters of the NPN/PNP output pair meet here. This node drives the speaker load and feeds the global feedback resistor network back to Node A.
Component Specification and Design Walkthrough
For a 50W RMS output into an 8-ohm load, you need roughly ±35V DC rails. We use ON Semiconductor's MJL21193 (NPN) and MJL21194 (PNP) for the output stage due to their 16A continuous collector current and high safe operating area (SOA). Below is the exact bill of materials for the critical bias and feedback paths.
| Reference | Component / Value | Spec / Tolerance | Function |
|---|---|---|---|
| R_E1, R_E2 | 0.22 Ω | 5W, Wirewound, 5% | Emitter degeneration; limits output device current sharing and stabilizes bias. |
| R_BIAS1 | 2.2 kΩ | 1/4W, Metal Film, 1% | Upper leg of VBE multiplier; sets baseline bias voltage. |
| R_BIAS2 | 500 Ω (Trimpot) | Cermet, Multi-turn | Adjustable leg of VBE multiplier; allows precise quiescent current trimming. |
| R_BIAS3 | 1.5 kΩ | 1/4W, Metal Film, 1% | Lower leg of VBE multiplier; prevents bias voltage from collapsing if pot wiper fails. |
| C_MILLER | 68 pF | C0G/NP0 Ceramic, 50V | Dominant pole compensation between VAS collector and base. |
| R_FB1, R_FB2 | 22 kΩ / 1 kΩ | 1/4W, Metal Film, 1% | Global negative feedback network; sets closed-loop gain to ~23dB (14.1x). |
Topology Selection: Why Class AB Over Class D?
While Class D amplifiers dominate the portable and home-theater markets due to >90% efficiency, the discrete Class AB topology wins in specific bench and audiophile applications. The choice depends entirely on your constraints regarding heat dissipation versus high-frequency noise.
| Criteria | Discrete Class AB | Class D (PWM Switching) |
|---|---|---|
| Efficiency | 50% - 65% (Requires large heatsinks) | 85% - 95% (Minimal heatsinking) |
| Output Filter | None (Direct speaker connection) | Required LC low-pass filter (Adds phase shift) |
| EMI / Noise Floor | Extremely low; no switching artifacts | High-frequency PWM noise; requires careful PCB layout |
| Load Dependency | Stable into reactive/complex loads | Can oscillate or alter frequency response with reactive loads |
Choose Class AB when: You are driving complex, highly reactive loudspeaker loads, require an ultra-low noise floor for near-field studio monitoring, and have the physical space for adequate thermal management.
Choose Class D when: You are constrained by enclosure size, battery power, or strict thermal limits, and can implement a proper LC output filter and EMI shielding.
Behavior Table: Failure Modes at the Extremes
Discrete amplifiers lack the internal protection logic found in ICs like the LM3886. If a single passive component fails or a solder joint cracks, the result is often catastrophic. Understanding what breaks when one element changes is critical for designing robust protection circuits.
| Component | Failure Mode | Circuit Behavior | Physical Consequence |
|---|---|---|---|
| VBE Trimpot Wiper | Open circuit (loss of contact) | Bias voltage spikes to maximum rail voltage. | Output stage enters heavy conduction; thermal runaway destroys output BJTs in < 5 seconds. |
| R_E (Emitter Resistor) | Open circuit (burnout) | One half of the output stage cuts off. | Severe crossover distortion; massive DC offset at Node D; voice coil burns out. |
| C_MILLER (Comp Cap) | Short circuit | VAS gain drops to zero; amp loses all high-frequency response. | No output signal; amp remains stable but non-functional. |
| R_FB2 (Feedback to GND) | Open circuit | Global feedback loop breaks; amp operates at open-loop gain (>10,000x). | Output instantly saturates to the positive or negative DC rail; speaker destroyed. |
Step-by-Step Breadboard and Bench Testing
Never apply full mains voltage to a newly assembled discrete audio power amplifier circuit without current limiting. A single wiring error will result in exploding transistors and shrapnel. Follow this exact bench-testing sequence to verify stability and set the quiescent bias safely.
Phase 1: The Dim-Bulb Tester and Dummy Load
- Build a Dim-Bulb Tester: Wire a 60W incandescent light bulb in series with the mains hot line feeding your amplifier's power supply. If the amplifier draws excessive current (due to a short), the bulb lights up brightly, dropping the voltage and saving your components. If the circuit is healthy, the bulb flashes briefly as filter capacitors charge, then dims to a dull orange.
- Connect a Dummy Load: Do not use a real speaker for initial testing. Connect an 8-ohm, 50W non-inductive power resistor across Node D and ground. This prevents back-EMF from complicating your oscilloscope readings.
- Verify DC Offset: Power on via the dim-bulb tester. Immediately measure the DC voltage between Node D and ground with your multimeter. It must read < 20mV. If it reads > 100mV, power off immediately and check the input diff pair and feedback network for solder bridges.
Phase 2: Setting Quiescent Bias Current
The goal is to set the idle current through the output devices just high enough to eliminate crossover distortion, but low enough to prevent overheating. For the MJL21193/4 pair, 50mA to 75mA per device is the sweet spot.
- Measure Across R_E: Connect your multimeter (set to mV DC) across one of the 0.22 Ω emitter resistors (R_E1).
- Adjust the Trimpot: Using a non-conductive ceramic alignment tool, slowly turn the VBE trimpot. You are looking for a reading of 11mV to 16mV. (Using Ohm's Law: I = V/R. 11mV / 0.22Ω = 50mA).
- Thermal Tracking Check: The bias transistor must be physically bolted to the same heatsink as the output devices, using thermal paste. As the heatsink warms up over the next 15 minutes, the VBE of the bias transistor will drop, automatically reducing the bias voltage to compensate. Re-measure the mV drop across R_E1 after 20 minutes. It should remain stable within ±2mV of your target. If it drifts upward, your thermal coupling is inadequate, and the amp will eventually run away.
Phase 3: Signal and Stability Verification
Once bias is stable and the dim-bulb tester remains dim, you can remove the bulb and apply full mains voltage.
- Inject a 1kHz Sine Wave: Feed a 1V RMS 1kHz sine wave into Node A. Observe Node D on an oscilloscope. The output should be a clean 14.1V RMS sine wave (based on our 14.1x gain) with no visible ringing on the zero-crossings.
- Square Wave Test: Switch the function generator to a 10kHz square wave. Inspect the leading edges on the scope. A slight rounding is expected due to the 68pF Miller capacitor, but if you see high-frequency ringing or parasitic oscillation (>100kHz), your PCB layout has too much stray inductance in the feedback path, or your compensation capacitor needs to be increased to 100pF.
Building a discrete Class AB amplifier is an exercise in thermal and electrical balance. By respecting the node topology, calculating the VBE multiplier with fixed safety resistors, and rigorously testing with current-limited supplies, you will achieve a robust, high-fidelity audio power amplifier circuit that outperforms commercial silicon alternatives.
