Lessons and Canonical Structure#

The evolution from single-transistor Class A to modern Class D is a case study in how engineering design actually progresses. Each stage solved one problem and revealed the next. Looking back, several themes recur throughout the arc.

Recurring Engineering Themes#

Efficiency vs. Linearity#

This is the central tension. A transistor operating in its linear region provides smooth, faithful amplification — but it dissipates the voltage difference between supply and output as heat. A transistor operating as a switch wastes almost no power — but a switch is inherently nonlinear.

Every amplifier class is a different position on this tradeoff curve:

Class A: Maximum linearity, minimum efficiency (~25%)
Class B: Much better efficiency (~78%), but crossover distortion
Class AB: Nearly as efficient as B, nearly as linear as A
Class D: Efficiency above 90%, linearity recovered through feedback and filtering

The progression isn’t a straight line from “worse” to “better.” Each design makes a different trade.

Feedback as the Universal Stabilizer#

Negative feedback appears at every level of the design:

Emitter degeneration resistors — local feedback in the output stage for thermal stability and current sharing
Miller compensation — frequency-shaping feedback in the VAS for loop stability
Global negative feedback — from output back to input, setting gain and reducing distortion
Dead time correction — error-correcting feedback in Class D to mitigate switching artifacts

Feedback doesn’t just reduce distortion. It makes the amplifier’s behavior predictable and repeatable despite component variations, temperature changes, and aging. It’s the single tool that makes precision amplification possible with imprecise components.

Thermal Behavior Shapes Design#

V_BE drops with temperature. R_DS(on) rises with temperature. Power dissipation is non-uniform across a heatsink. Thermal time constants range from milliseconds (junction) to minutes (heatsink). Almost every amplifier failure mode has a thermal component.

Successful amplifier design isn’t just about small-signal analysis — it’s about understanding and managing heat. The V_BE multiplier, emitter ballasting resistors, thermal tracking, SOA derating, and the move to MOSFETs are all responses to thermal reality.

Separation of Voltage Gain and Current Delivery#

This is perhaps the most important architectural insight. The jobs of voltage amplification and current delivery are different and conflicting:

Voltage gain requires high impedance, small signals, and careful linearity
Current delivery requires low impedance, large signals, and raw power handling

The canonical linear amplifier separates these cleanly: the differential pair and VAS handle voltage gain, the output stage handles current. Neither compromises the other. This separation of concerns is what makes the architecture so scalable — you can change the output power by swapping the output stage without redesigning the front end.

The Canonical Linear Amplifier#

After 50+ years of refinement, the standard discrete linear amplifier has settled on a remarkably consistent structure:

Input ──→ Differential Pair ──→ VAS ──→ Driver ──→ Class AB Output ──→ Speaker
                ↑                                         |
                └──────── Feedback Network ←──────────────┘

Stage 1 — Differential input pair: Converts the single-ended input and feedback signals into an error current. Provides common-mode rejection. Often loaded with a current mirror for maximum gain.

Stage 2 — Voltage Amplifier Stage (VAS): Common-emitter stage that converts the error current into a large voltage swing. Miller-compensated for stability. This is where the open-loop gain lives.

Stage 3 — Driver: Emitter followers or complementary pairs that provide current gain to drive the output transistors. Isolates the VAS from the output stage’s nonlinearity.

Stage 4 — Class AB output: Complementary push-pull with bias network, emitter resistors, and thermal management. Delivers current to the speaker.

Feedback network: Resistive divider from output to the inverting input of the differential pair. Sets the closed-loop gain.

This architecture has proven so effective that it appears in virtually every discrete power amplifier design — from the LM3886 IC amplifier (which integrates this exact topology on a chip) to the most exotic audiophile designs. Variations exist in the details, but the structure is canonical.

What the Evolution Teaches#

The story of power amplifier evolution is really a story about engineering constraints:

You can’t have everything. Efficiency, linearity, simplicity, and cost are in tension. Every design chooses a point in that space.
Problems beget solutions beget problems. Solving crossover distortion created thermal runaway. Solving thermal runaway required feedback. Feedback required stability analysis. Each solution enlarges the designer’s toolkit but also enlarges the problem space.
Mature designs converge. The Class AB linear amplifier topology converged decades ago. Class D is converging now on a few dominant architectures (half-bridge with feedback, self-oscillating sigma-delta). Once the fundamental tradeoffs are understood, the design space narrows.
Abstractions earn their keep. The separation of voltage gain and current delivery isn’t just convenient — it’s what makes the whole architecture tractable. Good abstractions reduce the problem to manageable pieces.

Understanding this evolutionary pressure — not just memorizing a table of amplifier classes — is what builds real design intuition.