Beyond Power Ratings: A Qualitative Benchmark of Output Stage Topologies in Armorly Reference Amps

Why Power Ratings Obscure Real-World Amplifier Performance

The headline wattage number on an amplifier spec sheet tells you very little about how that amplifier will sound, drive complex loads, or maintain stability under dynamic conditions. In my two decades of audio system design and evaluation, I have repeatedly encountered cases where a 100-watt per channel amplifier was outperformed in every audible dimension by a 50-watt design using a different output stage topology. This is not an anomaly—it is a direct consequence of how output stages handle current delivery, thermal recovery, and feedback loop stability. Power ratings are measured under static, resistive loads with continuous sine waves, conditions that bear almost no resemblance to music reproduction. Real music signals are transient-rich, with peak-to-average ratios exceeding 15 dB, and speaker impedances that can dip to 2 ohms or swing wildly with frequency. A traditional push-pull Class AB stage may deliver its rated power into 8 ohms but current-limit or thermally sag into a 4-ohm load during a drum hit. Meanwhile, a well-designed single-ended Class A stage, despite lower continuous power, can deliver full current instantly because its output devices are always biased in their linear region. The qualitative benchmarks that matter—transient fidelity, load invariance, noise floor, and distortion profile—are determined by topology choices far more than by the maximum unclipped sine wave power. This guide provides a structured comparison of four major output stage topologies used in Armorly Reference Amps: single-ended Class A, push-pull Class AB, Class D, and current-feedback. Each topology is assessed using criteria that reflect real listening and measurement conditions, not marketing claims.

The Deceptive Simplicity of Continuous Power Measurements

Continuous power testing (FTC standard) uses a 1 kHz sine wave at 1% THD into a resistive load. This masks several critical behaviors: the amplifier's ability to recover from voltage clipping, its current delivery under complex reactive loads, and the thermal dynamics that cause long-term drift. In one composite scenario, a 200-watt Class AB amplifier measured perfectly on the bench but exhibited audible distortion when driving a 4-ohm bookshelf speaker with a 30 Hz bass note followed by a 10 kHz transient. The issue was not power headroom but output stage topology: the slow recovery of the complementary Darlington pair from the high-current bass transient caused intermodulation with the high-frequency signal. A 50-watt current-feedback amplifier, with its inherently fast slew rate and minimal phase shift, reproduced the same passage without artifacts. This illustrates why qualitative benchmarks—not power numbers—should guide amplifier selection for critical listening.

Qualitative Benchmarks Defined

We define five benchmarks for this evaluation: transient accuracy (rise time, slew rate, and overshoot), load stability (behavior into 2–8 ohm reactive loads), spectral noise (including power supply ripple rejection and idle noise), distortion profile (harmonic structure versus amplitude), and thermal resilience (how output stage biasing drifts with temperature). Each topology is scored qualitatively based on typical implementations in Armorly Reference Amps, using composite observations from system integration projects. The goal is to provide a decision framework for matching amplifier topology to application, whether that is studio monitoring, high-end home audio, or portable high-fidelity systems.

Single-Ended Class A: The Purity of Constant Current

Among output stage topologies, single-ended Class A holds a unique position: it is technically inefficient, thermally demanding, and limited in continuous power, yet it is revered for its linearity and absence of crossover distortion. The core principle is that a single output device (or a bank of devices in parallel) conducts current for the entire 360-degree signal cycle. This eliminates the nonlinearity that occurs at the zero-crossing point in push-pull stages, where one device turns off as the other turns on. In practice, this means the single-ended Class A stage operates in its most linear region at all times, producing a distortion profile dominated by low-order harmonics that are perceptually benign. The trade-off is severe: maximum theoretical efficiency is 25%, and real-world implementations often achieve only 10–15%. A 50-watt single-ended Class A amplifier may dissipate 300–400 watts of heat, requiring massive heatsinks, forced cooling, or both. In Armorly Reference Amps, the single-ended Class A topology is reserved for preamplifier stages and low-power headphone amplifiers where thermal management is feasible and the sonic benefits are most audible. One composite scenario involved a high-resolution headphone system where a single-ended Class A amplifier was preferred over a Class AB design for its ability to render microdynamics—the subtle decay of a piano note or the air around a cymbal—without the graininess introduced by crossover artifacts. Listeners consistently described the sound as 'liquid' or 'effortless,' terms that correlate with the absence of high-order harmonic distortion and the uniform transconductance of the output device over the signal range.

Thermal and Bias Stability

The Achilles' heel of single-ended Class A is thermal drift. As the output device heats up, its bias point shifts, potentially moving it out of the linear region or causing thermal runaway. Armorly Reference Amps employ sophisticated servo biasing circuits that monitor the temperature of the heatsink and adjust the bias voltage dynamically. In one design review, we observed that a passive biasing scheme allowed the idle current to increase by 30% after 20 minutes of operation, raising distortion and reducing device lifespan. The servo-corrected version held idle current within 2% over the same period. This is a qualitative differentiator: a well-regulated single-ended Class A stage can maintain its linearity over long listening sessions, while a poorly designed one may sound pristine at turn-on but degrade as the session progresses.

Load Sensitivity

Single-ended Class A stages are generally tolerant of load impedance variations because the output device is always conducting. However, they are current-limited by the quiescent current setting. If the load demands more current than the bias current (which is typically set to twice the peak output current), the stage transitions into Class AB behavior, introducing crossover distortion. Designers must set the bias current high enough to accommodate the lowest expected load impedance. For an 8-ohm load, a 25-watt amplifier requires a bias current of about 1.8 A, which is manageable. For a 4-ohm load, the same power requires 3.5 A, doubling heat dissipation. This makes single-ended Class A practical only for high-impedance, high-sensitivity speakers or headphone loads. In a studio monitoring scenario, a pair of 300-ohm headphones driven by a single-ended Class A amplifier delivered exceptional clarity and spatial imaging, while the same amplifier struggled with a 32-ohm headphone, producing audible distortion at moderate levels. The topology's strength is its linearity into benign loads; its weakness is its inability to adapt to demanding loads without excessive heat or distortion.

Push-Pull Class AB: The Workhorse with a Crossover Problem

Push-pull Class AB is the most common output stage topology in consumer and professional audio amplifiers, balancing efficiency (up to 60%) with reasonable linearity. It uses two complementary output devices (NPN and PNP, or N-channel and P-channel MOSFETs) that conduct for slightly more than 180 degrees of the signal cycle, with a small idle current to smooth the transition between them. This 'standing bias' reduces but does not eliminate crossover distortion—a nonlinearity that occurs when one device hands off to the other. The quality of a Class AB amplifier is largely determined by how well this crossover region is managed. High-end implementations employ bias servo circuits, emitter resistors, and sometimes local feedback to linearize the transition. In Armorly Reference Amps, the push-pull Class AB topology is used across a wide power range, from 50-watt integrated amplifiers to 500-watt monoblocks. The trade-off is that even with sophisticated biasing, the amplifier's distortion profile changes with output level: at low levels, crossover artifacts dominate; at high levels, the devices approach their saturation limits, introducing clipping and higher-order harmonics. One composite scenario involved a multi-channel home theater system where a 200-watt Class AB amplifier was driving a center channel speaker. During a movie with heavy bass and dialogue, listeners reported a 'harshness' on sibilant sounds. Measurement revealed that the crossover region was being excited by the high-frequency content of the dialogue, adding 0.1% third-harmonic distortion that was not present in a pure sine wave test. The solution involved increasing the bias current, which reduced the artifact but raised idle dissipation from 30W to 50W per channel. This exemplifies the inherent tension in Class AB design: bias must be set high enough to avoid crossover distortion but low enough to prevent excessive heat and power waste.

Bias Optimization Trade-offs

Setting the optimal bias current is a balancing act. Too low, and crossover distortion becomes audible, especially at low listening levels. Too high, and the amp runs hot, reducing component lifespan and potentially triggering thermal shutdown. Many manufacturers ship amplifiers with conservative bias settings to ensure reliability, sacrificing a small amount of linearity. Armorly Reference Amps use a temperature-compensated bias circuit that adjusts the bias current dynamically based on heatsink temperature. In a typical listening session, the bias might start at a moderate level and increase as the amp warms up, reaching an optimum after 15–20 minutes. This approach yields the benefits of higher bias without the thermal penalty of a fixed high bias. However, it requires careful loop design to avoid oscillation or bias drift. I have seen systems where the bias servo reacted too slowly, causing a 'thump' when the amp transitioned from standby to active. Proper compensation ensures a smooth warm-up and consistent performance.

Load Tolerance and Protection Circuits

Push-pull Class AB amplifiers are generally robust with load variations, but they are prone to secondary breakdown in the output devices if driven into high-current, low-impedance loads for extended periods. Armorly Reference Amps incorporate SOA (Safe Operating Area) protection circuits that limit current and voltage in real time, ensuring the devices operate within their safe bounds. These circuits can introduce nonlinearities if they activate during musical peaks. In one installation, a subwoofer amplifier with aggressive SOA protection would 'soft clip' on deep bass notes, producing a compression effect that was audible as a loss of punch. Adjusting the protection threshold and implementing a soft-knee limiter resolved the issue without sacrificing reliability. The qualitative benchmark for Class AB is not just maximum power but how gracefully it handles overload conditions. A well-designed Class AB stage will limit smoothly, preserving the sound's character, while a poor design will produce hard clipping or protection latch-up.

Class D: Efficiency and the Challenge of Filtering

Class D amplifiers, often called switching amplifiers, have become dominant in portable, automotive, and high-power applications due to their efficiency (typically 80–90%) and compact size. Instead of using output devices in linear mode, Class D stages switch the output devices on and off at a high frequency (typically 200 kHz to 1 MHz), using pulse-width modulation (PWM) to represent the audio signal. The output is a high-frequency square wave that is low-pass filtered to recover the audio. The theoretical efficiency is high because the output devices are either fully on or fully off, dissipating minimal power in the transition. However, the practical implementation introduces several qualitative challenges: switching noise, filter linearity, and electromagnetic interference (EMI). In Armorly Reference Amps, Class D stages are used in subwoofer amplifiers and portable units where heat dissipation is limited. The topology's ability to deliver high power without massive heatsinks makes it ideal for these applications. One composite scenario involved a subwoofer amplifier rated at 1000 watts RMS but weighing only 5 kg, compared to a Class AB amplifier of similar power that weighed 20 kg. The Class D amplifier performed flawlessly in extended bass reproduction, with no audible artifacts, because the switching frequency (500 kHz) was far above the audio band, and the output filter was designed with low-inductance, high-current components. However, when the same amplifier was used with a full-range speaker, listeners noted a slight 'haze' on high-frequency content, likely due to residual switching noise or intermodulation between the switching frequency and the audio signal. This highlights the importance of output filter design: a poorly designed filter can introduce phase shifts, ringing, or noise that degrades audio quality, especially in the treble region.

Switching Artifacts and Feedback Topology

Class D amplifiers use feedback to correct for non-idealities in the switching process, such as dead-time distortion, supply ripple, and filter nonlinearities. The feedback can be taken from the output of the switching stage (before the filter) or from the output of the filter (post-filter feedback). Post-filter feedback is more accurate but introduces stability challenges because the filter's phase shift can cause oscillation. Armorly Reference Amps use a proprietary feedback scheme that combines pre-filter and post-filter signals, achieving low distortion (