Class a Amplifier
A Class A amplifier biases its single active device — a bipolar transistor, MOSFET, or vacuum tube — at the exact midpoint of its load line so current flows continuously through the full 360° of the input cycle, never cutting off and never saturating, which buys the lowest distortion of any amplifier class at the cost of burning power even when no signal is present at all.
The 25% ceiling is a geometry problem, not an engineering failure. For a single-ended Class A stage driving a resistive load, the theoretical maximum efficiency is 25%, rising to 50% with a transformer-coupled or inductive load that lets the device swing to twice the supply rail — either way, most of the DC power drawn from the supply turns into heat in the transistor rather than signal in the load, and real designs with safety margin typically land at 15-20%. A 25-watt Class A amplifier like Pass Labs' XA-25 draws roughly 300 watts continuously from the wall and needs a heatsink sized for that number, not for 25 watts, which is why Class A gear is heavy, hot, and expensive to run even idling on a shelf.
Bias current never changes, which is exactly the point. Because the quiescent current Icq is set well above the peak signal current and the device stays in its active region across the whole swing, there is no crossover region where a second device hands off to another — the nonlinearity that defines Class B and AB simply does not exist in a single-ended Class A stage, which is why audio designers like Nelson Pass and Krell have built entire product lines around accepting the efficiency penalty in exchange for the lowest measured distortion achievable from a single gain stage.
It still dominates where efficiency never mattered. Small-signal RF front ends — low-noise amplifiers ahead of a receiver's first mixer, microwave test instrumentation, and most op-amp input and driver stages — run Class A because the signal power is trivial compared to the DC bias budget, so linearity and noise performance win every tradeoff; a cellphone LNA biased Class A might draw a few milliamps, where efficiency is irrelevant next to preserving a weak signal's integrity before it gets buried in amplifier noise.
Thermal design is the real constraint, not circuit topology. Because dissipation is highest at idle and barely changes with signal level, a Class A amplifier has to be designed to survive its worst case continuously — heatsink mass, fan cooling, and enclosure ventilation are sized for the no-signal condition, which is the opposite of Class AB designs where heatsinks are sized for sustained high-power operation that happens only intermittently in real listening or use.
Read a Class A amplifier through a *bias-as-insurance* lens rather than an *efficiency* lens: the designer is spending supply power to buy headroom, keeping the device away from cutoff and saturation at every point in the cycle so distortion stays near the noise floor, and that insurance premium — 75-85% of the DC power turned straight into heat — is the deliberate price of the lowest distortion a single active device can deliver.