Amplification: the quietest watt is the important one

1 — The question

Music signals spend almost all of their time at a tiny fraction of rated power. Why do most amplifier designs optimise for full output, and how should a design look if it optimises for the first watt instead?

2 — Why it matters

At 1 W into a 86 dB speaker you already play loudly at listening distance. Everything below that — noise floor, low-level distortion, supply ripple — lives inside the dynamic range of real music. A spectacular full-power THD number can coexist with a mediocre listening experience if the quiet end is neglected.

3 — Our approach

• Class-AB output stage with generous bias: the first watts come from the most linear region of the transfer curve.
• Dual-mono secondaries so channels cannot modulate each other through the supply.
• Noise-floor spectrum as a headline metric, not a footnote.
• Thermal behaviour verified on the bench over hours, not on paper.

4 — What we have learned

The DSP board’s digital supply was coupling into the input stage until a dedicated local regulation chain was added — a problem found only by the noise-floor sweep, not by any THD test. Measuring the quiet end is what found it, which is why the sweep stays in the routine.

5 — Open questions

• Optimal bias versus thermal budget in a fan-less chassis.
• How much the DSP bypass relay topology costs in measured crosstalk at HF.

6 — Related work

• Measurement archive: Amp One amplifier (sample data)
• Learn: Active vs. passive speakers