These are beautiful measurements, of which there are many. However, there is a discrepancy between the measurements and subjective perception. In your posts, you describe a lack of power. Now, technically, with continuous excitation such as sine waves analyzed with FFT, there will be no difference whether your switching power supply is oversized or not. The linked report also measures with an ohmic load, which is standard but also silly, as passive speakers almost always present complex loads and therefore the amps are differ in how well they can handle them. In this sense, these measurements are not very indicative of how 'powerful' an amplifier is perceived to be. When it comes to noise, it's a similar issue. We hear impulses and not tones - that's the current scientific opinion on hearing in the frequency range from about 1kHz. So, the noise in a circuit only presents itself as very slight distortion of the signal and is often hardly detectable in the continuous signal through the feedback circuit of the amplifier - in the measurement often represented in the form of the background noise detected by the FFT, usually without attributing the background noise of the measurement chain without the amplifier. This means that while measurements can demonstrate the function of an amplifier, we do not have suitable measurement methods to measure the quality criteria that we hear subjectively. To test my thesis, I did as always: in simulation, it is possible to make disturbing electrical effects visible and suppress them through electrical filters. The simulation can also simulate impulse behavior, as well as very complex noises. Based on these simulations and objectives, I optimized the circuit and built several variants, testing them on different speakers, sources, and listening rooms with different subjects to minimize the psychological effects of listening. After all, humans have a learning ear and therefore a fundamentally subjective individual interpretation. Nevertheless, this means that general effects are described quite similarly by different individuals.
