My amplifier is Rotel RMB-1075, THX certified. It has app the same power when 2 channels are played. Here are the specs:
THD (20-20,000Hz)
cont. rated power 0.03%, maximum
one-half rated power 0.03%, maximum
one watt per channel 0.05%, maximum
IM Distortion (at rated ower, 60Hz:7kHz = 4:1) 0.03%, maximum
Damping Factor (8 ohms) 180
So, if I turn the amp to 0.01W of power it will play quietly, somewhere between 60-65dB with speakers of 87dB/1m sensitivity. How am I supposed to hear the distortion of 0.05% which corresponds to -66dB when it is lower than the SPL of the test tone?
In answer to your first question, the fundamental does not fully mask the harmonics that are, almost always, at significantly lower levels. If they were masked, all musical instruments would sound like tea kettles (producing what would sound like pure sine waves). Harmonics are audible because they are separated by frequency.
In further answer to your question, the THD+N produced by your amplifier will not continue to go down at lower levels.
This can be seen from the fact that the one watt THD+N is higher than the half-power THD+N. This means that the 0.01 W THD will be much higher than 0.05%. Here are some calculations:
At 1 W, 0.05% is -66 dB relative to 1 W.
At 0.1 W, the THD+N will still be about -66 dB relative to 1 W or -56 dB relative to 0.1 W. THD+N will measure 0.16% at 0.1 W from your amplifier.
At 0.1 W, the THD+N will still reproduced at the same acoustic level (87 dB - 66 dB = 21 dB SPL).
At 0.01 W, the THD+N will still be about -66 dB relative to 1 W and the acoustic level of the THD+N will still be 21 dB SPL.
At 0.01 W the THD+N will measure -46 dB which is 0.5 %.
What I cannot tell from your specifications is the spectrum of the THD+N at low power.
The limitation may be noise (the N portion of THD+N) or it may be THD produced by crossover distortion.
An FFT of the low-power output will show the difference between noise and THD. This is an important test, but it rarely shows up on a spec sheet.
Unfortunately, low-power THD+N is rarely limited by random noise. It is usually crossover distortion plus power supply noise (line-related hum and buzz).