I always attack these issue from the recommendation once given by French speaker manufacturer Cabasse
in a very well written leaflet from the '70s.
Considering that a baby grand piano can produce around 115 dB peak SPL at 1 m meter (for brief milliseconds) when playing
fortissimo, Cabasse advocated to compute the power of amplifiers in order to achieve this sound pressure level, which is realistic at home. For that, it is necessary to take into account the sensitivity specification of the speaker (hoping it is accurate and fairly measured...), the peak power output corresponding to the mean power specification of the amplifier (for sinus signal, peak power is double that of mean power) and a 3 dB gain due to the use of two speakers in stereo, speakers whose combined outputs will roughly add up to 3 dB in a room.
Application in musica's case :
Gain (in dB power) required to obtain 115 dB : 115 (aim) less 3 dB (gain of use of 2 speakers in stereo) less 91 dB (sensitivity of the actual speaker used), that give 21 dB of gain (115-3-21).
If 1 W set 0 dB, 21 dB will be around 130 W (10 log 130 = about 21).
Since we speak of a peak power requirement, that means that the power of the amplifier, which is specified with sinus signal, is half 130 W, that is 65 W.
For the reproduction in high fidelity of a realistic home instrument (a baby grand piano) at home with speakers of 91 dB/1W/1m sensitivity, the amplification power required is thus 65 W per channel.
Or course, the faithful reproduction of more robust instruments (church organ, percussion, orchestra) or music (electronic music with deep bass) would required more power to be listen to undistorted (if well recorded and produced).
Hence, 65 W per channels is a bare minimum, not a limit!
This way of thinking excludes some caveats (behaviour of specific amplifier, speaker power handling, crest factor of different musical genres, listening habit...) that are thoroughly addressed in the above mentioned leaflet, but I think that, in this day and age, when power is cheap and speakers have became more robust, it would add unnecessary complexity just to give a simple advise.
