How Much Difference Does a More Powerful Amp Make?

According to perceived loudness calculation

k^(1/3.5) where k is ratio of powers

your loudness will increase by 15.7%.

Amp rated at lower continuous power might have higher momentary power. Average power needed for music is only few percent of peak power unless you listen to sinewaves.

Power specifications give only general guidance but all depends on particular amp design. Chances are, when comparing amps from the same line/company, that higher rated amp will be louder but 15.7% difference in loudness might not be worth extra money.

Comparing to a car is a little misleading since 10x power is only 2x louder (exponential scale).

Maril555 - Imagine 100W class AB transistor monoblock that dissipates 100W of continuous power into heatsink while driven with the sinewave test signal.

Now, get rid of this large expensive heatsink and replace it with 25W heatsink and use saved money to oversize power supply and up the voltage. You'll get an amp that is rated only 25W continuous but is louder with better control of the speakers at the expense of continuous power rating that is useless since average music power is only few percent of peak power. Peak power is what counts.

Most likely it will not sell very well since most of people, you including, believes that higher continuous power rating means better sound. Tubes might be different story - we need to ask Atmasphere (he knows tubes).

"The other downside of powerful amps is that they feature many output devices (by "output device", I mean output transistors in the case of solid-state amps and output tubes in the case of tube amps). All of those devices muck up the sound in systems that are otherwise high resolution because, among other reasons, it is very difficult to match output device pairs, every device adds noise, and they make the circuit more complex."

Great sounding Coda Technologies S5, 50W/8ohm class A (no global feedback) has 60 output devices. As for noise it is exceptionally quiet with S/N=120dB.

State of the art Atmasphere MA-3 (also no global feedback) amplifier uses even more output devices (about 40 tubes per channel).

In both cases a lot of output devices doesn't "muck up the sound" but makes it better (at the expense of complexity).

Metman - that's true, but I would rather have lower power amp that can perfectly drive very low complex loads. 50W/8ohm Coda S5, I mentioned, can output 200W at 2 ohm and according to Stereophile test can even drive 0.47ohm.

Half of the power loss in crossover, that you mentioned, is only 22% of loudness loss.

Raquel - I've never heard either amp that I mentioned but specified S/N=120dB was confirmed by Stereophile measurement and audition of Coda S5. It is dead quiet and very resolving. As for Ralph amps - reviews praise them just for being quiet and very resolving. Putting two tubes in parallel doesn't make it less resolving or noisier than one tube. Part of SixMoons review of Atmasphere MA-1:

"Compared to my BAT VK-75, the Atma-Spheres just offered more of everything: More detail, more extension, more transparency, better soundstaging, greater micro and macro dynamics, more extended and luscious high frequencies"

You assessment of matching output devices is not accurate because it is easier to get average characteristic by connecting many output transistors in parallel. Linearity also improves since each transistor works at smaller current range.
Noise voltage that limits resolution of bipolar transistor is defined by resistance of the Base and resistance of the Emitter that can be reduced by reduction of emitter current (inversely proportional) while Base resistance become dominating reason for the noise. Placing transistors in parallel not only reduces emitter current but also reduces resistance of the Base (in parallel). Placing many bipolar transistors in parallel is well known noise reduction technique in microphone preamps.

As for negative feedback - It is not the complexity that screws up the sound but lack of bandwidth. Limited bandwidth amp delays the signal (shifts phase). When input signal changes fast, feedback loop is momentarily open creating overshoots (or odd harmonics in frequency domain). Our ears are very sensitive to odd harmonics (especially higher order) because loudness clues lie in them. It is called TIM (transient intermodulation) and is known since 70s.

One way is to compromise THD, IMD, DF and make an amp without global feedback but the other is to use just enough feedback to reduce THD to about 0.2-0.5% and then limit input bandwidth (before feedback summing stage) to one that amp had without feedback applied. There are many great amps designed with or without feedback - there is no simple answer. Some of these great amps are very simple some are very complex.

I just realized that Ralph Karsten has the same last name as first name of the Icepower (that I use) designer Karsten Nielsen. Karsten people design good things.