Dynamic Headroom

In addition to Al's comments, the other problem with "peak" power is there is no standard definition of how to measure it. One amp may be measured at the power level that can be sustained for 1 second while another amp may measured for 20 milliseconds.

That makes comparing the true performance of one amp to another based on that number unreliable.

Still another issue is that music does not follow any set rules as to what constitutes a peak. The peak power that may be helpful for a drum strike that has a peak duration that is a fraction of a second long is going to be useless for a sustained organ peddle note.

If you need a certain amount of power to reach a certain volume, the only way to make sure it is there is to have an amp that is capable of sustained output at the desired level.

This is without getting into the issue of what is actually on most recordings in terms of dynamic range. (Most people think the dynamic range in recordings is much greater that what is typically present.)

In plain English
...means an ability to reproduce a large interval between 'quiet' and 'loud'.
The sound system (mainly defined with amplifier and speakers for dynamic headroom) with large dynamic headroom does not compromise the sound quality with large increase or decrease of sound.

Eldartford wrote: "Almarg...I don't agree with your description of the power supply as "weak". It may have been designed that way."

I'm not sure what you disagree with. Nobody designs power supplies for class AB amps to be able to sustain sinewave at full power for unlimited time - otherwise it would require heatsinks and transformer of the size of class A amp. This would not make any sense since average music power is only few percent of max power. Power supplies being much smaller are already compromised (weak). Al is just saying that within given power supply size (wattage) you could design for higher output voltage (getting better headroom) sacrificing output current and making it weaker for average power demand.

What I do know is that I had once a Luxman R107 receiver. Maybe 55 watts a channel and (back in the day, with my Cerwin vega D9 speakers) the system sounded extradorinarilly good (for the price). I had to investigate further. What I found out was the Luxman had 4db of dynamic headroom and my prior rig only 2dbs (same watts and, in those days, THD (remember THD?). Many of my engineer friends agreed that was the reason for the Luxmans great sound. Me? I just know it sounded better.

Kijanki...Agreed that "Unlimited time" would be more than "Continuous" in the definition of power. I don't know how the spec is exactly defined but I think that "continuous" could be 5 seconds or so. This would be long enough to reflect any power supply sag, but not long enough to get into heat sink issues.

Since the average power while playing music is a handfull of watts, but the peak may be hundreds (I have made measurements) I do believe that ability to follow peak voltage is very important. Power supply design is influenced by the need to advertize high "continuous rms power".

Using my MG1.6 speakers playing LOUD I measured average rms voltage in the 5 to 10 volt range but I still saw brief peak voltage corresponding to almost 600 watts. I was using 600 watt amps so I was able to see this voltage.
In the end I decided that while 600 watts for a MG1.6 is not ridiculous, 350 would probably be OK (so I bought CT D200 monoblocks).