One thing to look at is minimizing thermal compression.
As a ballpark rule of thumb, a driver will exhibit about 1 dB of thermal compression at 10% of its rated power. Now this "rated power" would be according to some engineering standard. Usually "music program power" is twice the actual rated power. And "recommended amplifier power" is who knows what, but probably at least twice the actual rated power. "Peak power handling" is utterly useless. So you want to take whatever the manufacturer is claiming and transcribe that back to "rated power".
Next, you want to consider the "crest factor" in music. The crest factor may well be 20 dB or more on good recordings, which means that you have peaks 20 dB louder than the average volume. So if you're going to listen at 85 dB average, you'll probably want to be able hit 105 dB peaks with negligible (1 dB or less) thermal compression.
And look at how the speaker's efficiency is rated. Anechoic? In-room? Is it relative to one watt, or 2.83 volts? Based on broadband average SPL, or a single peak? What protocols a manufacturer follows can make a big difference on paper, and you need to correct back to 1 watt (at the listening position) efficiency to make apples-to-apples comparisons for the sake of examining thermal compression. I know of an instance where one speaker was rated 8 dB louder than another, yet their real-world efficiency was essentially identical.
Okay, we didn't talk about amps much, but consider that the speaker's impedance curve may well work against the amp's rated power output. Impedance curve dips and/or peaks (especially above the bass impedance peaks), along with nasty phase angles, can all reduce the real-world power than a speaker delivers. Plenty of reserve power seems to always contribute to improved dynamics, so err on that side.
How far back are you sitting? How big is the room? Do the speakers approximate a point source, or a line source? Wide pattern or narrow pattern? All of this will affect the SPL at the listening position, which is what you want to be looking at. As a ballpark approximation, you can probably expect the room's reflections to add back about 3 dB at the listening position from a typical point-source speaker in a moderately reverberant room. If the manufacturer already did that when he calculated the speaker's rated efficiency or sensitivity, you don't get to add that back in again. But you do get to add the second speaker's output, which will increase the SPL at the listening position by about 6 dB relative to a single speaker. So suppose you have honest 88 dB/1 watt/1 meter point-source-approximating speakers, and you listen from about 13 feet away. Why thirteen feet? Because that makes the math for this example easier - you see, thirten feet is about 4 meters. So expect about 12 dB of theoretical anechoic falloff going from one meter back to 4 meters (6 dB per doubling of distance from a point source), then add back in 3 dB for the room's reflections, plus another 6 dB because you're listening to two speakers, and you have 85dB/1 watt at the listening position. In order to deliver uncompressed peaks of 105 dB at the listening postition, your speakers would have to be rated at about 1000 watts (not "music program", not "recommended amplifer power" and certainly not "peak" - but rather, rated consistent with an accepted engineering standard).
If you listen at about 9 feet, then you need 3 db less output from your speakers to hit your target of 105 dB uncompressed at the listening positon. So you can get there with honest 500 watt rated speakers.
This of course is not all there is to dynamics, but it's a start. A little arithmetic will show that efficient speakers with moderate (doesn't need to be extreme) power handling are good candidates. Subsitute an honest 98 dB speaker into the example above and you can get there with an honest 100 watts rated power handling at 13 feet, or 50 watts at about 9 feet.
And I'm NOT saying these numbers are absolutes; rather, I'm trying to raise awareness of thermal compression and emphasize the role that speaker efficiency and power handling play.
Duke
