Dynamic Headroom

Edartford - It is class D amp with full H-bridged Mosfets. There is practically no voltage drop on them (possibly a volt total).

As for amps being design for "huge peak-to-average" - that's true but this average might vary. Some amps will handle Jazz nicely but give up at heavy orchestral piece (much higher average).

Also note that headroom is not necessarily "how much punch is available above maximum rated power". No one (virtually) listens at maximum rated power anyway (well perhaps a few have done so, but speaking realistically it is not an acceptable practice).
Anyway: your *available* headroom is the difference between the actual power level at which you're listening to, compared to the maximum available peak.

20dB, or more, is considered optimal headroom.
10dB is 10 times the power; 20dB is 100 times the power. So if you're listening at 2 watts then you need 200 watts available if you want that 20dB of headroom.

My rig is setup so that it virtually never clips. With high effeciency speakers (105dB SPL @1 watt 1 meter) if I'm listening at 2 watts per channel it's already quite uncomfortably loud for most people. So you need a decent sized amp and high efficiency speakers to stay 100% of out clip.

Class of operation has everything to do with headroom. A class A amplifier has by definition 0 db of headroom.

I have always looked at the headroom spec as a way to make cheaper amps look better due to the term itself. IOW 'headroom' seems to imply that an amp that *has* headroom is better than one that does not, but IMO/IME the opposite is true. Theoretically and often in the real world, class A is as good as it gets.

"A class A amplifier has by definition 0 db of headroom."

I'm not so sure of that. Class A is often designed with bias current in order of 150% o max output current to guarantee that transistors won't become nonlinear - hence some headroom might be possible.

In a Class A amp the output devices do not go into cutoff within the amp's linear region. Sure, you have the amp biased heavily, but that does not affect output voltage.

Many people are surprised to find out that the bias point does not affect the output power (that is a function of the power supply- class AB amps usually have a higher Vcc or B+ than class A amps do so you don't roast the output devices with too much dissipation). Bias affect the *distortion* of the amplifier- class A allows a push-pull amplifier to cancel harmonics at any point in its operating region. In the case of a single-ended circuit, it puts the operation at the most linear portion of the output device's curve.