At the risk of further complicating the already complex, I've read in psychoacoustic research that for listeners to report one audio stimulus (most likely a single tone in lab situations) as "twice as loud" as another there must be a 10-decibel difference in terms of power delivered to the transducer (probably earphones). If that relationship holds for real-world listening situations, "twice as loud" as a 1-watt average power level to the speakers will be 10 watts, "twice as loud" as that will be 100 watts, and 99.9% of us can't afford "twice as loud" as that.