Calculate Frequency Spectrum has odd distribution

I’m using this node from the built-in audio visualization plugin and seeing something odd with it: the low end of the spectrum seems to have more energy—higher values—than it intuitively should. See the screenshot below (low end on the right)—there’s sort of a trumpet curve to it, which is present whatever music I feed in. Even near-silent parts of a song show a distinct increase in volume at the low ends of the spectrum. Anyone noticed this before? Is there some normalization factor I should be applying?