I always assumed it was a sine waveform that spent more of its time at the peak.  But I never did scope it out in a waveform editor, so this is pure speculation on my part.