Then maybe you don't need to equalize it? Run your experiment with both sources at a wide range of volumes. If there's actually a noticeable difference, it should come across in the data.