An interesting fact: as volume increases, low pitches seem to get louder than high pitches do. To put it another way, turning up the volume on the stereo enhances the bass, without touching any other controls. Conversely, turning down the volume on the TV late at night to not disturb someone sleeping can make the bass go away, and the audio sound tinny.
The general effect is called equal loudness contours or, more often and little confusingly, Fletcher-Munson curves. This graph is copied around the web a lot, because the actual standards document isn’t reproducible for free:
The first time I saw this graph I had to stare at it a while to understand. The solid lines here are levels which are perceived as the same volume. They zoom upwards at the high and low ends for frequencies we can’t hear well. There are also some wiggly bits in the midrange that have to do with how we hear speech, but none of that is what I want to focus on.
Here’s an easier way to understand the data, from NIH, that inverts the first graph, showing directly how perceived volume changes with sound pressure. It also normalizes by A weighting, which gets rid of the swoopy effects at each end.
The lines show how loudness is perceived as volume is increased. The key thing is that they aren’t parallel: increasing sound pressure makes the low end feel louder faster than it does at the high end.
Suppose your hand is on the volume knob and you are listening to a note two octaves above middle C (1khz). You turn the knob and hear it go from the equivalent of a whisper (30dB) to level of a normal conversation (60dB). Now you play the lowest B note on a piano (31Hz) and set the volume to be at a whisper. The graph says this level is objectively at about 45dB, and by turning the volume knob the same as before (+30dB more), you’ll make it seem as loud as city traffic (as loud as 85dB at 1khz).
Now forget the graphs and listen to the effect yourself. A good test is a movie with a sound track containing big explosions played at both high and low volumes. At low volume the explosions will be wimpy, but you’ll be able to hear speech fine.
I was so surprised when I first got serious about listening and ran into this substantial and under appreciated effect. It has all kinds of consequences, such as that perfect EQ is impossible, even for a perfect sound system playing perfectly recorded music in a perfect listening room. It isn’t actually possible to once and for all correctly balance highs and lows, except for a phrase of music having particular dynamics.
Worse yet, the effect varies between individuals. Perhaps this contributes to why some people often listen to music with volume turned way up – they are unconsciously EQing.
Of course, if I’m selling you audio equipment I’m not going to focus on how relative everything is. The important thing in marketing is how people feel, and showing flat dB graphs make people feel good about their purchases.
The question is why should this effect exist at all? What could be the evolutionary advantage of such a huge perceptual bias? I’ll muse about that in the next post.