The auditory system

The auditory system can be thought of as an extremely complex and highly developed local warning system. It is extremely sensitive, and it has an elaborate, three-dimensional localizing feature. This system can locate an object in space, and it does this both by (1) direct observation of the high-frequency components of the sound at each ear, and by (2) analysis of the assorted room reflections of that sound, including their various delays and the above mentioned angles of arrival. Furthermore, this amazing system also locates and establishes the limits of the room we are in by observation of the phase differences of low-frequency reflections. Such a highly complex and multi-stage echo-location system works mainly in the pre-conscious realm, and our conscious perception of sound is in fact a highly developed construct that has comparatively little relationship to the original physical sound pressure waves. (The system, as an aid to survival, evolved to give us useful information which we could trust and act on quickly, without thinking! The merging of early reflections with the direct sound is a clear example of this: in just 50 milliseconds, the brain receives and processes enough data to clearly present to the conscious mind an explicit and uncluttered image of a specific sound coming from a specific point in a given space!)

If we think about the actual information in visual terms (an exercise I always find useful and entertaining when I’m trying to think about these things), the sound and its early reflections are the equivalent of a light bulb in the mirror room in an amusement park fun house. We see dozens of reflected lights that are pretty much indistinguishable from the real one, and are faced with an illusion that is chaotic and confusing. To help us survive, the ear evolved to sort the audible equivalent of this fun-house chaos out and tell us, almost instantly, which sound is the real one.

Now here comes the really weird part. We have a paradox, a conflict in what we know. On the one hand, if we recombine a sound and its delay, we get comb filtering, which really changes the frequency response (and timbre) of the sound, in major-league ways. On the other hand, I’ve just been telling you (and you’ve probably noticed this on your own as well), we don’t hear the early reflections as such; they combine with the direct sound to create a richer version of the sound, but without obvious or major damage to the quality of the sound.

How can this be? Is there comb filtering? Yes. Do we hear it? Well, yes and no (I told you this is weird). Here’s the scoop: if the original sound and its reflection have the same angle of arrival, we hear comb filtering and a dramatic change in timbre. If the original sound and its reflection arrive from different directions, we barely hear the comb-filtering. There is no appropriate explanation yet as to why this is so. The comb filtering is still there and can be easily observed or measured with a real time analyzer. You can also demonstrate this for yourself if you have a delay line. Take a noise signal and run it to one speaker. Split the signal and run it through the delay and mix it with original signal going to that speaker. Voila! Comb filtering! Big time! Now send the delay to at the other speaker instead. Voila again! No comb filtering! The signal sounds like itself. Awesome!

So, the general rule is: when sound arrives from a single direction, any delays inherent in that sound are considered to be information about timbre. When the same sound arrives from multiple directions, any delays are considered to be information about space. In a real room, where there never is a single sound or even just one reflection, the multiplicity of reflections from different directions reduce even further the audibility of comb filtering. The fact that we have two ears and are therefore observing multiple points in space probably helps as well. So, mostly, in the real world, we don’t hear the effect of comb-filtering, even though it is there in spades.

Nevertheless, comb filtering is a big problem in a crucial area of audio: the mono compatibility of stereo recordings. When we sum stereo sounds to mono, comb filtering is clearly heard. Multi-directional delays become single-source delays. What was effective and powerful spatial information in the stereophonic mix becomes unintended timbral information in the monaural mix. So, good engineering practice demands a careful consideration of the impact of the use of time delays on the monaural mix, particularly if the mix may be broadcast, as monaural summing is a normal feature in the reception of radio and TV.