Reference

A quick look at the treble really shows this up. For most purposes, the 5kHz range from 15kHz-20kHz may be discarded, and very often is, by a combination of source spectrum, microphone, filters, and loudspeakers, many of which are quite ineffective above 12kHz. Compare that 5kHz half-octave bandwidth of upper harmonics and noise with just 50Hz worth in the bass range below 100Hz. This latter span is a whole octave of fundamental bass playing, in most works representing the foundation for musical composition. Rock could hardly exist without it. Every 10Hz of good, even, controlled bass extension matters in the design and specification of loudspeakers, and makes their design and evaluation such a subjective matter.

Perception of bass loudness
To understand why reproduced sounds are not always heard as one might expect in the bass, a closer look must be taken at the curves for perceived loudness. These are often used in an oversimplified way (see fig.2, which shows the free-field equal-loudness contours developed by Robinson and Dadson). In general, a discussion about audible loudness steps concentrates on the midrange, a region of naturally high sensitivity. This can be misleading where the bass region is concerned. There is a wealth of fascinating detail in these curves, and it is important to arrive at a better understanding of their implications. This part of the discussion was promoted by a casual aside made by Floyd Toole, of Canada's National Research Council, during a recent lecture on room acoustics at a meeting of the London AES: "Of course, we are more sensitive to small variations in bass level than we are in the midrange." In a related discussion, he also noted that a little bass appears to go a long way subjectively, or words to that effect. These two observations have serious implications.

Fig.2 Robinson and Dadson free-field equal-loudness contours (after The Audio Encyclopedia). Low- frequency sounds have to be played at a much higher SPL to sound as loud as mid-frequency ones.

First, it should be pointed out that our absolute sensitivity to bass sounds is much poorer than in the midrange. Evolution has provided us with a tailored aural sensitivity tuned to the real world: twigs breaking, rustles in the grass, the calls of animals and birds. The threshold of audibility is defined as a sound pressure of 0.0002dyne/cm at 1kHz, and "0dB" is assigned to this level. Between 3kHz and 4kHz, our hearing is still more acute; many younger listeners can detect sound levels down to -10dB in this region. Moving into the treble range, by 10kHz some loss in sensitivity is normal (though such loss increases with age). Nevertheless, levels down to 12 or 15dB above the threshold at 1kHz are still audible, if very quiet. However, if we look at the low-frequency region, the picture is very different. Consider that dotted threshold sensitivity line.

Even by 150Hz the typical sensitivity has fallen below that for 10kHz. To be audible, a 75Hz tone in the midbass must be 30dB louder than the 1kHz, 0dB threshold, while 50Hz requires 40dB, 100 times the pressure. The required level for audibility at 25Hz is over 60dB, one thousand times, although it is worth noting that the test results are fundamentally based on headphone listening and that the whole-body excitation which results from room-propagated bass, which can tend to enhance bass sensitivity, is absent.

As a direct consequence of these curves, there are also changes in tonal quality which occur with music according to the volume level at which it is reproduced. The subjective loudness curves represent a complex dynamic variable in which an envelope representing a musical section must exist. That envelope bounds the frequency response and loudness variation of that passage or whole work, and it must be located somewhere in the hearing response curves. As Peter Walker has concisely stated, "there is only one correct volume level for any particular piece of music." Taking into account listening-room acoustics, the original recording technique, etc., in theory there is just one volume level which places the dynamic music envelope correctly in the characteristic curves. A particular instrument was played at a particular loudness, and a matching tonality should be reproduced at a matched level relative to the listener and his expected location. Only then will the instrument sound natural, and the normal dynamics of musical expression be reproduced in the expected range.

Thus the perfectionist user of a volume control should learn to reproduce the correct, natural playback level for each and every recording.

Setting any other volume level, such as for the comfort of one's family or neighbors, may be judged a distortion of the truth. One is making a convenience of the reproducing system and partially devaluing the musical message. Conversely, one could argue that in a concert hall one can personally arrange for a different performed loudness and resulting tonal balance by selecting seats nearer or farther away from the performers. Here, as with reproduced volume, taste and preference also play their parts. In recording practice, the taste and judgment of the balance engineer is a dominant influence where the distance between the microphone and the performers is the major variable.

It is an accepted part of our experience that sound sources not only get quieter the farther away they are, but also that distance lends an appropriate change in tonal quality. Only a small part---a degree of treble softening---is due to specific high-frequency losses incurred over the intervening air path. The bulk of the change in perceived frequency response is due to one's ears, the placement of the sound envelope further down in the set of characteristic curves. Take, as a reference, an 80dB level; this is typical of loud nearby speech. Now consider the subjective frequency-response changes in the frequency range below 800Hz, the latter approaching the highest practical compass of a soprano. Comparing the 80dB level at 1kHz, the sensitivity at 180Hz is also on the line. However, when the overall sound level is set 20dB lower to 60dB, this 180Hz frequency is now heard 2-3dB more quietly than it should be, while at a 40dB intensity the relative loss has increased to 4dB. By a whisper-quiet 20dB, the relative loss has increased to 8dB, additionally accompanied by several dB of associated loss in the lower-mid frequency range centered on 500Hz.