FREQUENCY RANGE REQUIRED FOR GOOD TRANSMISSION OF SPEECH AND MUSIC*

As a result of the work of a number of authors, it had long been known that, for high-quality, so-called “natural” transmission of speech and music, it is necessary to transmit frequencies in the range from 30 to 15 thousand hertz. However, in 1945 a rather carefully performed study by Chinn and Eisenberg¹ was published, from which it follows that in the case of single-channel (monaural) sound reproduction, most listeners prefer a “shortened” range, limited at the top by a frequency of 5000 Hz. To explain this result, three assumptions may be put forward.

* H. F. Olson, Journ. Acoustical Soc. Amer. 19, 549 (1947).

1. The average listener, who has listened for a number of years to radio broadcasts and to gramophone music and who has not listened to natural music, is so accustomed to a truncated frequency range that he prefers it.

2. Musical instruments are not made with sufficient precision. Therefore, in the region of high frequencies they produce inharmonic components that greatly alter the timbre.

3. Distortions in the reproduction of natural music are less noticeable in the case of a truncated frequency range.

These distortions and deviations from the actual original may be of the following origin:

1. Frequency distortions. 2. Nonlinear distortions. 3. Incorrect spatial distribution of sound: a) the relative smallness of the dimensions of a loudspeaker in comparison with the dimensions of the actual sound source; b) separation of the sources in two-channel reproduction; c) dependence of the directional characteristic on frequency. 4. Absence of spatial perspective in reproduction. 5. Phase distortions. 6. Transient-response distortions. 7. Incorrect choice of the microphone location. 8. Influence of the acoustic coupling of two rooms (the broadcast studio and the listening room). 9. Limited dynamic range. 10. Difference between the levels of the original and reproduced sound. 11. Presence of a noise level.

Fig. 1.

Fig. 2.

In order to understand this question better, a series of observations was undertaken, during which the high frequencies were cut off in the natural sound of musical instruments reaching the listeners.

The experiment was arranged as follows (Fig. 1). In the corner of a room whose dimensions were \(7.3 \times 6.1 \times 2.9\) m\(^*\), a small orchestra of several instruments was placed. In the opposite corner were the listeners. The orchestra was separated by a partition, in which acoustic filters in the form of louvers were mounted.

The filters were made in the form of elements 8 feet by 1 foot in size. A cross-section of such an element is shown in Fig. 2; from it one can see that the filters consisted of three-layer perforated partitions placed at a small distance from one another. In the same figure is shown the equivalent circuit of such a filter. The filters were mounted in such a way that they could either be turned perpendicular to the plane of the partition, in which case the listener “received” the entire range of frequencies emitted by the musical instruments, or

\(^*\) The dimensions of the room were chosen so that it would correspond to an ordinary living room for listening to a radio receiver.

with one turn of a handle the filters assumed a position parallel to the partition, and then the frequency range reaching the listeners’ ears was determined by the pass band of the filters shown in Fig. 3. In order that the listeners could not see the position of the filters, the partition

Fig. 3.

was covered with a curtain opaque to light, the frequency characteristic of whose sound transmission is shown in Fig. 4. The listeners knew that some changes had occurred in the character of the music, but they never knew exactly what they were. This was achieved with the aid of an index pointer on which the letters A and B lit up, and the correspondence of these indices with the position of the filters was changed several times during the experiment. The reverberation time of the room was about 0.8 seconds at all frequencies.

Fig. 4.

The orchestra consisted of the following instruments: piano, trumpet, violin, clarinet, double bass, drum, and percussion instruments. In carrying out the experiment, light music was used chiefly as the musical material. The sound level was maintained at 70 to 80 db.

The listeners taking part were both laboratory staff members and random visitors of the most varied professions and cultural levels. The listeners were required to indicate which of the performance variants, A or B, they liked better.

About 1000 listeners were passed through the installation. 69% of them preferred reproduction with the full frequency range, and only 31% with the shortened one. In Fig. 5 the results are shown in greater detail, taking into account the age of the listeners (average of 500 persons). It is evident that the percentage of listeners preferring the shortened range is greatest between the ages of 14 and 20 and between 40 and 65. Apparently, the listeners of the first group had already managed to cultivate their hearing on radio listening and gramophone records instead of natural music, and therefore they preferred the shortened frequency range.

As for the last group of listeners (40–65 years), the somewhat higher percentage of persons preferring the shortened range evidently can be explained by the fact that at this age the sensitivity of the ear to the perception of high frequencies becomes dulled.

The indicated results pertain to the transmission of light music. Since genuine classical music is difficult to reproduce with so small an orchestra, a small control test was carried out with the participation of 150 listeners, using “semi-classical” music. These results proved to be very close to the preceding ones (66 and 34%).

Experiments with announcer speech showed that a considerable majority prefer the full range. With the shortened range the sound was, in the listeners’ words, “not so intelligible,” “muffled,” “muddy,” etc.

The results of these experiments thus showed that, as was to be expected, when listening to natural sound the full frequency range is preferred.

□ Full frequency band
■ Shortened frequency band

Fig. 5.

As for the three possible reasons cited above for preferring the shortened range when listening to reproduced sound, it follows from the experiments that the principal reason is the third, i.e., the presence of a large number of distortions when listening to reproduced sound, and, to some extent (for the youngest listeners), the first: habituation to listening to music in the shortened frequency band created by the radio receiver and gramophone.

L. D. Rozenberg

CITED LITERATURE

1. Proc. J. R. E. 33, 571 (1945).