Projection engineering (Sept 1929-Nov 1930)

Projection Engineering, September, 1929 Page 19 Speech Interpretation in Auditoriums The Relation of Frequency to Articulation, the Masking of Tones and the Importance of Adequate Absorption Qualities By E. C. IVente* THAT different auditoriums of approximately the same size vary greatly in their acoustic properties is apparent even to the casual observer. In some halls it is difficult to follow a speaker who might be heard easily in a room with better acoustics. An auditorium which is acoustically bad for speaking purposes usually has one or the other of two defects ; either the speaker's voice is too faint at remote points, or. on account of excessive reflections from the walls, the spoken syllables do not reach the ears of the listener as articulate sounds but as a chaos of 100 90 60 70 60 I z IU 50 o a. 30 20 0.010 0.009 0.008 0.007 >O006 O OCO.OOS u UJ0OO4 B / I A \ 0 A 0 1000 2000 3000 4000 5000 FREQUENCY (C.P.S.) Fig. 1. (A) Relative energy distribution of speech at each frequency; (B) Per cent energy below given frequency. tones, from which he can extract the meaning only by a tiring effort. In the open air, words reach flic listener directly from the speaker ; in a closed room, however, they are reinforced by reflection from the enclosing walls. This gives rise to the phenomenon that in a room a spoken syllable is heard for some time after it Is uttered, and the greater the refleeting power of its walls, the longer is this time of reverberation. If the room has hard walls the loudness of one syllable .'it some distance from the speaker may still be great enough to Interfere with the Interpretation of the succeeding syllable. There are I bus two extreme conditions: If the walls are highly absorbing, the loudness of the speech at remote parts of the room may be Insufficient : If the walls are hard, the loudness will be almost the same at all parts of the room but the excessive revcrboral ion will make it difficull to distinguish the individual syllables of the speech. Obviously in the design of an auditorium these two extreme conditions must be avoided. * Research Department, Bell Telephom Laboratorit s. Reverberation Time The late Professor Sabine of Harvard conducted a series of noteworthy experiments in which he set up as a measure of the acoustics of an auditorium the time required for the average sound-energy density to fall to one millionth of (60 db. below) its initial value. This time is technically designated as the reverberation time. Sabine found that for a given size of room it has an optimum value. Most of his data, however, refers to sounds of a frequency of 512 cycles, which lies near the middle of the musical scale. The absorption of sound by materials, however, varies greatly with frequency. Hence even rooms having the optimum reverberation time at 512 cycles may yet vary widely in their acoustic characteristics. Sabine recognized this fact, but it was not sufficiently stressed, especially with reference to the interpretation of speech. Speech sounds may be regarded as composites of pure tones of different frequencies and intensities. The distribution of energy among the component tones of representative English speech sounds throughout the frequency range from fifty to five thousand cycles per second has been determined by Crandall and MacKenzie.1 Their results are shown in 90 eo 70 60 z o a40 hJ a. 20 10 0 1000 2000 FREQUENCY (C.PS.) Fig. 2. Syllable articulation when energy below given frequency is suppressed. Fig. 1. Curve A gives the relative amount Of energy in speech corresponding to ih«' frequency shown as abscissa. In the region around two hundred cycles I here is approximately forty limes as much energy as around fifteen hundred cycles. Curve B gives the energies plotted in a little differ cnt way. Here the ordinate gives the fractional pari of the energy I hat lies below the corresponding frequency plotted BS abscissa. This curve shows thai Bixtj per cent of the energy in speech lies heiow five hundred and '/'// »«('<■«/ it, r lew, w\. p 221, March, thirty-five per cent below two hundred cycles. Frequency Suppression and Articulation In view of the preceding data it might seem that the low frequency components in speech are relatively important. However, it has been shown that, although the components in speech lying below five hundred cycles are of value in preserving the naturalness of a speaker's voice, they contribute relatively little to the interpretation of speech sounds. The curve shown in Fig. 2 is from a paper by Fletcher.2 This curve gives the percentage of syllables found in the English language which are correctly understood when all the energy below the frequency given as abscissa is suppressed by the transmitting system. For example, if all the energy below five hundred cycles is suppressed, the articulation is still within two per cent of the maximum, although sixty per cent of the energy in speech lies below this value. This fact points to the importance of considering the absorption characteristic's of materials that are used in rooms for damping purposes. In Fig. 3 is shown the absorption curve for a layer of hair felt such as is frequently used for deadening rooms. This material is merely given as an example: most other types of porous materials used for this purpose have very similar absorption characteristics. This material is seen to have huge absorption in the upper but small absorption in the lower frequency region. The average intensity throughout a room is. to a lirst approximation, inversely proportional to the absorptivity Of all the surfaces. Hence, in a room with a relative large amount of porous materials, the Intensities of the tones of (ho higher frequencies, which we have seen to he important for articulation, are very greatly reduced Journal of the VrankUn Inst., June, 1022. 100 90 60 70 1-60 z s' a a 40 30 / / / / y ' 0 c, C, C, c4 c, c, MUSICAL SCALE 64 126 256 512 1024 2046 FREQUENCY (C PS) Fig. 3. Per cent absorption of hair felt at different frequencies.