Volume 18, Issue 1, July 1946
Index of content:
- PROGRAM OF THE THIRTY‐FIRST MEETING OF THE ACOUSTICAL SOCIETY OF AMERICA
18(1946); http://dx.doi.org/10.1121/1.1902419View Description Hide Description
The sound spectrograph is a waveanalyzer which produces a permanent visual record showing the distribution of energy in both frequency and time. This paper describes the operation of this device and shows the mechanical arrangements and the electrical circuits in a particular model. Some of the problems encountered in this type of analysis are discussed, particularly those arising from the necessity for handling and portraying a wide range of component levels in a complex wave such as speech. Spectrograms are shown for a wide variety of sounds, including voice sounds, animal and bird sounds,music, frequency modulations, and miscellaneous familiar sounds.
18(1946); http://dx.doi.org/10.1121/1.1902420View Description Hide Description
Two translators are described which display instantaneously on a moving external phosphor the essential characteristics of sound waves, such as speech. There is thus presented to the eye a parade of speech patterns as though a printed line were passing by. The pattern of intensity vs. frequency vs. time remains in view long enough to permit the eye to observe pattern groups as an aid to speech interpretation. Typical applications are portraying the phonetics of speech and aiding the deaf to understand speech and to build up their own speech. The goal in the speech pattern portrayal is to show the stronger speech components simply but accurately as a function of frequency and time. The speech is equalized to bring out the stronger components over the frequency range. Next automatic volume control is applied to iron out level differences on a time basis. Then the 3500 cycle speech band is analyzed by twelve filters of about 300 cycles each and the resultant component powers applied to the excitation of a phosphor. The smaller translator employs twelve incandescent grain‐of‐wheat lamps for phosphor excitation, generating a pattern 1″ high and 5″ long and the larger translator sets up a pattern 7″ high and 47″ long. The patterns remain in the field of view for about two seconds.
18(1946); http://dx.doi.org/10.1121/1.1902421View Description Hide Description
A system has been developed whereby speech analysis patterns are made continuously visible on the moving luminescent screen of a special cathode‐ray tube. The screen is a cylindrical band that rotates with the tube about a vertical axis. The electron beam always excites the screen in the same vertical plane. Because of the persistence of the screen phosphor and the rotation of the tube, the impressed patterns are spread out along a horizontal time axis so that speech over an interval of a second or more is always visible. The upper portion of the screen portrays a spectrum analysis and the lower portion a pitchanalysis of the speech sounds. The frequency band up to 3500 cycles is divided into 12 contiguous sub‐bands by filters. The average speechenergy in the sub‐bands is scanned and made to control the excitation of the screen by the electron beam which is swept synchronously across the screen in the vertical direction. A pitchdetector produces a d.c. voltage proportional to the instantaneous fundamental frequency of the speech and this controls the width of a band of luminescence that the electron beam produces in the lower part of the screen. The translator has been used in a training program to study the readability of visible speech patterns.
18(1946); http://dx.doi.org/10.1121/1.1902422View Description Hide Description
This paper is concerned with the portrayal of speech in the form of visible patterns that one might learn to read as readily as one learns to understand spoken speech. The portrayal is to be accomplished automatically from the sound waves without a serious time lag between spoken word and visible patterns. In principle, the method attempts to perform a frequency‐intensity‐time analysis of the sound somewhat analogous to that performed by the ear, and to present the results in an orderly manner to the eye. The problem of obtaining legible patterns is one of choosing suitable frequency and time intervals in which to measure the intensity and suitable frequency, intensity, and time scales for the visual portrayal. The choice involves the characteristics of both the eye and the voice. The paper discusses objectives and requirements based on these characteristics, describes the patterns in current use and explores the possibilities and implications of certain modifications that might be made in the present portrayal.
18(1946); http://dx.doi.org/10.1121/1.1902423View Description Hide Description
The phonetic principles basic to the legibility and interpretation of visible speech are discussed in this paper along with the unique characteristics which differentiate the patterns of individual sounds and sound groups. The visible patterns are interpreted in terms of physiological phonetics and the types of modulation used by the speaker in producing the sounds. A threefold comparative classification is developed, including physiological, audible, and visible descriptions of the sounds used in American speech. From the analysis and interpretation of the visible patterns, suggestions are made concerning the inclusion of several sound variations that heretofore have not been of interest generally to phoneticians, and a reconsideration of the classification of some sound units. In analyzing or reading speechpatterns, problems arise from the innumerable and varied patterns formed by sound combinations. A key and system for the recognition and interpretation of these varied patterns which result from the transitional movements between sounds are presented in some detail. Similarities and differences apparent in the audible speech of different individuals are portrayed in the present visible patterns to a limited extent. While the limitations in the translation of the range and variation of frequency and intensity, and of vocal quality, simplify the patterns, they also tend to obscure the phonetic factors which are essential in differentiating individual speakers. The terminology developed for visible speech is introduced as it is required in the course of this discussion of the basic phonetic principles of visible speech.
18(1946); http://dx.doi.org/10.1121/1.1902424View Description Hide Description
The study of jungle acoustics was carried out during the wet season in Panama. Measurements permit the following conclusions to be drawn: Within a jungle the temperature and wind velocity gradients are so small that the sound refraction they produce may be neglected for all practical purposes. Humidity increases the transmission loss at high frequencies and field measurements of the loss agree with laboratory values reported by others. Terrain loss, measured in db, between any two specified distances from the sound source is defined as the transmission loss between these points less that caused by the geometrical divergence of the sound beam. Terrain loss in the jungle was found to increase linearly with distance. The terrain loss coefficients, measured in db per foot, were measured for various types of jungle and were found to be a function of frequency and of the density of the terrain, the density of terrain being measured by the difficulty of penetration and the distance a foreign object may be seen. The level of the ambient noise in the wet season jungle is very low especially for the quiet periods between animal calls. At night the low frequencies decrease as the light breezes cease and the high frequencies increase as the insects begin their nocturnal chorus. A jungle is a difficult place in which to judge the direction of a sound—a probable error of 20° is to be expected. The error is found to be smallest when the sound comes from a direction near the axis passing through the two ears, and in the range studied the error decreases as the sound source moves farther away. Reverberation and scattering are the cause of part of the error of judgment, but an improved technique of listening suggested may increase the observer's accuracy.
18(1946); http://dx.doi.org/10.1121/1.1902425View Description Hide Description
The sound field of a point source near a plane boundary (complex impedance) cannot be obtained by an acoustic ray approach. In fact such an approach, which utilizes the reflection coefficient for plane wavesleads often to completely contradictory results. The procedure which must be followed is exactly similar to that initiated by Sommerfeld to derive the electromagnetic field of a vertical dipole situated near a conducting plane. The results of such an analysis as applied to an acoustic point source are presented. The solution forms the basis for the explanation of hitherto anomalous results. For convenience further discussion will be restricted to cases in which the sound source is at the boundary although the solution is given for all source heights. The solution shows that when the bounding medium has a high real specific acoustic impedance, non‐zero fields are obtained at all points along the boundary. For bounding media adequately described by simple porosity theory, the results are especially interesting in that they show that at reasonably large distances and at low frequencies the acoustic pressure at the boundary is inversely proportional to the square of the distance and the square of the frequency. Some calculations of the sound pressure as a function of height above an absorbing material (Quietone) show the presence of a minimum occurring some distance above the boundary.
18(1946); http://dx.doi.org/10.1121/1.1902426View Description Hide Description
The absorption of sound in artificial fogs or clouds has been determined by measuring the rates of decay of sound in a reverberation room, first when there is no fog in the room (with the relative humidity nearly 100 percent) and then when an artificial fog of known density and particle size is in the room. For a fog containing 2.0×10−6 g/cm−3 of water and a mean droplet diameter of 0.0012 cm the attenuation owing to the fog increased from 0.018 db/m at 512 cycles to 0.032 db/m at 8 kc. For a fog of larger droplet size the attenuation was somewhat less. The attenuation is, to a first approximation, in agreement with the value predicted by a theoretical formula developed by Sewell, and independently by Epstein. It is possible therefore to predict approximately the attenuation of sound in fogs or clouds if the density (or visibility) and the particle size of the fog or cloud are known. Conversely, the composition of fogs and clouds can be investigated by measurements of audibility.
18(1946); http://dx.doi.org/10.1121/1.1902427View Description Hide Description
The atmosphere is referred to as being “uniform” when there are in it no large scale temperature or wind gradients. Usually under these conditions the variation of the ultrasonic intensity with distance from a small source, such as a whistle, can be explained simply in terms of the inverse square law, together with a negative exponential law which accounts for losses due to absorption. This is indicated by the rectilinearity of so‐called Δ curves, obtained by subtracting the loss due to assumed spherical divergence from the total transmission loss and plotting the difference Δ in decibels against distance in rectangular coordinates. The slope of such a curve, when it is rectilinear, equals the absorption coefficient in db/ft. At times in unforested areas Δ curves are not rectilinear, and in dense forests they never are. This may be explained in terms of scattering. A theory of propagation in scattering media has been developed which yields a method of computing the separate effects of scattering and absorption. Values obtained for absorption coefficients are consistent with those reported by other workers.
18(1946); http://dx.doi.org/10.1121/1.1902428View Description Hide Description
In the atmosphere, within ten feet of the earth's surface, there are often intense temperature and wind gradients. Such gradients, as well as their fluctuations in time have been measured by sensitive micrometeorological instruments, and their effects upon ultrasonic propagation investigated, both in and out of dense forests in Panama and in Pennsylvania, over different types of ground coverings, under various weather conditions, both day and night. These studies have revealed definite correlations between micrometeorological and ultrasonic phenomena and their relation to large scale meteorological conditions, and to the terrain.
18(1946); http://dx.doi.org/10.1121/1.1902429View Description Hide Description
Generalized specifications for the optimum design of highly efficient linings for free‐field (anechoic) sound chambers assuming the use of Fiberglas PF insulating board formed into wedges are presented in terms of either (a) lowest frequency at which 99 percent or better absorption is desired or (b) maximum depth of treatment which may be installed in the room. The application of these specifications to the construction of two rectangular rooms is shown and inverse square law measurements performed in two completed chambers, one large and one small, are presented. In the larger chamber, the deviations from inverse square law are within ±0.3 db out to 10 feet and ±1.0 db out to 30 feet from a point source of sound. In the smaller, the deviations are within ±1.0 db out to 10 feet.
18(1946); http://dx.doi.org/10.1121/1.1902430View Description Hide Description
This paper discusses a method for measuring the effectiveness of lightweight wall structures and thin windows in attenuating transmitted sound. Samples approximately 18 inches square are used. The results of measurements on airplane side wall structures (including acoustical treatment) are mentioned briefly.
18(1946); http://dx.doi.org/10.1121/1.1902431View Description Hide Description
The acoustics of rectangular rooms whose walls have been covered by the non‐uniform application of absorbing materials is treated theoretically. Using appropriate Green's functions a general integral equation for the pressure distribution on the walls is derived. These equations show immediately that it is necessary to know only the pressure distribution on the treated surfaces to predict completely the acoustical properties of the room, such as the resonant frequencies, the decay constants, and the spatial pressure distribution. It was found useful to introduce a new concept, that of “effective admittance,” to express the results for the resonant frequency and absorption for then the amount of computation is reduced and the accuracy of the results are increased. The absorption of a patch of material was found as a function of the position of the absorbing material and was checked experimentally for a convenient case, an absorbing strip mounted on the otherwise hard walls of a rectangular room. Particular attention is given to the case where the acoustic material is applied in the form of strips. Approximate formulas are obtained which permit estimates of the diffusion of sound in a non‐uniformly covered room. In agreement with experience, these equations show that diffusion increases with frequency and with the number of nodes on the treated walls. The “interaction effect” of one strip on another is shown to decrease with an increase of the number of nodes.
18(1946); http://dx.doi.org/10.1121/1.1902432View Description Hide Description
There are many examples of acoustically coupled rooms where ordinary reverberation formulas do not apply, for example: theaters, auditoriums, or churches connected to foyers, anterooms, hallways, or organ chambers. The theory of coupled rooms has been given by Davis and by Eyring based on the assumptions usually made in deriving reverberation‐time formulas: diffuse distribution of the flow of sound in the room and continuous absorption at the boundaries. Eyring showed experimentally that these conditions are not always fulfilled, indicating that the problem should then be considered from the wave point of view. Experimental data are presented showing how the frequency and damping constant (a measure of the absorption) for a number of normal modes of vibration of a model room vary with the size and position of a coupling window between two chambers. These results are then correlated with wave theory, treating the acoustics of coupled rooms as a boundary value problem.
18(1946); http://dx.doi.org/10.1121/1.1902433View Description Hide Description
The widespread use of underwater acoustical devices during the recent war made it necessary to obtain precise information concerning ambient noise conditions in the sea. Investigations of this subject soon led to the discovery that fish and other marine life, hitherto generally classified with the voiceless giraffe in noisemaking ability, have long been given credit for a virtue which they by no means always practice. Certain species, most notably the croaker and the snapping‐shrimp, are capable of producing noise which, in air, would compare favorably with that of a moderately busy boiler factory. This paper describes some of the experiments which traced these noises to their sources and presents acoustical data on the character and magnitude of the disturbances.
18(1946); http://dx.doi.org/10.1121/1.1902434View Description Hide Description
The word‐articulation curves of normal listeners were determined with a communication system, known as the “Master Hearing Aid,” of uniform frequency characteristic from 100 to 7000 c.p.s. and a wide range of amplification. The characteristic could be flat or else “tilted” to give a slope of either 6 db or 12 db per octave toward either the high or the low frequencies. The acoustic output was limited sharply at 112 db re 0.0002 dyne/cm2 maximum instantaneous pressure (abrupt peak clipping). Maximum word articulation scores of 95 percent or better were attained by normal listeners with all five tilts of the frequency characteristic. With the patterns that accentuated low tones, the scores fell rapidly toward zero when the input to the limiting stage was increased above the clipping threshold. Intelligibility was best maintained with 6 db per octave accentuation of high tones and next best with 12 db per octave upward tilt. With 6 db upward tilt the scores remained above 80 percent correct, even with peak clipping so severe (80 db) as to reduce the speechpattern to a series of square waves of uniform amplitude. With the same degree of distortion and a flat frequency characteristic, ordinary connected speech was still intelligible, but less so than with the rising characteristics.
Twenty‐five hard‐of‐hearing ears with various degrees and types of hearing loss were similarly tested, with a maximum instantaneous acoustic output of 124 db re 0.0002 dyne/cm2. Performance was judged on the basis of (a) maximum articulation score attained and (b) the useful operating range, i.e., the input range to the limiter over which word articulation was 50 percent or better. By each criterion the best performance is given by either the flat or the 6 db‐per‐octave upward tilt. The 6 db‐per‐octave pattern was best for the group as a whole and also for all but two of the hard‐of‐hearing subjects individually, regardless of the shapes of their audiograms. The addition (acoustically) of a moderate background of “static” noise lowered the articulation scores of most of the hard‐of‐hearing subjects more than those of normals, but the relative performance of the different frequency patterns was unchanged.
A single hearing aid with a frequency characteristic adjustable between flat and the 6 db‐per‐octave upward tilt should provide the best compensation for all patients. If a frequency characteristic approximating tilt and range of the 6 db‐per‐octave upward tilt of the “Master Hearing Aid” is incorporated in a hearing aid, “fitting” would be based primarily on the maximum acoustic gain required and the maximum acoustic output tolerable to the patient, and not on the frequency characteristic of the instrument. The idea of individual selective amplification is fallacious. For our group we have been unable to devise any simple rule for “fitting” the individual audiogram that leads to any better result than the arbitrary selection of the 6 db‐per‐octave upward tilt for everyone.
Our tentative acoustic design objectives for hearing aids, based on these experiments and the general experience of the Psycho‐Acoustic and Electro‐Acoustic Laboratories, are as follows:
Frequency response and range: Moderate high‐tone emphasis (4 to 6 db‐per‐octave); otherwise uniform, without marked resonant peaks or valleys, from 300 to 4000 c.p.s. Sharp cut‐offs below and above this range are desirable.
Tone control: Not essential. If provided it should allow selection between a flat, a rising 3 db, and a rising 6 db‐per‐octave frequency characteristic.
Limiting of output: Preferably by compression, alternatively by simple symmetrical peak clipping.
Maximum output: Semi‐permanent adjustment (or separate models) at 114, 120, 126, or 132 db instantaneous re 0.0002 dyne/cm2.
Maximum acoustic gain: Separate models probably desirable, lowest powered to have at least 40 db acoustic gain available, highest powered a maximum of 80 db.
Gain control: Smoothly graded on approximately logarithmic scale over a 40‐db range.
Intrinsic noise: Must not mask speech delivered to the instrument at an input level of 30 db SPL.
18(1946); http://dx.doi.org/10.1121/1.1902435View Description Hide Description
Because of the extreme variety of conditions under which a hearing aid may operate when in use, it is desirable to determine the character and ranges of magnitude of differences between the performance of an instrument when worn and its performance as measured by the conventional laboratory procedures. The principal differences are due to the following factors: (1) differences between the 2‐cc coupler and human ears, (2) the snugness of fit of the molded earpiece, (3) the size and shape of the acoustical channel of the earpiece, (4) the acoustic baffle effect of the wearer's person, (5) clothing covering the hearing aid, and (6) the nature of the sound field. These effects have been investigated by techniques described in an earlier paper. In the order listed, it is found that, (1) in many cases the 2‐cc coupler does not represent the ear as well as might be desired, (2) a loosely fitting earpiece will reduce the response at low frequencies and will allow some earphones to distort low frequency signals, (3) differences among about 40 commercial earpieces from several manufacturers did not cause important differences in the response, (4) the body‐baffle effect consists of several characteristic peaks and valleys in the response, the effect being less marked in a random field than in free field, (5) the effect of clothing over the instrument is to reduce the response in the frequency range 800–3000 c.p.s. and (6) differences between field and pressure sensitivities of the ear are of sufficient magnitude to be considered when comparing performance of aided and unaided ears.
18(1946); http://dx.doi.org/10.1121/1.1902436View Description Hide Description
The use of the conventional radio range in blind flying depends fundamentally upon an auditory discrimination. The off‐course signal is heard only when the pilot can discriminate between the signals in the overlapping sectors. Any factor adversely affecting the pilot's auditory discrimination widens the beam. Interfering noise, such as set noise or atmospheric static, is shown markedly to reduce the ability to perform the necessary discrimination. Under special circumstances filters will offset some of this loss. The pitch of the signal, on the other hand, has little effect over a moderate range of frequencies. Some patterns of signals are discriminated better than others. Finally, the possibility of fatigue will be discussed.
18(1946); http://dx.doi.org/10.1121/1.1902437View Description Hide Description
The variation of the sound pressure along the auditory canal was determined experimentally on a number of subjects, male and female, placed in a progressive sound field. This was accomplished by insertion of a small flexible probe microphone at various positions along the length of the auditory canal. The subjects were placed in the sound field of a loudspeaker in a room free from acoustic wall reflections. The measurements were carried out over the significant range of audio‐frequencies at various orientations in azimuth of the subjects with respect to the sound source. The free‐field sound pressure was also determined at the subjects' location. The sound pressure at the eardrum is greater than the free‐field pressure. The average ratio of these two quantities is a function of frequency and reaches values of about 20 db in the vicinity of 3000 c.p.s. The human ear is thus an effective acoustic “amplifier.” This amplification is a combination effect of diffraction around the head and pinna and resonance in the auditory canal. The measurements of the sound pressure at some of the other positions along the auditory canal serve to separate these two phenomena to a certain extent and to furnish additional information about the pressure distribution in the auditory canal.
18(1946); http://dx.doi.org/10.1121/1.1902438View Description Hide Description
Early in 1941 the writer designed an ear shield which should enable soldiers to hear conversation and commands about as usual but which should stop sudden pressure or suction pulses. Occasional improvements were made and more than a year ago a patent application was made. The first models of the ear shield had a hard plastic shell containing a thin stiff plate floating between two valve seats, so as to permit alternating air flow of small amplitude around its edges. A sudden pulse forced the plate against the appropriate valve seat, thus protecting the ear drum. About 1942 the rigid shell was modified to fit into a soft adapter similar to the present N.D.R.C. Type V‐51R Ear Warden. Other soft models had a very small spot made quite thin at the middle of a thicker closing septum. The thin spot kept the attenuation low for sounds of low intensity. The decrement ranged from 5 to 10 db for frequencies below 1000 c.p.s. Beyond this frequency it reached a maximum of 20 db in the range up to 12,000 c.p.s. Most other ear protectors impede all sounds as much as possible.