NOISE AND METHODS OF ITS MEASUREMENT *

G. W. C. Kaye, London

The struggle against noise is of outstanding interest from the most varied points of view—from the improvement of urban amenities to problems of defense. The article printed below belongs to the director of the National Physical Laboratory in Teddington (near London), G. Kaye. Although the author operates exclusively with examples from the life of England and America and pays insufficient attention to noises connected with various kinds of production, his article is of great interest as a very complete and clearly presented survey of the present state of the physics of noise.
Ed.

THE RANGE OF AUDIBILITY

As in many other branches of science, in the field of acoustical research the modern precision and simplicity of measurements are connected with the development of electrical methods of measurement. The key to the use of these methods was provided by the invention of the thermionic valve: it is to this that technical acoustics owes the rapidity of its development, long delayed.

Before passing to the question of methods for measuring noises, let us recall the basic physical facts concerning hearing, since, inasmuch as the matter concerns noises, auditory perception is the final criterion. At the foundation of the study of hearing lie experiments with pure tones. Here we must note that for our knowledge in the field of hearing and speech we are to a considerable extent indebted to the remarkable

* Supplement to Nature, No. 3224, p. 253, 1931.

studies by Harvey Fletcher and his collaborators, as well as the work of the Bell Telephone Laboratories in New York.

As for frequency, the average ear perceives as sound vibrations whose frequencies cover the range from 20 to 20,000 hertz, with the upper limit of audibility decreasing with age. In various problems of applied acoustics, a considerably narrower frequency range is of interest—from 50 to 5000 hertz for speech and from 35 to 7000 hertz for music.

Fig. 1.

As for the intensity of sound, as was shown by Wegel and others on the basis of experiments with pure tones, for each frequency there exists a certain minimum amplitude below which the average ear ceases to perceive sound. At the same time, the ear is considerably more sensitive in the region of the middle frequencies than to high and low tones. In fact, with a continuous change in frequency, the threshold, or lower limit, of audibility passes through a minimum at about 2000 hertz.

In exactly the same way, for each frequency there exists a maximum amplitude, above which the ear perceives, instead of sound, painful pressure and even intense pain. Here too the ear functions best in the middle part of the range of audibility (about 500 hertz), where the upper

the audibility boundary (the threshold of pain) passes through a maximum as the frequency is gradually varied. Fig. 1 shows the auditory-perception region of the average ear, according to the data of Fletcher and Wegel. The boundaries of this region are formed by the curves of the two thresholds; the dotted line indicates those parts of the curves where accurate measurements are difficult. We see that the audibility region is widest near 1000 hertz and that it narrows considerably toward both ends of the musical scale. This maximum corresponds to a range of intensities differing by approximately a million million times, or to a range of sound pressures differing by a million times, approximately from 0.0005 to 3000 dyn/cm². Let us note that, for the perception of very high and very low sounds, their intensity must be very great and that the audibility region here is quite narrow. A vivid illustration of this is the sound of a 10-meter organ pipe, which we feel rather than hear.

Perception of an Increase in Sound Intensity and the Masking Effect

It has been established that the normal ear, only under the most favorable conditions, can detect a difference of 10% in the intensity of two pure tones of medium loudness sounding alternately without interruption. This figure more than doubles if the two tones alternate with a small interval, even one of 0.5 sec. Under ordinary conditions, however, the smallest change in energy level perceived by the normal ear is, on average, of the order of 26% for sounds of medium frequencies and intensities. This number increases for weak sounds and decreases for very strong sounds. Likewise, for very high and very low tones this number is considerably greater than in the region of middle frequencies.

From the fact that the ear associates steps of loudness not with absolute but rather with percentage increases in sound intensity, it follows that, while the steps of the loudness scale increase in an arithmetic progression, the physic—

... the force of sound grows in a geometric progression; the relation here is the same as the relation between numbers and the scale of numbers on a slide rule. We have here, therefore, yet another illustration of the Weber–Fechner law, according to which the physiological effect is approximately proportional to the logarithm of the energy that produces the corresponding stimulus. To determine the dependence between sound intensity and loudness in different frequency ranges, Kingsbury carried out a series of experiments on the discrimination of the intensity of pure tones. These experiments show that, for frequencies between approximately 700 and 4000 hertz, the relation between loudness and sound intensity does not depend on frequency. For lower tones, as the sound intensity increases, loudness increases proportionally more rapidly. Kingsbury’s results are presented in Fig. 2, which shows a set of curves of equal loudness lying within the region of auditory perception. As can be seen from the figure, these curves are approximately parallel to one another in the region of middle frequencies above 700 hertz.

Fig. 2. I—threshold of pain sensation, II—decibel scale, III—loudness scale, IV—threshold of audibility, V—curves of equal loudness.

An illustration of the curves of equal loudness may be furnished by the following experiment: if the frequency of a pure tone of constant intensity is gradually raised, it will be heard that the loudness passes through a maximum at about 2000 hertz;

this will become evident if one can draw in Fig. 2 a horizontal straight line at a level of approximately \(0.1\) dyne/cm\(^2\). The masking action of one pure tone upon another is usually measured by the elevation of the threshold of audibility of the masked tone. In the general case, a given tone is masked best of all, first, by a tone of approximately the same frequency; secondly, a lower tone masks better than a higher one—in any case for loudnesses corresponding to conversational speech and greater ones.

The situation is not so simple with loud complex sounds, such as, for example, noises, since here the total masking effect may be complicated by other factors, for instance by the masking of separate individual components and by the formation of subjective tones (see Fletcher’s book Speech and Hearing).

Decibel

We now have an idea of the dimensions of the region of auditory perception; we must choose a unit that will serve us as a “meter” or “degree.” It is necessary to try to connect loudness with the strength or energy of sound; we have already seen that, while loudness increases in an arithmetic progression, the energy of sound increases by enormous jumps on a scale having almost astronomical dimensions. Such a relation of quantities greatly complicates measurements, and it is therefore clear that it is in fact far more convenient to base measurement on a scale of ratios of sound energies. A similar need, which arose in the field of telephone measurements, was satisfied by the introduction of the “bel”—a unit whose name was chosen in honor of the inventor of the telephone, Alexander Graham Bell. One “bel” represents a tenfold increase in power or energy; in other words, two intensities related to one another as \(r:1\) differ from each other by \(\lg_{10} r\) bels. The bel was then also adopted as the basic acoustic unit; however, the “decibel” (db) is more often used, since the bel is too large—

a smaller unit for acoustic purposes. Thus we obtain the following table of relations:

It should be noted that the decibel scale has no physiological basis and is constructed entirely on measurements of intensities by physical methods. Nevertheless, such a scale has two advantages: first, in a first approximation it coincides with the scale of subjectively perceived loudnesses; second, as experience shows, the decibel in the definition given above corresponds approximately to the smallest change in loudness perceptible under ordinary conditions (in the region of medium loudnesses). In reality, the perceptible increment of loudness sometimes proves to be somewhat greater, and sometimes somewhat less than a decibel, varying within the limits from 0.2 to 9 db, depending on the frequency of the tone and its position in the region of auditory perception.

We are now in a position to determine the loudness level of a pure tone of a definite pitch in physical terms. The decibel will serve as our “degree,” and our “zero” will be the threshold of audibility for the given frequency. The loudness level of a pure tone may, in accordance with what has been said above, be defined as the intensity, expressed in decibels above the threshold of audibility for the given frequency.

Experience shows that for pure tones of medium pitch the entire region of auditory perception, from the threshold of audibility to the threshold of pain, is covered by approximately 130 db. For high and low tones this number is considerably smaller. The situation, however, becomes more complicated if one poses the broader question of comparing the loudness of pure tones of different pi-

physiological intensity. We shall immediately see that, with such a broader formulation of the question, it is impossible to establish a simple relation between physical intensity and loudness. Two pure tones of different pitch, generally speaking, do not produce the same sensation of loudness even when their physical intensities are equal, or exceed by the same number of times the intensities corresponding to the thresholds of audibility, i.e., when the loudness levels of both tones are the same.

Further, if the levels of two pure tones of different pitch and equal loudness are increased by the same number of times, then the tones, generally speaking, will cease to seem equally loud.

Thus, for tones of different pitch, or for complex sounds, neither intensity nor level of sensation can serve as a measure of loudness. For practical measurements one must choose an arbitrary scale as the standard, the most convenient for this purpose being the scale of loudnesses of a pure tone in the frequency region around 1000 hertz. Then the loudness of any simple or complex sound (for example, noise) is defined as the level of sensation (in decibels above the threshold of audibility) of a standard tone perceived by the ear as a sound equally loud with the given one. Let us note that the zero point of this scale, or the threshold of audibility for a tone at 1000 hertz, corresponds to a pressure approximately equal to \(0.001\ \text{dyn}/\text{cm}^2\).

We repeat that the choice of this standard scale is arbitrary, and that equal intervals on the scale in general do not too closely coincide with an equal number of steps of loudness. For example, an increase of the level from 0 to 10 db on the standard scale approximately corresponds to an increment in loudness barely noticeable under ordinary conditions; an increase from 50 to 60 db corresponds to 10 distinct gradations, and an increase of the level from 100 to 110 db—about 15. In other words, we may measure the range of intensities lying between the two thresholds either by a decibel scale with equal intervals, or by a loudness scale with steps that are at first very large, then more gradual, but that correspond to a uniform increase in loudness (Fig. 2).

Let us add that, in establishing a practical loudness scale, one may allow a certain freedom of choice, since in the region of medium frequencies (above 700 hertz) the relation between loudness and sensation level remains constant. For example, the National Physical Laboratory, in much of its work, takes as its standard tone one of 800 hertz, while in the United States a scale with a standard tone of 1000 hertz is widely used.

Speech and Noise

Of primary interest is the masking effect of a noise background on conversational speech. Conversation becomes difficult when the noise reaches 70–80 db, and at 90 db, even with loud shouting, conversation becomes simply impossible.

Let us recall that the greater part of the energy of the human voice lies in the region of low frequencies. About 60% of the energy lies below 500 hertz, and about 85% below 1000 hertz. It is known, however, that the intelligibility of speech depends chiefly on the transmission of high-frequency consonants (approximately above 1000 hertz), and not on vowels, which are characterized by low-frequency vibrations (approximately 120 hertz for the male voice and 240 hertz for the female), despite the fact that the low-frequency components contain the main part of the energy. Davis and Evans at the National Physical Laboratory observed that speaking in an enclosed room becomes easier if the high-frequency noises are eliminated. Fortunately, eliminating high frequencies is much easier than eliminating low ones; this is especially important in cases where the walls of the room cannot be made sufficiently massive, for example, the walls of an aeroplane cabin. Furthermore, high frequencies are more strongly absorbed by sound-absorbing materials mounted on the inner walls of the room.

Investigating the influence of noise on the audibility of conversational speech, Knudsen in 1925 found that, if the interfering sound is a pure tone of approximately the same loud-

…that is, the interfering effect almost does not depend on frequency; at high intensities, however, the interfering effect of low tones is greater than that of high ones. He also established that noise interferes more strongly than a pure tone of any pitch. Fig. 3 gives a summary (made by Fleming) of Knudsen’s results on the question of the influence of extraneous noises on the intelligibility of speech, measured by the percentage of articulation of conversational speech of normal loudness (50 dB). As can be seen from the figure, even weak noise substantially disrupts the perception of speech; noise with a loudness of 30–40 dB reduces the intelligibility of speech to inadmissible values.

Fig. 3. 1 — good audibility, 2 — satisfactory audibility (with attentive listening), 3 — unsatisfactory audibility.

Noise and the Irritation Caused by It

Recently, the question of irritation caused by noises has been subjected to experimental investigation. Precise measurements in this area are difficult to expect, but it is clear that among the factors responsible for this phenomenon, frequency and loudness play a considerable role. As for pitch, it is probable that, for most people, shrill sounds have a more irritating effect than low ones. For example, the high tone of an automobile horn with its staccato, so common among Parisian taxicabs, is more unpleasant than the low sound of the horn usually used in England. This impression is confirmed by the work of Laird and Coye, who found that the irritating effect is a function of loudness and pitch, with high tones irritating incomparably more strongly than low and medium ones, and high sounds of great loudness being the most unpleasant.

For tones with a frequency below 500 hertz the effect of irritation depends on loudness. In this connection it is interesting to note that the range of frequencies characterizing normal human speech apparently causes the least irritation. The irritating effect of certain tenor voices and soprano, in the opinion of Laird and Coye, is in agreement with their investigations.

Apparently, the unpleasant sensation caused by complex noises, for example by an automobile horn, depends to a considerable extent not only directly on loudness, but also on the presence of strong high and inharmonic components.

Absolute Measurement of Acoustic Energy

The standard method of measuring the energy of a wave in absolute units, generally speaking, consists in having the wave absorbed by some suitable material and then measuring the amount of heat released. But even if it were possible to make sound waves be completely absorbed, the absolute amount of energy of speech and of a large part of the other everyday sounds is so small that it lies at the limit of what can be measured with the most sensitive instruments. For example, the average sound power of conversational speech is approximately \(10\,\mu\mathrm{W}\). For a loud shout the power rises to \(1000\,\mu\mathrm{W}\); for quiet speech it falls to \(0.1\,\mu\mathrm{W}\), and for a soft whisper to \(0.001\,\mu\mathrm{W}\). To illustrate these figures, let us give the following example: a crowd of one hundred thousand people in the stadium at Wembley, speaking continuously and loudly, would produce such sound power that, if converted into heat, it could make a small electric bulb glow for the duration of the entire contest. The acoustic energy expended before the end of the contest, if converted into heat, would be enough to boil a cup of tea. An especially excited crowd, shouting all the time at the top of its lungs, could boil 10 cups of tea—the amount [[unclear: continuation cut off at bottom of page]].

enough, perhaps, to fill a prize cup.

One can, however, find one or two acoustic phenomena connected with the radiation of a considerable amount of energy. Measurements above steamship sirens in New York showed a power of about \(6\ \mu\mathrm{W}/\mathrm{cm}^{2}\) at a distance of \(34.5\ \mathrm{m}\), so that the total amount of acoustic energy emitted by the siren comes out to be approximately \(1/3\) h.p.

It is clear, however, that for sounds of ordinary intensity thermal methods of measuring acoustic energy promise absolutely nothing, and we must turn to some other properties of sound waves: one may use the oscillations of air pressure in the passing sound wave, then the small oscillations of temperature and refractive index connected with them, the velocity of the oscillating air particles, or the pressure exerted by acoustic radiation on a sound-reflecting surface. Conversational speech corresponds to an effective pressure of approximately \(1\ \mathrm{dyne}/\mathrm{cm}^{2}\),* in other words to a pressure oscillation equal to one millionth of an atmosphere. The changes in refractive index for such pressure oscillations are approximately \(10^{-9}\), the temperature change is \(0.001^\circ\mathrm{C}\), the particle velocity is approximately \(1/40\ \mathrm{cm}/\mathrm{sec}\), and the pressure of sound radiation amounts to only a few \(10^{-1}\) fractions of an atmosphere. The corresponding power is approximately \(0.001\ \mu\mathrm{W}/\mathrm{cm}^{2}\); let us recall that power (or energy) varies in proportion to the square of the pressure and amplitude.

For carrying out absolute measurements we must choose such a measuring instrument whose readings would not depend on the frequency and shape of the sound wave. One of the most suitable instruments is Rayleigh’s disk, by means of which the velocity of air particles is measured. However, this is a very fragile instrument, which should be regarded above all as a means for calibra-

* Street noise in New York gives, on the average, a pressure of about \(5\ \mathrm{dyne}/\mathrm{cm}^{2}\) and may even reach \(20\ \mathrm{dyne}/\mathrm{cm}^{3}\).

...tion; it is used only in laboratory measurements. For practical purposes, more suitable and robust instruments are needed, and therefore electrical microphones are usually resorted to, preferably of the non-resonant type. Microphones transform acoustic oscillations into electrical ones, which are then amplified by a suitable tube amplifier and, under proper conditions, can be easily and accurately measured. A whole series of other devices (chiefly technical ones, of the resonant type) have also been proposed for measuring or recording sound; however, for a number of reasons they have been almost completely displaced by electrical methods, mainly by condenser (or electrostatic) microphones.

A condenser microphone, in its essential features, consists of a tightly stretched metallic membrane placed parallel to a metallic plate at a very small distance from it, so that the gap between them is approximately \(0.025\) mm. The thin layer of air in this gap increases the elasticity of the system, so that the natural frequency of the membrane becomes very high—usually above the normally transmitted range of acoustic frequencies. Across the condenser, consisting of the fixed electrode and the membrane, a potential difference of about \(200\ \mathrm{V}\) is applied through a large resistance. Sound waves, acting on the membrane, cause it to vibrate, thereby producing changes in the capacitance of the condenser; at the terminals of the series-connected resistance an alternating electromotive force arises. The latter can easily be amplified by a tube amplifier and measured by means of a rectifier (for example, a thermocouple) and a microammeter. At the same time, the waveform can be determined with the aid of a cathode oscillograph. The condenser microphone is, to be sure, not very sensitive, but its advantage is that it operates almost equally well over the entire range of acoustic frequencies, and therefore represents a convenient standard instrument which can be calibrated in absolute measure.

Measurement of Noises

It seems Lord Kelvin said that if a method of measurement has been found, then one can begin to study the phenomenon. It is clear, however, that the question of measuring noise is rather complex. Besides physical questions, physiological and psychological questions are also involved here. From the physical point of view, it is evidently desirable to come to an agreement concerning the choice of a system of physical quantities. It is desirable that this system be absolute, so that there would be the possibility: a) of translating qualitative subjective judgments and sensations into the language of facts and figures; b) of clarifying the causes and characteristic features of noises; c) of comparing the results of observations by different investigators; and d) of establishing arbitrary standards suitable from the point of view of social, technical, and legal requirements.

The practical measurement of noise reduces to one or several of the following operations:

1. Physical measurement of the total power or energy of noise, so that in the final analysis the result is expressed in absolute units (for example, in dynes or microwatts per square centimeter).

2. Physical analysis of noise and obtaining the acoustic spectrum of the noise components. Such an analysis can be obtained by recording individual noises, in particular the noise of machines.

3. Physical determination of the form of the noise wave, although its quantitative interpretation is sometimes difficult, especially in the case of aperiodic noise.

4. Measurement of the loudness of noise by ear in some suitable units—in other words, a subjective-physiological evaluation of “noisiness” with the aid of the ear.

Physical Measurement of Noise

We have already spoken of measuring sound energy by means of a condenser microphone and an amplifier. The amplified current is fed either to a rectifier and microammeter

(calibrated as desired in decibels) for the purpose of measurement, or simultaneously also to a cathode oscilloscope, if it is desirable to determine the waveform. In view of the fact that the readings of the microammeter serve as a measure of the physical intensity, and not of loudness (which depends on the differing sensitivity of hearing to tones of different frequencies), correcting circuits are sometimes included in the measuring circuit, with the aim of bringing the measurement results into agreement with the data of normal auditory perception. The frequency characteristic of the apparatus should be chosen with reference to a loudness level of 30–40 db above the threshold of audibility on the scale with the standard tone of 1000 hertz.

If necessary, noise analysis may be carried out by including in the circuit electrical filters that pass a narrow band of frequencies (“band” filters). In this way it is possible to determine the energy falling on individual components or frequency bands. True, by this means it is not always possible to achieve sharp selectivity; however, on the other hand, it must be acknowledged that our knowledge in this area is insufficiently extensive and does not allow an exact correspondence to be established between the overall loudness of a noise and the energy or loudness of its individual components.

Noise Analysis by the Method of the “Probing” Tone

The problem of analyzing noises of a more or less periodic nature has been simplified to a considerable extent by the introduction of the method of the “probing” tone, a method which makes possible easy analysis and permits far higher selectivity over a broad frequency range than is possible when working with “band” filters.

The method of the “probing” tone consists in the following: the noise being analyzed is picked up by a microphone; the microphone current is amplified and “mixed” in a tube rectifier or modulator with a probing sinusoidal

by a current of constant amplitude obtained from a heterodyne generator; the frequency of the probing tone is varied smoothly. Thus the modulated current contains not only the probing tone, but also sum and difference tones combining the probing tone with the individual noise components. For example, if the frequency of the probing tone is equal to \(S\), and the frequency of one of the components is \(C\), then the frequencies of the sum and difference tones will be \((S + C)\) and \((S - C)\).

One of the ways of detecting the existence of these tones consists in exciting, by the modulated current, some mechanical resonator with sharp selectivity, for example a steel rod capable of performing longitudinal vibrations. With a smooth change of the probing frequency the rod resonates each time that \((S + C)\) or \((S - C)\) becomes equal to the natural frequency of the rod. Since \(S\) is known, \(C\) can be determined; on the other hand, the amplitude of the resonator oscillations, observed by some suitable method, makes it possible to judge the energy falling on the given component. In practical conditions \(S\) may vary, say, from 11000 to 16000 hertz; the natural frequency of the rod is taken to be of the order of 16000 hertz if the sum tone is being sought, and 11000 hertz if the difference tone is being sought. Grützmacher, instead of a mechanical resonator, uses a choke filter passing only frequencies below 30 hertz. Then in most cases both the probing and the sum tones will be filtered out, and only difference tones with frequencies less than 30 hertz will pass through the filter and act on the amplifier and detector. Thus, with continuous variation of the probing frequency, the detector will respond only when this frequency differs by less than 30 hertz from one of the noise components. The magnitude of the reading after the detector is a measure of the intensity of the corresponding component. In the installations of the National Physical Laboratory the frequency of the probing tone is varied, approximately, from 30 to 10000 hertz by rotating an air capacitor through \(180^\circ\).

G. V. Kay

Measurement of Noise by Ear

As we have already seen, the loudness of any pure tone, or in general of any complex sound, can be assessed by ear by comparing it with the loudness of a standard pure tone of medium pitch (above 700 hertz), the intensity of which can be varied at will within limits determined by calibration by physical methods. On the other hand, for this purpose one can determine the masked loudness of the standard tone, masked or drowned out by the sound being measured. The standard tone may be obtained from an electric membrane buzzer (similar, for example, to the buzzer in the Siemens-Barkhausen audiometer, giving a frequency of about 800 hertz), from a tube generator (for example, the generator in the Western Electric audiometer, giving 8 different frequencies), or from a gramophone record (with pure or “howling” tones), using the appropriate attenuators.

When working with audiometers of various types, the most accurate results are obtained when, in a series of successive observations, we approach the critical value alternately from both sides. In general, experience shows that most people, after a little practice, can obtain excellently agreeing results even with the simplest form of audiometer, at least when measuring more or less prolonged noises. Generally speaking, it is easier to work by the method of determining masked loudness than by the method of comparing loudnesses. In the United States the method of determining masked loudness is preferred, on the grounds that the results obtained in this way give an idea of the degree of “deafening” or of the increase in the threshold of audibility for the various frequencies of the band used.

Experiments in Measuring Noises with the Aid of a Buzzer

In 1929 I carried out a series of rough measurements of noises with the aid of a buzzer consisting of a bent steel strip (clicker), such as are sometimes used in lectures. The tone

of such a buzzer is very high and is heard surprisingly far away; in quiet surroundings it can be heard even at a distance of 300 m. The masking effect of the buzzer was observed under the most varied conditions. In the cabin of an airplane or near an operating pneumatic riveting machine, the tone of the buzzer is heard at a distance of 60–90 cm, while next to an airplane engine it is heard only at a distance of a dozen centimeters. Applying, however, not quite justifiably, the inverse-square law, one may say: experiments with the buzzer confirmed that the noise in subway cars (75–80 db) is considerably louder than the noise in an express train traveling at a speed of 90 km/hour, even in a corridor with several open windows (70 db). Indeed, it is well known that in a subway car it is difficult to talk and to listen, whereas on a train, in a car with closed windows, it is easy, especially in first-class cars with soft upholstery.

The noise in the cabin of an airplane during a flight across the Channel proved to be at least a thousand times (by 30 db) stronger than the noise of an express train, although the cabin walls with wooden lining reduced the noise of the engine a hundredfold (by 20 db). The preference shown by knowledgeable passengers for seats at the rear of the cabin, as compared with seats near the side propellers, was justified, since a difference of about 10 db was found between them. It was also found that the customary practice among airplane passengers of plugging the ears with cotton reduces the noise by approximately 10 db.

Measurement of Noises with the Aid of a Tuning Fork

A very convenient and portable instrument for measuring noise was developed by Davis at the National Physical Laboratory. Having set a tuning fork in vibration by some suitable means (one may simply strike it against the heel of a shoe, without taking the usual precautions), the tuning fork must be brought as close as possible to the ear, but without touching it. The instant at which the tuning fork is set in vibration is noted, and the time is observed during which the loudness of the tuning fork falls to the level of the surrounding noise. If desired,

TABLE I

Loudness of various noises

Noise and Methods of Its Measurement

Continuation of Table I

Continuation of Table I.

LITERATURE. Davis, Journal Roy, Amer. Soc., 1931; Free, Journ. Acous. Soc. Amer., 1930; Parkinson, Journ. Acous. Soc. Amer., 1930.

It is also possible to measure the time after which the sound of a tuning fork begins to be masked by noise. The rate of attenuation of the tuning-fork sound is calibrated in decibels with the aid of a buzzer or an audiometer of some other type. Since the decrease in the loudness of the tuning-fork sound proceeds practically according to a logarithmic law, the calibra-

of plotting “decibels—time” is approximately rectilinear. Making the count is facilitated if, when comparing the strengths of sounds, the tuning fork is alternately brought closer to the ear and moved away from it, so that the sound becomes now louder, now weaker than the noise being investigated. The tuning fork used by Davis had a frequency of 640 cycles. Its loudness immediately after being struck was approximately 90 db, and the rate at which the loudness decreased was about 1.5 db/sec. Noises with a strength of 110 db were measured by means of observations of the masking effect.

Davis applied this method to determining the loudness of a great variety of noises within the limits of audibility, and obtained results which, as is seen from Table I, agree well with the results of American investigations carried out with the aid of an audiometer.

Relation between loudness and the masking action of noise

The masking action of noise depends on its composition, and theoretically, in the general case, the loudness of a noise (determined by physical methods or by ear) cannot be connected with the masking action determined by ear with the aid of one audiometer or another. In reality, however, it turns out that for the majority of complex and sufficiently prolonged noises of everyday life the loudness of the noise differs from the value obtained by the method of determining masked loudness by an almost constant amount, which tends to increase somewhat for louder noises and for noises of an intermittent character.

The New York Noise Abatement Commission established that for ordinary street noises and noises inside buildings the value of the loudness for the middle range of frequencies exceeds the value obtained by the method of determining masked loudness on the average by approximately 15 db. For very strong sounds (90 db), for example for the loud noise of a large crowd, the difference is obtained, apparently according to the Commission’s report in City Noise, as 20 db. Davis in his

in experiments with a tuning fork obtained the same value of the difference for a loudness of 110 db. He also established an almost rectilinear dependence between the values obtained by the methods of masked loudness and comparison of loudnesses for noises of medium strength.

In measurements made by the Aeronautical Research Noise Sub-Committee on noises of very great intensity produced by airplane propellers, it was found that these two values differ by 20–30 db, although it must be admitted that measurements made under such conditions are inevitably somewhat crude.

Examples of Noise Measurement

To illustrate what has been said, it will be useful to give a few simple examples of the measurement of everyday noises.

Conversational speech has a loudness of from 40 to 60 db. If, however, the lips of the speaker are brought close to a person’s ear at a distance of 1 cm, the loudness of the speech for the listener increases to approximately 100 db. An ordinary automobile horn sounds at a distance of 6 m with a loudness of about 80 db. It is curious that the simultaneous cry of two twins is only 3 db louder than the cry of one of them; the noise increases by 3 db when the listener approaches the source of the noise by 30% (in the open air); approaching by another 20% (i.e., reducing the distance by half) gives an increase in loudness of 6 db.

Figure 4 shows a kind of “noise thermometer”—a scale of noises with loudness reaching up to 100 db; exceeding this limit under ordinary conditions is unlikely. The noises are divided into several broad groups (the left part of the figure); the part of the scale above 50 db is set off separately, since it contains noises with which it is, insofar as possible, necessary to contend. We may regard this figure as a kind of “normal” level of our noise thermometer.

Table I gives the loudness of various noises (to an accuracy of up to 5 db) according to English data (National Physical Laboratory) and according to American determinations, carried out for the most part by the Noise Abatement Commission (the figures corresponding to experiments with a standard tone of medium pitch were selected). Let us now turn to a consideration of those of these data which are of general interest.

Decibels above the threshold of audibility:

1 — in an airplane (very noisy),
2 — in a train (noisy),
3 — in the street (noise of medium intensity),
4 — at home (quiet),
5 — outside the city (very quiet),
6 — airplane cabin,
7 — pneumatic drill,
8 — underground railway (London),
9 — busy street traffic (London),
10 — train carriage (window open),
11 — normal speech (1 m),
12 — closed automobile (45 km/hour),
13 — suburban street,
14 — country garden,
15 — quiet whisper (1.5 m),
16 — threshold of audibility.

Fig. 4.

In Table I a whole series of traffic noises in London and New York is presented; if one may speak of an exact comparison of them, then apparently it must be acknowledged that the streets of New York are on average 10 decibels noisier than the corresponding London streets. Although in London there are places where at a distance of ten paces a dog’s barking cannot be heard, in New York, on the other hand, according to the commission’s statements, there is a square where even a tiger could roar continuously without attracting the attention of passers-by. It is asserted that in New York there are intersections at which, so far as it has been possible to investigate, the noise is stronger than anywhere else in the world; such, for example,

corner of 6th Avenue and 34th Street, where three main streets meet, three tram lines pass, as well as a double-track elevated railway and the subway. The chief violator of silence is the elevated railway; I consider it unlikely that London will permit anything similar to an elevated railway passing over the streets to be built overhead.

The New York commission found that the “tides” and “ebbs” of noise during the day run parallel to changes in the density of street traffic, at least up to the figure of 50 carriages per minute. It is natural to expect some reduction of street noise on the upper floors of buildings, but this effect is much weaker if tall buildings stand on both sides of the street. In such a case, even in “skyscrapers,” a reduction in noisiness is observed only on floors lying above the building opposite. In some New York and Chicago hotels it is recommended that rooms be chosen no lower than the 20th floor.

Fig. 5. I — Lindbergh’s passage, II — brass band, III — silence.

Fig. 5, given by Galt, shows the values of masked loudness obtained with the aid of an audiometer with three frequency bands (vocal tones); the curves relate to the noise of the crowd during Lindbergh’s arrival in New York after his flight across the Atlantic Ocean. The observers were placed on the 5th floor, at approximately a distance of 33 m from the street. The masking action of the noise of the crowd greeting Lindbergh upon his appearance is sharply expressed; indeed, this noise is capable of masking the sounds of a brass band located at a short distance. As Galt notes, here we have quanti-

...a natural method by which artists and other favorites of the public can periodically check their popularity.

As for the individual components of street noise, judging from the limited data available concerning English and American streetcars, the latter differ little from one another in noisiness. The same observation, apparently, is also true for automobiles. It is interesting to note that a modern automobile at moderate speed produces less noise than a horse-drawn carriage on a paved road.

A considerable part of street noise is due to automobile horns. Therefore, by order of the Ministry of Transport, the National Physical Laboratory carried out a small investigation of the noisiness and shrillness of horns. The observations were made in an enclosed room with strongly damped walls. The noise was measured both by physical methods and by ear; with the aid of a cathode oscillograph, oscillograms of the sounds under investigation were recorded. So far as can be judged from the small number of observations, the shrillness of a sound depends to a considerable degree directly on its loudness, although other factors also have some importance: the presence of intense high components, a clearly expressed dissonance, and features of the initial period. It is not yet possible to establish the dependence between the shrillness of a sound and the form of the oscillation.

The New York Noise Study Commission arrived at approximately the same conclusions. It considers that signals producing a sound louder than 90 db at a distance of 7 m are unnecessary, and that objections should be raised against their use. The Commission also finds that complaints about the shrillness of a sound will cease if the fundamental tones of the signal lie between 200 and 300 hertz, with all components being harmonics of the fundamental tone and with a uniform distribution of energy among them, and if the sound contains no strongly pronounced high frequencies.

As for trains, in terms of the loudness of noise in cars of a similar type, express and suburban trains in England and America, apparently, differ little from one another.

The American method of dividing a Pullman car by curtains, in addition to other shortcomings, apparently produces a higher noise level than in the English cars with more perfectly separated compartments. True, much here depends on other factors, for example on the proper fitting of doors and windows that prevent the penetration of noise. As for subways, the New York subway, as everyone who has ridden on it can attest, is especially noisy; the London subways are quieter by at least 10 db, although, of course, the question cannot be settled without taking speed into account.

Among the loudest sounds one has to encounter are the noises of riveting, of the pneumatic drill, the sound of a steamship siren, and the noise of printing presses. Still louder are the roar of a lion and the noise of Niagara Falls—noises of approximately equal loudness (85 db). The loudest and most unpleasant is the noise of an airplane engine (110 db). The noise in the cabins of flying airplanes reaches 80–110 db, depending on the type of machine. The dominant factor is apparently the noise of the propellers, although the motor exhaust and the overall noise of the machine are almost not inferior to them in strength; the struggle against noise must be conducted in all three directions. It should be noted that at the present time the Aviation Commission for the Investigation of Noise expects to reduce considerably the noise in airplane cabins (possibly to the level of the noise of a railway train). Propellers with lower speed will be used, as well as more perfect exhaust silencers; it is proposed to attenuate the motor noise by enclosing it in a casing; the cabins are to be built with double, well-insulating walls.

Protection from Noise

The best method of protection from noise is to remove the causes contributing to its occurrence. This is far more effective than attempts at subsequently combating noise. For example, machines should be protected with closed casings, better balancing and assembly should be ensured;

mount them on insulating materials. As for the noise of street traffic, in most cases the most unpleasant noise is that produced by improperly loaded and worn-out vehicles. On noisy streets with heavy traffic, protection of buildings from noise can often be achieved by creating “sound shadows” and through the architects’ ingenuity in arranging “buffer” rooms, examples of which we see in the new buildings of the British Broadcasting Corporation (Butisch Broadcasting Corp.).

The protective action of buildings is excellently illustrated by the quietness of enclosed quadrangular buildings, such as the Inns of Courts,* located in immediate proximity to noisy streets. Bedrooms whose windows face the inner courtyard of a hotel are usually considerably quieter than rooms situated on the outer side of the building.

There are two principal methods of insulating interior rooms from noise: rigid, non-porous walls or partitions, and multilayer partitions, the separate layers of which should, as far as possible, be independent of one another and separated by a layer of air or of some loose filling.

For solid massive walls the principal factor is weight; sound absorption (in decibels) is proportional to the logarithm of the mass of the wall per square meter of its surface. In Fig. 6 are given the results of measurements by Davis and Littler at the National Physical Laboratory. Their results, together with the results of Knudsen, the Bureau of Standards, and others concerning the sound absorption of ordinary partitions in the region of medium frequencies (512 hertz), are collected in Table II. Generally speaking, low frequencies are absorbed less than high ones. The transmissivity of partitions is connected chiefly with their vibrations (the partition acts as a membrane); therefore, in the region of low frequencies resonance may appear, which under normal conditions has only secondary significance.

* Inns of Courts—four English societies of barristers (Inner Temple, Middle Temple, Lincolns Inn, Grays Inn). Ed. note.

TABLE II

Sound insulation of solid single-layer partitions

In practice, for an approximate calculation it may be assumed that doubling the mass increases the insulation by approximately 5 dB, although resonance effects may disturb this relation.

Fig. 6.

1 — brick wall (12 cm),
2 — double fiber partition,
3 — double fiber partition with an insulating layer,
4 — double fiber partition on frames,
5 — redwood,
6 — fiber boards,
7 — sailcloth,
8 — [[unclear: item not visible]],
9 — paper.

For porous flexible materials, Knudsen finds that sound insulation is proportional rather to mass, while—

* Brick wall 12 cm thick.

...per 1 m² of wall than to the logarithm of this quantity. It often proves advantageous to combine porous materials with dense, rigid partitions. Let us mention that for sound-motion-picture theaters, sound insulation from 50 to 70 db is chosen.

As for multilayer partitions, they should not be pierced through with nails; that is, the layers must be completely isolated from one another, so that, with the same overall thickness, composite partitions absorb (as they should) better than ordinary ones. Whether it is expedient to make an interlayer between the partitions out of porous material, and whether such an interlayer is needed at all, must be decided by experiment. On the one hand, the gasket may absorb and damp vibrations; on the other hand, it may itself serve as a coupling.

Attention, one hopes, will be directed to the question of protection against sounds transmitted through the walls of a building by the new acoustical laboratories being opened at the National Physical Laboratory. Up to now, little systematic research has been carried out on this question. Here, apparently, heterogeneity and discontinuities are important, while loose fillers may protect against the rattling of resonating partitions and walls, which often accompanies such vibrations and sometimes leads to considerable sound radiation.

In many houses, windows play the principal role with respect to the admission of outside noise. The construction of sound-impermeable buildings would be greatly simplified if windows could be eliminated. This, to be sure, is not worth discussing; but in any case windows must be made of thick glass. A window, even slightly ajar, almost nullifies the advantages obtained by protection from noise with the aid of absorbing devices. Within known limits, the amount of sound energy admitted through a crack or a partly opened window is proportional to the area of the opening. For a door or window giving, say, sound insulation of 30 db, a crack with an area equal to 0.001 of the area of the door may admit as much sound energy as passes through the entire window or door.

Fig. 7 (given by Norris) shows the increase in the loudness of noise in a room as a window facing the street is gradually opened. Here it is evident that a small opening can produce a large effect. Finally, a considerable improvement can be achieved by not allowing the noise that has penetrated into the room to grow to a high intensity owing to reverberation, which can be reduced by lining the walls and ceiling with sound-absorbing materials. Many banks and City institutions are now successfully using this means. I fully understand that many houses in New York line the vestibules of entrance doors with absorbing materials, which, they say, gives considerable advantages.

Fig. 7.

Fig. 8. I—undamped walls (reverberation time 5 sec.). II—damped walls (reverberation time 1 sec.).

Fig. 8 (given by Galt) depicts the influence of lining the walls with an absorbing material on the results of measurements of the masking effect in a room into which city noise reaches. As can be seen from the figure, the masked loudness decreases on average by approximately 7 db when the reverberation time is reduced from 5 to 1 sec.