THE MEASUREMENT OF NOISE

by R.P.S Horn PhD

CURRICULUM VITAE: R.P.S. HORN

Graduated from Natal University College in 1947. Practical experi-ence in SAR and ESC. Then on the staff of the University of Natal as a lecturer and senior lecturer for 16 years. During this time obtained PhD in radio wave propagation and later developed an interest in acoustics. He played a major part in the establishment of Electroacoustics Laboratory in the new building for the Department of Electrical Engineering at the University of Natal. Appointed to the position as head of Department of Engineering at Univer-sity of Durban-Westville in 1969. Member of South African Bureau of Standards Working Groups on Acoustics. Participated in ISO meetings overseas. Recently attended 8th International Congress on Acoustics in London where he presented a paper on Speech Intelligibility.

INTRODUCTION

Noise is one of those 'facts of life' that we have come to accept. We are never truly free from sound pressure varia-tions of one form or another during each day of our lives. Our activities at work and at home are influenced to a great extent by the noise around us.

This paper will be concerned almost exclusively with the work situation, with particular reference to factory noise. Fig. 1 shows some of the interrelations which exist between engineers and factory management on the one hand and between many of the other auxiliary activities on the other.

At this stage, it is necessary to give a definition of terms. Measurement is easy: this we could describe as expressing the magnitude of a quantity in clear terms, preferably to which a numerical scale can be provided. Noise seems equally simple. Consider the definition: an unwanted sound. Indeed, a simple statement but soon the trouble starts. We must be clear on what is unwanted and why? The sound may be unwanted because of its loudness, e.g. sound from a pneumatic drill just outside an office window, or for its particular character e. g. a coloratura soprano practising scales, or for its information-carrying content, e. g. the cry of a mall baby at 2 o'clock in the morning.

It is obvious that response to noise is a personal matter and there are many psychological factors involved. Thus, we can add a further question namely, 'unwanted by whom?' Re-turning to the example of the cry of the baby, the reaction to this noise would be different for the parent whose turn it was to get up and attend to the infant, compared with the one who could turn over and attempt to ignore it. Psychological fac-tors are very important when considering the character of noise. One's tolerance to noise is to a large extent dependent on one's pre-conditioning. What subjective reaction does the noise incur? The characteristic sound of breaking glass would engender different reactions if heard by the production foreman in a glassworks, a householder in the middle of the night or an air traveller returning home with a bottle of duty-free whisky.

For a number of reasons, the character of the noise also has a marked effect in assessing the extent of the disturbance likely to be caused by it. Is the noise continuous or regular? Does it have strong tones, such as a whistle caused by escaping steam? Is it impulsive in form? Is it regular, as from a rotary punch or random such as from construction tools using explosive cartridge? Apart from all the other annoying characteristics that the noise may have, its ability to startle and interrupt a chosen task may be of great significance in a factory operation requiring concentration and skill, such as precision machining or a delicate assembly line operation.

It is with this background of mounting confusion that the meaning of the word measurement must also be considered more carefully. First is the measurement accidental or delib-erate? The extent to which the song of a bird is lost in the clamour of the surrounding activities is a measure of the noise present (both natural such as wind in the trees and the barking of dogs, and man-made such as transportation noise).

The tape recording in Fig. 2 becomes a direct deliberate measure of the noise recorded in the dwelling. Neither of these measurements could be classed as highly reproducible. At least one could ask for a subjective opinion after playing the recording to one or more trained observers. The second set of recordings taken at some other time could produce differ-ent results unless such care was taken to eliminate secondary factors such as varying distances between microphones and wanted and unwanted noise sources, varying settings of the tape recorder, both during recording and during playback etc. The problems of accurate comparison become much worse when diff rent microphone, tape recorders, and lis-tening situations are used and a common standard of response is required in different town, cities or countries. For no is measurements to have any objective meaning and value very clear definitions have to be drawn up and accepted by all prospective users. This has been the work for both national and international bodies for many years. Today we have documents produced by bodies such as the I.E.C., I.S.O., and various National Standards Bureaux. The aim is to pro-duce uniformity in regard to many of the basic problems of noise measurement.

Activities such as these are bound to be expensive and time-consuming as they cover all aspects of the problem from basic research in Institutes and Universities to the techniques of handling measurements in factory workshops and produc-tion lines. The driving force behind this is economic necessity: for the costs of separate construction techniques, modifications to suit special buyers must be avoided, as must be the artificial restriction of sales by strings of conflicting and half-understood rules and regulations. Even with the basic documents now available areas of disagreeing are still encountered. The E.E.C. (or common market) authorities are greatly concerned with the problem and are actively engaged with the I.S.0. in reducing the divergent requirements of the various member countries on the acceptability of each other's products.

We are dealing with sound pressure magnitudes ranging from the threshold of pain at 100 Pa to the threshold of hearing at 20 μ Pa. As this pressure range is over a million to one it is virtually impossible to accommodate it with a linear scale. It has become common practice to use a relative scale of sound pressure based on the logarithm to the base p of the ratio of two amounts of sound power.

where p is the measured value at the test microphone the Po is widely accepted reference value of 20 μ Pa.

A pressure range of 1 000 000 is now represented by a scale which goes from 0 to 120 - a far more convenient range even if it is hard to visualise a non-linear seal. We are dealing here with alternating quantities i.e. small pressure variations about the static value of atmospheric pressure. Hence in considering the temporal variations, we need the concept of instantaneous and 'root-mean-square' values, familiar to both mechanical and electrical engineers. Fig. 3 explains some of the necessary concepts. The first graph shows a random variation in noise intensity taking place over a period of T of several minutes typical of a workshop, with a large number of noise sources operating under varying load conditions. The second graph shows the noise next to a grinding machine that operates on a definite cycle. During each period of grinding the noise is high but reasonably constant. In between grinding operations, the level represents the general background noise at the measuring point.

In graph three the measuring device shows an almost steady reading. This could correspond to the r.m.s. value of the single-frequency noise shown in graph four. It would also correspond to the average value of a randomly varying noise if the time constant of the measuring device is suitably chosen.

We often need to consider the signal from a statistical viewpoint and hence take note of the extent to which the pressure variations lie above distinct values e.g. for 70% of the time the value lies above one level or for 90% of the time they lie above another level. The time constant of the measur-ing system is obviously of importance here. Rapid response systems are needed if there is an interest in the fine structure of the noise. Slow response systems are required when the average value is needed. The frequency range of interest has to be specified. While this may cover the normal hearing range of 15 Hz to 20 kHz a particular measuring instrument may have a marked departure from this range for special reasons e. g. because of limitations in microphone response or for a deliberate intention to exclude a certain range of frequencies from the measurement.

Whatever the frequency characteristics of the measuring device, be it ear or instrument, the sound source radiates a definite noise spectrum. The energy distribution with fre-e frequency contains an infinite number of points. We can only compromise by representing the energy output in discrete frequency bands. The type of diagram which results from the measurements I knew as a bar-chart or histogram.

In all the fine detail of measurement, we must not 10 e sight of the fact that we are attempting to replace the human ear by a measuring instrument. Far more than that we are attempting to replace the ear and the incredibly complex computing facilities of the human brain by an indicating instrument which provides us with a numerical value of a particular quantity. We have scales for loudness, for sound pressure, for sound powers and even for annoyance in a number of different circumstances. It is more than likely that the wealth of measurement concepts is confusing to the acoustician let alone to the practising engineer, now looking glumly at the latest set of regulations for the control of noise in factories for the conservation of hearing.

At the risk of boring those familiar with the hearing pro-cess, I intend to state briefly those attributes which define the unique characteristics of the human ear. Not unnaturally, as our understanding of the hearing process has improved, we have been able to design better measuring instruments to substitute for the ear. Even so, we cannot replace the wide dynamic range of the ear by a single device. Even taking into account the logarithmic response, a precision sound level meter must be provided with a variable attenuator to set the instrument such that the amplifiers have the best signal to noise ratio.

The construction of the human ear consists of three main parts, outer, middle and inner ear. The outer ear serves the acoustical purpose of providing a good 'match' between the impedance of the eardrum and the surrounding air. The effectiveness of the matching is good at the speech frequen-cies between 400 and 8 000 Hz but deteriorates outside this band particularly at the lower frequencies.

The middle ear provides mechanical coupling to the inner ear in such a manner that mechanical advantage is obtained and the larger amplitude variations of the eardrum are trans-formed to higher pressure variations in the lymph-filled inner ear. Finally, at the basilar membrane of the cochlea, the pressure variations are perceived by the nerve fibres in the organ of Corti. The magnitude of the incident sound and corresponding vibrations is assessed by the number of nerve impulses transferred to the brain. The basilar membrane also performs a rough frequency analysis of the sound. It would seem that a final analysis is made by the brain. Fre-quency discrimination is excellent in human hearing - particularly with respect to small changes about some reference value. The sense of absolute pitch is less well developed in the average subject. Musicians are much better at this task and it is probably partly an inborn facility and partly the result of their training.

We have considered the sensitivity, frequency range and selectivity of the ear. Another interesting property is known as masking. Thus, a loud noise is able to mask a second sound completely by causing a change in the threshold of hearing at the second frequency.

To further complicate the problem of providing a measur-ing device which can be substituted for the human ear, there is an additional property of discrimination which is often of considerable importance when working with noise. The brain is able to concentrate on certain sounds which form acceptable patterns to the exclusion of other sounds: hence the 'cocktail' effect whereby one can listen to a particular speaker during a simultaneous conversation by several other speakers. The basic components of a sound level meter are as fol-lows: a microphone with suitable reference standards for calibration purposes, an input amplifier, a frequency weight-ing network, an output amplifier and an indicating device. The frequency weighting networks have been provided in an attempt to adjust the meter to the characteristics of the ear whose frequency dependence is also a function of noise intensity. Although four sets of scales are internationally recognised there is now a strong tendency to use the meter on one scale exclusively for simple noise measurements irres-pective of noise intensity. For very many noises this gives a reasonably good measure of the noise. If frequency analysis is required, this may be obtained by the use of external filters of octave or one-third octave-band type. Alternatively, the signal obtained from the output amplifier may be recorded for later frequency analysis using frequency spectrometers. Fig. 4 shows a frequency analysis using constant bandwidth filters of different width. It is obvious that very narrow band filters are required if great frequency detail is required. The objection to narrowband filters is that the time of analysis is correspondingly increased. This difficulty has led to the development of the so-called' real-time' analysers which give a greatly enhanced speed of operation. These instruments are now available with both narrowband displays and the octave or third-octave band displays. The additional information obtained by frequency analysis is vital for the correct applica-tion of sound reduction techniques.

The present-day precision sound level meter is indeed a valuable tool in the hands of a trained observer. Nevertheless, it is easy to make inaccurate measurements if certain precau-tions are not taken. One must be careful to take into account noise sources other than the particular one of interest. This is of great importance when the character of the noises involved is such that the ear readily distinguishes between them. Under these circumstances, the operator may unconsciously underrate the importance of the sound from surrounding noise sources. The sound level meter is unable to make any distinction between wanted and unwanted sound and give a reading representing the total noise. The increase in reading due to background noise may be taken into account fairly readily if a background noise reading is taken before and after the total reading when the wanted noise source is operating. Perhaps the greatest cause of error in sound level measure-ment is caused by reflecting objects. These may be operating personnel, the machines themselves, or the walls of the surrounding enclosures. The range of environments encountered in practice extends from virtually free-field t highly reverberant. Examples of the former occur in noise measurements in open plant e.g. in oil-refineries while engine- rooms aboard ship are good examples of the latter. Perhaps the bulk of noise measurements are carried out under semireverberant conditions. The I.S.O. is currently producing a set of documents defining the measurement techniques required for all noise situations.

While perhaps not being strictly justified by the title of this paper some remarks on hearing conservation are not com-pletely inappropriate since the ultimate aim of much of the noise measurement carried out in the factory situation is in this direction. (Regulation B. 17 of the Factories Machinery and Building Work Act (1941) deals with hearing conserva-tion. Section 2a(iii) of B. 17 clearly defines the responsibility of the employer to apply noise reduction measures to reduce the equivalent noise level to a value below 85dB A or to a level as near as possible thereto. In the latter case where Leq exceeds 85dB A the area so affected shall be declared a noise zone in which employees shall be provided with ear pre tee- tors. It is also the employers, responsibility to endure that the ear protectors are worn by the employees. It is clear that considerable advantages are to be obtained by carrying out the first alternative of section 2a(iii), thereby avoiding the necessity for the maintenance of noise zones. Again, it is beyond the scope of this paper to describe the detailed methods for noise reduction and/or source isolation which is required by industry. However, attention should be given to some fundamental concepts. Noise isolation requires vib-ration damping and partition structures of considerable mass. Materials suitable for sound absorption such as fibreglass, rock wool and polyurethane foams are too light for applications involving sound isolation. This simple fact is often ignored in practical applications, as is the difference between resonant structures and those suffering forced oscil-lations. The techniques of using vibration damping materials in the two situations are quite different.

Returning to the subject of measurement it is appropriate to consider what is meant by the term equivalent noise level, Leq dBA This is defined as follows in the SABS Code of Practice, 083-1970, "The assessment of noise-exposure during work for hearing conservation purposes":- The level of a steady noise (i.e. a noise without impulsive or fluctuating character) that is reputed to cause a stated amount of hearing impairment to individuals who are exposed to it over a 40 hour week.

It may be evaluated in a number of ways dependent on the noise character: -

(a)    From readings of steady noise

Leq = LA + Ci

where Ci is an impulse correction of 10 dB for repetitive noise e. g. riveting, or for single bur t such as the noise from a drop-forge hammer.

A reasonably steady sound is one where the level fluctuates through a total range of less than 8 dBA as measured on the 'slow' response setting of a sound level meter.

(b)    From readings of fluctuating noise

Leg = LA(av) + Ci

where LA(av) is the statistical average noise level.

LA(av) =

where LAi = noise level at the midpoint of the ith class dB(A)

duration of the ith class sound level expressed as a % of the total analysis time (normalised to a 40-hour total period).

(c)    Leq can also be calculated by using a nomogram and calculating the fractional exposure for each noise level.

(d)    An approximate method of estimating the equivalent noise level is given in Fig. 5.

The validity of the various methods depends on the charac-ter of the noise. Where the fluctuations are of an unpredicta-ble nature, the safest method is undoubtedly the accurate analysis of a sample recording of noise using a statistical analyser. An equally acceptable alternative is the use of a sound level measuring device designed to give the Leq value directly for a set time interval. Such instruments are com-mercially available. But as with all measurement techniques, a common-sense approach to the problems involved is in- valuable. Complicated analyses are seldom required and would be justified only when operating very close to the 85 dB A limit, in an unproclaimed area. In the interests of operating personnel generally, the aim should be to provide working conditions which seldom exceed an 80 dBA limit.

While this may be difficult to achieve at the present time future designs of working environments must certainly be carried out with noise (or rather the lack of it) as a vitally important criterion.