THINK QUIET -G. M.

Diehl
PART l : SOUND WAVES AND DEFINITIONS Page 2
Properties of sound waves-3; Traveling, standing, and complex waves-3; Sound velocity in air and various media-3; Plane and spherical
waves-3; Longitudinal and transverse waves-3; Definitions-4; Sound pressure, sound power, and sound intensity-4; Decibels and levels,
reference quantities-4; Bars, millibars, microbars, dynes/cm
2
, newton/m
2
-5; Phons, sones, PNdB-5.

PART II: INSTRUMENTATION AND MEASUREMENT OF SOUND Page 6
Instrumentation for sound measurement-6; Sound level meter, weighting networks dBa-6; Microphones (condenser, crystal dynamic)-6;
Octave-band analyzer-6; Preferred frequencies, old octave bands-6; Narrow-band analyzers (half-octave, third-octave, tenth-octave,
constant percentage band width, constant-band width)-6; Analyzer corrections-6; Spectrum level-7; White noise, pink noise-7; Calibrators-
7; Windscreens-7; Recorders-7; Sound-measurement technique-7; Microphone placement-7; Data and sketches-7; Calculations-8;
Subtracting background noise-8.

PART III: SOUND TEST EVALUATION Page 9
Criteria-9; Hearing damage criteria-9; Calculation of dBA from octave-band data-10; Partial exposure calculations-l0; CA GI-
PNEUROP Sound Test Code-10; Comparison of IEC 179 and ANSI S1.4-10; Octave bands of interest-11; Decibel drop-off criterion-11;
Effect of waving microphone-11; Speech Interference Levels-11.

PART IV: NOISE CRITERIA Page 12
Annoyance criteria-12; Residential criteria-12; Level rank-12; Prediction of neighborhood response-12; Noise criteria-12; ISO noise rating
numbers-12; NEMA standards-12; Precautions when using dBA levels-12.

PART V: SOUND WAVES UNDER VARIOUS CONDITIONS Page 14
Free-field (Inverse square law-14; Nearfield, far-field-14; Correction for atmospheric conditions-14; Sound power level calculation-14);
Reverberant field-14; Semireverberant field (Sound absorption coefficient-15; Room constant-15; Reverberation time-15; Decibel
drop-off with distance-15; Directivity index due to machine 16; Directivity due to location-16; Sound power level calculation-16;
Sound power level by calibrated sound source-16).

PART VI: CONTROLLING THE GENERATION OF NOISE Page 17
Machinery enclosures and mufflers-17; Sound transmission coefficient-17; Transmission loss-17; Mass law-17; Double wall-17; Effect
of absorption inside enclosure-18; Effect of damping-18; Effect of vibration isolation-18; Importance of small openings-18;
Reprinted fromCompressed Air. Copyright 1971
1








































Acoustic shields and barriers, partial enclosures-18; Wave coincidence-19; Pipe noise radiation, pipe lagging-19; Large-diameter
pipes-19; Small-diameter pipes-19; Pipe noise transmission-19; Demonstration of effects of absorption, transmission loss,
damping, vibration isolation-20-21.
PART VII: MUFFLING GENERATED NOISE Page22
Dissipative mufflers-23, Insertion loss (dynamic insertion loss, transmission loss, SPL difference, noise reduction, end difference,
attenuation)-23; Reactive mufflers 23; Helmholtz resonators-23; Quincke tube-23.
PART VIII: COMPONENTS OF A MACHINE AND DESIGN PARAMETERS THAT AFFECT NOISE
GENERATION Page 24
Industrial machinery noise, general-24; Effect of horsepower, rotation speed, dynamic balance, structural resonances, flow
velocity in pipes-24; Cavitation-24; Bends and restrictions-24; Valves and metering orifices-24; Critical flow valves -25; Interior
finish-25; Jet noise-25; Gear noise-25; Bearings (ball, roller, sleeve)-25; Coupling noise-25; Electric motor noise-25;
Center of gravity mounting-25; Noise due to improper installation-25; Areas and dimensions that radiate certain frequencies-25.
PART IX: INDUSTRIAL MACHINERY NOISE-CENTRIFUGAL MACHINES Page 26
Aerodynamic noise, turbulence-26; Impeller rotation speed-27; Impeller tip speed-27; Blade-passing frequency-27; Blade-rate
frequency-27; Impellerdiffuser combinations-27; Impeller vibration induced by steady force-27; Number of stages-28; Blade
loading-28; Impeller-diffuser distance-28; Impeller blade unequal spacing-28; Effect of gas molecular weight-28; Effect of head,
capacity operating point-28; Inlet and discharge-28; Turning vanes-28.
PART X: INDUSTRIAL MACHINERY NOISE-RECIPROCATING MACHINES Page 29
Inlet, exhaust, combustion noise-29; Mechanical noise-29; Inertia forces-29, Piston impact-29; Torsional vibration-29; Fuel
system noise-29.
PART XI: SOUND-CONTROL RECOMMENDATIONS Page 30
Summary of recommendations to be considered when purchasing sound-control materials and components-31.
ABOUT THE AUTHOR Page 32
2



THINK QUIET Part I-Sound waves and definitions
Sound is what we can hear. It exists in many ranges.
Dogs and bats, for example, can hear or react to sound waves
pitched far above those that humans can discern. Sounds may
be roughly classified as speech, music, and noise. Speech is
usually pleasant to the ear. Music may or may not be agreeable,
depending upon the opinion of the hearer and his musical
appreciation. Noise is disagreeable or unwanted sound.
Society during the past 2 years has decided to take
positive action against excessive noise. "Unlike air and water
pollution," reports The New York Times, "noise pollution has only
recently been recognized as a major threat to human well-being.
Consider these facts. Eighty decibels is about as loud as a sound
can get without making most people uncomfortable. Any rating
above 80 produces physiological effects. Long exposures at
much above 100 decibels threaten permanent impairment of
hearing.
"Against this background, here are some common noise
values: ordinary conversation (60); heavy traffic (80); a cocktail
party (90); a moving subway train (100); a riveting gun (130); a
hard-rock band (100-138); and, a jet plane heard at close range
(150). At 140 decibels or more, acute pain is experienced."
There were many early warnings that social and legisla-
tive action against noise was imminent, but few people took them
seriously. Since 1958, the number of sound specifications made
by purchasers of machinery has increased steadily, and today,
there are noise restrictions on practically every type of machinery.
It is much better to design machinery to be quiet than to
try to quiet it after it has been built. Sometimes it is necessary
to do the latter, even though acoustic principles have been carefully
followed in the design. Nevertheless, it is always better to start
with a quiet machine than a noisy one.
The objective of this series is to attack the noise problem
at its source-with the design engineers.* In many


'Editor's Note-Previous articles have been published in Compressed Air dealing with sound
pollution and its control. This current series of articles is intended primarily for design engineers.
3
instances, it is no more expensive to make a quiet design
than to make a noisy one. In others, the additional cost of built-in
sound control is in sight. However, if enough design engineers
keep thinking about the problem, understand acoustic design
principles, and keep applying them whenever they can, the end
products are bound to come out quieter. In short, think quiet.
Sound waves are one form of a general class of
elastic waves that can occur in any medium having the properties
of mass and elasticity. Mass, or inertia, enables displaced
particles to transfer momentum to adjacent particles.
Elasticity tends to restore a displaced particle to its original
position, like a spring.
Air has both mass and elasticity. At ambient
temperature and pressure, it weighs about 0.075 pound per cubic
foot. Its elasticity can be demonstrated by pushing a piston
into a closed cylinder, as in a bicycle pump, with the air
outlet closed. As the air compresses, it offers resistance, just
as a spring would. Because air has both of these necessary
characteristics, it can transmit wave motion.
Whenever an object moves back and forth, or
vibrates, it disturbs the air next to it, and causes the air to
move back and forth too. This movement produces a variation in
normal atmospheric pressure, and the disturbance propagates
from particle to particle, causing alternate layers of compression
and rarefaction to move away from the vibrating source.
When such waves reach the ear of a listener, the sensation of
"sound" is produced.
Sound waves traveling from a source to a listener, or
receiver, are called forward-traveling waves. A wave coming
from the opposite direction is called a backward-traveling
wave. These may be produced by a second vibrating source,
or by placing a reflecting object in the path of the original
forward-traveling wave. This turns the wave around, and it
seems to originate from the reflecting object or wall.
If the original wave is sinusoidal, the reflected one
will also be sinusoidal, but traveling in the opposite direction.
When the two waves meet, a somewhat surprising, but com-
monly encountered, effect is produced.
Assume the original sound wave is produced by a
piston moving back and forth in a cylinder in simple harmonic
motion. If a sensitive pressure transducer capable of detecting
instantaneous pressure levels is located at a fixed distance
from the source, it will indicate all the pressure peaks and valleys
in the wave as it moves past the transducer. The pressure
increases to a maximum, decreases to zero, goes to a
negative maximum, returns again to zero, and then repeats the
cycle.
A transducer located slightly farther away from the
source will go through the same instantaneous pressure
variations, but see them a little later in time. That is, any sound
pressure that exists at one location will occur at a more distant
location, but at a later time.
Now, if the wave strikes a wall and is reflected, this
reflected wave is also a traveling wave, but in the opposite
direction. When the backward-traveling wave and the forward-
traveling wave are combined, the resulting wave no longer
travels. At each point in space, the sound pressure increases
and decreases with time, but the maximum value of the
variation is different at each location, and at certain places it
is zero. When this happens, the result is called a standing
wave. If a microphone is moved along a line between the sound
source and the reflecting surface, a series of maxima and
minima in sound level is found.
This effect is most pronounced when the noise source
produces sinusoidal waves, or discrete frequency components
are present. Complex sounds, composed of many different
components, average the effects so that standing waves are not
noticeable. However, if they are present and are not handled properly,
they can cause large errors in sound measurements.
The point to be remembered is that sound is a wave motion
and a knowledge of the behavior of sound waves is important to the
proper application of sound control. This must be considered in muffler
design, the design of acoustic enclosures, in structure-borne sound, and
even in the proper placement of sound-absorbing material.
When a piston is moved back and forth in a cylinder in
simple harmonic motion, it produces a variation in atmospheric pressure
and generates sound. The displacement of the piston determines the
strength of the sound wave. The number of times per second it is moved
back and forth, from the midpoint to the far end of the tube and then
back to the opposite end of the tube, and then returned again to the
mid-position, determines the frequency of the sound wavethat is, the
number of cycles per second. The time required for one complete
cycle is called the period, and is the reciprocal of the frequency.
The speed of sound in air depends upon the temperature, and is equal to

C = 49.031!R
where,
C is the speed of sound in feet per second (fps), and
R is the temperature in degrees Rankine (459.7 plus the
temperature in degrees Fahrenheit).
For example, at 70 F, the speed of sound in air is

C = 49.031!459.7 + 70
= 1128 fps.

The wave length of a sound wave is the distance the wave travels
during one cycle, and speed (frequency) and wave length are related
by
C = f"
where,
C is the speed of sound in feet per second,
f is the frequency in cycles per second, now Hertz (Hz), and
" is the wave length in feet.
At 70 F, when the speed of sound is 1128 fps, the wave
length of a 1000-Hz sound wave is 1.128 feet.
Sound travels about 4700 fps in water, 16,500 fps in steel, and
13,000 fps in wood. Its velocity is known for a great variety of
solids, liquids, and gases. An understanding of this is important
because a different technique must be used for control of long-wave-
length, low-frequency sound waves than is used for short-wave-length,
high-frequency waves.
The sound waves thus far described were generated by a
piston moving back and forth in a tube. The tube was assumed to be
infinitely long, and the wave consisted of alternating sections of
compression and rarefaction, bounded by the walls of the tube. This plane
wave, as it is called, is not the only type of sound.
Consider a sphere, such as a balloon, expanding and
contracting radially in simple harmonic motion. This generates a
spherical wave that propagates in the same manner as the plane
wave. The speed of propagation is the same as that of the plane wave,
and the same relations exist between frequency and wave length.
Although the wave now has a spherical front, it is still one-
dimensional since all of its parameters can be referred to one
measurement, the radial distance of the wave front to the center of
the sphere.
Many other complex waves can be generated, too. For
example, a cylindrical tube can vibrate in its radial direction,
generating a cylindrical wave. If the ends of the cylinder vibrate in the
axial direction, at the same time as the cylinder vibrates radially, a
more complex, two-dimensional wave will be produced. Many other
types can be imagined.
Nevertheless, in all sound waves, there is longitudinal
compression and rarefaction. That is, the air is alternately
4

THINK QUIET
compressed and rarefied in the direction of travel by the wave. This
type of wave differs from light waves in that light consists of
transverse waves, or waves that vibrate in a plane normal to the
direction of propagation.
Certain definitions must be explained preliminary to a
discussion of sound and its control.
Sound pressure is the instantaneous pressure measured
in a sound wave; that is, the variation from atmospheric pressure.
Sound power of a source is the total sound energy radiated
by the source per unit of time.
Sound intensity in a specified direction is the average
rate of sound energy transmitted in that direction through a unit
area normal to the direction.
When a sound source radiates power uniformly in all
directions, all of the power must pass through a surface enclosing
the source. If we move away from the source, and the radius of
the enclosing surface is increased, the power per unit area must
decrease. The total sound power remains the same, but the
enclosing area is increasing and, therefore, the sound intensity
must be decreasing. That is
W=IS
where,
W is the total sound power,
I is the intensity, and
S is the total surface area.
At present, there is no instrument that can read
directly either sound power or sound intensity. The human ear
and microphones respond to sound pressure, and most acoustical
instruments are designed to measure the effective sound
pressure level, or the rms (root-mean-square) sound pressure level.
The term decibel (dB) may be thought to be uniquely
related to sound. Actually, it originated in electrical engineering. By
definition, a decibel is 10 times the logarithm, to the base 10, of
a ratio of two powers. For example, if the input power to an
electronic amplifier is W
1
,

and the output power is W
2,
the gain of the
amplifier in decibels is

10 log

If W
1
is 1 milliwatt and W
2
is 10 milliwatts, the power gain is

10 log

or 10 dB. If the input power is 2 w and the output power is 20 w,
the power gain is still 10 dB. Note that the decibel gain does not
indicate the actual power developed by the amplifier, nor even what
the input or output values actually are. It shows only the ratio of
the output to the input. Note also that a 10-dB gain means an
increase of 10 to 1. A 20-dB gain means an increase of 100 to 1, and a
30-dB gain, an increase of 1000 to 1. In other words, each time 10
dB are added, the power ratio has been multiplied by 10.
In acoustics, the reference level W
t
has been set at 10
-12
w. That is,

Sound Power Level = PWL = 10 log

If the output power of a sound source is 10
-5
w, the sound power level
is

THINKING QUIET-
For making snow, Ingersoll-Rand Whisperized portable compressors, (1200 cfm) operate with reduced
dB levels at Duchess Ski Area, Beacon, N.Y.
Paving breaker, once a symbol of excessive noise, is equipped with a silencer; and supplied with air by a
Whisperized Gyro- Flo compressor.
W
2

W
1

10
1
W
10
-12
5


If the output power is 1.0 w



It can be seen that a range of 70 to 120 dB, often found in
industrial machinery, represents a range in actual sound
power of 0.00001 to 1.0 w, a range of 100,000 to 1. A soft
whisper may produce only about 0.000 000 001 w, while a jet
engine can radiate 10,000 w. This indicates a range of
10
13
to 1. The advantage of the decibel scale, then, is that
small numbers can be used in calculations instead of large,
unwieldly ones.
Some references, articles about sound, and
specifications refer to 10
-13
w as the reference level, instead
of the new level of 10
-12
w. This should always be checked
when discussing sound power levels. It makes a difference
of 10 dB.
In electrical engineering, it can be shown that when
the impedance is constant, the power ratio is equal to the
square of the voltage ratio. That is,
where, E
2
and E 1 are the output and input voltages.
Voltage is electrical pressure and corresponds, in acoustics,
to sound pressure. Since
where,
SPL is sound pressure level, in decibels,
P
2
is sound pressure in newtons/m
2
, and
P
1
is reference pressure, equal to 2.0 X 10
5
newton/m
2
.
A bar is a pressure of one atmosphere. A millibar is one
thousandth of an atmosphere, and a mi crobar i s one millionth of an
atmosphere. This is by definition. Specifically, a microbar is 1
dyne/cm
2
, or 0.1 newton/m
2
. In acoustics, a reference sound pressure has
been established at 0.0002 microbar (dynes per square centimeter). In
international standards, it is written as 2.0 X 10
-5
newton/m
2
, which is
exactly the same as 0.0002 microbar.
The phon is a unit of loudness level. Loudness level, in phons, is
defined as the sound pressure level of a 1000-Hz tone that sounds as loud
as the noise being rated.
The sone is a unit of loudness. By definition, a 1000-Hz tone, 40
dB above a listener's threshold, produces a loudness of 1 sone. The
loudness of any source that is judged by the listener to be n times that of
the 1 sone tone is n sones.
Perceived noise level, or PNdB, is defined as the sound
pressure level of a band of noise from 910 to 1090 Hz that sounds as
"noisy" as the sound under comparison. It is used in a manner similar to
phons, except that where phons are used in equal-loudness contours,
PNdB are used to obtain equal-annoyance contours.


The Super-Breaker has been designed by Ingersoll-Rand with a casing that surrounds the working barrel and provides internal
baffling for exhaust air.
Gas turbine-driven centrifugal compressor, set rated at 15,500 hp, has an inlet silencer. exhaust silencer, and an acoustic
enclosure over the gas generator and power turbine.


6
The basic instrument in all sound measurement is the
sound level meter. It consists of a nondirectional microphone,
a calibrated attenuator, an amplifier, an indicating meter, and
weighting networks. The meter reading is in terms of root-mean-
square (rms) sound pressure level, re 2.0 X 10
-5
newton/m
2
. Thus,
the instrument cannot measure the peak level of high-speed
sounds, as those produced by hammer blows, punch presses, or
gunshots. Special instrumentation is required for such
applications.
Usually, three weighting networks are included in these
instruments-A, B, and C. Their purpose is to give an
indication of the total loudness level. The ear is not as
sensitive to low-frequency sounds as it is to high-frequency
sounds, and this effect is more pronounced at low sound levels
than at high.
The original intent of the weighting networks was to use
the A network for sounds below 55 dB, the B network for sounds
between 55 and 85 dB, and the C network for levels above 85
dB. Recent studies have shown that the A network gives a very
good indication of the loudness of sounds, regardless of the
level, and now the B network is hardly used. Figure 1 shows
the frequency response curves for the A, B, and C networks.
An important distinction is made between the numbers
read on the A and C scales. The C network is essentially flat,
and sounds read with it are called sound pressure levels. All
frequency analyses must be measured on the C scale.
The A network falls off sharply at low frequencies,
to correspond to the response of the ear. Sounds read with it
are called sound levels and are referred to as dBA. The A
scale gives the most important overall measurement.
In addition the A, B, and C networks, newer sound
level meters have an amplifier with a flat frequency response from
20 to 20,000 Hz. Overall response when using this network
depends on the microphone being used.
The microphone is the most important part of the sound-
measuring instrumentation because the accuracy of the
measurement can be no better than that of the mike.
Frequency response, sensitivity, directionality, and range


are primarily determined by the microphone. Three general types are
available. Crystal microphones of the early type were relatively
inexpensive, rugged, and highly sensitive. Their frequency response
was not good, however. Modern crystal microphones retain all the
good features of the older models, and in addition, have good
frequency response. They can operate at temperatures to about
200F. The dynamic microphone has a coil moving in a magnetic
field. It has good frequency response and can operate at
temperatures to 180 F. These microphones are affected by
magnetic fields and should not be used around motors, generators,
or transformers. Condenser microphones have the best frequency
response of any today. Some can go as high as 200,000 Hz. Their
disadvantages are high cost, limited humidity range, and the need of
an additional preamplifier.
The octave-band analyzer is the most common analyzer for
industrial noise measurement. As its name implies, it separates the
complex noise into frequency bands one octave in width, and
measures the level in each of the bands.
An octave has an upper cutoff frequency twice the lower cutoff
frequency. In noise studies, specific frequency bands have been
established. The older octaves were 37.5-75, 75-150, 150-300, 300-
600, 600-1200, 1200-2400, 2400-4800, and 4800-9600 Hz. Some
instruments had slightly more than an octave at the low and high
ends, and these octaves were shown at 20-75 and 4800-10,000 Hz.
The center frequency of each band is ! 2 times the lower
cutoff frequency. Octave-band analyzers usually provide sufficient
information about industrial noise to solve noise control problems.
Newer octave-band analyzers, instead of identifying the
octaves by the lower and upper cutoff frequencies, such as 300-
600 Hz, identify them by the center frequencies of the new,
preferred octave-band series. The octaves are now shown as 31.5,
63, 125, 250, 1K (1000), 2K, 4K, and 8K. As before, the lower
and upper cutoff frequencies are equal to the center frequency
divided by and multiplied by ! 2, respectively.
Narrow-band analysis must be performed when the source
of a noise component must be identified for purposes of sound
reduction, or some other reason. Half-octave analyzers have an
upper cutoff frequency of ! 2 times the lower, while third-octave
analyzers have a ratio of the cube root of 2, or 1.26. Tenth-octave
analyzers have a band width equal to the tenth root of 2.
Constant-percentage analyzers have a band width that is
always a fixed percentage of the frequency to which it is tuned.
For example, a 2 percent band-width analyzer would have a band
width of 2 cycles when tuned to 100 Hz. Its band width would be 200
cycles when tuned to 10,000 Hz.
A constant-band-width analyzer has a band width that is
independent of its setting; a 2-cycle band-width analyzer will
always have a band width of 2 cycles. This type is difficult to
use when there are slight changes in the measured frequency,
especially at the high end of the spectrum. For this reason, they are
not often used for machinery noise analyses.

When noise is fairly evenly distributed over the spectrum, the
level measured with one analyzer can be corrrected to give a


THINK QUIET
Part II-Instrumentation and measurement of sound


Frequency-response characteristics in American Standard
for Sound-Level Meters, S 1.4,1961
Figure 1
20 50 100 200 500 1K 2K 5K 10K
FREQUENCY (Hz)
7
close approximation to what it would be with an analyzer of a
different band width. When a full octave is being measured,
more sound energy is being considered than if only a third-
octave band were admitted to the analyzer. Therefore, 4.86 dB
should be subtracted from the octave-band reading to obtain
the equivalent third-octave reading. Similarly, other
corrections would be

Octave Third-Octave 8 Percent 2 Percent
Band Band Band Band

Octave band to ---- 0 4.86 9.46 15.48

These corrections do not apply in the case of discrete fre-
quency components because all the readings would be the
same.
A convenient method for making these calculations
is first to convert all readings to what is called spectrum level.
This means correcting a reading to an equivalent sound
pressure level for a band 1 c/s wide; that is, it is the
pressure-squared-per-cycle-per-second level. The correction
for this case is

where,
f
2
f
1
is the upper frequency minus the lower
frequency in the band being measured, and
B is the band width, where level is required.
Then. if



In certain measurements, it is convenient to use
"white noise," or noise that has equal energy per cycle. An
octaveband analyzer measuring white noise will show a slope
of plus 3 dB per octave. This is because each octave admits
twice as much sound energy as the previous lower octave.
That is, the octave from 300 to 600 cycles contains 300
cycles, while the 600 to 1200 octave contains 600 cycles.
Since 10 log 2 equals 3, the slope will be 3 dB per octave.
An electrical filter having a slope of minus
3dB per octave inserted in the circuit will produce a flat
response. White noise converted from equal energy per cycle
to equal energy per octave is called "pink noise."
An acoustic calibrator is an essential part of a
soundmeasuring system. It fits over the microphone and
calibrates the entire system of microphone, attenuator, ampli-
fier, and meter. Various types are available, but a good
unit to use is one that generates accurate sound pressure levels
at several different frequencies. An elaborate, but
expensive, calibrator, which can perform absolute calibration
of laboratory standard microphones, is called a reciprocity
calibrator. These are not used for usual industrial sound
measurement or field work.
Air currents blowing across a microphone produce
erratic low-frequency noise. If the wind velocity is high enough,
the turbulence noise can exceed the noise being measured
and cause errors in the readings. When it is necessary to
take data under such conditions, the microphone should be
shielded by a wind screen. A suitable one may be made be
stretching a thin silk or nylon cloth over a wire frame that
completely encloses the microphone. The wind screen
should be of fairly large volume and not tightfitting. The attenuation
of the screen may be measured by reading the same noise, with and
without the screen, in a location where there is no wind.
Recording noise for analysis in the laboratory has a definite
advantage. It can be analyzed with various bandwidth analyzers,
displayed on a graphic chart, and retained for other purposes if desired.
However, when magnetic tape recorders or graphic level recorders
are used, they should have a frequency response as good as the other
components so distortion is not introduced. The usual stereo recorder is
not good enough, even though it may sound perfect for music playback.
In addition to flat-frequency response over a wide range, the recorder
should have low hum and noise level, low distortion, and constant-speed
drive. Data should be recorded at about 15 ips.
Remember, while recording has definite advantages, it
serves best when it supplements, rather than replaces, directly
measured data. Directly read data means that you know the answer
sooner. It removes the possible disappointment of having nothing
recorded on the tape when you return to the laboratory.
Microphone placement depends upon the objective. If the
noise at a worker's ears is to be measured, the microphone should be
placed about where the worker's ears would be. If a particular noise is
to be evaluated, such as that produced by a compressor inlet, then
the microphone is located there.
For machinery noise evaluation, the recently developed CAGI-
PNEUROP Test Code For The Measurement Of Sound From
Pneumatic Equipment states, "A preliminary survey shall be taken all
around the machine at a distance of 1 meter from the nearest major
surface of the machine, and at a height of 1.5 meters to locate the point of
maximum overall sound level (A scale). This is the primary micro-
phone location. Additional microphone locations shall be established at
each end of the plant and at the center of the sides of each casing. All
these microphone locations shall be at a horizontal distance of 1 meter
from the nearest major surface of the machine, and at a height of 1.5
meters above the floor or above the walk level."
At each of the microphone locations, the following data should
be taken with the machine operating:
1. overall sound level using the A-weighting network;
and,
2. octave-band sound pressure levels using the flat re
sponse or C network. (Overall sound on the C net
work is not really necessary, but is usually taken as
a check to see if the octave-band data look right.)



Noise monitor with microphone provides noise-exposure information.
8
THINK QUIET
A similar set of data should be taken at one of the
locations with the machine shut down. This gives background
noise. A sketch should be made showing the machine, locations of
other machinery, building walls, and microphone locations. A
description of the machine should be given, and the operating
conditions stated. Test data also should include the make, model,
and serial number of the sound test instruments used.
Certain calculations are usually required to interpret
data taken in sound tests. For example, it is often necessary to
combine sounds made by different machines, or even octave-
band data measured on the same machine. The following is an
example of how this may be done:

Example: Assume that three sounds of different fre-
quencies are to be combined to obtain the total
sound pressure level. Let the three sound pressure
levels be (a) 90, (b) 88, and (c) 85 dB.

Solution: 1. Mean-square sound pressure ratio of



2. Mean-square sound pressure ratio of



3. Mean-square sound pressure ratio of



4. Total mean-square sound pressure ratio =

(10.0 X 10
8
) + (6.3 X 10
8
) +






5. Total sound pressure level =
10 log 19.46 X 10
8
= 92.9 dB
A simpler, graphical method of doing this is shown in
Figure 2. When using the chart, any two sounds are combined and
the sum added to the next sound. It can be seen that when the 90 dB
and the 88 dB sounds are to be combined, the difference of 2 dB
indicates that 2.1 dB should be added to the higher level, making 92.1
dB. Then, this is combined with the 85 dB sound. The difference is
now 7.1 dB and the number of decibels to be added to the 92.1 dB is
0.79, making the total 92.9 dB.
This procedure cannot be used to combine two sounds of the
same frequency since the calculation must now include the phase
angle of the two sounds. The equation for doing this is:

where,
P is the rms pressure, total,
P 1 is rms pressure of first sound,
P
2
is rms pressure of second sound, and
is the phase difference.

If any two of the sounds have the same frequency, they must first
be combined by this equation before using the chart to combine with
others. Two sounds of equal magnitude and different frequencies
combine to equal 3 dB more than one alone. Two sounds of equal
magnitude and the same frequency can combine to produce 0, or 6 dB
greater than one alone, or anything between 0 and 6 dB, depending upon
the phase angle.
Background noise often must be subtracted from the total
noise to obtain the sound pressure level of a machine alone. This can
be done mathematically in a manner similar to that for combining
sounds. However, a more convenient way is to use the chart shown
in Figure 3. Note that this chart should be used only when the total
noise exceeds the background noise by 3 dB or more. If the difference
is less than 3 dB, a valid sound test cannot be made.




Figure 3
DIFFERENCE BETWEEN TOTAL AOUND AND
BACKGROUND SOUND (dB)
Figure2
9








A PNEUMATIC LEITMOTIV
When the word "sound" is heard today, the first reaction is to think of
raucous sounds that besiege our ears. But there are other sounds, such as
the pleasant tinkling of a wind-bell as it sways in a warm summer breeze. We
have selected this incongruity for a leitmotiv to mark off each part of the
current series of "Think Quiet" articles.
It is understandable that our readers show deep interest in sound-abatement
articles there is a great need for solid information on the subject, and a
piteous lack of it. Interestingly, there was also curiosity shown in the design
selected for the leitmotiv. For this reason, we asked its creator, our art
director, to turn it into a cover for this issue. It serves, as does this series,
to remind the reader that the compressed air industry is cognizant of its
responsibility to curb noise pollution. Although for space limitations these
articles will be listed henceforth as a Department on the contents page, this in
no way diminishes their value.
--the editors
THINK QUIET
-Editor's Note-Because of the intense interest in sound control, we feel many readers will be saving this
series. Reference to figures are therefore continuous through the series.
Part III-Sound test evaluation
An excellent reason for conducting sound tests is because
it is becoming increasingly difficult to sell noisy machinery.
While no one will dispute the fact that engineering and
manufacturing excellence, and the resulting reliability of
machinery, are of first importance, it is equally true that
customers are not only asking for quiet operation, they are
insisting on it. They compare reliability, efficiency, cost,
service, and all the other factors that enter into final selection,
and do this with the understanding that the machine will meet a
certain sound specification.
Most large industrial firms have their own acoustical
engineers and the ability to evaluate acoustical performance of
machinery, both in the manufacturer's plant and after final
installation. It is therefore very important that machinery
manufacturers be capable of understanding and handling noise
problems too.
The main reason for insisting on quiet machinery is to
prevent hearing damage. A report by the National Research
Council Committee On Hearing, Bioacoustics & Biomechanics,
of the National Academy Of Science, provides what is probably
the most accurate method for the assessment of hazardous
exposure to intermittent and steady-state noise. This so-called
CHABA report presents data in the form of eleven graphs.
Unfortunately, it is too complicated for general use.
Figure 4* shows octave-band data taken from the report.
It can be seen that higher sound levels are permitted at the low-
frequency end of the spectrum than at the high-frequency end
because the ear is more sensitive to high frequencies than to
low. Sounds must be at higher level at low frequencies to seem
as loud as high-frequency sounds.
A study of the CHABA data and other investigations
about hearing damage levels has revealed that a simpler ap-
proach can be made. It is based on A-weighted overall levels
instead of the more complicated octave and third-octave band
method. (The shape of the A-weighted response in the sound
level meter is the reverse of that shown on hearing damage
curves.) This was the basis for the Walsh-Healey regulation
that permits 90 dBA for 8 hours per day continuous exposure.
Potential hearing damage is a function of both
decibel level and time of exposure. On an equal-energy basis,
this means that the permissible level could be increased 3
dBA each time exposure time is halved. The Walsh-Healey Act
permits a 5 dBA increase, instead of 3 dBA, when the time is
halved to allow a reasonable correction for intermittency of the
sound.
The Walsh-Healey Act also states that in no case
should a worker be exposed to more than 115 dBA, however
short the exposure; and not to impulsive or impact noise in
excess of 140 dBC peak sound pressure level-that is, when
measured with the C-weighting network. This latter level re

Figure 4
10

THINK QUIET
quires special instrumentation, such as an impact noise analyzer or a
cathode-ray oscilloscope. It cannot be read on an ordinary sound
level meter.
Even at 90 dBA, there is a 14 to 16 percent risk of
noiseinduced hearing loss. At 85 dBA, the risk is about 5 to 7
percent. The risk at 80 dBA is about zero. The indications are
that present limits will remain in effect for 5 to 6 years, and
possibly 10. At the end of that time, the permissible level will
probably be reduced to 85 dBA, for 8 hours exposure.
To determine dBA levels, the simplest way is to measure
them directly. Occasionally, octave-band levels are available, but
there is no dBA reading. The dBA level can be calculated from
octave-band sound pressure levels by the following method:


Octave
Band Band
Corrected dB
Antilog
dB
Band SPL Correction SPL 10 10
63 88 -26 62 6.2 0.02 X 10
8

125 88 -16 72 7.2 0.16 X 10
8

250 94 -9 85 8.5 3.16 X 10
8

500 96 - 3 93 9.3 19.95 X 10
8

1K 96 0 96 9.6 39.81 X 10
8

2K 92 1 93 9.3 19.95 X 10
8

4K 89 1 90 9.0
10.00 X 10
8

8K 76 -1 75 7.5 0.32 X 10
8

93.37 X 10
8

10 log 93.37 X 10
8
= 10 (9.97) = 99.7 = 100 dBA
The sound pressure levels are first corrected to obtain the A
response levels. The corrected levels are then combined

Figure 5
either mathematically or graphically to obtain the dBA sound level.
An alternate procedure is offered in the Walsh-
Healey regulation. The octave-band sound pressure levels are
plotted on a chart of equivalent sound level contours (Figure 5). The
point of highest penetration into the sound contours determines
the A-weighted sound level. This dBA level is used to evaluate the
exposure limit, even though it may differ slightly from the actual
A-weighted sound level.
If an employee is exposed to noise of different levels for
two or more different periods of time, the combined effect must be
considered, rather than the individual effect of each. That is, if the
sum of the following fractions exceeds unity, then the mixed
exposure is considered to exceed the limit value
where,
C
n
is the total time of exposure at a specified level, and
T
n
is the permissible time at that exposure.
For example, in a work area where the noise levels are found to
be 95 dBA for 3 hours, 90 dBA for 4 hours, and 85 dBA for 1 hour,
the permissible time for 95 dBA is 4 hours; for 90 dBA, 8 hours;
and for 85 dBA, no limit. The equivalent steady value is then

This exceeds the permissible limit of 1.0.
Acoustical characteristics of a machine may be required for a
variety of reasons. Manufacturers may want to verify that it
conforms to a customer's sound specification. They may want to
compare it with other machines of the same type. Purchasers
may want to compare the sound produced by various machines. To
keep all sound tests and comparisons on the same basis, as far
as compressed air machinery is concerned, the Compressed Air
& Gas Institute (CAGI) in the United States and the European
Committee Of Manufacturers Of Compressed Air Equipment
(PNEUROP) in Europe have developed a Test Code For The
Measurement Of Sound From Pneumatic Equipment.
Users of this code will find one of the requirements is that
the sound level meter must meet the standards of IEC 179, the
International Electrotechnical Commission Publication On
Precision Sound Level Meters. This is in line with recent
trends where specification-writing committees are insisting on
greater accuracy of instrumentation.
The General-Purpose Sound Level Meter Specification

Impacts and impulses in excess of 140 dB can be measured with a
precision analyzer from B & K Instruments, Inc. Unit incorporates a
sound level meter with an octave-band filter.
11

ANSI SI.4 is more lenient with sound level meter characteristics.
Briefly, the difference is that below 100 Hz, the general-
purpose specification has tighter tolerances than the precision
specification. Between 100 and 800 Hz, the tolerances are
identical. Above 800 Hz, the precision tolerances are tighter
than the general-purpose ones, and extend to 12,500 Hz. The
general-purpose specification extends only to 10,000 Hz.
Sound tests should be made in a location where
background noise is 10 dB less than the noise with the machine
running. In many cases, this is influenced by low-frequency noise,
and this can rule out an otherwise acceptable area. Further,
sound pressure levels below 50 dB are of little or no interest in
many industrial machinery sound tests.* Therefore, the CAGI-
PNEUROP Sound Test Code states that "the readings at each
location with the machine running should exceed the
background levels by at least 10 dB in each octave band of
interest."
Octave bands of interest are those where the sound
pressure levels produced by the machine are within 40 dB of the
highest measured octave-band level. In no case are levels below
50 dB considered important. For example, if the highest
measured level, in any octave band, were 120 dB, levels below
80 dB would not be of interest. However, if a machine were very
quiet to start with, and the highest measured octave-band level
were 70 dB, then levels below 50 dB would not be important.
When sound measurements are made at a distance of 1 meter
from the machine, the levels are not affected very

*This is not true when talking about residential annoyance levels, a subject of its own andnot thesameas
industrial machinerysoundtestswheremeasurementsaremade1 meter fromthe machine.
This test code has been submitted to the International Standardization Organization for approval, andwill
probably beaccepted: however. ISOthinks that octavebands of interest should be those within 50 db of the
highest one instead of 40 db.




Figure 7
much by sound reflected from walls and other machinery, even
though there is some contribution from those sources. The CAGI-
PNEUROP Code states "the test area should be large enough so
that the sound pressure levels at the prescribed microphone
locations are at least 6 dB greater in each octave band of
interest than the sound pressure levels measured at more distant
points in the same direction from the source as the microphone
locations." Calculations show that when this criterion is met,
the reflected sound contribution will not be more than 1 dB.
The CAGI-PNEUROP Code also describes a method for
reducing the effects of noise reflected from the floor. It states,
"because of the interference between direct sound waves and
those reflected from the floor, large errors may occur when
strong discrete frequency components are present. Tests shall be
made by rapidly moving the microphone vertically approximately
plus or minus 0.3 meter from each location."
The interference effect is not related to side walls,
nor is it to other nearby machinery. It can occur in a free field
over a reflecting plane with no reflections from walls. The main
interferences occur when the difference between the direct path
length between sound source and microphone, and that of the
reflection from the floor (that is, source to floor to
microphone), is an odd multiple of a half wave length. Figure
6 shows the effects at three different microphone locations, and
how it can be eliminated by moving the microphone.
The objective in a sound control program may be to
avoid levels that will interfere with speech communication. Speech
Interference Level (S.I.L.) is defined as the arithmetic average of
the sound pressure levels in the three octave bands: 600-1200 Hz;
1200-2400 Hz; and, 2400-4800 Hz.
In terms of the new octave bands, it is called P.S.I.L.,
and is the arithmetic average of the levels in the 500- , 1000-
, and 2000-Hz octaves. The resulting number is a good indication of
how the noise interferes with speech.
Speech interference levels may also be given in terms
of dBA, as read on a sound level meter. Figure 7 shows the
levels that will interfere with conversation at various
distances.
No matter which criterion is used, the point to be remem-
bered is that the ability to communicate depends upon the
intensity and frequency of the message as compared to the
intensity and frequency of the background noise.
Figure 6
12









Preventing community complaints about noise is a
growing problem for industry. Criteria to prevent annoyances are
more restrictive than criteria to prevent hearing damage or speech
interference. More factors enter the problem.
Annoyance depends upon the level of the offending noise
compared to the pre-existing background level, its absolute value,
its frequency, how it varies with time, and whether it occurs during
the day or night. Many cities are writing antinoise ordinances,
and these will include permissible residential noise levels. The
report of the mayor's task force on noise control in New York City
recommends a daytime limit for noise in wholly residential areas of
40 dBA, and below 30 dBA at night.
While it is impossible to predict exactly the response
from any particular neighborhood to any specific noise, a fairly
reliable method has been developed by Stevens, Rosenblith, and
Bolt. One can plot the measured octaveband sound pressure
levels on Figure 8 to determine the initial "level rank." The
highest zone into which any of the octave-band levels penetrates is
the level rank of the noise.
Corrections are next applied as follows:

CONDITION CORRECTION
Pure-tone components + 1
Wide-band noise 0
Impulsive + 1
Not impulsive 0
Continuous exposures, to 1 per minute 0
10-60 exposures per hour -1
1-10 exposures per hour -2
4-20 exposures per day -3
1-4 exposures per day -4
1 exposure per day -5
Very quiet suburban +1
Suburban 0
Residential urban -1
Urban near some industry -2
Area of heavy industry -3
Nighttime 0
Daytime only -1
No previous conditioning 0
Considerable previous conditioning -1
Extreme conditioning -2

The sum of the various corrections then is applied to the original
level rank to obtain the corrected level rank. The expected
community response can then be predicted by Figure 9.
A widely used set of noise criteria for various offices,
conference rooms, residences, and the like was developed by Dr.
Leo Beranek (Figure 10). The number of the curve is numerically
equal to the sound pressure level in the 1200-2400 Hz octave. The
International Standardization Organization also has
recommended a similar set of criteria, called NR (Noise Rating)
curves.
The National Electrical Manufacturers Association
(NEMA) has produced a set of noise criteria for gas turbine
installations. These curves are shown in Figure 11. It can be seen
that NEMA "A" is approximately the same as NC25; NEMA
"B" is about the same as NC30; NEMA "C" corresponds to
NC35; and NEMA "D" is about equal to NC40. Also, if the various
octave-band sound pressure levels are combined, it can be seen
that NEMA "D" would be equivalent to about 51 dBA; NEMA
"C," equal to 49 dBA; NEMA "B," to 44 dBA; and NEMA "A," to
40 dBA.
Precautions should be taken when using dBA levels. The
popularity of A-weighted sound levels is increasing rapidly as a
means of expressing all types of sound criteria, whether it be
hearing damage, speech interference levels,
THINK QUIET
Part IV-Noise criteria
Figure 8
Figure 9
Figure 10
13

Figure 11
community annoyance levels, or machinery acoustic perfor-
mance. However, it should be remembered that dBA levels
must be used with caution; whatever noise control must be
engineered, octave-band data are needed. In some cases,
still narrower band analyses are necessary to identify the
source of objectionable noise; dBA levels are inadequate for
noise control because different techniques are required to
reduce low-frequency noise than those needed for high
frequencies.
Figure 12 shows octave-band sound pressure levels
for two noises that have exactly the same dBA level as
read on a sound level meter, but the distribution of the noise
is entirely different in the two cases.
Other examples of erroneous conclusions that
can be drawn by relying too heavily on dBA levels alone are
shown in Figures 13 and 14. It is usually stated that when

Figure 12
Figure 13 Figure 14

the background noise is 10 dB or more below the noise of the
machine itself, the test area is perfectly satisfactory. Figure
13 shows a case where the machine noise is 88 dBA and
the background noise is 77 dBA. Even though there is 11 dB
difference, the test data would be absolutely useless,
except in the 4800-9600 Hz octave.
Figure 14 shows another case where the machine
alone is 88 dBA, and the background noise alone is also 88
dBA. It would ordinarily be assumed that the test data were
no good, when actually the data are perfectly valid, except
in the highest octave.
Thus, there is an important rule. Sound should be
combined on an octave-band basis only, and not by adding or
subtracting overall levels.
14

Part V-Sound waves under various conditions
A free sound field is defined by American Standard
Acoustical Terminology as a homogeneous, isotropic medium, free
from boundaries. In practice, it is a field in which the effects of the
boundaries are negligible over the region of interest.
A nondirectional source of sound suspended in a free
field will emit sound uniformly in all directions. As the spherical
wave front progresses farther and farther from the source, its area
increases with the square of the distance since the area of a
sphere is 4! times the square of its radius. Therefore, the intensity
decreases as the square of the distance from the source. If the
distance is doubled, the sound energy will be spread over four
times the area, and the intensity will be one-fourth its original
value. This is called the "inverse square law."
Sound pressure varies as the square root of the intensity,
so it will be one-half its original value; that is, it will be reduced by
6 dB each time the distance from the source is doubled.
Sound pressure in the vicinity of the usual sound source
often shows an appreciable variation with position, even when
the source is in a free field. In what is called the "near field," the
sound pressure level does not decrease 6 dB each time the distance
from the source is doubled. The extent of the near field depends
upon the frequency of the sound, on the dimensions of the source,
and on the phase relations of the various radiating parts of the
source. A frequently used criterion assumes that the near field
ends at a distance of about twice the largest dimension of the
source. If the source rests on a hard floor, the near field extends
to about four times the largest dimension. Beyond this distance, in
the "far field," the sound pressure level then decreases 6 dB for
each doubling of distance.
Atmospheric conditions usually cause the measured
octave-band sound pressure levels to be less than the theoretical
values, as shown in Figure 15. The increased attenuation is
greater in the higher frequencies than in the lower. Note that at 1000
feet, the additional attenuation in the 300-600 Hz band, beyond the
inverse square law, is only 1 dB, but in the 4800-9600 Hz octave,
it is about 12 dB. Atmospheric temperature and pressure also
affect sound level measurements, as shown in Figure 16, but
these are not usually appreciable.
Sound power level cannot be measured directly. It must be
calculated from sound pressure level measurements. If a point
source radiates sound uniformly in a free field, a single
measurement of the sound pressure level at a known distance is all
that is needed to calculate the sound power level:

PWL= SPL + 20 log r + 0.5
where,
PWL is the sound power level re 10
-12
w,
SPL is the sound pressure level re 2.0 X 10
-5
newton/
m
2
, and
r is the distance in feet from source to measurement point.
(If r is measured in meters, this equation becomes
PWL = SPL + 20 log r + 10.8.)
If the sound pressure level is measured over a hard reflecting
surface, instead of in free space, the sound power level can be
calculated by the following:

PWL = SPL + 20 log r - 2.5.

(Where r is in meters, PWL = SPL + 20 log r + 7.8.) Once the
sound power level is known, the sound pressure level at any other
distance can be calculated by the same equation.
A reverberant field exists when the sound from the source is
reflected back and forth between relatively hard surfaces in a room.
Ideally, these multiple reflections create a sound field that is uniformly
diffuse with respect to position in the room, and the sound is made
up of the directly radiated component plus that which is reflected
from many directions. Under these conditions, the sound pressure does
not decrease according to the inverse square law, but is the same
everywhere in the room.
If the sound source is stopped suddenly, the sound persists for
some time. This persistence is called reverberation. If the walls are
very hard and practically all the sound is reflected, again and
again, the sound persists longer than if some of the sound is absorbed
each time it strikes a wall. Certain types of sound measurements are
performed very conveniently in a reverberant room. For example,
sound power level determinations on air conditioning equipment are
usually done in this way.
Since the sound pressure level is the same everywhere in
the room, a single measurement is all that is needed to calculate the
sound power level:

PWL = SPL + 10 log R - 16.5
where,
PWL is the sound power level re 10
- 12
w,
SPL is the sound pressure level re 2.0 X 10
-5
newton/m
2
,
and
R is the room constant in feet squared.
(If R is expressed in terms of meters squared, the equa-
tion becomes PWL = SPL + 10 log R - 6.2.)
Note that the relation between sound power level and
sound pressure level in the reverberant room depends upon the room
constant, R.
Sound measurements are made frequently under actual
operating conditions that are neither free field nor reverberant
field-that is, the walls and ceiling are neither completely
absorbing nor completely hard, although a hard floor is almost always
present. These locations are called semireverberant. Tests made
under these conditions are not as accurate as those made under
free-field or reverberant conditions. In spite of this, practically all
industrial sound


THINK QUIET
Figure 15
15
Figure 16
tests, when machinery is operating normally, are conducted in
semireverberant environments.
When a sound wave falls on a surface, its energy is
partially absorbed. The sound-absorbing ability of a material is
given in terms of an absorption coefficient designated by #.
Absorption coefficient is defined as the ratio of the energy
absorbed by the surface to the energy incident upon the surface.
Therefore, # may be anywhere between 0 and 1. When # =
0, all the incident sound energy is reflected; when it equals 1,
all the energy is absorbed.
The absorption of a material depends to a certain
extent on how it is used and how it is mounted. Further, the
absorption coefficient is different for different frequencies.
Therefore, when specifying the absorbing qualities of a
material, either a table or curve showing # as a function of
frequency is required.
The table at the right shows absorption
coefficients of various materials used in construction. (For
comparison, the absorption coefficient of Fiberglas, 6
pounds per cubic foot density, is given in Table 1.)
Sometimes a supplier will state, for simplicity, the
acoustical performance of a material at 500 Hz only, or by a
Noise-Reduction Coefficient (NRC) that is obtained by
averaging, to the nearest multiple of 0.05, the absorption
coefficients at 250, 500, 1000, and 2000 Hz.
The absorption coefficient varies somewhat with
the angle of incidence of the sound wave. Therefore, for
practical use, a statistical average absorption coefficient at
each frequency is usually measured and stated by the manufac-
turer. It is often better to select a sound-absorbing
material on the basis of its characteristics for a particular
noise rather than by its average sound-absorbing qualities.
The sound absorption of a surface is given in
"sabins," which is the absorption of 1 square foot of a
perfectly absorbing surface. A surface area, S, that has an
absorption coefficient of #, has a total absorption of S x #
sabins. Average absorption coefficient, #, is calculated as
follows:
Room constant is defined by American Standard Acous-
tical Terminology as "the product of the average absorption
coefficient of the room and the total internal area of the room
divided by the quantity 1 minus the average absorption
coefficient." That is,

where,
R is the room constant in square feet,
# is the average absorption coefficient, and
S
t
is the total area of the room in square feet.
(If the area of the room is expressed in square meter,
instead of square feet, then R will be expressed in square
meters.)
Reverberation time of a room is "the time that would be
required for the mean-square sound pressure level therein, originally in
a steady state, to decrease 60 dB after the source is stopped."
Reverberation time is important when determining whether
additional sound absorption would be effective in reducing the sound
in a room, or in calculating the amount of sound absorption
already in the room. It is usually measured by means of a high-
speed recorder, showing sound pressure level versus time. A 60
dB decrease in sound pressure level means that it has decreased to one-
thousandth of its original amount.
Room constant may be calculated from reverberation time
by the following method:

where,
R is the room constant in square feet,
T is the reverberation time in seconds,
St is the total room surface in square feet, and
V is the volume of the room in cubic feet.
This method is frequently used for industrial areas where it
is impractical to calculate the total absorption.
Most industrial sound measurements are made in semi-
reverberant locations, and usually the machine, or sound source, is
installed on a hard, reflecting floor or surface.
If the source is relatively small, radiates sound uniformly in all
directions, and is in a very large room, the sound pressure level at first
decreases with distance from the source almost the same as it does in a
free field. When the boundaries of the room are approached, the
sound is reflected from the walls and adds to the directly radiated
sound, preventing further reduction. The sound pressure level stays
almost constant for the rest of the distance to the wall that is, the
direct sound pressure level decreases, but the reflected component
increases and the total remains the same.
If the walls of a somewhat smaller room are treated with
acoustical material that has a high absorption coefficient, a similar
effect will be observed. This is because the room constant is
relatively high. If the room is smaller, or the

MATERIAL 125 Hz 250 Hz 500 Hz 1000 Hz 2000 Hz 4000 Hz
Brick. Unglazed 0.03 0.03 0.03 0.04 0.05 0.07
Bri ck. Ungl azed,
Painted 001 0.01 0.02 0.02 0.02 0.03
Concrete Block 0.36 0.44 0.31 0.29 0.39 0.25
Concrete Block,
Painted 0.10 0.05 0.06 0.07 0.09 0.08
Concrete 0.01 0.01 0.015 0.02 0.02 0.02
Wood 0.15 0.11 0. 10 0.07 0.06 0.07
Glass, Ordinary,
Window 0.35 0.25 0.18 0.12 0.07 0.04
Plaster 0.013 0.015 0.02 0.03 0.04 0.05
Ply wood 0.28 0.22 0.17 0.09 0.10 0.11
Ti l e 0.02 0.03 0.03 0.03 0.03 0.02
6 lbs/ft
3

Fiberglas 0.48 0.82 0.97 0.99 0.90 0.86

Table 1
where,
# is the average absorption coefficient,
#
1
, #
2
, #
n
are the absorption coefficients of materials
on various surfaces, and
S
1
,

S
2
, S
n
are the areas of various surfaces.
16
THINK QUIET
walls harder and more reflective, the room constant is smaller, and
decrease of sound pressure level with distance does not follow the
inverse square law. Figure 17 shows a plot of relative sound pressure
level in decibels versus distance from the source for various room
constants in a semireverberant field above a reflecting plane.
Most practical sound sources do not radiate sound uni-
formly in all directions. At a fixed distance away from the source,
the sound pressure level in any octave band generally will show
different levels for different directions. A polar plot of the levels gives
the directivity pattern of the source. This pattern is really three-
dimensional, on a hemispherical surface. (Many sources are
nondirectional in the low frequencies, but become more
directional in the higher frequencies.)
Directivity index in any particular direction is defined as the
octave-band sound pressure level in that direction minus the average
octave-band sound pressure level around the source, with all
measurements taken at the same distance from the source-that is:

where,
DI
!
is the directivity index in a particular direction,
SPL
!
is the octave-band sound pressure level in that
direction, and
SPL is the average octave-band sound pressure, level
around the source.
The correct way to obtain the average of a number of
decibel readings is first to determine the individual sound power
values, calculate the average power, and convert this back to a
decibel level. This is laborious and time-consuming. A simpler
way, but less accurate, is to calculate the arithmetic average
directly. The arithmetic average is always lower than the average
obtained by averaging watts.
In most cases, the following rule can be used in practical
problems of industrial sound measurement:
1. When the spread of decibel values is between 0 and 5
dB, take the arithmetic average directly (the average will
never be in error by more than 1 dB);
2. 2. When the spread of decibel values is between 5 and 10 dB,
take the arithmetic average and add 1 dB (the error
probably will not be more than 1 dB).
Directivity factor, which is dimensionless, is defined as:


Figure 18

Even when a source radiates sound uniformly in all directions, it
may exhibit directional characteristics due to its location in a
room. If it is suspended in the geometric center of a very large
room, or outdoors, it radiates uniformly into a sphere, and Q = 1.
If the source is placed on a hard, reflecting floor, half the sound is
radiated into a hemisphere, the other half is reflected from the
floor back into the same hemisphere. In this case, Q = 2.
Similar reasoning may be applied to other locations, as
shown in Figure 18. This shows that if the source is located at a
wall-floor junction, the directivity factor is 4, and when it is
placed at a corner, where two walls meet the floor, the directivity
factor is 8.
Octave-band sound power levels of a machine in a semi-
reverberant room may be calculated from the measured octave-
band sound pressure levels by:

where,
PWL is the sound power level re 10
-12
w,
SPL is the sound pressure level re 2.0 X 10
-5
newton/m
2
,
Q is the directivity factor, and
R is the room constant in square feet.

A convenient, alternate, method of determining sound
power level in a semireverberant room is as follows:
1. Octave-band sound pressure levels of the machine to be tested
are measured over a hemisphere around the machine;
2. A nondirectional, calibrated sound source is set up in the
room (the octave-band sound power levels of the source have
been determined previously by the manufacturer);
3. Octave-band sound pressure levels are then measured in the
hemispherical array around the reference source; and,
4. Octave-band sound power levels of the machine under test are then
calculated by the equation:
where,
PWL
x
is the sound power level of the machine under test,
PWL
s
is the sound power level of the standard source,
SPL
x
is the average sound pressure level over the test
hemisphere measured with the machine under test, and
SPL
s
is the average sound pressure level over the test
hemisphere measured with the standard source
.


EFFECT OF LOCATION IN ROOM
Figure 17
17
THINK QUIET
Part VI - Cont rol l i ng t he generat i on of noi se

The best sound-control technique is to prevent the
generation of noise, or at least to reduce the level of generated
noise. This is possible in many cases, but it is important to
understand the factors that influence the level of generated
noise so that proper consideration can be given to them when
designing machinery.
It is theoretically possible to reduce sound almost to the
level of ambient noise, but often this becomes quite ex-
pensive and results in compromises on size, weight, accessibility
for maintenance, and convenience in operation. It is desirable,
therefore, to set realistic noise-control goals and to give proper
importance to all the factors involved.
Sometimes it is impossible to reduce sound levels of
high-horsepower, high-speed machinery to acceptably low levels
by internal design changes alone. In these cases, additional
sound reduction will be necessary after the machine has been
built. When additional sound-control measures must be applied,
the first step is to find where the objectionable noise is
coming from and the path it takes to the ear of the listener. The
next step is to decide on how to reduce the sound to the
design level.
Remember, noise may go directly from a vibrating
machine part to the ear of the listener, or it may excite
vibration in some other part, which in turn sends out additional
sound waves. These may actually be amplified by some structural
resonance, set in motion by either airborne sound or by structure-
borne vibration.
Acoustic enclosures of some form, combined with inlet
and discharge silencers, or mufflers, provide a very effective
form of sound control for many types of machinery. A knowledge
of certain fundamental acoustic principles becomes necessary in
order to apply these devices. This involves an understanding
of the various techniques of sound control.
In general, there are four basic methods used to reduce
sound: sound isolation; sound absorption; vibration isolation;
and, vibration damping. These four methods do not overlap. The
most effective sound control, at minimum expense, can be
accomplished if they are understood.
Sound isolation and sound absorption are not the
same; it is extremely important to know the difference between
them, and to understand the difference between the kinds of
materials used for isolation and those used for absorption.
In Part V, the sound-absorbing ability of a surface was
expressed in terms of its absorption coefficient, defined as








the ratio of the energy absorbed by the surface to the energy incident
upon the surface. Materials that have high absorption coefficients
usually have soft, porous surfaces. When sound waves strike these
surfaces, air flows in and out of the minute pores in the material because
of the pressure changes produced by the sound. Frictional forces
convert the sound energy into heat, although the actual amount of
energy is small.
The sound level in a room can be reduced by adding ab-
sorbing material to the walls. Energy is taken out of the sound
waves every time they pass through the material at each reflection.
This reduces the reverberant sound level. However, it does not reduce
the sound that travels directly from the source to a listener's ear, or to a
microphone.
The sound level at a point away from the source also can be
reduced by providing a barrier between that point and the source. This is
what is meant by sound isolation. In general, the loss provided by a
uniform porous material is directly proportional to the thickness of the
material.
In the case of nonporous materials, the loss increases as
the logarithm of the mass, and the two most important factors
determining the sound isolation of a partition, dividing barrier, or
enclosure, are the weight per square foot of the material and its
airtight sealing. Heavy partitions are better than light ones, provided
they are not porous and all cracks and openings are sealed. It must be
kept in mind that most sound-absorbing materials are very ineffective
sound isolators. They have the wrong characteristics. They are
porous, instead of airtight, and are lightweight, instead of heavy.
Sound transmission coefficient is defined as the fraction of
incident sound energy transmitted through a partition, or enclosure.
Sound transmission loss is defined as ten times the logarithm of the
reciprocal of the transmission coefficient, or,

T.L. = 10 log

Sound transmission loss is a measure of sound-isolating ability and is
equal to the number of decibels by which sound energy is reduced in
transmission through a partition or enclosure.
Transmission of sound through a rigid partition, or solid wall,
is accomplished mainly by the forced vibration of the wall; that is, the
partition is forced to vibrate by the pressure variations in the sound
waves.
The transmission loss of a partition increases with its
weight, and also increases with frequency. At very low frequencies, the
stiffness of a wall determines the amount of transmitted sound. The
stiffer the wall, the less transmitted sound.
1
T
18
THINK QUIET
As the frequency increases, the stiffness and the mass of the wall
resonate and a series of resonant frequencies occurs.
If the barrier is very flexible, or limp, the T.L. varies with
weight per square foot and frequency, according to the equation:

T.L. = 20 log W + 20 log F 33

where,
T.L. is transmission loss in decibels,
W is the surface weight in pounds per square foot, and
F is the frequency in Hertz.
This means that the T.L. increases 6 dB each time the weight is
doubled. Actually, limp panels do not exist in practice; they all
have a certain amount of stiffness. The average transmission
loss increases about 4 to 5 dB each time the weight is doubled.
A double wall, or enclosure, separated by an air space, or
by sound absorbing material, with no rigid connection between
the two walls, has a transmission loss greater than that predicted by
the "mass law."
When sound-absorbing material is added in a large room,
or enclosure, it reduces the reverberant sound, as has been said.
The maximum theoretical effect is to reduce the sound level to
that which would exist in a free field. In some instances, this
may amount to as much as 6 to 10 dB in the room.
When sound-absorbing material is placed inside a
smaller machinery enclosure, it reduces the sound outside the en-
closure because it absorbs sound inside and prevents sound buildup.
Energy is absorbed each time the sound waves are reflected by the
enclosure wall and they pass through the sound-absorbing
material.
Damping materials, such as mastics, impregnated felt,
asphalt, lead, and sand, are very effective in reducing vibration,
particularly in highly resonant members, such as panels, webs,
sound-isolating enclosures, and springs. The choice of vibration
damping materials depends upon many factors, including the type
of structure, temperature, and the frequency of vibration.
The damping material should have a stiffness comparable
to the stiffness of the material being damped so it can remove
the maximum amount of vibration energy. It should be at least as
thick as the metal section to which it is applied, and
preferably several times thicker. Most benefit can be obtained
by applying it to relatively thin sections where maximum
bending occurs.
Isolating a vibrating part from the rest of a structure can,
in many instances, substantially reduce airborne sound. This may
be done by means of steel springs, cork, felt, rubber, plastics,
or dense fiber glass.
Steel springs can be calculated quite accurately and can do
an excellent job of vibration isolation. However, they also can
have resonances, and high-frequency vibrations can travel through
them readily, even though they are effectively isolating the
lower frequencies. For this reason, springs usually are used in
conjunction with elastomers or similar materials.
Elastomers, plastics, and materials of this type
have high internal damping and do not perform well below about
15 Hz. This is below the audible range, however, and therefore
does not limit their use in any way for effective sound control.
Small leaks or openings in machinery acoustic en-
closures have an importance that cannot be overemphasized. Figure
19 shows how the overall transmission loss may be determined for
an enclosure made of different materials. The abscissa indicates
the difference in transmission loss of two parts of the composite
wall. The various curves show the percent of the entire partition
having the smaller transmission loss. The left ordinate gives the
decibels that must be subtracted from the higher transmission loss
to obtain the net T.L.
It can be seen that a very small area of low
transmission loss can greatly reduce the effectiveness of an
enclosure. For example, if the enclosure would normally have a
transmission loss of 50 dB, and if its wall dimensions were, say, 5
feet long by 3 feet high, a crack 0.1 inch wide would reduce its T.L.
to about 25 dB.
Effects of absorption, transmission loss, damping, and
vibration isolation are graphically demonstrated in the illustrations
on pages 20 and 21.
Controlling the transmission of noise can take the form of
an acoustic shield or barrier. These may be used to reduce noise
when placed between a source and a receiver. The attenuation is
due to the diffraction of sound waves around the barrier, and the
effect is greater at the higher frequencies than at lower frequencies.
This has been confirmed experimentally, although not much
theoretical work has been done on the subject.
The approximate reduction may be calculated by the
following equation (shown on Figure 20):


where,
H is the height of the barrier, in feet,
R is the distance from source to wall, in feet, and
" is the wave length of sound at the particular frequency.
The equation can be used only when the distance from
the wall to the receiver is much greater than the distance from the
wall to the source, and if the distance from the wall to the
source is greater than the height of the wall. For example, if
the height of the barrier is 40 feet, the distance from the source
Figure 19
Figure 20

19
to the wall is 200 feet, and the wave length of sound is 0.5 foot at a
frequency of 2000 Hz, then,


It would be expected that a high amplitude of
vibration would be induced in a panel, or the wall of a machinery
enclosure, if one or more of the frequency components in a
sound wave matched one or more of the normal modes of
vibration of the panel. It may be somewhat surprising though to
find that for every frequency above a certain critical
frequency, there is a particular angle of incidence at which the
panel vibrates as though it were at resonance.
The reason for this behavior is that above the
critical frequency, the wave length of the bending wave in the
panel ("B) can become equal to the wave length of the sound
wave in air, projected on the panel, and a high degree of
coupling between the panel and the air is achieved. This
condition is called wave coincidence.
When a sound wave is incident on a panel at the
frequency and angle where wave coincidence occurs, the panel
vibrates just as though it were at resonance. Figure 21 shows
how this is brought about. The wave length of the bending wave
in the panel is "B. If the wave length of the sound wave in air is ",
and it impinges on the plate at an angle $
0
, so that "/sin $
0
is
equal to "B, the intensity of the transmitted wave approaches the
intensity of the incident wave. The frequency for which " = "B is
called the critical frequency.
Under this condition, the panel vibrates at an
amplitude almost equal to the amplitude of the air particles in
the incident wave. In turn, the panel radiates a transmitted wave
with almost the same amplitude, also at the coincidence
angle $
0
.

In other words, the panel radiates a wave that is
almost as intense as the exciting wave.
The bending wave in the panel is a traveling one, and is
not the same as the usual type of panel vibration where the
waves are stationary and fixed nodes can be located. Note that
wave coincidence can occur only when the wave length of the
sound in air is less than the wave length of sound in the
plate.

Figure 22 shows critical frequencies (f
c
) for various
materials, plotted as a function of their thickness (h). Note that for
steel or aluminum ! inch thick, the critical frequency is 1000
Hz. Wave coincidence can occur at all frequencies above this. If the
plate thickness is only 0.1 inch, the critical frequency is raised to 5000
Hz.
This indicates that machinery enclosures should be made of
thin material, with as high a weight per square foot as possible to
keep the coincidence frequency high. Then damping material
should be applied. This of course does not agree with the previously
discussed relationships concerning transmission loss, where it was
found that T. L. increases with weight per square foot. It is
obvious that once again, as in many other engineering problems, a
compromise must be made to obtain the optimum design.
Piping noise radiation, in many instances, exceeds noise
radiated from machinery. In almost every case involving large
compressors, pumps, and similar equipment, piping must be treated
along with the machine in order to detect the expected noise reduction.
Remember, if piping noise equals machine noise, complete removal of
the machine noise would reduce the original level by only 3 dB.
Flow noise in pipes increases with the velocity. Increasing
the pipe diameter therefore reduces the noise. However, radiated
noise also increases with the area of radiation, so this step increases
the noise. Effective noise control can be accomplished by proper
lagging of the pipes. As previously stated, this cannot be done by
sound-absorbing material alone. A dense cover or jacket must be
included. A good combination is 2- or 3-inch thick fiber glass of 3- to
6-pound-per-cubic-foot density, with a jacket weighing about 1
pound per square foot. This is equivalent to about #24 gauge steel.
Small-diameter, heavy-wall pipes usually do not require
lagging. In many instances, they do not require any treatment. In
critical applications where they are found to be vibrating, the best
approach is to use vibration-isolation techniques to keep them from
inducing vibration and noise somewhere else in the system.
Pipe noise may be transmitted from a pump or compressor
into the piping system either mechanically or through the fluid, or
both. An investigation or experience usually will point out the correct
solution to such problems. If noise is transmitted mechanically, a
vibration break in the piping can be very effective. When noise is
transmitted through the fluid, the vibration break does not help.
Either vibration isolation of the piping or pipe lagging must be used.
Fi gure 21
Figure 22
Bolt Beranek &Newman, Inc.
20
THINK QUIET




General Radio Corporation




21






22
















Figure 23
THINK QUIET
Part VII-Mufflinggenerated noise
Reactive mufflers, on top of Ingersoll-Rand ECM and CM Crawlair towers, reduce noise and move it
away from operators. Exhaust air is brought up through channels in the towers, then into the chamber-type
discharge muffler. Crawlairs, powered by an I-R portable compressor, operate at a Lincoln, R.I., road job.
Delta P. Inc

23
Silencers are usually divided into two categories-
dissipative and reactive. Actually, all mufflers accomplish
noise reduction by combining both effects, to a certain
extent, so that the distinction is really arbitrary.
Dissipative mufflers have relatively wide-band,
noisereduction characteristics, and are usually applied to
noise control problems associated with continuous noise
spectra, such as fans, centrifugal compressors, jet engines,
and gas turbines. They are also used in cases where a
narrow-band noise predominates, but the frequency varies
because of a wide range of operating speeds.
The acoustic performance of a dissipative-type
muffler, illustrated in Figure 23, depends upon its
dimensions, internal arrangement, and kind of absorptive
material used. It is extremely important to correlate the
acoustic performance with the flow rate through the
muffler. Therefore, the muffler manufacturer must be given
accurate information concerning flow velocity, pressure drop
permitted, gas temperature, etc.
Reactive mufflers have a characteristic acoustic
performance that does not depend to any great extent on the

Figure 24
presence of sound-absorbing material, but utilizes the
reflection characteristics and attenuating properties of
conical connectors, expansion chambers, side branch
resonators, tail pipes, etc., to accomplish sound reduction. Figure
24 shows a typical reactive muffler.
A Helmholtz resonator is a vessel containing a
volume of air that is connected to a noise source such as a piping
system as shown in Figure 25. When a pure tone sound
wave is propagated along the conduit, the air in the vessel
expands and contracts. By proper design of the area, length
of the neck, and volume of the chamber, phase cancellation can
be produced to reduce the tone. This type absorber can be
designed to produce maximum absorption over a very narrow
frequency range, as shown in the figure.
It is possible to combine several Helmholtz
resonators on a piping system so that not only will each






Figure 26

cancel out at its own frequency but one will be out of phase with
the other two, or vice versa. Then, instead of getting a sharp tuning
point, the noise is attenuated over a wider range.
The Quincke tube illustrated in Figure 26 shows that if a
pipe were divided into two sections with one leg a half wavelength
longer than the other, the two sound waves would come together
out of phase. Phase cancellation would occur, provided that the same
amount of flow goes through both sections.
For example, if the speed of sound in a liquid is 4720 fps,
a 100-Hz sound will have a wavelength of 47.2 feet. If the loop
section of the Quincke tube is 47.2 divided by 2, or 23.6, feet longer
than the straight section, then total cancellation of the 100-Hz tone
will occur. Cancellation will also occur at 300 Hz and 500 Hz.
Cancellation is sharp, but slight changes in the speed of
sound, due to changes in temperature and pressure, make the Quincke
tube impractical in most cases.
Figure 25

24





Sound produced by a machine is related, generally
speaking, to the horsepower input to the machine. This does not
imply that in the case of an inefficient machine the excess
horsepower all goes into noise. Nor does it mean that if the
machine had been designed more efficiently, it would have been
less noisy. However, in practically every case, there is a direct
relation between horsepower input and generated noise.
Somewhat less obvious is how the noise varies with the
horsepower. It is not the same for all types of machinery, and the
problem is complicated by the fact that when horsepower input
changes, other parameters change too. If it is assumed that the
percentage of input horsepower going into noise remains constant,
then doubling

the horsepower should double the sound power, and
this would mean an increase of 3 dB. In most cases, this is not true.
Doubling the horsepower of a centrifugal compressor or pump
results in an increase of about 4 to 5 dB in the sound.
The effect of horsepower on generated noise for a par-
ticular class of pumps or compressors may be obtained ex-
perimentally by conducting sound tests under controlled conditions
and developing an empirical equation. With no test data available,
an approximate equation for the increase in sound pressure level is:

Increase in dB = 17 log hp ratio.

High-speed machines are noisier than low-speed machines.
For any particular design, the sound will increase anywhere from
20 log speed ratio to 50 log speed ratio. At lower speeds,
centrifugal compressors will increase about 20 log rpm ratio. At
high speeds, the increase will more nearly equal 50 log rpm
ratio. Pump noise increases in a similar fashion, except not quite
as rapidly at high speed. There, the increase is about 40 log rpm
ratio. These same relations hold when only impeller tip speeds are
considered.
The increase in sound with speed usually applies to the
overall noise and to the component of highest level, usually blades
times revolutions per second. The increase at other frequencies is
not as great, and may be in the order of 10 to 15 times the log of the
speed ratio.
Improving the dynamic balance reduces the noise levels
over the entire spectrum, but the amount of the decrease is
relatively small, unless the dynamic unbalance was very great to
start with. The effect of dynamic balance is more pronounced at
high speed than at low speed, but it is impossible, with present
information, to relate sound pressure level change to ounce-inch
unbalance change.
Structural resonances often are responsible for many
components in the sound spectrum, which are not attributable to
any of the obvious forcing frequencies. These resonances can be
excited by unbalance, by impacting parts, or by sliding or
rubbing contacts. Not infrequently, the vibrating part is internal.
To locate it and correct it can be a major project.
In the design of quiet machinery, it is almost mandatory that no
major structural resonances be coincident with any strong forcing
frequency. Electromechanical shaker tests can usually detect
vibration frequencies and mode shapes when any doubt exists
concerning their presence.
Flow in piping systems can be laminar or turbulent. In
laminar flow, individual particles move along parallel paths in the
direction of motion. In turbulent flow, there is irregular, random
motion.
Factors that determine whether the flow is laminar
or turbulent are pipe diameter D (in feet), fluid density " (in
pounds per cubic foot), absolute viscosity (in pounds mass per
foot-second), and flow velocity V (in feet per second). The
variables are related by a dimensionless quantity, R, called
Reynolds number.



When the Reynolds number is below 1200, the flow
is laminar. When it is more than 2200, the flow is turbulent.
Between these two values, the flow can be either laminar or
turbulent; 2000 is considered the dividing point.
In practically every industrial machinery piping system,
the flow may be considered to be turbulent. This is probably the
most important source of noise in centrifugal compressor and
pump piping systems.
Cavitation in fluid systems can be a major source of
noise. This occurs at flow velocities that are quite critical, but
that are difficult to predict because of the influence of geometry,
temperature, and pressure.
Cavitation is the result of the collapse of vapor bubbles
that have been formed by some unstable static pressure
condition in the fluid system. The most typical case is where a
flow restriction increases the velocity and decreases the static
pressure. When the static pressure falls below the vapor
pressure, bubbles form. As they proceed downstream, past the
restriction, the velocity decreases and the pressure increases.
When pressure increases to a certain point, bubbles collapse to
cause pressure fluctuations.
When cavitation is not present in fluid systems, the flow
noise increases with velocity approximately as:
60 log V
2
/V 1.
This means that if the velocity is doubled, the noise
increases about 18 dB. When cavitation occurs, the noise may
increase about:
120 log V
2
/V
1
.

Sound pressure levels due to turbulence can be quite
high, even in straight pipes. Sharp bends and restrictions in the
pipe create additional turbulence, and more noise. The radius of
bends should be at least five times the pipe diameter.
Abrupt changes in direction caused by small-radius bends or
by rapid changes in pipe diameter can produce both turbulence
noise and cavitation noise. To avoid this, there should be a
smooth transition between diameters.
Concerning valves, almost all will cavitate when the flow
rate is great enough. Cavitation will occur at a reduced flow
rate when the valve is closed partially because of the greater area
restriction. Therefore, throttling valves
THINK QUIET
Part VIII-Components of a machine and design
parameters that affect noise generation
Consideration has been given to various sound-control procedures that
can be applied to a machine to reduce noise after it has been built. It
has been pointed out that in many instances, it is not practicable to
design high-horsepower, high-speed machinery to be as quiet initially
as some sound specifications may now require, and in such cases, ad-
ditional noise control can be applied externally. This eighth part of the
Think Quiet series is directed toward the components of a machine
and the design parameters that affect the noise-generating capacity of
a machine.

25
should be located as far away as possible from a machine that
is to operate quietly, or when a sound test is made on it.
It s often quite difficult to reduce the effects of noisy
valves. In water, sound is attenuated about 0.00003 dB per
foot, so there is very little loss between the noise source and the
point of measurement. Adequate pipe lagging can reduce the noise
considerably. Turbulence noise transmitted through the piping may
be interrupted by installing a vibration break to prevent metal-to-
metal contact.
In most valves, the increase in noise is proportional to the
seventh or eighth power of the velocity. Therefore, the way to
reduce it is to decrease the velocity through the valve. One method
of achieving this is to introduce restrictions in series with the valve
seat so that throttling is gradual.
High-pressure valves in gas piping systems are usually
noisier than valves in fluid systems because the velocities are
higher. The noise increases with flow and with pressure reduction
until the critical pressure ratio is reached. After this, the noise
level increases only with an increase in flow Acoustic velocity is
reached when the critical pressure ratio is approximately 2:1.
Maximum noise is generated because of the turbulence and
sonic shock waves. A series of pressure reductions, each below
the critical ratio, would generate less noise. In low-pressure piping
the highest noise is generated on the downstream side of the valve.
Gear noise is related to design, manufacturing toler-
ances, and operation. The rolling and sliding metal-to-metal
contact of surfaces that are neither perfectly smooth nor
geometrically correct causes both noise and vibration. Accuracy
of tooth form and spacing must be maintained for quiet operation.
If the tooth form is not correct, there will be a small change in
acceleration from one tooth to the next. These speed changes
result in both tangential and radial forces, which in turn can
cause torsional vibration and excitation of other parts. If the tooth
spacing is not correct, there will be rough contact on the next tooth.
The resulting small impacts cause noise.
Friction is one of the major sources of gear noise,
and probably the most important one. At the pitch line, the
radial force suddenly changes direction from pushing to pulling, at
a rate equal to the tooth-passing frequency that is, the number
of teeth times revolutions per second. It should be noted that this
major source of noise is not dependent upon the accuracy of the
gear, and even if the gear were 100 percent perfect, noise still
would be generated by friction.
In sum, gear frequency noise consists of the tooth-passing
frequency and a number of higher harmonics, plus other
components associated with the impacts and structural
resonances.
Ball bearing noise comes from a number of sources:
1 . the fundamental rotational frequency that is generated by
any unbalance or eccentricity;
2. rotation of the train of rolling elements (noise will be
produced by an irregularity in either a rolling element or the
cage);
3. the spinning of the rolling elements (a rough spot or indentation
may strike the inner and outer races alternately);
4. an irregularity on the inner raceway; and,
5. an irregularity on the outer raceway.
The frequencies of all these components can be calculated from the
dimensions of the bearing, the number of rolling elements, and the
speed of rotation.
Sound tests are usually made to select bearings for
quiet operation in cases where bearing noise is likely to be
an important factor. These tests are made on an Anderometer as
the bearing rotates at constant speed. The number of
"anderons" is a measure of the radial displacement, with respect
to angular displacement, and its units are micro-inches per radian
per second. In plain bearings, friction is the major source of noise,
and eccentricity is another factor.
A very important role is played by design and manufacture in
producing quiet bearings, but it must be emphasized that about 50
percent of bearing noise problems are caused by incorrect installation.
This is particularly true in the case of roller-contact bearings.
Alignment is extremely important.
Couplings produce noise due to windage. This can usually
be reduced effectively by installing a sheet metal enclosure or
guard, lined with fiber glass or some other sound-absorbing
material. Misalignment often produces noise and vibration in
addition to the windage noise. .
In compressors and centrifugal pumps, extra care in
producing fine interior finish on the casings to reduce noise is not
justified. Experience has shown that there is no detectable
difference when passages are hand-finished.
With jets, noise increases as the eighth power of the dis-
charge velocity. Therefore, substantial reductions can be obtained
by reducing jet velocities.
Electric motor noise often is a major factor in motor-
driven pump and compressor machinery. There are four major sources
of motor noise.
(I) Windage noise consists of noise generated by the cooling air
stream plus that caused by the fan. The principal cause of windage
noise is turbulence. Siren effects usually are caused by fan blades
rotating close to a number of stationary members. Fan noise can be
reduced by using backward-curved blades. This is particularly
important on 3600-rpm motors, since fan noise varies approximately
as the fifth power of the blade tip velocity.
(2) Rotor slot noise is caused by the rotating open slots and has a
frequency equal to the number of slots times the revolutions per
second. It may be reduced by filling the slots with epoxy or some
other material.
(3) The most important source of noise due to electrical design is
the combination of rotor bars and stator slots. This combination
should be selected to provide minimum noise due to rotor and stator
slot magnetomotive force interaction. Usually, the combination is
selected so that the number of exciting force-pole pairs is as high as
possible, since the motor frame is stiffer in these modes.
(4) Mechanical factors responsible for excessive noise include
misalignment of frame and end bells, which distort the frame and
bearings; dynamic unbalance; noisy bearings; and structural
resonances.
Mounting equipment so that the supports are symmetrical
through the center of gravity can reduce airborne and structure-borne
noise. A machine may be considered to have six degrees of
freedom-three translations and three rotations. When its supports are
unsymmetrical with respect to the center of gravity, certain
translation modes couple to rotational modes.
Improper mounting of equipment can be the cause of
fairly high sound levels. Heavy equipment should be mounted on
adequate foundations, or on massive, rigid parts of a structure. It
should never be mounted on lightweight beams or long, unsupported
spans.
It is better to place machines on separate foundations
rather than on the same foundation, and it is desirable to keep
machinery foundations separate from building floors or walls.
Vibration beats are often caused by machines on the same foundation
running at nearly the same speed.
Sound generated by a vibrating surface depends upon the
velocity of the surface motion and the area of the radiating surface.
In general, an area with one dimension greater than 1/4 wavelength
can effectively radiate sound at the frequency corresponding to that
wavelength in air. Therefore, low-frequency sound radiation is limited
to large surfaces, while any surface of more than a few square
inches can radiate sound at frequencies above 1000 Hz. Surfaces
radiating certain low frequencies can often be divided into smaller
areas, thereby reducing radiating efficiency.

26










In centrifugal pump and compressor systems, the most important
source of noise is turbulence; it has been pointed out previously how
this generates high noise levels in compressors, pumps, piping,
valves, and metering orifices. Although the following article is
concerned only with the centrifugal machine itself, turbulence is
still the major source of noise. What is called turbulence here is
actually a combination of two effects-vortex shedding and
upstream turbulence.
The boundary layer over each blade is turbulent by the
time it reaches the trailing edge. The turbulent layers on the top
and bottom surfaces produce a fluctuation in the lift, and this
fluctuation has a broad frequency spectrum. The application of a
fluctuating force to a fluid generates sound at the same frequency.
Therefore, broad-band noise is radiated. This is called vortex
shedding. For any particular blade design, the vortex shedding
establishes a lower limit to the broad-band noise produced by a
centrifugal compressor. It is difficult to reduce the generated
noise below this level, but it is very easy to exceed it when upstream
turbulence is present.

VON KARMAN VORTICES

Figure 27

If the flow is turbulent when it enters a blade row, lift
fluctuations again are produced, and the greater the turbulence, the
higher the noise. In addition to this, Von Karman vortices will be
shed from the trailing edge of the blade, alternately from
opposite sides of the blade, clockwise and then counterclockwise.
This also generates wide-band noise. Figure 27 shows that in the
case of cylindrical obstacles, the relation of frequency, diameter,
and velocity is given by

In the case of axial-flow compressors, the noise from this
source increases with the blade thickness. This is not necessarily true
with centrifugal compressors. Noise from Von Karman vortex
shedding will be much higher in the case of axial-flow compressors, if
the airflow is such that the vortices leaving one blade are struck by
the following blades.
In both centrifugal and axial compressors, noise due to
turbulence will be produced by any obstruction in the inlet or
discharge that interferes with the airflow. This state ment is also
THINK QUIET
Part IX-Industrial machinery noise-centrifugal machines

AUTHOR'S NOTE

The effect on noise of certain general design and
installation features has been discussed in the preceding articles
in this series. This article and the next will be concerned
with particular factors pertaining to centrifugal and
reciprocating machinery.
It is surprising how little information is available on
these subjects, compared to available data about architectural
acoustics. Furthermore, the needed information is difficult to
obtain, both analytically and experimentally. Small machines
can be taken to anechoic or reverberant rooms for sound tests,
but large machines, which really need sound control more
than the small ones, usually cannot be moved to such rooms.
Furthermore, acoustic conditions in the manufacturers' test
areas, or at the final installation site, where sound measure-
ments must be made, are almost always far from favorable.
Some machines cannot be tested in the manufacturers'
shops under rated full-load conditions because of power
limitations. The ones that can be tested usually cannot be held
on the test stand very long because of shipping schedules. It takes
time to obtain useful comparative data under various operating
conditions, and it is not feasible to vary impeller diameters,
number of stages, operating speeds, flow, gas mole weight,
impeller blade and diffuser vane combinations, and so forth.
A pure analytical approach to all these variables is
almost impossible, and on certain occasions, where analysis
has been tried, the answers have been entirely wrong. For
example, let it be assumed that radiated sound power varies
directly with input power. Then, doubling the input horsepower
should result in a 3 dB increase in the sound pressure level. This
does not work in practice where it is found that doubling the input
horsepower increases the sound pressure level 4 to 5 dB.
An empirical relationship could be developed with an
analytical approach if certain controlled sound tests could be
made to establish a base line. This procedure has worked very
well in practice.
Data presented in this and the next article are the
best available at the moment. It will obviously not enable anyone
to immediately redesign a machine to effect a 20 dB
reduction. That is not the purpose. The objective is, as
stated in the first article of this series, to encourage design
engineers to think about ways to apply acoustic design
principles so that the end products are quieter than they are
now.

27
true for pumps. In fact, improper inlet conditions is one of the
most important factors in centrifugal pump noise generation, and
in the production of pressure pulsations in the fluid.
Bends inside pump and compressor casings, or changes in
flow direction, are additional sources of turbulent noise. That
is, for minimum noise generation, the entire internal flow path
should be aerodynamically designed. This includes:
1. keeping the flow path as large as possible to decrease
velocities, and as short as possible to reduce friction;
2. removing all unnecessary obstructions;
3. maintaining smooth surfaces, but not necessarily hand-finished;
and,
4. providing gradual changes in flow path cross section.
High-speed machines are noisier than low-speed units. For
a particular design, sound will increase with speed anywhere from
20 log N
2
/N
1
to 50 log N
2
/N
1
, where N i s revolutions per minute.
At low speeds, the sound of both compressors and pumps
increases about 20 log N
2
/N
1
. High-speed compressor noise
increases about 50 log N
2
/N
1
, while in high-speed pumps the
increase is about 40 log N
2
/N
1
. This is probably because high
speed for pumps is usually fairly low speed for compressors, and
because compressors are usually more efficient noise generators
than pumps.
For impeller tip speeds, these same relations apply.
However, there is evidence that less noise will be obtained with
large-diameter, slow-speed units than with small-diameter,
high-speed machines, even though the impeller tip speeds are the
same in both cases. There are probably three reasons for this:
l. not all turbulence is produced by the impeller (even though the
tip speeds are the same, the slow-speed machine will have lower
velocities and less turbulence);
2. a slow-speed unit will have larger areas in internal passages
and less restrictions; and,
3. mechanical forces due to unbalance are proportional to the
square of the speed, and therefore will produce less structure-borne
sound and structural resonances when the speed is low.
As for blade-passing frequency, every time a blade passes
a given point, the air or fluid at that point receives an impulse.
Therefore, that point will receive impulses at a frequency equal to
the number of blades times revolutions per second. In axial-flow
compressors, this blade-passing frequency is the major component
in the sound spectrum. In centrifugal compressors, it is usually
strong, but not as pronounced as it is in the case of axial
compressors. In pumps, it is always there, but in many cases it is
not the major component, and often it is even rather difficult to
detect.
Blade-rate frequency is often the major noise component
in both pumps and compressors. This change in terminology is used
to differentiate it from blade-passing frequency, which is equal to
blades times revolutions per second, as indicated.
Blade-rate frequency is calculated as follows:

where,
f is the frequency in Hertz,
N
R
is the number of rotating (impeller) blades,
N
s
is the number of stationary (diffuser) vanes, and
K is the highest common factor of N
R
and N
s
.
For example, let N
R
= 6, N
s
= 8, and the speed equal 6000 rpm.
Then,



Figure 28 shows how various combinations of six and
nine vanes react. Any of the vanes may be considered to
rotate, with any others being stationary. By visualization, it is
obvious that with four impeller blades and six diffuser vanes,
there are twelve times in each revolution when impeller blades
line up with diffuser vanes, and that each time this happens, two
impeller vanes match two diffuser vanes. Therefore, the
frequency will be 12 times rps and the pulses will be of double
strength.
When there are six impeller vanes and nine diffuser
vanes (Figure 28), there are eighteen times when impeller vanes
and diffuser vanes line up. The equation is

Each pulse is three times as strong as for a single blade because
three rotating vanes match three stationary vanes each time. The
frequency is 18 times rps. It is obvious that combinations like
twelve and twelve are not recommended because of the many
points of coincidence and the strength of the pulses.
In the case of five rotating vanes and nine stationary
vanes, there are 45 points of coincidence, but each time only
one rotating vane lines up with one stationary vane. This
combination produces a frequency of 45 times rps.
Therefore, it can be concluded that (1) it is better to
use unequal numbers of rotating and stationary vanes and (2)
prime numbers are the best to use because they have no
common factor. It should be noted that it is easier to control
high-frequency noise than low-frequency noise, and that, in the
case of prime numbers, each pulse is only a single strength
one.
Impeller vibration can be induced by steady force. When
an impeller runs at one of its "disk critical speeds," it can


As the impeller vanes (color) rotate, they come into
coincidence with the stationary vanes (black). In this set
of diagrams, there are six rotating and nine stationary vanes.
The first coincidence is reached in the second illustration
at point A, as well as two other points, and is followed
by coincidence at point B, and again at two other
points, and so forth throughout the rotation. The
denominator in the equation in the text shows the number
of coincidences, in this case of six and nine vanes, to be
three. The same holds true for other combinations of rotating
and stationary vanes.
Figure 28


28

Centrifugal compressors are used for hydrocracker service in Martinez,
Calif. The first- and second-stage recycle units are Ingersoll-Rand Type
MTGB-224 and MTGB-324 compressors.



THINK QUIET
vibrate at large amplitudes. This usually cannot be measured outside the
machine because the vibration is balanced with respect to the impeller.
The sound corresponding to that mode of vibration may be detected, how-
ever. This type of vibration can be caused by a steady force on the
impeller, and a vibration force is not needed to excite it.
Usually, noise is not the most important consideration in these
cases, and noise control is not the solution. The impeller can fail in a
very short time, and although a discrete tone may be produced, there may
not be enough time even to isolate it. The solution is to change the
vibration frequency of the impeller, or move the operating speed away
from the disk critical speed.
Decreasing the work per stage may help reduce noise
generated by centrifugal pumps and compressors. This means, of course,
increasing the number of stages. According to theory, the sound is
reduced by 25 log N, where N is the number of stages.
Blade loading is a consideration. The relation between generated
noise and the number of impeller blades is not definite. If the number of
blades is very low, and they are narrow, then increasing the number will
decrease the noise. The number of audible higher harmonics will be
reduced. If the number of blades is doubled, the overall noise should
decrease about 3 dB, that is, by 10 log N
2
/ N
1
. On the other hand,
there is evidence that high solidity blade rows have greater surface area
and generate more noise.
No experimental work has been done in this area to determine
the effect on noise of varying the solidity, and there are very little data
published on the subject.
Concerning impeller-diffuser distance, there is substantial
evidence that increasing the radial distance between impeller blades and
diffusers reduces noise. It is particularly effective in reducing the blade-
rate frequency component. There are very little data to predict which
decreases more rapidly, noise or performance, as the radial spacing is
increased. This would be valuable information to have.
Regarding unequal spacing of impeller blades,
certain theories have been offered. Supposedly, unequal
spacing would suppress the generation of blade-passing or
bladerate frequencies. Unfortunately, they cannot be spaced
with enough inequality to be really effective, and other
problems are encountered, such as upsetting the dynamic
balance and the cost of manufacture.
Molecular weight of the gas in compressor systems
has a pronounced effect on the generated noise. Very little
test data are available about this, and mathematical relation
cannot be stated with present information. However, there are
enough data to show that more noise is produced with high-mole-
weight gas than with low. The effect is something that must be
considered when predicting the noise of centrifugal compressors
because the difference in sound is quite appreciable.
Mass flow and discharge pressure both have a profound
effect on the noise of a compressor. As the mass flow is
reduced, the noise decreases until a point near surge is
reached. Beyond this point, the noise increases rapidly.
Pump noise is less at the best efficiency point than at
minimum flow.
The effects of mass flow and pressure should be
evaluated so that a calculation procedure can be developed
to estimate and predict noise when sound guarantees must be
made.
Inlet and discharge to and from centrifugal pumps
and compressors, as experience has shown, have an extremely
important part in the noise-generating ability of the
machines, and in the pressure pulsations produced by them.
The noise generated by inlet and discharge piping was
discussed in a previous article, but the importance of
bringing the fluid to the machine through a properly designed
inlet system cannot be overemphasized. Nevertheless, the
actual machine inlet is equally, if not more, important.
It is sometimes requested that the inlet area and
discharge area be increased to reduce fluid velocities and
thereby reduce noise. If the suction opening is increased, it may
actually create more noise rather than reduce it. In order to
obtain proper entrance, the fluid should be as near as possible
to the impeller center line. This, of course, indicates a small
inlet. When the fluid enters near the center line on its way to the
impeller vane, there will be relatively low shock and turbulence.
If the suction opening is enlarged, the fluid is admitted farther
up on the impeller vane where the linear speed is higher. The
sudden change from low velocity in the large inlet to high
velocity part way up on the vane causes shock, turbulence,
and increased noise.
Reducing the flow velocity in the discharge piping is an
effective way to reduce piping noise. However, if the discharge
opening is increased by simply relocating it with respect to the
cutwater, then it has not really changed anything. It is
usually more satisfactory to accomplish the same thing by
using a properly designed pipe increaser to connect the pump or
compressor discharge to the discharge piping.
Turning vanes are of great importance in piping
systems. Their purpose is to maintain correct flow conditions in
going around bends. It must be emphasized that they are
often a source of self-generated noise. They can introduce
additional turbulence in the system, and they can vibrate at
their own resonant frequencies and radiate them as pure
tones.
When turning vanes are installed, the cord length and
the separation between the vanes should not be uniform. They
should be heavy enough to resist vibration, and they should not be
identical in construction.

29
Noise produced by reciprocating machinery is usually a multiple of
the piston movement. The noise is generated by both
aerodynamic, or fluid sources, and by mechanical sources.
Inlet and exhaust noises are major sources. The pul-
sating flow can be silenced effectively by a chamber-type
silencer or a snubber. Inlet noise is usually the second highest
noise source, and is exceeded only by that of the exhaust.
Engine exhaust noise usually occurs at the fundamental engine
firing frequency. It is usually 8 to 10 dB higher than the inlet
noise. Both inlet and exhaust noise increase with load, and
are affected considerably by the size of valves, timing, and
the way the ports are constructed. Inlet noise is also affected
by the construction of the exhaust system.
Combustion noise is predominant after the inlet and ex-
haust have been silenced. It is dependent upon the pressure
rise in the cylinders. The theory of how noise is produced by
the combustion process is not very well understood. The
actual pressure rise does not seem to be as important as the
rapidity of the rise, and it is known that combustion noise can
be varied by as much as 10 dB by changing the form of the
cylinder pressure rise. This is an appreciable change in
loudness, and indicates that investigations along these lines
could prove worthwhile.
The cylinder pressure wave form contains quite a
few higher harmonics in addition to the fundamental
frequency. The low-frequency components of the cylinder
pressure wave are determined mainly by the peak pressure.
The higher harmonics are affected more by the shape of the pres-
sure diagram.
Pressure curves of diesel engines differ from those
of gasoline engines, particularly from 800 to about 3000 Hz.
The difference is due mainly to the ignition process. In the
gasoline engine, combustion is initiated by a spark, and
the flame propagates comparatively slowly. In diesel engines,
the ignition is rapid, causing a rapid rise in pressure and a
rather broad peak from about 800 to 2000 Hz.
The spectrum of gasoline engines is different. The
components in the 800- to 2000-Hz range are usually lower in
amplitude and the greatest peaks are in the 400- to 600Hz
range. While the sound levels change with load, the shape of
the spectrum, that is, the components present in the total
noise, shows a marked difference between diesel and gasoline
engines. Diesel engine spectra change very little from no
load to full load. There is a considerable change in the
shape of gasoline engine sound spectra.
Strong mechanical noise, in many instances, masks the
combustion noise. These mechanical noise sources
include:
1. crank-connecting rod;
2. fuel injection system;
3. valves;
4. excessive clearances in crossheads;
5. inadequate lubrication;
6. misalignment;
7. bent parts;
8. excessively tight packing;
9. loose parts; and,
10. improper mounting.
In general, the combined effect of all the sources increases
engine noise in proportion to about ten times the logarithm of
the horsepower. That is, the increase is about 3 dB per
doubling of horsepower. The noise increases about 30 times
the logarithm of the speed ratio, that is, about 9 dB per
doubling of speed.
Inertia forces are a major exciting force causing
vibration and noise in reciprocating machinery. These forces
are due to the motion of the pistons and related parts, and the
imbalance of the connecting rod and crank mechanism. The
forces produced by unbalanced masses also appear in the rotating
parts of the machine, as both static and dynamic unbalance.
Impacts in the crank-connecting rod system,
particularly the knocking of the pistons against the cylinder liners
during crossover, have been found to be the most important noise
source in reciprocating engines. This source of noise is also a major
one in reciprocating compressors and pumps.
During each revolution of the crankshaft, the piston shifts from one
side to the other, several times, moving in the plane of the connecting
rod motion. The gap between the piston and the cylinder liner
permits the piston to move with a certain velocity in the transverse
direction, impacting against the wall of the cylinder. These knocks
produce an intense vibration of the cylinder walls at their resonant
frequency, resulting in a noise level of 3 to 9 dB higher than any
other internal noise.
Factors that determine the noise level of the impacts are:
1. speed;
2. the amount of clearance in the bearings;
3. weight of the piston and connecting rod;
4. the force acting on the piston;
5. the ratio of the crank radius to the length of the
connecting rod;
6. the material in the cylinder block and cylinder head;
7. the thickness of the cylinder block;
8. the number of cylinders; and,
9. the viscosity of the lubricant.
The mechanical noise produced by piston impacts can, therefore,
be reduced by reducing clearances, reducing piston and connecting
rod weights, and by reducing speed.
Torsional vibration can cause problems. Naturally,
machinery is designed to operate at speeds removed from any
lateral or torsional critical speeds. Certain engines, though, may
operate fairly close to a torsional critical speed. This can produce
additional bearing impacts and additional noise, usually of a low
frequency.
The fuel system of engines can be an additional source of
noise. The injection process, impacts on cams, vibration of
camshafts, impacts of valves on their seats, and driving gear
tooth contact noise all combine in this source. Usually this
noise is considerably less than the piston impact noise.

Reciprocating compressors, two Ingersoll-Rand. 4000-hp.
Type 8-HHE-4 units, supply air for in situ fire flood oil recovery
from a 2900-foot depth in Battrum, Sask.
THINK QUIET
Part X-Industrial machinery noisereciprocating machines

30
THINK QUIET
Part XI-Sound-control recommendations






This series of articles began with a discussion of the
properties of sound waves. Various definitions were given,
instrumentation for measuring and analyzing sound were explained and
compared, and sound measurement technique was explained. This was
followed by a review of various types of criteria for evaluating
sound levels, including hearing damage criteria, the Walsh-Healey
regulation, speech interference levels, and annoyance criteria. Noise
criteria curves, ISO noise-rating numbers, and NEMA standards
were compared. Also discussed were the environmental conditions
that affect the sound produced by machines and other noise sources.
Part VI discussed sound-control techniques that can be
applied to machinery after it has been built. Acoustic enclosures
were explained, along with various types of dissipative and reactive
type mufflers and silencers. The difference between sound absorption
and sound isolation was explained and illustrated. Transmission loss,
mass law, vibration isolation, and vibration damping were reviewed.
This was followed by a general discussion of the components of a
machine and the design parameters that affect the noisegenerating
capacity of a machine.
The last two parts discussed particular factors pertaining to
centrifugal and reciprocating machinery. This concluding article lists
some of the recommendations for machinery noise reduction.
-G. M. Diehl



31
SUMMARY OF RECOMMENDATIONS
(1) Reduce horsepower. Noise is proportional to horsepower,
therefore the machine should be matched to the job. Excess
horsepower means excess noise.
(2) Reduce speed. Slow-speed machinery is quieter than high-speed
equipment.
(3) Keep impeller tip speeds low. It is better to keep the revolutions
per minute low and the impeller diameter large than to keep the
revolutions per minute high and the impeller diameter small, even
though the tip speeds are the same.
(4) Improve dynamic balance. This decreases rotating forces,
structure-borne sound, and the excitation of structural resonances.
(5) Reduce the ratio of rotating masses to fixed masses.
(6) Reduce mechanical run-out of shafts. This improves the initial
static and dynamic balance.
(7) Avoid structural resonances. These are often responsible for
many unidentified components in the radiated sound. In addition to
being excited by sinusoidal forcing frequencies, they can be excited
by impacting parts and by sliding and rubbing contacts.
(8) Eliminate or reduce impacts. Either reduce the mass of
impacting parts or their striking velocities.
(9) Reduce peak acceleration. Reduce the rate of change of velocity
of moving parts by utilizing the maximum time possible to produce
the required velocity change, and by keeping the acceleration as
nearly constant as possible over the available time period.
(10) Improve lubrication. Inadequate lubrication is often the cause
of bearing noise, structure-borne noise due to friction, and the
excitation of structural resonances.
(11) Maintain closer tolerances and clearances in bearings and
moving parts.
(12) Install bearings correctly. Improper installation accounts for
about half of bearing noise problems.
(13) Improve alignment. Improper alignment is a major source of
noise and vibration.
(14) Use center-of-gravity mounting whenever feasible. When
supports are symmetrical with respect to the center of gravity,
translational modes of vibration do not couple to rotational modes.
(15) Maintain adequate separation between operating speeds and
lateral and torsional resonant speeds.
(16) Consider the shape of impeller vanes from an acoustic
standpoint. Some configurations are noisier than others.
(17) Keep the distance between impeller vanes and cutwater, or
diffuser, vanes as large as possible. Close spacing is a major source of
noise.
(18) Select combinations of rotating and stationary vanes that are not
likely to excite strong vibration and noise.
(19) Design turning vanes properly. They are a source of self-generated
noise.
(20) Keep areas of inlet passages as large as possible and their length as
short as possible.
(21) Remove or keep at a minimum any obstructions, bends, or abrupt
changes in fluid passages.
(22) Pay special attention to inlet design. This is extremely important in
noise generation.
(23) Pay attention to discharge design, although the inlet is more
important than the discharge, from an acoustic standpoint.
(24) Maintain gradual, not abrupt, transition from one area to the next in
all fluid passages.
(25) Reduce flow velocities in passages, pipes, and the like. Noise can be
reduced substantially by reducing flow velocities.
(26) Reduce jet velocities. Jet noise is proportional to the eighth power of
the velocity.
(27) Reduce large radiating areas. Surfaces radiating certain frequencies
can often be divided into smaller areas with less radiating efficiency.
(28) Disconnect possible sound-radiating parts from other vibrating parts
by installing vibration breaks to eliminate metal-to-metal contact.
(29) Provide openings, or air leaks, in large radiating areas so that air can
move through them. This reduces pressure buildup and decreases
radiated noise.
(30) Reduce clearances, piston weights, and connecting rod weights in
reciprocating machinery to reduce piston impacts.
(31) Apply additional sound-control devices, such as inlet and discharge
silencers and acoustic enclosures.
(32) When acoustic enclosures are used, make sure that all openings are
sealed properly.
(33) Install machinery on adequate mountings and foundations to reduce
structure-borne sound and vibration.
(34) Take advantage of all directivity effects whenever possible by
directing inlet and discharge openings away from listeners or critical
areas.
(35) Use quiet components, such as motors, gears, and auxiliary
equipment, purchased to meet your company's sound specifications.

32














ABOUT THE AUTHOR



George M. Diehl, P.E., is manager of the Sound & Vibration Section of
Ingersoll-Rand Research, Inc., Princeton, N.J. He joined Ingersoll-Rand in 1942
and has become one of industry's most knowledgeable men in the area of sound
control.
A member of several International Standards Organization committees,
Diehl is chairman of the American National Standards Institute working group
for the Measurement Of Noise From Compressors, Pneumatic Tools, &
Pneumatic Machines. In addition, he is a member of the Compressed Air & Gas
Institute International Standards Coordinating Committee and Technical
Committee, American Petroleum Institute Task Force On Mechanical Equipment
Sound Control, Acoustical Society Of America, New York City Mayor's Sub-
Council On The Environment Noise Task Force, and the ASME National
Committee On The Measurement Of Industrial Sound. He also represents CAGI
on the PNEUROP (European Committee Of Manufacturers Of Compressed
Air Equipment) committee on sound.
Diehl is a graduate of Lafayette College, holding a bachelor of science
degree in electrical engineering. He has been a frequent contributor to Compressed
Air, and numerous other society and technical journals.
Born in Easton, Pa., Diehl makes his home in Phillipsburn. N.J. He is
married and has three children.

33
I f you would like to receive complimentary
copies of Compressed Air monthly, through
the courtesy of Ingersoll-Rand Company, you
will be joining some 100,000 management a
and supervisory readers ... readers who are
finding the magazine a valuable tool in
their work. For your convenience, use the
cards on the flap.

34