TURBOMACHINES
Wolfgang Neise and Ulf Michel
DLR Internal Report 2231494/B5 (1994)
German Aerospace Center (DLR)
Institute of Propulsion Technology, Department of Engine Acoustics
Previous affiliation:
Institute of Fluid Mechanics, Department of Turbulence Research
MllerBreslauStr. 8
10623 Berlin,
Germany
Contents
1 INTRODUCTION
1.1 GENERAL REMARKS
1.2
11
1.2.1
FAN TYPES
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
11
1.2.2
14
1.2.3
17
20
1.2.4
1.3
INDUSTRIAL COMPRESSORS
........................
20
1.4
AIRCRAFT ENGINES. . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
21
1.5
PROPELLERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
24
1.6
HELICOPTERS
.................................
26
1.7
BIBLIOGRAPHY OF CHAPTER 1 . . . . . . . . . . . . . . . . . . . . . .
26
31
2.1
INTRODUCTION .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
31
2.2
LIGHTHILL EQUATION . . . . . . . . . . . . . . . . . . . . . . . . . . ..
32
2.2.1
32
2.2.2
34
2.2.3
35
2.2.4
.....
37
37
2.3.1
KIRCHHOFF INTEGRAL. . . . . . . . . . . . . . . . . . . . . . ..
37
2.3.2
INTEGRAL OF CURLE. . . . . . . . . . . . . . . . . . . . . . . ..
38
2.3.3
FARFIELD SOLUTION. . . . . . . . . . . . . . . . . . . . . . . ..
39
2.3
2.4
2.5
40
2.4.1
40
2.4.2
AUTOCORRELATION FUNCTION
41
2.4.3
POWERSPECTRAL DENSITY . . . . . .
41
BIBLIOGRAPHY OF CHAPTER 2 . . . . . . . .
43
.................
45
45
46
49
50
51
53
56
57
59
59
60
63
63
64
64
64
64
67
67
68
68
68
71
71
72
73
73
75
77
79
80
81
81
81
82
5.3.3
5.4
5.5
NOISE
GENERATION
BY
ROTOR/STATOR
ROTOR/CASING INTERACTION . . . . . . . . . . . . .
AND
85
5.3.4
IMPULSIVE NOISE . . . . . . . . . . . . . . . . . . . . . . . . . . .
86
5.3.5
ROTATING STALL . . . . . . . . . . . . . . . . . . . . . . . . . . .
86
5.3.6
87
5.3.7
89
91
........................ .
91
5.4.1
GENERAL REMARKS
5.4.2
............ .
92
5.4.3
92
5.4.4
.................... .
93
5.4.5
94
5.4.6
96
QUADRUPOLE NOISE . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
98
5.5.1
RANDOM NOISE . . . . . . . . . . . . . . . . . . . . . . . . . . . .
98
5.5.2
DISCRETE NOISE. . . . . . . . . . . . . . . . . . . . . . . . . . ..
101
5.6
CONCLUSIONS
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.7
BIBLIOGRAPHY OF CHAPTER 5
. . . . . . . . . . . . . . . . . . . . . . 103
DUCT ACOUSTICS
107
6.1
INTRODUCTION
107
6.2
6.3
110
6.3.2
112
6.3.3
114
6.3.4
115
...
6.4
SOUND PROPAGATION IN HARDWALLED CYLINDRICAL OR ANNULAR DUCTS IN THE ABSENCE OF TEMPERATURE GRADIENTS AND
MEAN FLOW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6.5
121
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
122
6.6
CONCLUSIONS
6.7
BIBLIOGRAPHY OF CHAPTER 6
3
.............. .
123
145
145
146
148
153
9 PROPELLERS
9.1 INTRODUCTION
9.2 EFFECT OF HELICAL BLADETIP MACH NUMBER .. .
9.3 EFFECTS OF INFLOW CONDITIONS . . . . . . . . . . . .
9.4 EFFECTS OF NONUNIFORM BLADE DISTRIBUTION . . . . . . .
9.5 EFFECTS OF COUNTERROTATION . . . . . . . . . . . . . . . . .
9.6 NOISE REDUCTION MEASURES. . . . . . . . . . . . . . . . . . . .
9.7 BIBLIOGRAPHY OF CHAPTER 9 . . . . . . . . . . . . . . . . . . .
157
157
158
159
161
165
165
167
169
169
170
170
171
172
174
174
178
182
185
.. . . . . . . . . . . . . . . . . . . . . . . . . . . .
185
187
189
190
......
..............................
196
196
199
201
201
....................
201
203
203
206
. . . . . . . . . . . . . . . . . . . . 207
211
214
12.8 CONCLUSIONS
..............................
215
216
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
221
222
13.2 CASING MODIFICATIONS TO REDUCE THE STRENGTH OF THE INTERACTION FORCES BETWEEN IMPELLER FLOW AND THE CUTOFF222
13.2.1 INCREASING THE CUTOFF CLEARANCE. . . . . . . . . . ..
222
.. . . . . . . . . . . . . . . . . . . . . . 226
234
234
235
235
13.4.1 TRANSITION MESHES AT THE INNER AND OUTER CIRCUMFERENCE OF THE BLADE ROW . . . . . . . . . . . . . . . . . . . 235
13.4.2 ANNULAR CLEARANCE BETWEEN FAN INLET NOZZLE AND
IMPELLER MOUTH. . . . . . . . . . . . . . . . . . . . . . . . . . . 237
13.4.3 INFLUENCE OF THE BLADE NUMBER. . . . . . . . . . . . . . . 238
13.4.4 ROTATING DIFFUSER. . . . . . . . . . . . . . . . . . . . . . . . . 238
13.4.5 AIRFOIL BLADES. . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
13.4.6 COMPARISON OF DIFFERENT IMPELLER DESIGNS
. . . . . . 240
243
................................
247
247
273
273
274
275
277
281
293
294
283
288
290
294
294
294
297
297
298
307
307
307
308
308
310
310
312
19 BIBLIOGRAPHY
Chapter 1
INTRODUCTION
Contents of Chapter 1
1 INTRODUCTION
1.1 GENERAL REMARKS
1.2 INDUSTRIAL, VENTILATING AND AIR CONDITIONING FANS.
1.2.1 FAN TYPES . . . . . . . . . . . . . . . . . . . . . . . .
1.2.2 BASIC FLUID MECHANICS OF FANS . . . . . . . . .
1.2.3 AFFINITY LAWS AND NONDIMENSIONAL FAN PERFORMANCE PARAMETERS . . . . .
1.2.4 TYPICAL FAN NOISE SPECTRA
1.3 INDUSTRIAL COMPRESSORS
1.4 AIRCRAFT ENGINES.
1.5 PROPELLERS . . . . . . . . . .
1.6 HELICOPTERS . . . . . . . . .
1.7 BIBLIOGRAPHY OF CHAPTER 1
1.1
9
11
11
14
17
20
20
21
24
26
26
GENERAL REMARKS
In turbomachines the working fluid is compressed and moved by the dynamic action of
the rotating blades of one or several impellers. The rotating impeller transfers mechanical
energy from the fan shaft to the working fluid in the form of increased velocity and/or
pressure. There is a wide variety of machines that fall under the above definition: fans,
blowers, compressors, turbines, pumps, propellers, wind energy converters, helicopter rotors,
etc. Only turbomachines handling gaseous media are treated in this course, which excludes
all sorts of pumps and ship propulsors. Of the various kinds of compressors, only turbocompressors are treated in this chapter, because both the mechanism of energy transfer from
the machine to the working fluid and the mechanism of noise generation in reciprocating
compressors, roots compressors or rotary compressors are entirely different from those in
turbomachines.
The field of applications of fans and compressors handling gases or vapors is extremely
wide, and in nearly all cases the noise emitted by these machines is of considerable annoyance.
9
The dominant part of the total noise emitted by these machines is the aerodynamically
generated noise. Only this noise contributor is treated here, other noise sources like gear
noise, bearing noise, vibrational noise and noise of drive motors are excluded.
Turbomachines are usually categorized with respect to the mean flow direction of the
working fluid. The main types are axialflow machines, radialflow (centrifugal) machines,
mixed flow fans and tangential flow fans. Examples of axial flow machines are the propellers
of airplanes and hovercrafts, helicopter rotors, the fans, turbines and most compressors of
modern aircraft engines, and the fans installed in all kinds of ventilating and air conditioning
systems and industrial plants. The dimensions of industrial axial flow fans range from a
few centimeters in diameter, for example the cooling fans in electronic equipment, over
the medium size fans used for ventilation purposes, over windtunnel fans with impeller
diameters of the order of 10m, up to machines as big as several ten meters which operate in
the coolingtowers of electric power plants. Rotor diameters of as much as 100 m are in use
for wind energy converters.
The range of sizes of radial flow machines is not quite as big as that of axial flow machines.
Small fan units are in household appliances such as vacuumcleaners or electric heaters,
medium sizes provide airflow in ventilating and airconditioning duct systems, and impellers
up to several meters in diameter are employed in the mining industry, steel making plants and
chemical industry. Radial compressors operate in small aircraft turbo engines, refrigerating
equipment, compressed air installations, automotive vehicles (turbo chargers) and again in
various industrial plants. Radial turbines are used for turbo chargers.
Although the basic mechanisms of aerodynamic sound generation are the same for axial
and radial flow machines, by far more detailed knowledge has been established for the axial
type, mainly because of its wide application in aviation and the demand for quieter aircrafts
which has given the research work on turbomachinery noise problems a tremendous impact.
It is the aim of this course to provide some fundamental understanding of the basic aerodynamic sound generation by rotating blades and to give some insight into the theoretical
background relevant to turbomachinery noise rather than to present detailed solutions and
results for particular applications. After the presentation of the basic acoustic theory developed by Lighthill (1952) in chapter 2 and its application to jet noise in chapter 3 , the
influence of solid surfaces on the sound generation is derived in chapter 4. This leads to
a discussion of the sound generation mechanisms relevant to turbomachinery in chapter 5.
An important aspect is the generation of acoustic modes by subsonic and supersonic rotors,
by the interaction between rotor blades and stator vanes, and by the interaction between
adjacent rotors. Sound propagation effects are described in the chapters on duct acoustics
and free field radiation of open rotors. The next three chapters 8, 9 and10 are concerned
with propeller, helicopter, and windturbine noise. Methods of noise control for axial and
centrifugal flow machines are discussed on the basis of the sound generation mechanisms
in chapters 12 and 13. The remainder of the course is devoted to the noise of industrial,
ventilating and air conditioning fans covering effects of acoustic loading, installation effects
on fan noise, fan sound power measurement procedures, acoustic similarity laws, fan sound
power estimation, and fan selection with respect to noise.
In this first chapter, some general aerodynamic and acoustic characteristics of various
turbomachines are discussed.
10
1.2
There is no definite distinction between compressors, blowers and fans. Broadly speaking,
fans and blowers are mediumpressure or lowpressure compressors. A boundary between
fans and compressors can be derived from the international standard on fan performance
testing of industrial fans ISO 5801 (1994), where an upper limit for the fan work per unit
mass of 25 kJ jkg is set, corresponding to an increase of fan pressure approximately equal
to 30 kPa for a mean density of 1.2 kgjm 3 . This is equivalent to a ratio of inlet to outlet
pressure of 1.3, assuming standard atmospheric pressure of 100 kPa.
1.2.1
FAN TYPES
Different fan applications require fans with different performance characteristics which are
determined primarily by the design of the rotating impeller and the fan casing. The types of
fans described here are normally used in air conditioning systems, in industrial ventilating
systems and industrial process applications. Typical designs of axial fans, centrifugal fans,
halfaxial fans (mixed flow impeller in axial type casing), half radial fans (mixedflow impeller
in centrifugal type casing), and crossflow fans are schematically shown in Figures 1.1 to 1.7.
Axialflow fans are categorized into the three following major groups: propeller fans (no
casing), axial fans with casing and no guide vanes (tubeaxial fans), and axial fans with
outlet guide vanes (vaneaxial fans). Inlet guide vanes are rarely used nowadays for axial
fans, primarily for noise reasons.
The designs of the two types of mixed flow fans (halfaxial and halfradial fans) lie
somewhere between those of axial and centrifugal fans, and so does their performance which
is the technical reason for these constructions.
Centrifugal impellers are normally operated in volute casings, most often without outlet
guide vanes with the casing spiral being the only means to guide the flow and to convert
kinetic flow energy to potential energy (increased static pressure). Centrifugal fans are
designed with single inlet or double inlet. Depending on the application, various impeller
designs and blade shapes are employed. The best design from an aerodynamic point of
view is that with backward curved airfoil blades with a nicely curved shroud. Radial or
radial ending blades are used for pneumatic transport applications or when large centrifugal
stresses are encountered. Centrifugal fans incorporating a drum type impeller with a large
number of forward curved blades of short radial extent are often called scirocco blowers or
simply blowers.
The crossflow fan is a unique type of fan, operating in a fundamentally different way than
axial or centrifugal flow machines. The impeller is of the drum type with forward curved
blades which are usually simple circular arcs, similar to the centrifugal fan impeller with
forward curved blades. Unlike the scirocco blower, however, the crossflow impeller is closed
at both ends, and the working fluid enters the impeller perpendicular to its axis of rotation
over its full axial length. The fluid passes through the blade row in the radially inward
direction, through the interior, then through the blade row a second time, this time in the
radially outward direction, to discharge on the other side of the impeller. For this reason, the
crossflow fan is, in principle, a twostage fan. The flow inside the impeller is characterized
by the formation of an eccentric line vortex parallel to the rotor axis which rotates in the
same direction as the impeller. The presence of this vortex is of utmost importance for the
11
0)
Support
b)
Motor support
c)
Impeller
Guide vanes
Tail fairing
Figure 1.1: Axialflow fans a) propeller fan; b) vaneless axial fan (tubeaxial fan); c) axial
fan with outlet guide vanes (vaneaxial fan).
operation of the crossflow fan. Other names of this fan type are tangential fan, lineflow
fan, and transverseflow fan.
Typical performance characteristics of axial and centrifugal fans, all of D = 600mrn
impeller diameter, are shown in Figure 1.8. The impeller speed is n = 3000/min for the axial
and the centrifugal fan with backward curved blades and n
600/min for the scirocco blower.
12
Impeller
Tail fairing
Halfradial fan
The operating points of maximum efficiency are marked by circles. Generally speaking, axialflow fans are characterized by large flow rates and moderate fan pressures, while the inverse
applies to centrifugal fans. The examples shown in Figure 1.8 were chosen purposely to
illustrate this general behavior, i.e., a high pressure centrifugal fan design and a low pressure
axial fan. If the performance characteristics of a low or medium pressure centrifugal fan and
a high pressure axial fan were compared, the difference would not be quite as drastic.
Because of their weak structural design, scirocco blowers and crossflow fans are normally
run at much lower tip speeds than the other fan types. Due to their blade shapes, however,
they handle large flow rates against relatively high back pressures. On the other hand, while
the peak efficiencies of well designed axial fans and centrifugal fans with backward curved
blades are in the range 80 to 90%, the optimum efficiency of scirocco blowers lies between
60 and 70%; the values reported for crossflow fans are scattered in the range 30 to 70%,
obviously depending on both the fan design and the test method.
13
Single inlet
Double inlet
~ Discharge nozzle
+..::.
Diffuser
Impeller
Volute
Impeller backplate
1.2.2
The transfer of mechanical energy from the impeller to the working fluid results in a change of
the thermodynamic steady state values of the medium when flowing from the inlet (subscript
i) to the discharge (subscript d). Assuming reversible processes and a compressible working
fluid, the fan work per unit mass is governed by the following equation (see Baehr (1962)):
d
Yt
dps
,. p
1 2
2
+ (Cd
 Ci) + g(Zd  Zi),
2
14
(1.1 )
Sheet metal
blades
Airfoil blades
!'.
@!)
a)
b)
@@G
c)
e)
d)
Figure 1.6: Blade shapes of centrifugal impellers; a) backward curved blades; b) backward
inclined blades; c) radial ending blades; d) radial blades; e) forward curved blades.
rvlotor
II


     f
'
"
~~~
.<
Velocity profile
at outlet
15
7000
Po
6000
Cl
5000
<J
4000 3000
2000
Scirocco fan
0:: 600 mm
n:: 1000/min
U::32m/s
~J~
1000
00
p :: 1.2 kg/m3
Centrifugal fan
b.c. blades
0:: 600 mm
n:: 3000/min
U 95 m/s
y;t = Psd
po
+ Cd
2
+ gZd  (PSi
po
+  + gZi ) .
Ci
(1.2)
In ISO 5801 (1994) the effect of compressibility is accounted for by means of a "Mach
Factor". In equation (1.2) the fan work per unit mass for the case of a reversible process
is expressed as the energy difference of the mass flow between inlet and outlet. It is also
customary in fan engineering to describe the transfer of mechanical energy in terms of the
fan work per unit volume (1m 3 ) instead of unit mass (lkg); this is the increase of the total
flow pressure between inlet and outlet:
C~
+ pOZdg + p02) 
(Psi
c;
+ pOZig + p02)'
(1.3)
The influence of the geodetical height is normally neglected when gaseous media are
considered. Moreover the difference in heights between fan inlet and outlet is normally
small. For this case one has
(1.4)
The aerodynamic power of a fan is equal to the product of volume flow Q and total
pressure difference !:::'Pt :
(1.5)
The total efficiency of a fan is given by the ratio of aerodynamic power to the mechanical
power transmitted by the impeller.
(1.6)
16
The impeller shaft power is transmitted in the form of torque Timp times angular shaft
speed Wimp' The impeller torque is balanced by the torque exerted by the working fluid,
which is caused by the deflection of the mean flow between intake (subscript 1) and outlet
of the impeller (subscript 2). The condition of angular momentum conservation leads to
the following expression for the impeller torque:
(1. 7)
where Tl and T2 are the radii of the blade leading and trailing edges, respectively, and Cl u
and C2u the circumferential components of the absolute velocities at Tl and T2.
In most practical cases the flow entering the impeller is without swirl (Cl u = 0); for this
case the velocity triangles of axial and centrifugal blade rows are sketched in Figures 1.9 and
1.10.
Outlet
guide
vanes'
Impeller
~
u
Figure 1.9: Velocity triangles for the inlet and outlet of an axial fan for the case of swirlfree
intake flow (adapted from Schreck (1961)).
1.2.3
The affinity laws are based on the geometric similarity of the members of one fan family,
i.e., fans of different size but identical dimensional design, and on the kinematic similarity
of the interior flow fields at different impeller speeds and sizes. In this case, the following
relationships apply:
(1.8)
(1.9)
where D, n, and U are impeller diameter, speed and peripheral velocity (tip speed). Based on
the affinity laws, the following nondimensional fan performance parameters can be defined:
17
Forward curved
Radial ending
w~
1
\\
Figure 1.10: Velocity triangles for the inlet and outlet of a centrifugal fan for the case of
swirlfree intake flow (adapted from Schreck (1961)).
'P
'ljJt
'ljJs
,\
7r
4Q
fl
ffi .
2!:lpt
uz
total pressure coefficient
po
2!:lps.
ffi .
statIc pressure coe Clent
poU
'P . 'ljJt
.
  power coeffiCIent
'lJt
2
(1.10)
(1.11)
(1.12)
(1.13)
Note that different definitions of the flow coefficient are sometimes used for centrifugal
fans ('Pcentr
Q/7rbDU; b impeller width) and crossflow fans ('Pc!! = Q/bDU). Typical
dimensionless fan performance curves are shown in Figure 1.11, where the flow coefficient is
defined as in equation (1.10), except for the crossflow fan which is defined as above. In this
example the performances of axial and centrifugal fans differ not as much as in Figure 1.8.
The affinity laws do not provide an exact description of the behavior of fans of different
size and speed, because of the influence of viscosity and compressibility on the aerodynamic
performance. These effects are accounted for by two additional parameters,
UD
Re =   Reynolds number
1/
18
(1.14)
_  0 
4~
c
OJ
0   
u
.~
4_
4
ill
u
ill
'
::J
(f)
(f)
OJ
'
0..
~~.71.~2.~3.~4.~5.~6~
Flow coefficient
++Axial fan,
DGIf, airfoil blades
++Axial fan,
OGV, airfoil blades
If<Axial fan, no GV, airfoil blades
"*Axial fan,
OGIf, airfoil blades
Gt>Centr. fan, b.c. airfoil blades
'i'VCentr. fan, b.c. airfoil blades
~Centr. fan,
b. C. blades (f lat)
~Centr. fan,
flat radial blades
<1  <:>Sc irrocco blower (f.c. blades)
G8Mixedflow fan (halfradial)
G
oCrossflow fan,
n
1/min
3000
3000
2925
2943
1820
1600
1600
1800
700
1000
1320
m
0.45
0.45
0.60
0.50
0.60
0.51
0.51
0.51
0.50
0.45
0.10
m/s
70.7
70.7
91.9
77 .0
57.2
42.7
42.7
48.0
18.3
23.5
6.9
z
24
12
8
12
12
12
12
15
38
8
36
Ma = 
Mach number
ao
(1.15)
where v is the kinematic viscosity of the fluid, and ao is the speed of sound. In reality
the nondimensional fan performance curves are weak functions of these two parameters,
'ljJ = f(r.p,Re,Ma), and 'TI = f(r.p,Re,Ma). For the present purpose, however, the influence of
Reynolds number and Mach number is disregarded.
The total efficiency was already defined in equation (1.6), and the static efficiency can
be defined accordingly:
'TIs =
(1.16)
Pimp
In addition to the dimensionless flow rate, fan pressure, and power, the following parameters are useful to characterize the performance of different fan types:
r.pl/2
(J"
(1.17)
'lj;1/4
(1.18)
The speed coefficient indicates how much faster a fan has to run than a fictitious reference
1 and 'IjJ = 1, and the diameter coefficient indicates how much
fan characterized by cp
bigger the impeller diameter is than that of the reference fan. Speed coefficient and diameter
coefficient are usually defined only for the best efficiency points of each fan. Their usefulness
in characterizing the performance of different fan designs is demonstrated best in the socalled
Cordier diagram. Cordier (1953) showed that in a plot of specific diameter versus specific
speed coefficient, the data for all fans of good aerodynamic design (efficiency) collapse in
single curve with reasonable scatter, see Figure l.12. Here it is obvious that centrifugal fans
are characterized by small values of the speed coefficient and large values of the diameter
coefficient, which is equivalent to high pressure and low flow rates, as was stated before.
Similarly for the other fan types. Also shown in Figure l.12 are the ranges of total fan
pressure, flow coefficient and total fan efficiency, each at the operating point of maximum
efficiency. This presentation is a valuable help for the initial selection of a fan type for a
given pumping task. A further overview over the performance characteristics of the various
fan designs in terms of their dimensionless performance parameters is given in Figure l.13.
1.2.4
In Figures l.14 to l.19 are shown examples of the onethird octave band sound power spectra
of most of the fan types discussed above. The spectra were measured at the respective best
efficiency points and represent the sum of inlet and outlet sound power (total sound power
level). A general observation is that the spectral distribution of the broadband noise of axial
fans is characterized by a level maximum at medium frequencies, with a fall off towards higher
and lower frequencies, while the random noise spectra of centrifugal fans of all kinds, of halfradial fans and crossflow fans show a typical maximum at the very low frequency end with a
continuous drop in level with increasing frequency. The blade passage frequency component
is very distinctly visible in the spectra of axial fans and centrifugal fans with backward
curved and radial blades. The spectrum of the scirocco blowers is mainly broadband in
nature with a weak tone component. No published information is available for crossflow
fans, and one has to rely on the assumption that the noise spectrum of this fan type is similar
to that of scirocco blowers, because of the similar blade shapes and numbers. As will be
shown later, see chapter 16 on acoustic similarity laws, the general spectral shape of each
spectrum is shifted with increasing fan speed (towards higher frequencies) and increasing
fan size (towards lower frequencies).
1.3
INDUSTRIAL COMPRESSORS
Axial and radial compressors are used in aircraft engines, compressed air installations, refrigeration systems, industrial process applications, turbochargers, etc. The tip speed of
compressors is typically much larger than that of fans because of the high pressure rise requirements. Typical designs of multistage axial and centrifugal compressors are shown in
Figures l.20 and l.2l. The axial compressor has a set of inlet guide vanes and a series of
stages consisting of impeller blades and stator vanes. Note that the last stage has a centrifugal impeller. Flow rate control is achieved by adjustable guide vanes in the first few
stages. The centrifugal compressor has two intakes and two discharge ducts. On both sides,
the working fluid leaving the first impeller passes a vaned outlet diffuser to decelerate the
flow and thereby increase the pressure before it is guided radially inward to the intake of the
second stage, etc.
Different designs of single stage centrifugal impellers are schematically shown in Figure
20
Lowpressure
Scirocco blowers
(S8)
Tubeaxial
4.0
tV t
0.4
0.2
2.0
1.0
0.1
0.5
ljl
0.05
0.25
0.01
0.9
Ilt 0.7
0.5
20
1.0
0 10
OjO
5
0.5
~~
"~
IW"~9, 9,"~'~5"O51 j
0, /0;:;;; 0.25
Figure 1.12: Cordier diagram (adapted from VDr 3731 Blatt 2 (1990)).
1.22. The two designs on the left have twodimensional impeller blades (2D), i.e., curved in
only one direction, whereas the blades of the others have a 3Dcurvature. The two designs
shown on the right are open impellers (with no shroud) which run at substantially higher
speeds than the shrouded ones. The change of the characteristic performance parameters
with the design are described relative to a reference impeller, which was chosen to be the
second from the left.
1.4
AIRCRAFT ENGINES
The noise emission of industrial compressors is normally of not so much concern because of
the heavy ducting connected to them. The situation is entirely different in the case of the
aircraft engine compressors which radiate sound almost directly into the environment. On
the other hand, while the aeroacoustics of axial fans and compressors have been investigated
21
q;
~
L+
;~
1,0
2 .. 4
2 ....!
2 .. 3
2 .. 3
0,3
O,i5
0,225
0,68
,,>
0,6
0,12
0,657
1,065
0,13
1,0
0,13
0,:J61
2,72
57,1
0,03
1,1
0,033
0,162
5,02
26,6
6,3
0,35 .. 0,6
1,14 .. 1,19
40 .. 95
69 .. 93
\ I
_:1
!0,438.
  "
~
ohj
1,7
107,5
101
_._
'8
J.
3~a_
_._          .
otd
l'
O,ISTa
I'
I,
I
I
0,00185 !
1,1
0,00203
0,04
24,4
0,005 .. 0,02
1,6 .. 3,8
1,0 .. 1,78
250600
~
I
I
0,3
0,5
0,15
0,924
1,535
146
0,3
0,7
0,21
0,715
1,62
113
Figure 1.13: Ranges of the dimensionless fan performance parameters for different fan types
(after Eck (1972)).
thoroughly because of their aeronautical use, hardly any information is available on centrifugal compressor noise. Examples of early, contemporary, and future aircraft engine designs
are sketched in Figures 1.23 to 1.26.
In Figure 1.27 is shown how the noise emission of aircraft engines has changed with the
design concept. As a result of the increased bypass ratio (fan mass flow to core engine mass
flow) the jet velocity has been lowered and with it the noise level as well. It should be
pointed out that the increase in bypass ratio was not made for acoustic considerations but
for reasons of fuel economy. While jet nOIse was the dominating contributor in the early
22
110
100
m
D
90
Axial fan, no guide vanes
0 = 600 mm,
80
Z=8
n = 2925/min
Q = 5.1 m'/s
7031
63
125
250
500
li<
21<
81< A lin
41<
f. rlz
Figure 1.14: Total sound power spectrum of an axial fan with no guide vanes.
120
Axial fan, outlet guide vanes
0= 500 mm
Z = 12
n = 2843/min
)\
Q = 3.3 m'/s
\
110
.,
100
1H\
\
I
I
'""
...
*,*~)I...;r"A,(
",
\
"
.i,
90
..*
)"
80
"' "'"'"
7031
63
125
250
500
1k
f, Hz
2k
4k
8k A Iln
Figure 1.15: Total sound power spectrum of an axial fan with outlet guide vanes.
80
!
J
70
m
D
60 
_J
Z=8
n = 1000/min
Q=1.1/'m'/s
63
125
250
500
lk
f, Hz
21<
~I<
81< A Iln
engine designs, the noise emitted by the fan and compressor have become important with
the reduction in jet speed.
23
100
90
co
C1
,",
80
70
60
3:
63
125
250
500
1k
f, Hz
2k
4k
8k A 1in
Figure 1.17: Total sound power spectrum of a centrifugal fan with backward curved blades.
100
o:J
90
"CO
80
Centrifugal fan, flat radial blades
D = 510 mm
Z = 15
n = 1800/min
O=1.24m'/s
70
6031
63
125
250
500
lk
f, Hz
2k
8k A lin
4k
Figure 1.18: Total sound power spectrum of a centrifugal fan with radial blades.
100
Q
90
<7(>.>$0o:J
80
'<X
\
"CO
,
1>
"'",
0",
~~
"
\
<&
70
60
'(,.0
\
0B",
Scirocco fan
D = 500 mm
Z = 38
n = 700/min
0= 1.8t. m'/s
1&
..,.0'&
Q
\
\
oS>~
5031
63
125
250
500
1K
f, ilz
2k
ilk
8k A
lln
Figure 1.19: Total sound power spectrum of a centrifugal fan with forward curved blades
(scirocco blower).
1.5
PROPELLERS
Typical propeller noise spectra are depicted in Figure 1.28. The levels of the blade tone
fundamental are almost the same for clean inflow and disturbed inflow, However, while the
24
Radial impeller
\,~.~=..,~
,
~~v.T~==========~
Discharge
Intermediate
discharge
Intake
harmonics fall off sharply when the propeller operates under clean flow conditions, high harmonic tone levels are generated by disturbed flow conditions. Note that tone level differences
of more than 30dB exist, and the raise in random noise level is of the order of 20dB. The
influence of the helical tip flow Mach number on the noise emission is shown in Figure 1.29.
25
  Shrouded impellersi
~
Unshrouded
impellers
4.00
~ 3.00 1+.IIt7;;;wZZZ~
..c
2.00
f
O~ lOO
o
0.16
0.12
9
/
~
/
0.08
0.04
/
!
o~
~~~8~ LJu'mlsJcM
i ~~~ 10vuM+"nU"+"'7n,~ t:1
ZO(min)
ZO(mox)
30(norm)
30(mox)
1.6
HELICOPTERS
Helicopters are much needed aircrafts in civil as well as in military life, but they are notorious
for their noise. A typical narrow band spectrum of helicopter noise showing the influences
of the various noise sources involved is shown in Figure 1.30. Contributors to the overall
noise are the main rotor, the tail rotor, the gear box, and the engine noise. The spectrum
shown does not include the impulsive blade slap noise which occurs when the main rotor
blade chops through the wake from the preceding blade.
1.7
BIBLIOGRAPHY OF CHAPTER 1
337340.
26
CombuSl!on
Applications
Air inlel
1950's design
Comet, Caravelle
early B707 & DCS
Business jets
(NB: Concorde uses purejet
wilh atterburning)
1960's design
T urbino
Compressor
NOlzle
~~t<=
Purejet
single shaft
Low Prossure
HIgh Pressure
Comprossor
Compressor
f ..1!xcr
Exhaust velocities
Bypass up
    4 0 0 mls
Bypass
2shaft
Core up
10
10
600 mls
(or mixed
al 500 m/s)
Bypass Flow
=;;:::::::::t_Fanslream up
10
_ _ _ _ __ Core up
Turbofan
2 or 3 shaft
300m/s
10
450m/s
Figure 1.23: Aircraft engine designs from 1950 to 1960 (from Smith (1989)).
,Fan :,
"OGV"
,
Inlet
~ Bypass (fanductl
~airflow
"
..... : .
.... :. ,',
airflow~
:i!:::::l:!:::~::!::+~
, _ _ . Engine
centreline
.       _. _.
0
D

airflow
Rotating blades
Static vanes
Airflow dueling lines
Figure 1.24: Front end of a highbypass turbofan engine (from Smith (1989)).
27
'""
Free power turbine
Gas generator

1\
1m
Figure 1.26: Ducted counterrotating propfan (MTUCRISP; from Heinig, Kennepohl &
Traub (1992)).
ECK,
HECKL, M.
28
o o
" )
5'
10
00
~.or'\o~ 1';:
Unsuppressed
\):
mUltistage
compressors
 %:
Noise
(EPNdB)
Modern
suppressed
turbofilns
15
20
25
Aircraft noise
data corrected
to constant thrust
30
Iy.
olse 'fl o
Or Set
b ;,.::..._ __
Y let mIxing noise
35~~~~~
10
20
40
Bypass ratio
Figure 1.27: Aircraft engine noise as a function of bypass ratio (from Smith (1989)).
110
ill
>
ill
I
.6.f =3.1 Hz
dB
E
::J
.....U
I
90
ill
0.
!J)
ill
80
lf)
'0
z 70
<ill
ill
0.
0
60
l
n.. 50
0.4
0_8
1.2
kHz
1.6
Frequency
Figure 1.28: Propeller noise spectra for clean and disturbed infiow conditions (after Dahlen,
Dobrzynski & IIeller (1988)).
Aachen, Germany DGLR/ AIAA 9202137, Deutsche Gesellschaft fiir Luft und Raumfahrt e.V. (DGLR), Bonn, Germany.
ISO 5801, 1994. Performance testing of fans using standardized airways. International
Standard, in preparation, International Organisation for Standardization.
LEVERTON, J.
W., 1989. Twentyfive years of rotorcraft aeroacuostics: Historical prospective and important issues. J071rnal of S071nd and Vibration 133, 261287.
29
Figure 1.29: Influence of the helical blade tip Mach number on propeller noise (after Dobrzynski et al in Heckl & Muller (1994)).
30
20
CD
10
o
<:
""
20
Transmission noise
<:
;.
Compressor noise
100
1000
Frequency (Hz)
10000
30
Chapter 2
2.1
31
31
32
32
34
35
37
37
37
38
39
40
40
41
41
43
INTRODUCTION
Although the first papers on rotor noise were published at the beginning of this century and
the first propeller noise studies appeared in the thirties (for a historical survey see the review
papers by Morfey (1973b) and Cumpsty (1977)), most of the recent progress in understanding aerodynamic sound generation by turbulent flows and by rotating blades is based on
the acoustic analogy developed by Lighthill (1952), Curle (1955), and Ffowcs Williams &
Hawkings (1969a).
The basics of aeroacoustic sound generation were presented in many papers and short
courses. The following presentation is based on the work of Fuchs & Michalke (1973),
Goldstein (1974), Fuchs (1976), and Michalke & Michel (1979).
31
The Lighthill equation is derived in this chapter and a general solution for the sound
radiation into an unbounded space is presented which includes the effects of stationary
surfaces. These equations are applied to the problem of jet noise in the following chapter 3.
The influence of stationary and moving surfaces is treated in chapter 4.
2.2
2.2.1
LIGHTHILL EQUATION
DERIVATION OF LIGHTHILL EQUATION
The basic equations of aeroacoustics can be derived from the equations for the conservation
of mass and momentum. The conservation of mass is given in differential form and cartesian
space fixed coordinates Xi by
ap
apci _ 0
(2.1 )
at + aXi  ,
where p is the fluid density and
conservation of momentum is
Ci
(2.2)
Here, p is the pressure and Tij is the viscous stress tensor in the fluid.
Many aeroacoustic wave equations have been derived based on these two equations. The
first work was published by Lighthill (1952). He derived an inhomogeneous wave equation
for the description of the aeroacoustic noise generation of a small region of turbulent flow
which is embedded within a homogeneous fluid at rest. The procedure is as follows.
Take the momentum equation (2.2) and add Ci times the continuity equation (2.1). This
yields
apc<
__
t
ap
+ apc<c
__ + __
<
t_J
at
aXj
aXi
aT< <
o.
aXj
(2.3)
(2.4)
and subtract the time derivative of the continuity equation (2.1),
(2.5)
which results in
ap ap
2
at 2
axt
(2.6)
To obtain a wave equation for the pressure one has to add aC; 2a 2p/at2  a 2p/at 2 on both
sides of equation (2.6) which yields
(2.7)
32
Similarly, one obtains a wave equation for the density by adding 8 2p18x;  a682 pi 8x; on
both sides of equation (2.6) which yields
2
28 P
82
2  a08x2 = 8 .a .(pCiCj  Tij)
8t
,
x, X J
82 P
82
+ 8
2(P Xi
a~p).
(2.8)
(2.9)
r:"]~
'J
8t 2
(p~)
a6'
(2.12)
The first term of equation (2.12) is identical to the first term of 82Tij I ox i 8x j on the right
hand side of equation (2.9) when equation (2.10) is evaluated.
For q = 0, equation (2.11) describes the sound wave propagation in a uniform medium
at rest. Within a turbulent flow region, q ::J O. Thus, the right hand side can be interpreted
as a source term which is produced by the turbulent flow, but which is zero, or at least
quadratically small in the fluctuations, outside the region occupied by the turbulent flow.
The situation is described in Figure 2.1.
33
turbulent
flow r,egion
q=O
Figure 2.1: Turbulent flow region as a source of sound surrounded by a fluid at rest.
is the turbulent velocity vector.
Ci(Xi,
t)
~(t)
....
U'I
Figure 2.2: Turbulent flow region as a source of sound surrounded by a fluid in uniform
motion. Ui(Xi, t) = Ci(Xi, t)  Ui is the turbulent velocity vector relative to the ambient flow
velocity, Ui .
2.2.2
The situation of a medium at rest outside the source region is a rare condition in the case
of aerodynamic noise generation. A more realistic assumption is that the fluid flows past
a noise generating region. In such a situation, the source term q in equation (2.11) is not
quadratically small outside the noise generating region. This case is better approximated by
the assumption of a uniform flow speed outside the turbulent flow region as shown in Figure
2.2. The convective form of the Lighthill equation (2.11) can be derived for this situation as
was first shown by Ribner (1959) for the study of flight effects on jet noise.
1 ([)
a6
fJt
[) )
+ Ui [)Xi
[)2p
P
[)x;
(2.13)
= q.
Xi denotes the coordinates in a coordinate system fixed relative to a point within the source
region (e.g., the center of the nozzle in the case of jet noise or a position on the rotor axis of
an axial fan), Ui is the velocity vector of the ambient flow relative to this coordinate system,
and ao is the sound speed in the ambience.
The source term q for equation (2.13) is given by
[)2
( [)
[)t
34
[) )
+ Ui [)Xi
2 (
P
a6P ) '
(2.14)
where Ui = CiUi is the local velocity vector relative to the velocity vector U i in the ambience.
Equation (2.13) has the property that the left hand side describes the sound wave propagation in a medium with uniform velocity Ui and sound speed ao. The right hand side is
produced by the turbulent flow and is zero, or at least quadratically small in the fluctuations,
outside the region occupied by the turbulent flow. It was shown, e.g., by Michalke & Michel
(1979) with the assumption that the entropy is conserved along particle path lines (viscous
stresses Tij = 0, no heat conduction) that equation (2.14) can be approximated by
(2.15)
where
(2.16)
and
qi
= p/~ (Po) .
OXi
(2.17)
pi = P _ Po is the difference between the local pressure p and the ambient pressure po. Terms
of order p'2/ (p o a6) were neglected, which might not be permissible in expansion cells of
supersonic jets or in rotor flows with higher pressure ratios.
The first term on the right hand side of equation (2.15) which is defined by % in equation
(2.16) describes a quadrupole source distribution following the argumentation for the term
Tij of Lighthill (1952).
The second term on the right hand side of equation (2.15) describes a dipole source
distribution which was overlooked by Lighthill and is even rarely considered in more recent
studies of aerodynamic noise generation. It can be concluded from equation (2.17) that this
unsteady density source term qi will only be significant if the local density p differs from the
ambient density po. This unsteady density source term was discussed by Ffowcs Williams
(1969) but it was first shown by Morfey (1973a) that these terms are important for hot jets.
Dipole sources are generally more efficient than quadrupole sources and it will be shown
later in chapter 3, that this dipole source term is an important contribution to jet mixing
noise if the density of the jet is different from the density of the ambient fluid.
2.2.3
Elementary sources, e.g., monopoles, dipoles, and quadrupoles, play important roles in acoustic theory. The sound field of a monopole source can be created by a pulsating sphere as
shown in Figure 2.3. The directivity of a monopole is uniform in all directions. A good
approximation of a monopole source is an oscillating diaphragm in an infinite baffle if the
diameter of the diaphragm is small compared to the acoustic wave length.
The sound field of a dipole source can be created by two monopoles with opposite sign
(the source rate of one monopole is absorbed by the sink rate of the second) whose distance
is very small as shown in Figure 2.4. The dipole strength M = hQ is defined as the monopole
strength Q times the distance h between the two monopoles. The distance must be very
small compared to the acoustic wave length. The directivity of a dipole is proportional to
cos (), where () = 0 is alined with the axis of the two monopoles (derivation in Fuchs &
Michalke (1973)).
The sound field of a quadrupole source is equal to the sound field of two closely spaced
dipoles of opposite strength or four closely spaced monopoles with opposite strength as shown
35
directivity
streamlines
pulsating sphere
diaphragm in infinite
baffle
monopole
Figure 2.3: Sound field of a monopole source created by a pulsating sphere or an oscillating
diaphragm in an infinite baffle.
directivity
streamlines
oscillating sphere
dipole
Figure 2.4: Sound field of a dipole source created by two adjacent monopoles in antiphase
or an oscillating sphere.
longitudinal
quadrupole
lateral
quadrupole
Figure 2.5: Sound field of quadrupole sources created by four adjacent monopoles located in
a linear array (qii,i = 1,2,3) or in a square (qij,i I j).
in Figure 2.5. A longitudinal quadrupole (qii, i = 1,2,3) is formed when the two dipoles or
the four monopoles are located on a line. A lateral quadrupole (qij, i I j) corresponds to
two dipoles located side by side or four monopoles located in a square. The distance between
the sources must be very small compared to the acoustic wave length. The directivity of a
longitudinal monopole is proportional to cos 2 0, where = 0 is the axis of the two dipoles.
The directivity of a lateral dipole is proportional to cos 0 sin e(derivation in Fuchs & Michalke
36
(1973) ).
2.2.4
The basic assumption in Lighthill's analysis is that the source distributions on the right
hand sides of equations (2.9), (2.11), or (2.13) are known. This is generally not the case,
because a turbulent motion is the result of the same equations for the conservation of mass
and momentum that were used for the derivation of the above mentioned wave equations.
The importance of Lighthill's approach comes from its capability for deriving scaling laws
when certain scaling assumptions are made for the turbulent source terms. This capability
is based on the existence of closed form solutions for the Lighthill equation. The first, now
famous scaling law was the result of Lighthill (1954) that the acoustic power of a free jet is
proportional to the eighth power of the jet's speed. It will be shown in the next chapter,
that this is only valid for a jet with constant density because the dipole source terms lead to
a sixth power law which agrees with measured data from hot jets like those of jet engines.
The right hand side q of equation (2.13) on page 34 includes the influence of all deviations
of the mean velocity from the velocity Ui in the ambience (e.g., due to the mean velocity
profile in a jet) and of all deviations of the local speed of sound from the ambient value ao
(e.g. in areas with different temperatures) on sound propagation. Through this mechanism,
the effects of convection and refraction of sound waves in high speed jet flows are considered
as source effects despite being the result of propagation effects. Sources have to be known a
priori while propagation effects can act only on the generated sound waves.
To resolve this contradiction, other acoustic wave equations have been derived for mean
velocity profiles which more resemble the actual mean velocity distributions in the flow.
It was assumed that the right hand sides would better describe the "true" sources in the
flow. One of these equations was derived by Lilley. He assumed that only terms that are
nonlinear in the fluctuations are true source terms. Therefore, all terms that are linear in
the fluctuations were moved to the left hand side of the equation. The final differential
equation has the order three. Unfortunately, solutions for this partial differential equation
can be obtained only numerically. An additional complication originates from the fact that
the equation has nontrivial solutions for the homogeneous equation (when the source term
on the right hand side of the equation, q _ 0). The great advantage of Lighthill's equation
in comparison to these other aeroacoustic wave equations is that closed form solutions exist.
2.3
2.3.1
Supposed the source function q in equation (2.13) is known as function of time and space.
Then, general solutions are available from classical theories. For the case of a source region
surrounded by an infinite space which may contain solid surfaces, aI,ld for Ui = 0 (only
to simplify the following equations for these notes, the external motion will we included
in chapter 3) and for a uniform sound speed ao in the ambience, we obtain the Kirchhoff
integral for the pressure perturbation p'(Xi' t) = p(Xi' t)  Po in a field point Xi
pi (Xi, t)
47r
J1r {[ OYiOYj
[J2%] + [Oqi]}dV()
0Yi
Yi +
37
~
L/s
turbulent
region
source point
field point
J
s
Yi,
field point
Xi
[op] +[p]+1 or
1 or [op]}
dS(Yi).
{1 r
an
r2
an
aor
an at
hi.
(2.18)
The notation is explained in Figure 2.6. The volume integral can be limited to the volume
Va that contains the turbulent flow. r = Iril = IYi  xii is the radiation distance between the
source point Yi and the field point Xi. The brackets indicate that the enclosed terms have
to be evaluated at the retarded time tr = t  r / ao which considers the time needed by the
sound wave to propagate the distance r from the sorce position Yi to the observer position
Xi with a sound speed ao. The integral (2.18) is valid in any position, even within the source
region Va. Later, far field approximations will be presented that require that Xi is far outside
Va.
It can be seen in equation (2.18), that the pressure perturbation pi in the field point Xi
is made up by a volume integral involving the source terms qij and qi which are described
by equations (2.16) and (2.17), and by a surface integral involving the pressure p.
is the gradient in the direction of the outward normal of a surface element. The volume
integral was first studied with respect to aerodynamic noise generation by Lighthill (1952),
the surface integral by Curle (1955).
a/an
2.3.2
INTEGRAL OF CURLE
A more convenient form of the solution (2.18) can be derived according to Curle (1955).
(2.19)
It can be concluded from this equation that the sound field in the field point Xi is created by
a volume distribution of quadrupoles qij, a volume distribution of dipoles qi, and a surface
distribution of dipoles ii, where ii are the forces per unit area acting on the fluid at each
surface element dSi . The brackets indicate again, that the source distributions have to be
38
evaluated at the appropriate retarded times. The integral (2.19) is valid in any observer
position Xi.
It must be noted, here, that the solution for the sound radiation into a duct is different
from this solution into free space. A solution for a circular duct was discussed by Goldstein
(1974).
2.3.3
FARFIELD SOLUTION
In almost all practical applications (e.g., the sound generation by jets, propellers, wind
turbines, and helicopters) one is interested in the sound radiation at large distances from
the source region. This allows two simplifications:
Field points Xi in the acoustic far field (far in comparison to the sound wave lengths):
the space derivatives 0/ OXi and 0 2 / (OXiOXj) in equation (2.19) can be replaced by time
derivatives .
Field points Xi in the geometric far field (far in comparison to the dimensions of the
source region): the factor l/r in equation (2.19) can be considered identical for all
source points Yi and can be determined with the distance r r between a reference position
Yri within the source region and the field point Xi.
With the reference position arbitrarily defined as the origin of the coordinate system,
Yri = 0, equation (2.19) simplifies to
(2.20)
with r = IXil.
This equation can be written in a form that is physically easier to understand. The three
integrals shall be studied, separately:
(2.21 )
The first contribution is described by a volume integral over a quadrupole source distribution
I
1
[02qq]
(2.22)
Pq(Xi, t) = 47ra6r
ot 2
v
where the quadrupole source term is described by
(2.23)
This can be approximated for small pi to
(2.24)
39
Here, U r is the component of the velocity Ui in the direction to the observer point Xi. The
integral (2.22) describes the sound due to those velocity fluctuations in the source region
that are directed toward the observer. The brackets indicate evaluation at the retarded time,
(2.25)
The second term on the right hand side of equation (2.21)
integral over a dipole source distribution,
IS
described by a volume
(2.26)
qd
,a
= P aYr
(po)
P .
(2.27)
a / aYr is the component of the gradient in the direction toward the observer. The integral
(2.26) describes the sound due to the pressure fluctuations in a density gradient toward the
observer.
The third term in equation (2.21) is described by a surface integral over a dipole source
distri bu tion,
pj(x/, t) =
(2.28)
where ir is the component of the surface force per unit area acting on the fluid, ii, in
the direction to the observer. The integral is the result of fluctuations of the surface force
component directed toward the observer.
The volume integrals will be studied in more detail in the next chapter about jet noise.
The influence of surfaces is then studied in chapter 4.
2.4
Many aeroacoustical studies are content with a solution for the sound pressure in the time
domain. However, in acoustics, one is generally not interested in time histories of pressure
fluctuations but in the mean square of the pressure fluctuations and their frequency spectra.
2.4.1
The mean square is independent of the integration boundaries t1 and t2 only if p'(Xi, t)
is stationary random. This means that the integration volumes and integration surfaces in
40
equations (2.22), (2.26), and (2.28) must be stationary or at least periodic (for the cases of
rotors) .
The sound pressure level is then defined by
(2.30)
where
2.4.2
Pref
= 2 . 10 5 Pa in air.
AUTOCORRELATION FUNCTION
The distribution of the meansquare sound pressure in the frequency domain is expressed
by the powerspectral density of the pressure fluctuations which can be derived from the
autocorrelation function Rpp( Xi, T) of the pressure fluctuations in the farfield point Xi
(2.31 )
2.4.3
POWERSPECTRAL DENSITY
The power spectral density Wpp ( Xi, 1) of the pressure fluctuations in the far field point Xi
can be obtained by Fourier transforming equation (2.31). This yields
00
Wpp (Xi, 1)
(2.33)
00
Like the autocorrelation function, the powerspectral density function Wpp(.Ti, 1) consists
of contributions from the three source distributions. Under the condition that these are
uncorrelated, we obtain
(2.34)
The power spectral density Wppqq ( Xi, 1) of the pressure fluctuations in the far field point
Xi due to the quadrupole source distribution is given by the double integral
W ppqq (Xi, 1) =
(47r:a6)2j j WQQqq(Yi,1]i,1)exp(i~R)dV(1]i)dV(Yi)'
v v
41
(2.35)
toward observer
1]i
and
1]r.
J
00
WQQqq(Yi' 1]i, 1) =
=Qq('Yl"t,)=Q,q(,Yl+1]i,

t+T') exp(i21f IT) dT,
(2.36)
00
J
J
v v
(2.39)
. t)
Qd (Yn
= aqd(Yi, t)
at'
(2.40)
(41f:ao)2
JJ
(2.41 )
s s
42
(2.42)
The powerspectral densities Wppqq(Xi' j), Wppdd(Xi, j), and Wppff(Xi, j) include the effects of interference between the sources. The importance of these interference effects will
be shown in the next chapter.
2.5
BIBLIOGRAPHY OF CHAPTER 2
ASME
CURLE, N., 1955. The influence of solid boundaries upon aerodynamic sound. Proceedings
of the Royal Society (London) A 231, 505~514.
FFOWCS WILLIAMS, J. E. & HAWKINGS, D. L., 1969. Sound generated by turbulence and
surfaces in arbitrary motion. Philosophical Transactions of the Royal Society (London)
A 264, 321 ~342.
FFOWCS WILLIAMS, J. E., 1969. Hydrodynamic noise. Annual Review of Fluid Mechanics
1,
197~222.
FUCHS, H. V. & MICHALKE, A., 1973. Introduction to aerodynamic noise theory. Progress
in Aerospace Science 14, 227~297.
FUCHS, H. V., 1976. Basic aerodynamic noise theory. AGARDVKI Lecture Series 80 on
"Aerodynamic Noise".
GOLDSTEIN, M. E., 1974. Unified approach to aerodynamic sound generation in the presence of solid boundaries. Journal of the Acoustical Society of America 56, 497~509.
LIGHTHILL, M. J., 1952. On sound generated aerodynamically. I. General theory. Proc.
Royal Soc. London A 211, 564~587.
LIGHTHILL, M. J., 1954. On sound generated aerodynamically. II. Turbulence as a source
of sound. Proc. Royal Soc. London A 222, 1~32.
MICHALKE, A. & MICHEL, U., 1979. Prediction of jetnoise in flight from static tests.
Journal of Sound and Vibration 67, 341~367.
MORFEY, C. L., 1973a. Amplification of aerodynamic noise by convected flow inhomogeneities. Journal of Sound and Vibration 31, 391~397.
MORFEY, C. L., 1973b. Rotating blades and aerodynamic sound. Journal of Sound and
Vibration 28, 578~617.
RIBNER, H. S., 1959. New theory of jetnoise generation, directionality, and spectra. The
Journal of the Acoustical Society of America 31, 245~246.
43
44
Chapter 3
3.1
45
45
46
49
50
51
53
56
57
59
59
60
63
63
64
64
64
64
INTRODUCTION
The sound generated by a free jet when it mixes with the ambience is generally not important
for turbomachinery noise generation. However, this sound generation process was the first
application of Lighthill's acoustic analogy and it is a remarkable example for the capabilities
of analytical considerations concerning aerodynamic sound generation. In addition, it is an
important contribution to the noise of aircraft jet engines.
The now famous scaling law for the sound power of a jet (though only valid for a jet with
constant density) was derived from equation (2.9). The sound power was expected to be
45
proportional to the eighth power of the jet's speed (Lighthill (1954)). In addition, the theory
explained why the sound pressure level in the far field of a jet is larger in the rear arc (in
the direction of the jet's mean velocity vector) than in the forward arc. This was possible,
despite of the fact that the source term q on the right hand side of the inhomogeneous wave
equation (2.9) was not known. Lighthill only assumed that the sources are convected with
the flow and made certain assumptions on how the sources scale on mean flow quantities.
However, other experimental findings could not be explained with Lighthill's results, e.g.,
the deviations from the eighth power law for low jet speeds and for hot jets, the additional
noise generated by supersonic jets, and the unexpected high sound pressure levels of aircraft
in flight. In addition, the frequencies found experimentally in the rear arc are not higher
than in the forward arc as was expected from the assumption of moving sources and the
frequencies in flight are not reduced as expected from the reduced relative jet speed.
It will be shown in this chapter, that all these experimental findings can be explained
and described with the acoustic analogy provided the correct source model is chosen and the
dipole source term is included.
3.2
We start with the convective form of the Lighthill equation (2.13) which is repeated here,
(3.1 )
and use definition (2.15) for the source function q,
(3.2)
in which the original Lighthill source function EPTij / oxJix j of quadrupole type is replaced
by a quadrupole source function 02%/OXiOXj and a dipole source function Oqi/OXi. qij and
qi are defined by equations (2.16) and (2.17), respectively. It was already concluded in the
previous chapter that the dipole source term qi may be important when density gradients
are present in the jet flow.
Equation (3.1) has the property that the left hand side describes the sound wave propagation in a medium with uniform velocity Ui and sound speed aQ. The right hand side is
produced by the turbulent flow and is zero, or at least quadratically small in the fluctuations,
outside the region occupied by the turbulent flow.
A solution of equation (3.1) for an unbounded field and for the static case Ui =
{Uj, 0, o} =
was already derived as equation (2.18). In jet noise, the surface integrals
playa role for some installation effects, but this influence shall not be considered, here.
The influence of the external velocity vector Ui cannot be neglected when the important
effects of flight speed on jet noise are to be studied. This situation is described in Figure
3.1 in a coordinate system fixed on the nozzle (wind tunnel coordinate system) in which the
flow is stationary random, i.e., all time averaged flow quantities are independent of the start
time of the averaging process.
The solution of equation (3.1) for an unbounded field that includes the effects of Ui
46
uI
y.I
Yi
quadrupole sources
~~vr
dipole sources
This solution for the pressure fluctuations is valid in any position Xi inside or outside the
integration volume V which must include the turbulent flow field of the jet. With qi
0 and
flight speed Uf = 0, the solution corresponds to that discussed by Lighthill (1952) for the
density fluctuations, qij being very similar to Lighthill's stress tensor Tij . The dipole term
qi in the second integral was overlooked by Lighthill but is important for hot jets.
In comparison to equation (2.19), the uniform motion of the external fluid is now considered while the surface integral is neglected. The external motion is considered by replacing
the distance l' in the denominators in the two integrands by the term roD f. D f is the
Doppler factor defined by
D f = 1 MafcosB o
(3.4)
with the flow (or flight) Mach number Ma f = Uf / ao. The emission distance 1'0 and the
emission angle Bo are explained in Figure 3.2. The Figure shows a source point Yi within
the jet flow field and a field (or observer) point Xi in the external stream. l' is the distance
between these two points and B the corresponding geometrical angle relative to the upstream
direction. 1'0 is the distance the wave has to propagate with the speed of sound relative to
the moving medium to reach the observer. Maf 1'0 is the distance by which the wave is
convected downstream by the external flow during the time of propagation, 1'0/ ao. Bo is the
angle normal to the wave fronts of the propagated waves. It is therefore also called wavenormal angle (Morfey (1979)). Here, we use the term emission angle because it is the angle
into which the sound is emitted from the source. The bracketed terms have to be evaluated
at the retarded time ir which is defined by
(3.5)
where t is the time of observation. The choice of emission coordinates (Xi) rather than
observer coordinates (1', B) is causal for the compact form of equation (3.3). The observer
47
source point
      
8o
8
III
(3.6)
and
cos
eo =
cos
e [Jl 
+ Maj.
(3.7)
The near field of jets is of interest when the acoustic loads on structures must be determined which is a problem for some high performance aircraft. Generally, one is more
interested in the sound received in positions far away from the jet. The solution (3.3) can
then be simplified .
Acoustic far field  emission distance "'0 far in comparison to the wave length A: the
space derivatives O/OXi and fJ2/(OXiOXj) can be replaced by time derivatives, e.g.,
aoo / OXi = a/at.
Geometric far field  emission distance "'0 far in comparison to the dimensions Yi within
the source volume: the Doppler factor D j can be considered to be identical for all
source positions and one emission distance "'0 relative to a reference position can be
applied to all source positions except for its influence on retarded times.
It was shown by Michalke & Michel (1979) that the sound pressure in the acoustic and
geometric far field point Xi is given by the following integral over the source volume V:
where
po) P/
P
48
(3.9)
(po)

I
qd = p
fJ 
fJYra
(3.10)
eo
Source position 1
position 2
Observer position
Figure 3.3: Definition of source and observer coordinates in the geometric far field.
Yra and UTa are the components of Yi and Ui, respectively, in the direction of Bo. fJ / fJYra
is the gradient in that direction.
The retarded time for the terms in brackets in equation (3.8) is now defined by
iT = t 
/0
ao
+ ~.
aoDj
(3.1l)
The last term considers the retarded time difference between the source position Yi and the
reference position.
The far field solution (3.8) consists of two integrals. The first is Lighthill's quadrupole
contribution, the second is a dipole contribution. Note, that the exponent of the Doppler
factor D j is different for both integrals which means that flight speed has a larger effect on
the quadrupole noise of a jet (which is the only sound contribution of a jet with constant
density, e.g., many laboratory jets) than on dipole noise (which requires density gradients in
the jet flow, e.g., like in an engine jet). This explains why different flight effects were found
for pure jet engines which have a large dipole contribution and for highbypassratio engines
which have a relatively large quadrupole contribution.
3.3
One is generally interested in time averaged solutions, e.g., the mean square or the powerspectral density of the sound pressure in the far field. Onethird octave spectra or octave
spectra can then be computed from the powerspectral density. The meansquare value is
49
defined by equation (2.29) from which the soundpressure level can be derived with equation
(2.30). The autocorrelation function is the basis for the determination of meansquare values
as well as powerspectral densities.
3.3.1
The farfield solution of equation (3.8) is stationary random in the coordinate system fixed
on the nozzle. The autocorrelation function Rpp (Xi, T) is then defined by
(3.12)
The mean square value of the sound pressure is given for
= 0,
(3.13)
If only the quadrupole term of solution (3.8) is inserted in equation (3.12), we obtain the
autocorrelation due to quadrupole terms,
where the observer position Xi is expressed in terms of emission distance TO and emission
angle eo. With Yli = Yi and Y2i = Yi + T/i, and after exchanging the sequence of integrations
we obtain
R
(x T) =
ppqq~,
16 7r 2 Toa
2 4D6
j
o
JJt
V V
 t1
t1
[8
t
2qq
(Yi
+
T/i, t + T)]
8t2
} dt d . d .
t
T/~
Y~
~v~
Rqq(Yi,7)i,r)
(3.14)
The integrand of the double integral over the source volume is the crosscorrelation
function Rqq (Yi, T/i, T) of the quadrupole source terms between the positions Yi and Yi + T/i
The separation vector T/i between the two source points is illustrated in Figure 3.4.
Equation (3.14) can then be abbreviated as follows:
(3.15)
The double integral over the source volume will be discussed later. Here, we shall discuss
the terms in front of the integral. The autocorrelation function of the sound pressure (as well
as the mean square value) is inversely proportional to the square of the wave normal distance
TO. This is expected because T5 p'2=T5Rpp ( Xi, 0) is independent of TO in the geometric far field
1 Maj cos eo in the
(spherical sound propagation). Notable is the Doppler factor D j
denominator which appears with a power of six. Since D j < 1 in the forward arc (toward
the flight direction) and D j > 1 in the rear arc, flight speed causes a dramatic increase of
the sound pressure in the forward arc and a corresponding decrease in the rear arc. This
explains the strong forward arc amplification of jet noise observed in flight which was not
understood for a long time.
50
Source position 1
Source position 2
Observer position
1]i
and
1]7'0'
(3.17)
Please note that the power of the Doppler factor D j has decreased to four in equation (3.16)
in comparison to six in equation (3.15). This means that the flight effect is larger for the
quadrupole source terms.
The evaluation of equation (3.12) by use of equation (3.8) also results in terms containing
the crosscorrelation functions between quadrupole and dipole terms. These will be neglected
here for simplicity which is equivalent to the assumption that quadrupole and dipole source
terms are not correlated.
The autocorrelation function Rpp( Xi, r) is then given by the sum of only the two equations
(3.15) and (3.16).
Rpp(Xi,r) = Rppqq(xi,r)+RpPdd(xi,r)
(3.18)
This equation demonstrates that the autocorrelation function (as well as the meansquare
value of the sound pressure, see equation (3.13)) in the far field of a jet consists of contributions from quadrupole sources and of dipole sources. The solution is exact, except for
the neglect of the crosscorrelation terms between quadrupole and dipole source terms and
the assumption that viscous stresses and heat conduction do not playa significant role in
aerodynamic sound generation (see page 35 in chapter 2).
3.3.2
POWERSPECTRAL DENSITY
The powerspectral density Wpp(Xi' f) of the pressure fluctuations in the far field point Xi
can be obtained by Fourier transforming the autocorrelation function as defined by equation
51
J
(Xl
(3.21 )
(Xl
eo
as
(3.24)
In addition to the source positions Yi and Yi + 1]i, Wqq depends on the frequency j, emission
and all other parameters that influence the turbulence within the jet.
angle
By introducing the coherence /12(Yi, Yi + 1]i) of Qq between the two source positions Yi
and Yi + 1]i, equation (3.20) can be reformulated as
eo,
W ppqq (Xi, 1) = (4
\ D3)2
7f'rOa O
JJ
JWqq1 W qq2
/12
'.'
~
source strength co erence
,:Xp {i ('l/;q
+ 'l/;R)},
. Y f
source mter erence
vr~'
(3.25)
W qq1 = Wqq(Yi, Yi) is the powerspectral density of the source quantity Qq (defined by equation (3.22)) in source position Yi and W qq2 = Wqq(Yi + 1]i, Yi + 1]i) is the corresponding value
in the second position Yi + 1]i. Both quantities describe the strength of the source quantity
Qq as a function of frequency.
The coherence /12 in equation (3.25) indicates that only the coherent part of the fluctuations in the two positions contributes to the sound in the far field. The coherence function
decays with increasing distance between two source positions as shown in Figure 3.5. The
rate of decay depends on the scale of the turbulence for the frequency in question. The
52
... TJ1
Figure 3.5: Decay of coherence function /'12 between two source positions as a function of
axial separation 'T/1 between the source points. The decay is gradual for large scale turbulence
and rapid for small scale turbulence.
integration over the coordinate 'T/i in equation (3.25) need only be carried out over the region
in which the coherence /'12 > O. This region is termed coherence volume in equation (3.25).
The source interference term in equation (3.25) describes the phase relationship between
the contributions from different source positions. The phase 'l/Jq considers the effect of the
propagation of the fluctuations in the flow. These fluctuations are governed by the equations
of motion which result in instability waves, whose phase variation in the axial 'T/1direction
is given by
(3.26)
'T/1 is the axial component of the separation vector 'T/i and Up(j, Yi, 'T/i) is the phase velocity
of the considered frequency component. 'l/Jq may also be a function of the lateral separations
'T/2 and 'T/3
The phase 'l/Jr considers the influence of the difference of the retarded times between the
two source positions and is defined by equation (3.23).
An equation corresponding to equation (3.25) can be derived for the contribution of the
dipole source term
(3.27)
It is given by
Equation (3.19) together with equations (3.25) and (3.28) is an almost exact solution of the
Lighthill equation. (See the end of the previous subsection for the limitations).
3.3.3
A typical polar directivity of the sound pressure level Lp of a static jet is shown in Figure 3.6.
It shows increased levels in the rear arc with a dimple close to the jet axis. The increased
53
Lp(Gol
Figure 3.6: Polar directivity of the sound pressure level of a jet with a subsonic jet exit speed
Uj The sound pressure level is smaller for 00 < 90 deg (so called forward arc) and larger for
00 > 90 deg (so called rear arc). A dimple is observed close to the jet axis (0 0 = 180 deg)
which is often attributed to sound refraction effects but may also be caused by helical modes
within the jet's turbulence (Michalke (1972)).
sound emission in the rear arc (0 0 > 90 deg) was attributed to source convection in early
papers on jet noise (e.g. by Lighthill (1954), and Ffowcs Williams (1963)). However, a
motion of the sources would also require that the frequencies of jet mixing noise observed
in the far field are higher in the rear than in the forward arc according to the Doppler shift.
This was not found experimentally.
The reason for the discrepancy between these theories of convected sources and the
experimental results is that the condition of stationary randomness of the source distributions
under the source integral (3.20) is violated if the coordinate system is convected with the
sources.
It was shown by Michalke (1970), Michalke (1972), and Michalke (1977), that a wave
model can be used to describe the sound generation of a jet. The coordinate system remains
stationary relative to the nozzle like in the work of Ribner (1959). The sources are modeled
by propagating waves with growing and decaying amplitudes with limited spatial coherence.
These waves within the jet flow are caused by the inherent instability of the shear layer of a
jet. The frequencies in the far field are independent of emission angle 00 in this model which
agrees with the experimental findings. The differerence between eddy convection and wave
model concerning the frequency is shown in Figure 3.7 for a normalized jet speed Uj / ao = 1.2
under the assumption that the waves propagate within the jet with a phase speed of 0.7 Uj
In terms of the wave model, the directivity can be caused by a directivity of the powerspectral densities Wqq or Wdd of the source terms (under the square root of equations (3.25)
and (3.28)), by a directivity of the coherence {12, or by source interference effects based on
the exponential terms exp {i ('l/Jq + 'l/Jr)} or exp {i ('l/Jd + 'l/Jr)}. The source interference effects
are sufficient to explain the experimental findings as shall be demonstrated. If the radial
extension of the source volume is neglected for simplicity, the phase angle 'I/J = 'l/Jq + 'l/Jr in
the interference function of equation (3.25) is given for a static jet by
'I/J
'l/Jq
+ 'l/Jr =
o
27[" f 'rll ( a
ao
0.7 Uj
+ cos 00 )
(3.29)
where a phase speed of Up = 0.7 Uj was used. Phase speeds of this order are observed in
isothermal circular jets. The term in the parenthesis is largest for 00 = 0 which is in the
54
OL~J~L~L~
30
60
90
120
Emission angle reI. inlet
150
180
Figure 3.7: Influence of emission angle on frequency of emitted sound. Wave model of
Michalke (1970), Michalke (1972), and Michalke (1977) (       ), eddy model of
Lighthill (1954), Ribner (1959), and Ffowcs Williams (1963) (~).
forward arc (opposite the jet flow direction) and smallest in the rear arc for eo = 180 deg.
The effect of the cosine term is small for small jet speeds Uj and large for large jet speeds.
The product of /12 exp {i ('ljJq + 'ljJr)} determines the effects of source interference in equation (3.25). This is plotted as a function of dimensionless axial separation
= 2nI'l1/ao
in Figure 3.8 for the two angles eo = 30deg (left side) and eo = 150deg (right side) and two
normalized jet speeds of Uj/ao = 0.5 (top) and Uj/ao = 1.2 (bottom). The chosen coherence
Ll,
OJ
OJ
o
c
o
c
~ 0.5
~ 0.5
OJ
.:s
OJ
OJ
a5 0.5 .
~ 0.5
.c
o
.c
03
OJ
1~~~~~
5
10
Normalized axial separation
15
_1L~~~~
5
10
Normalized axial separation
15
OJ
= 1.2, Theta =
03 0.5
Uj/aO
Integral = 0.02
J ~ ~:::~,~:;"'"
30 deg
't:
OJ
0:
..
150deg
OJ
OJ
ai 0.5
a5 0.5 .
OJ
OJ
o
c
Q;
.c
o
Integral = 0.02
OJ
Q;
.c
1L~~~
5
10
Normalized axial separation
15
_1L~~~
5
10
Normalized axial separation
15
Figure 3.8: Demonstration of source interference effects for the sound radiation of a static
jet in the upstream and the downstream directions.
/12 = exp( 0.05 Lle) determines the envelope of the curves. The effect of interference is an
oscillation of the product. The oscillation is smaller for high jet speeds in the rear arc. The
55
integral of the curve is given in each plot. It is a measure of the radiation effectiveness of the
source field. While the integral value is almost unchanged between forward and rear arcs for
Uj / ao = 0.5 it is very large for Uj / ao = 1.2. It is obvious that the integral in equation (3.25)
is strongly influenced by the oscillation.
3.4
The calculation of the sound pressure in the far field of a jet according to equation (3.8)
requires that the source functions defined by equations (3.9) and (3.10) are known as a
function of time in the whole source region. This can only be realized numerically. However, a
threedimensional simulation of the whole jet flow field for the actual flow Reynolds numbers
and Mach numbers is required which surpasses the computational capabilities available today.
Since instability mechanisms seem to playa dominant role in jet noise generation, it might
be sufficient to perform large eddy simulations.
If time averaged solutions are sufficient, one needs the powerspectral densities Wqq or
Wdd of the source terms for all emission angles
in the whole source region Yi, the coherence
/'12 between the source terms in different source positions for the whole source region Yi and
for all separation vectors 'f/i, and the phase angles 7/Jq and 7/Jd' The required data can be
measured in principal although this would be a tremendous task.
eo
The acoustic analogy can be utilized to derive scaling laws which eliminate these numerical and experimental problems. For this purpose, it is only necessary to know, how the flow
field scales with the change of basic flow parameters.
For circular jets it is assumed here that these basic parameters are the fully expanded jet
diameter D j (which may differ from the nozzle diameter when the jet is supersonic), the jet
density Pj, and relative jet speed !:lU = Uj  Ufo In the derivation of the flight effects scaling
laws of Michalke & Michel (1979), Michel & Michalke (1981a), Michel & Michalke (1981 b)
the influence of the external flow speed Uf on the mean and fluctuating flow field of the jet
was considered through a stretching factor (J t 1. It was assumed that the normalized flow
conditions in a static jet at a position Y1 downstream of the nozzle are more or less identical
to the normalized values in position (J Y1 in flight. The situation is depicted in Figure 3.9.
UJS      . 
Uf
tU. _. __ . J
'     L =0 L s     I
Figure 3.9: Influence of flight speed on the source and coherence length of a jet.
56
The quadrupole source term of equation (3.9) would then scale in normalized form as
follows,
(3.30)
where qqs is the normalized source term in the static jet. The tilde indicates normalization
with the basic flow parameters. a is defined by
(3.31)
where
(3.32)
is the normalized phase velocity of the flow disturbance which is often assumed to be close
to up
0.7 in a circular and isothermal jet. This assumption was experimentally shown to
be satisfied surprisingly well (Michel & Michalke (1981a)). Accordingly, the lengths of the
two integration volumes V in equation (3.25) and (3.28) are increased by this factor a.
A similar influence was found experimentally for the influence of flight speed on the
non dimensional frequency J of the fluctuations in the flow field in positions stationary relative
to the nozzle. It is generally called Strouhal number,
(3.33)
The scaling law for the frequency is given by
(3.34)
where Sis = fsDjs/U s is the Strouhal number for the static jet.
The stretching model according to these equations was also verified by a stability analyses
of Michalke & Hermann (1982).
3.5
A scaling law for the power spectral density of the sound pressure in the acoustic and
geometric far field can now be derived by utilizing the scaling laws for the flow field.
The terms in equation (3.25) and (3.28) are normalized with the above defined basic flow
quantities. This yields
(3.35)
where K, is the ratio of the specific heat capacities. The tilde indicates normalization with
respect to the fully expanded jet diameter D j , jet density Pj, and relative jet speed 6,U =
U j  Uj . The Gqq terms are defined by double integrals over the source region
Gqq =
:2 vJvJV
57
(3.36)
where Yi = y;jD j , iii = TI;jD j , and dV = dV/D]. The factor 1/0 2 in front of the integral
compensates the effect of increased source and coherence volumes in flight. For the derivation of scaling laws, it is assumed that Ggg is not explicitely influenced by the basic flow
parameters D j , Pj, LiU Uj  Uj , and flight speed Uj.
Ggg is the result of equation (3.36), evaluated for the crossspectral density of the
quadrupole source function

[j2
{2
po
Qg = at 2 uro Pj 
(3.37)
and G dd is evaluated for the crossspectral density of the dipole source function
at
(3.38)
The exponential term in the integrand of equation (3.37) describes the source interference.
The phase difference 'l/Jq in equation (3.36) is given by
(3.39)
where up is defined by equation (3.32) and the Strouhal number St as nondimensional frequency by equation (3.33). The phase 'l/Jr in equation (3.36) describes the retarded time
differences and is defined by
(3.40)
Tli, TIro' TIl, and TIn are shown in Figure 3.4. TIl and TIro are the components of the separation
vector Tli in the negative flight direction and the wave normal direction, respectively. TIn IS
perpendicular to TIl, both vectors combining to TIro'
While the noise of a hot jet might be dominated by dipole noise in the rear arc, it rnay
be dominated by quadrupole noise in the forward arc for high subsonic flight Mach
numbers .
In addition to the Doppler factor, the relation between quadrupole and dipole contributions depends on the ratio of pJl Po and on the relative jet speed t:"Ulao.
It can also be derived from equation (3.35) that the ratio of the two noise contributions
remains constant if the normalized effective jet speed
(3.41)
is kept constant. This also means that the ratio of the two terms for a jet in flight can be
expressed by a static jet with a different jet speed or a jet with a different flight speed, as long
as Ue according to equation (3.41) is identical. This conclusion was the basis of the flight
effects scaling method proposed by Michalke & Michel (1979). Ue is equal to the relative jet
speed !::.U for eo = 90 degrees and is larger in the forward arc (toward the flight direction)
and smaller in the rear arc.
The quadrupole and dipole contributions can be identified in Figure 3.10 in which the
sound intensity generated by model jets with different ratios 1j ITo between total jet temperature and ambient temperature are shown as a function of normalized jet speed Uj lao. The
dipole contribution with a (Uj lao)6 power law can be seen for small normalized flow speeds
in all cases. The range of dipole dominance is Uj I ao < 0.4 for the cold jet and Uj I ao < 0.6
for the hottest jet. Apparently, the density of the cold jet was not identical to the density
of the ambient air in these tests. A decreased gradient can also be observed for the hot jet
for Uj I ao > 1 which may already be an source interference effect.
3.6
3.6.1
SUPERSONIC JETS
MACH WAVE RADIATION
+ 'ljJr
'ljJ = 27r St
{~
up
iiI!!L
+ 6..U
+ at:,,~
[ih cos eo
(3.42)
where ijn is the component of ijro perpendicular to the jet axis and iiI is the component
parallel to the jet axis (see Figure 3.4).
Maximum sound emission according to equation (3.36) is achieved for the smallest possible values of 'ljJ = 'ljJq + 'ljJr. If the lateral separation component ijn in equation (3.42) is
neglected (like in equation (3.29)), 'ljJ vanishes for all axial separations iiI and all Strouhal
numbers St if the term within braces vanishes. This occurs for any flight Mach number Maf
for an emission angle defined by
cos eOM =  ~
U
1
6..U'
(3.43)
P ao
This condition leads to the socalled Mach wave radiation which can only occur if the value
of the right hand side of equation (3.43) is smaller than 1. This is the case for
t:"U
ao
>
~ ~
up
59
1.4.
(3.44)
0.2
D.
0.3
0.5
0,6
0,7
0,8
0,9
1.0
1.5
2.0
3.0
SYMBOLS:
130
120
+,.... +
100
J'
90
.!
80
70
I
+.,::'l'1i~7ir;~
:.:.'.~:!::::
___
.1.
"1
.. I
T
.. r,
1 Ci'""'
60 1_ _
.1...  .  ; . . .   )  ,~.~~.,.,
0.2
0.3
0.4
0.5
..1
0.6
0.7
0.8
0.9
JET VELOCITY,
.,"
1.0
v/o o
1.5
2.0
3.0
Figure 3.10: Sound pressure levels of model jets as a function of normalized jet speed Uj / ao
(from Lush & Burrin (1972)).
The jet speed of a static jet has to be at least 1.4 times the ambient speed of sound for that
to happen. For a given jet speed Uj , fiU = Uj  Uj is smaller in flight than during static
operation of the engine. This means that the Mach wave radiation angle due to the phase
velocity of the disturbances in the shear layer is closer to 180 degrees relative to the flight
direction in flight or that Mach wave radiation would disappear in flight.
A static jet that has the same Mach wave radiation angle like a jet in flight is given by
fiU
(3.45)
It is assumed here that the normalized phase speed Up is identical in both cases.
3.6.2
The pressure Pn in the exit plane of the nozzle is not equal to the ambient pressure po
if the jet is supersonic and the fully expanded jet Mach number Maj is not equal to the
design Mach number Mad of the nozzle. The mean flow field of the jet is then dominated
by an almost periodic expansion cell structure as shown in Figure 3.11. A first theory for
the mean flow field in the expansion cells in the case of small pressure differences Pn  Po
was developed by Prandtl (1904) for a jet simulated by a cylindrical vortex sheet. Tam,
Jackson & Seiner (1985) presented a multiple scales expansion that included viscous terms.
Their numerical results agree excellently with experimental data of Norum & Seiner (1982).
The effects of flight speed, finite shear layer thickness, and of temperature distribution were
studied theoretically by Michalke (1992).
60
P  Po =
(3.46)
n=1
00
2:: 2::
(3.49)
En(Yi) Em(Yi + "7i)Wqq .
n=O m=O
Introduction of equation (3.48) and use of equation (3.47) yields for the integrand of equation
(3.20):
Wqq =
+ VJ r)]
(3.50)
The number 1 after the opening brace in the second line stands for the undisturbed integrand.
The next two terms are additional contributions that modify jet mixing noise in the far field
in the presence of expansion cells. The modifications are limited because the sign of these
terms changes with position Y1. The two following lines contain an exponential function
of the axial separation 'Tl1 which combine with exp[i(7,Vq + 7,Vr)] in the first line to a new
interference function,
(3.51 )
A kind of Mach wave radiation occurs if conditions exist for which 7,V 0 over a large part of
the source volume. When 'Tln in equation (3.42) is neglected (just like before in the derivation
of equation (3.43)), we have Mach wave radiation for certain frequencies with the minus sign
in above equation. Based on the definition of the Strouhal number, St = i DJi llU, this
occurs for the frequencies
in = CinllU (Up
27f
(3.52)
f  CinUj
n 
27f
up
1 + up!!i cos 00 .
(3.53)
ao
The latter frequency scaling law agrees with the results of HarperBourne & Fisher (1973),
Ffowcs Williams & Kempton (1978), and Tam & Burton (1984). The frequency is Doppler
shifted corresponding to the phase velocity of the disturbances, i.e., the frequency is higher
in the rear arc than in the forward arc. The actual value of the frequency depends on the
normalized phase speed up Uj / ao of the disturbances in the jet shear layer and on the wave
number Ci n . The fundamental frequency is given by the lowest wave number Ci1 for the
description of the shock cell mean flow field according to equation (3.46).
The influence of flight speed in equation (3.52) is quite surprising since the frequency is
Doppler shifted although this equation is valid for the wind tunnel coordinate system.
In order to compute the actual frequency, an equation for the wave numbers Ci n is needed
which are inversely proportional to the cell length. Analytical results, like those of Michalke
(1992) can be a help for this purpose.
The actual level of broadband shock noise is the result of an integration of equation (3.20)
by including the broadband shock noise terms. The power spectral density is proportional
to the coherence /12. The shock noise peak frequencies are independent of /12. However,
these peaks are sharp (like a 8function) for a fully coherent motion and rather smooth for
a less coherent motion within the jet.
The following conclusions can be drawn from equation (3.50):
Broadband shock noise is proportional to the jet mixing noise.
Broadband shock noise increases with the shock expansion cell strength.
Broadband shock noise has the same relation between its quadrupole and dipole contributions as the underlying jet mixing noise provided the modulations of the quadrupole
and dipole source terms are equal.
Broadband shock noise generation is not conditioned on the existence of large scale
structures. Large scale structures cause narrow peaks in the frequency spectrum, small
scales wider peaks.
Shocks in the cells are not required for broadband shock noise.
62
3.6.3
SCREECH
Screech noise is radiated when instability waves are generated at the nozzle lip in a phase
locked loop through flow induced excitation. In a free jet without walls in its vicinity, the
broadband shock noise is the source of the excitation. In this case, the screech frequency is
equal to the radiation frequency of broadband shock noise in the upstream direction toward
the nozzle which can be calculated with equation (3.52) for
= o.
eo
in = a n6.U
(Up
1 + Up
211"
Maj)
(3.54)
ao
ns 
anUj
Up
211" 1 + up!!.i
(3.55)
ao
3.7
In the flyover case, the observer is not stationary relative to the nozzle, but moves with flight
speed U j relative to the nozzle. This is shown in Figure 3.12 The sound field and the sound
........1111   
wind tunnel
flyover
Figure 3.12: Comparison between wind tunnel coordinates and flyover coordinates.
pressure amplitudes are not changed by the motion of the observer. The motion has only an
effect on the frequency observed by the moving observer is. The observed frequency iFO in
the flyover case is Doppler shifted from the frequency iWT in the windtunnel case according
to
iWT
(3.56)
fro = .:1  Maj cos
eo
The sound pressure levels are unchanged in the flyover case as well as the levels in onethirdoctave bands provided the frequencies of the bands are also Doppler shifted according
to equation (3.56).
The powerspectral densities of the sound pressure in the moving coordinate system are
affected, because the frequency shift llas to be compensated by a level change to ensure that
the integral over all frequencies remains constant.
63
3.8
3.8.1
Static jet noise can be predicted with the SAE method (SAE ARP 867C (1985)). The
method incorporates some basic results of the Lighthill theory and uses extensive experimental results for empirical adjustments. The influence of the dipole sources is considered
through an empirical density exponent w which can alternatively be described by quadrupole
and dipole contributions according to Michel & Bottcher (1992). The predictions with the
SAE method are generally considered to be close to reality. The levels for engine jets are
slightly underpredicted.
3.8.2
The flight effects portion of the SAE method does neither consider the frequency increase
according to equation (3.34) nor does it consider the Doppler amplification of the amplitudes
according to equation (3.35). The effect of Doppler amplification is empirically considered
through incorporation of a relative velocity exponent n(Bo), p,2 ex: (6U/ao)n which is a
function of emission angle Bo. Relative velocity exponents were first derived from simulated
flight tests with the Aerotrain (Drevet, Duponchel & Jacques (1977)) which were carried
out with a pure jet engine with a strong dipole noise contribution. Aircraft engine makers
use their own relative velocity exponents for each class of engines. The prediction scheme is
tuned for the flight speeds expected during takeoff and landing and fails for high subsonic
flight Mach numbers.
A prediction scheme based on equation (3.35) and adjusted to sound emission data measured during flight tests up to flight Mach numbers Maj = 0.9, was proposed by Michel &
Bottcher (1992).
3.9
BIBLIOGRAPHY OF CHAPTER 3
DREVET, P., DUPONCHEL, J. P. & JACQUES, J. R., 1977. The effect of flight on jet noise
as observed on the Bertin Aerotrain. Journal of Sound and Vibration 54, 17320l.
FFOWCS WILLIAMS, J. E. & KEMPTON, A. J., 1978. The noise from the largescale
structure of a jet. Journal of Fluid Mechanics 84, 673694.
FFOWCS WILLIAMS, J. E., 1963.
HARPERBoURNE, M. & FISHER, M. J., 1973. The noise from shock waves in supersonic
jets. In Noise Mechanism. AGARDCP131, Proceedings of an AGARD Conference.
LIGHTHILL, M. J., 1952. On sound generated aerodynamically. 1. General theory. Proc.
Royal Soc. London A 211, 564587.
LIGHTHILL, M. J., 1954. On sound generated aerodynamically. II. Turbulence as a source
of sound. Pmc. Royal Soc. London A 222, 132.
LUSH, P. A. & BURRIN, R. H., 1972. The generation and radiation of supersonic jet
noise, Volume V, An experimental investigation of jet noise variation with velocity and
temperature. AFAPL TR7253Volume V, Air Force Aero Propulsion Laboratory.
64
MICHALKE, A. & HERMANN, C., 1982. On the inviscid instability of a circular jet with
external flow. Journal of Fluid Mechanics 114, 343359.
MICHALKE, A. & MICHEL, D., 1979. Prediction of jetnoise in flight from static tests.
Journal of Sound and Vibration 67,341367.
MICHALKE, A., 1970. A wave model for sound generation in circular jets. Deutsche Luftund Raumfahrt, DLR FB 7057.
MICHALKE, A., 1972. An expansion scheme for the noise from circular jets. Zeitschrift FiT
Flugwissenschaften 20, 229237.
MICHALKE, A., 1977. On the effect of spatial source coherence on the radiation of jet noise.
Journal of Sound and Vibration 55, 377394.
MICHALKE, A., 1992. On the effect of external flow and shearlayer thickness on the expansion cells of underexpanded supersonic circular jets. European Journal of Mechanics)
B/Fluids 11, 363382.
MICHEL, U. & BOTTCHER, J., 1992. Prediction of jet mixing noise for high subsonic flight
speeds. In DGLR/AIAA J 14th Aeroacoustics Conference, pp. 846853. DGLR/ AIAA
Paper 9202145, DGLRBericht 9203.
MICHEL, U. & MICHALKE, A., 1981a. Prediction of flyover jet noise spectra. AIAA Paper
812025. AIAA 7th Aeroacoustics Conference.
MICHEL, U. & MICHALKE, A., December 1981b. Prediction of flyover jet noise spectra
from static tests. NASA Technical Memorandum 83219.
MORFEY, C. L., 1979. Propagation from moving sources in flows. In Special course on
Acoustic Wave Propagation, vol. AGARDR686, pp. 111  1113.
NORUM, T. D. & SEINER, J. M., 1982. Measurement of mean static pressure and far field
acoustics of shockcontaining supersonic jets. NASA Technical Memorandum 84521.
PRANDTL, L., 1904. Uber stationare Wellen in einem Gasstrahl. Phys. Zeitschrift 5, 599601.
RIBNER, H. S., 1959. New theory of jetnoise generation, directionality, and spectra. The
Journal of the Acoustical Society of America 31, 245246.
SAE ARP 867C, 1985. Gas turbine jet exhaust noise prediction. Aerospace Recommended
Practice, Society of Automotive Engineers.
SEINER, J. M., DASH, S. M. & WOLF, D. E., 1983. Shock noise features using the
SCIPVIS code. AIAA Paper 830705.
TAM, C. K. W. & BURTON, D. E., 1984. Sound generated by instability waves of supersonic flows, Part 1: Two dimensional mixing layers. Part 2: Axisymmetric jets. Journal
of Fluid Mechanics 138, 249271 and 273295.
TAM, C. K. M., JACKSON, J. A. & SEINER, J. M., 1985. A multiplescales model of the
shockcell structure of imperfectly expanded supersonic jets. Journal of Fluid Mechanics
153, 123149.
65
66
Chapter 4
INFLUENCE OF SOLID
SURFACES
Contents of Chapter 4
4
4.1
67
67
68
68
68
71
71
72
73
73
75
77
INTRODUCTION
The basic aeroacoustic equations are applied here to the problems that are dominated by
the presence of surfaces. The influence of stationary solid surfaces on aerodynamic sound
generation was first considered by Curle (1955). Boundary layer noise is an example for this
kind of sound generation and determines the lowest possible sound level of the flow noise of
a body. In the case of moving surfaces, the aerodynamic sound generation is dominated in
all practical cases by the interaction of the flow with the moving surfaces. This situation
was first considered by Ffowcs Williams & Hawkings (1969a).
In this chapter, a general solution for the Lighthill equation (2.11) is derived which
includes the influence of a uniform motion of the ambient fluid. A coordinate transformation
into the system fixed to the rotor is then carried out as shown by Goldstein (1974) and the
various sound generation mechanisms are explained. Boundary layer noise of an infinite plate
and of finite plates in a uniform flow is then studied. Rotor noise itself will be discussed
later in chapter 8.
67
4.2
4.2.1
The aerodynamic sound generation in the presence of solid boundaries can be studied with
the solution of the inhomogeneous wave equation (2.11) on page 33. However, we shall
start here with the convective form of the Lighthill equation, equation (2.13), because it
includes the influence of a uniform motion of the ambient fluid, Ui = {Uj , 0, O}. A solution
for this equation was already derived as equation (3.3) for jet noise by neglecting the surface
integrals. Here, the surface integrals must be included and a motion of the surfaces must
also be permitted. The solution for the radiation into an unbounded space is then given by
~~vquadrupole
v,dipole
(4.1)
~v~
dipole
+ 4~
"
:t J [ro~~ncil
dS(Yi) .
S(t)
monopole
This equation was first obtained by Ffowcs Williams & Hawkings (1969a) except for:
1. the second integral with the dipole source term qi on the right hand side of the first
line which is included here for completeness,
2. the consideration of a possible external motion through the radiation distance ro and
the Doppler factor D j = 1  Maj cos
(see discussions in connection with equation
eo
(3.4)).
The factor C in the numerator of the integrands is a Doppler factor that considers the
motion of the volume element dV or the surface element dS in position Yi relative to the
observer in position Xi,
(4.2)
where Uro is the component of the velocity of the element directed in the wave normal
direction toward the observer position Xi.
4.2.2
The integration volumes and surfaces are unsteady in the case of rotors. Therefore, it
is tempting to introduce a transformation from the spacefixed coordinate system of the
observer into the Lagrangian coordinate system rotating with the rotor. This transformation
68
has the advantage that the surfaces and source volumes are stationary. The result for the
case with Ui = {Uj, 0, O}
0 (where 10 = I and D j = 1) was given by Goldstein (1974):
p' (Xi, t)
Il
dV(r/i)
~v~
dipole
( 4.3)
~v~dipole
In his paper, Goldstein (1974) has derived a more general equation. Equation (4.3) is the
solution for the sound radiated into an unbounded space. A similar equation for the radiation
into an infinite duct with circular cross section was also derived by Goldstein (1974).
The solution in equation (4.3) for the pressure pi in the field point Xi as a function of
time t contains integrals over three different domains (see Figure 4.1 in which a propeller is
shown).
(i) The volume integrals in the first line are carried out over the whole volume V
external to the surface S of the propeller.
(ii) The surface integral in the second line considers the contributions of forces on the
surface S of the propeller blades.
(iii) The volume integrals in the third line are carried out over the internal volume
Vc surrounded by the surface S. Note that the apparent monopole source of equation
(4.1) is now expressed by dipole and quadrupole terms.
Equation (4.3) was derived for assumptions that will be discussed with the help of Figure
4.1. A cartesian coordinate system "Ii(Yi, t) is used to describe the integrals. This system
is fixed to the rotating rotor in such a way that the surface S remains stationary in the "Ii
space. Thus, the solution is valid for a rotating solid body that does not change its form.
The coordinate system "Ii rotates with an angular velocity of ni ( t) in the Yi space. (When a
motion of the ambient fluid must be considered, the angular velocity vector ni would have
to be parallel to the ambient vector Ui .)
The velocity Wi in a point fixed in the rotating "Ii system in relation to the velocity W Oi
in the space fixed Yi coordinate system is given by
(4.4)
Each position "Ii in the rotating coordinate system suffers a centripetal acceleration bi that
is given by
bi = n~"Ii'
The integrals have to be evaluated under consideration of the Doppler factor
( 4.5)
69
112
O(t)
~
r'i
Yi,
Xi,
at emission
( 4.6)
70
3. The third integral shows that each surface area element dS(1]i) emits an elementary wave which is the same as that emitted by a moving dipole source of strength
 fi dS(1]i), where fi denotes the fluctuating force per unit area exerted by the boundaries on the fluid,
( 4.8)
Here ni is the unit vector normal to a surface element, and Tij is the viscous stress tensor
acting on the surface element. The third integral represents the sound generated by
the fluctuating forces of the solid surfaces on the fluid.
4. The fourth integral describes the result of moving dipole sources of strength POb i d V (1]k)
for each volume element within S.
5. The fifth integral describes a moving quadrupole of strength POWiWj dV(1]k) for each
volume element within S.
The last two integrals represent the sound resulting from the volume displacement effects.
They vanish for infinitely thin or stationary bodies.
The denominators of all integrands in equations (4.1) and (4.3) contain the Doppler
factor C which vanishes periodically for certain ranges of emission angles (Jo if the velocity
W is supersonic in any part of the source region. This introduces a singularity and causes
the emission of Mach waves. It was shown by Crighton & Parry (1991) that the sound from
supersonic propellers is dominated by this Mach wave emission.
4.3
4.3.1
FARFIELD APPROXIMATION
ROTATING SURFACES
In many practical applications (e.g., the sound generation of propellers, wind turbines, helicopters) one is interested in the sound radiation at large distances from the source region. On
assuming that the source region remains concentrated near a bounded surface S, equation
(4.1) can be simplified for the radiation into the acoustic and geometric far field (see previous
chapter on farfield definitions). The dipole source term is neglected, now, assuming that it
does not playa significant role in the sound generation of open rotors:
(4.9)
where
qq = pou;o
(1 + P:~6)  (1  ~)
p'.
( 4.10)
u ro is the component of the velocity Ui in the wavenormal direction to the observer point Xi.
fro is the component of the surface force per unit area, fi, in the wavenormal direction to
71
the observer and Un is the component of the velocity vector of the blade surface normal to
the surface. C 1  Uro / ao is a Doppler factor based on the component Uro of the velocity of
the volume or surface element directed in the wavenormal direction. The brackets indicate
evaluation of the term at the retarded time iT which (in spite of the farfield approximation)
is defined by a nonlinear equation as a consequence of source motion.
The corresponding farfield solution of Goldstein (1974) for equation (4.3) without consideration of the dipole source terms is given by
p' (Xi, i)
( 4.11)
dV(7]i).
4.3.2
STATIONARY SURFACES
When the surface is stationary, the last integral in equation (4.9) vanishes because Uro = O.
Similarly, the last two integrals in equation (4.11) vanish because the acceleration bi = 0
and the velocity Wi = O. In addition, the two coordinate systems 7]i = Vi, and the Doppler
factor C = 1. The retarded time according to equation (4.7) is then a linear function of i
and the derivatives with respect to the retarded times ir can be replaced by derivatives with
respect to time i. Equation (4.11) simplifies to
p'(Xi, i)
( 4.12)
which is the equation derived by Curle (1955) except for the second term which is added
here to account for the dipole source terms of a turbulent flow with density gradients. This
equation can be written in a form that is physically easier to understand,
p'( Xi, i)
( 4.13)
72
where the influence of a uniform motion of the ambient fluid is included, again (compare
equation (3.8)). The definitions of qq and qd of chapter (3) are repeated here,
(4.14)
qd
a (po)
P
= P aYro
(4.15)
iro is the component of the surface force per unit area, ii, in the wavenormal direction to
the observer point.
l. The first term in equation (4.13) describes the sound due to those velocity fluctuations
in the source region that are directed in the wavenormal direction toward the observer
position, Xi.
2. The second integral describes the sound due to the pressure fluctuations in the wavenormal component of the density gradient toward the observer.
3. The third term is the result of fluctuations of the surface force component directed in
the wavenormal direction toward the observer.
Curle (1955) has pointed out, that the surface integrals in equations (4.12) and (4.13)
will dominate the generated sound if the wavelength of the sound is long compared to the
dimensions of the source region. This dipole source term is the starting point for the analyses
of aerodynamic sound generation in the presence of stationary solid surfaces. Examples are
the calculation of the Eolian tones by Phillips (1956a) and the analyses of the broadband
noise from a small plate in a turbulent jet by Sharland (1964).
The first propeller noise studies were carried out with a different procedure which will
later be described in chapter 8. Goldstein (1974) has shown how to use equation (4.11) to
derive the propeller noise equation.
4.4
4.4.1
Equation (4.13) also describes the noise generation of turbulent boundary layers. Phillips
(1956b) concluded that no dipole sound is generated when one assumes that the turbulent
motion in the boundary layer can be described by the incompressible equations of motion.
Powell (1960) reached the same conclusion after applying the reflection principle to satisfy
the boundary condition of zero normal velocity on the wall. Only the first two integrals in
equation (4.13) would then contribute to the far field sound. However, the second integral,
which describes a dipole contribution, was not yet recognized at that time. It may be
important for the noise of highspeed boundary layers when the mean density gradients
within the boundary layers are large.
Powell's result can be derived from equation (4.13) with the help of Figure 4.2. The
integrations in equation (4.13) are carried out over all space with the assumption that qq,
73
plate
Figure 4.2: Mirror technique for the derivation of a solution for the sound generation of a
turbulent boundary layer on an infinite plate.
qd, and fro are mirrored on the plane. The source quantities 8 on the negative side of the
plane can then be described in terms of the quantities E8 on the positive side,
q~, for Y2 > 0,
qq , for Y2 < O. '
(4.16)
(4.17)
f~, for Y2
fro' for Y2
> 0,
< O.
'
(4.18)
The relation between the quadrupole terms on both sides of the plate is discussed in Figure
4.3. The normal velocity components in mirrored positions have an opposite sign on both
sides of the plane, whereas the tangential velocity components have the same sign. The source
term qq is defined according to equation (4.14) as the square of the velocity fluctuations in
the wavenormal direction which means that the sources in mirrored positions are equal
for an emission angle of
= 90 deg. Note that the time averaged farfield solution of the
quadrupole source terms is given by equation (3.25) which contains a double integral over
the source volume. The directivity in the far field may be determined by the coherence of
the source terms on both sides of the surface. E.g., the coherence between mirrored positions
is /12 = 1 for
= 90 deg .
Powell (1960) did not consider the dipole volume sources. If they are present (e.g. in
supersonic boundary layers), they would have an opposite sign for mirrored positions and
for an emission angle of
= 90 deg.
The normal components of the surface forces on both sides are identical in value and
opposite in sign. Therefore, they do not contribute to the sound of an infinite plate. The
surface integral is then determined by the viscous surface forces on both sides which radiate
in the upstream and downstream directions.
The volume integral over the quadrupole term is the only contribution when viscous
forces are neglected and no density gradients are present. This is the result of Powell (1960).
eo
eo
eo
74
+
Figure 4.3: Mirrored quadrupole source terms
infinite plate.
qq III
A scaling law similar to that derived for jet noise, equation (3.35), can be derived for
the sound generation of a surface element of an infinite plate. Without the dipole term,
omitting the jet stretching term 0', and by normalizing the source volume with the boundary
layer thickness 8 and the surface element S, we obtain the following scaling law for the
powerspectral density due to a surface element
( 4.19)
where K, is the ratio of the specific heat capacities. Gqq is the normalized double integral
similar to equation (3.36). The directivity of Gqq is not known. However, it can be seen
that the Doppler factor D f has a very strong effect and results in a strong forward arc
amplification of the sound.
An experiment in which the sound is without question generated only by a turbulent
boundary layer on a large plate has yet to be made. Measured data can easily be contaminated by other noise mechanisms such as the sound generation of a turbulent flow near an
edge which is discussed next.
4.4.2
This problem was treated by Ffowcs Williams & Hall (1970) who solved the first integral of
equation (4.12) for a semiinfinite plate and concluded that the sound emission by a turbulent
flow is dipole like for positions close to the edge which means that the effectiveness of sound
generation is much higher.
The situation is illustrated in Figure 4.4 for the cases of leading and trailing edges, as well
as side edges. The boundary condition of vanishing normal velocity components is satisfied
in this case by a fluctuating flow around the edge which requires a fluctuating pressure
difference between both sides. The small fluctuations are added to the large mean velocity
which means that the fluid does not flow around the corners but alters the flow direction
unsteadily.
75
,   ... Xi
X'I
~+'
side edges
Figure 4.4: Fluctuating part of the velocity component oscillates around edges which requires
fluctuating pressure differences between both sides.
The resulting normal surface force components fz near the edge have opposite sign on
both sides,
f z8 = fEB
(4.20)
z .
which means that the surface force integral does not vanish.
An analyses of the force integral results in the following scaling law for trailing edge noise.
(4.21 )
Gff(()o, Si) is the normalized double integral over the force fluctuations on the surface.
The value of Gff(()o,Si) is largest for ()0:::::J 90deg. The surface S in equation (4.19) is now
replaced by the product of boundary layer thickness 8 at the trailing edge and length L of
the trailing edge. The power of the Doppler factor has decreased to 4 and the power of the
flow or flight Mach number Uj / ao to 5. The effect of the Doppler factor is an amplification of
edge noise in the forward arc as indicated qualitatively in Figure 4.5 for the case of airframe
nOIse.
The integration of the powerspectral density of the sound pressure (equation (4.21)) over
the whole frequency range results in the following scaling law for the sound intensity I in
the far field.
j ) 6
p'2
8
4
(4.22)
1=  = "'poao (
)2D j
G(()o),
pao
411"10
ao
(U
where G(()o) is a directivity function. Here, it is assumed that the frequency within the
source region scales according to f <X Uj /8. Since 8 decreases with increasing Uj , the
resulting effective power of UJi ao on trailing edge noise is slightly less than 6.
This result for the power agrees with the value of 5.7 derived by King III (1989) for
boundary layer noise using a completely different procedure.
The experimental evidence of all attempts to measure boundary layer noise is a close
to UJpower law for the sound intensity. Doak (1960) proposed a simple equation for the
76
/\
\
standard assumption
for directivity
II
/
/~
//~
measured evidence
agrees with scaling law
Figure 4.5: Common assumption for directivity of airframe noise and measured evidence
which agrees with the Doppler amplification of equation (4.21).
intensity due to a surface area 5,
(4.23)
The overall sound pressure level is then given by
Lp = 10 log
(~) ,
(4.24)
where 10 = 10 12 W 1m 2
The equations (4.22) and (4.23) have the same power for Uf lao but differ in their dependence on the surface. While the first equation is proportional to the length L of the trailing
edge and the boundary layer thickness 6, the second is proportional to the surface S. The
Doppler amplification is not considered in equation 4.23.
Howe (1978) also derived a scaling law for trailing edge noise which can be written as
I ex "'Po ao L13
2
411"1'0
(Uf )

ao
1
Uf )
0.2ao
u'2
2'
h
( 4.25)
where 13 is the correlation length parallel to the trailing edge, U'2 IU~ is the normalized mean
square velocity fluctuations in the boundary layer, 13 ex 62 = momentum thickness, with
62 ex L(LUh lv)0.2 and chord length L. The smaller power for Uflao in Howe's equation
must be noted. The Doppler factor is also not considered in equation (4.25).
4.5
BIBLIOGRAPHY OF CHAPTER 4
D. G. & PARRY, A. B., 1991. Asymptotic theory of propeller nose part ii:
Supersonic singlerotation propeller. AIAA Journal 29, 12, 20312037.
CRIGHTON,
77
CURLE, N., 1955. The influence of solid boundaries upon aerodynamic sound. Proceedings
of the Royal Society (London) A 231, 505514.
DOAK, P. E., 1960. Acoustic radiation from a turbulent fluid containing foreign bodies.
Proceedings of the Royal Society (London) A254, 129145.
FFOWCS WILLIAMS, J. E. & HALL, L. H., 1970. Aerodynamic sound generation by
turvbulent flow in the vicinity of a scattering half plane. Journal of Fluid Mechanics
40, 657670.
FFOWCS WILLIAMS, J. E. & HAWKINGS, D. L., 1969. Sound generated by turbulence and
surfaces in arbitrary motion. Philosophical Transactions of the Royal Society (London)
A 264, 321342.
GOLDSTEIN, M. E., 1974. Unified approach to aerodynamic sound generation in the presence of solid boundaries. Journal of the Acoustical Society of America 56, 497509.
HOWE, M. S., 1978. A review of the theory of trailing edge noise. Journal of Sound and
Vibration 61, 437465.
KING III, W. F., 1989. Aeroacoustics of highspeed tracked vehicles. DLRIB 2221489/B4,
Deutsche Forschungsanstalt fur Luft und Raumfahrt e.V.
PHILLIPS, O. M., 1956a. The intensity of aeolian tones. J. Fluid Mech. 1,607624.
PHILLIPS, O. M., 1956b. On the aerodynamic surface sound from a plane turbulent baudary
layer. Proceedings of the Royal Society (London) A234, 327335.
POWELL, A., 1960. Aerodynamic noise and the plane boundary. Journal of the Acosutical
Society of America 32, 982990.
SHARLAND,1. J., 1964. Sources of noise in axial flow fans. Journal of Sound and Vibration
1, 302322.
78
Chapter 5
AERODYNAMIC SOUND
GENERATION MECHANISMS IN
TURBOMACHINES
Contents of Chapter 5
5 AERODYNAMIC SOUND GENERATION MECHANISMS IN TURBOMACHINES
5.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
5.2 BLADE THICKNESS NOISE . . . . . . . . . . . . . . . . . . . . . . . . ..
5.3 TONAL NOISE DUE TO STEADY AND UNSTEADY AERODYNAMIC
FORCES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
5.3.1 ROTOR IN SPATIALLY UNIFORM STEADY FLOW FIELD
(STEADY BLADE FORCES) . . . . . . . . . . . . . . . . . . . . ..
5.3.2 ROTOR IN SPATIALLY NONUNIFORM STEADY FLOW FIELD
(UNSTEADY BLADE FORCES) . . . . . . . . . . . . . . . . . . ..
5.3.3 NOISE
GENERATION
BY
ROTOR/STATOR
AND
ROTOR/CASING INTERACTION. . . . . . . . . . . . . . . . . ..
5.3.4 IMPULSIVE NOISE . . . . . . . . . . . . . . . . . . . . . . . . . ..
5.3.5 ROTATING STALL . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.6 NONUNIFORM ROTOR GEOMETRY . . . . . . . . . . . . . . ..
5.3.7 N ARROWBAND NOISE DUE TO A ROTOR OPERATING IN UNSTEADY FLOW FIELD . . . . . . . . . . . . . . . . . . . . . . . .
5.4 RANDOM NOISE DUE TO UNSTEADY AERODYNAMIC FORCES . ..
5.4.1 GENERAL REMARKS . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.2 TURBULENT BOUNDARY LAYER NOISE . . . . . . . . . . . ..
5.4.3 NOISE DUE TO INCIDENT TURBULENCE. . . . . . . . . . . ..
5.4.4 VORTEX SHEDDING NOISE . . . . . . . . . . . . . . . . . . . . .
5.4.5 FLOW SEPARATION NOISE. . . . . . . . . . . . . . . . . . . . ..
5.4.6 TIP VORTEX NOISE . . . . . . . . . . . . . . . . . . . . . . . . ..
5.5 QUADRUPOLE NOISE . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
79
79
80
81
81
81
82
85
86
86
87
89
91
91
92
92
93
94
96
98
5.6
5.7
5.1
98
101
101
103
INTRODUCTION
The discussion of noise generating mechanisms in turbomachines begins with a review of the
socalled Ffowcs Williams & Hawkings equation (4.3) which was derived in chapter 4 on the
influence of solid surfaces on the aerodynamic noise: For turbomachinery noise problems,
the source term qi, which accounts for the effect of different densities in the source region and
in the ambient, is commonly disregarded. With this simplification, equation (4.3) becomes
(5.1 )
As was mentioned before, this equation was first obtained by Ffowcs Williams & Hawkings
(1969a), but is written here in the notation used by Goldstein (1974). Ii denotes the ith
component of the force per unit area exerted by the boundaries on the fluid, C is the Doppler
factor, ri is the vector between the positions of the source, Yi, and the observer, Xi, (r = Iril),
and 'fJi are the source coordinates in a reference system moving with the rigid surface. liVi
and bi denote the velocity and acceleration of the rigid surface S with interior volume \1;;.
All terms in equation (5.1) within brackets have to be taken in the moving source coordinate
system and at the retarded time tT) i.e., the time of the emission of sound.
The main feature of equation (5.1) for the present purpose is that it allows to discuss
the various aeroacoustic noise source mechanisms involved when moving rigid surfaces are
present. The first term on the right hand side of equation (5.1) is roughly the same as the
Lighthill integral for an unbounded flow. The double derivative of the volume integral over
the turbulent shear stress tensor in the flow region represents the generation of sound due to
volume sources. This contribution is equivalent to a distribution of moving point quadrupoles
of strength qijdV, and therefore the sound due to turbulent shear stresses is often termed
'quadrupole noise'. The second integral in equation (5.1) describes the sound radiation due
to the forces exerted by the solid surfaces on the flow. This mechanism is called 'dipole
noise' because the forces can be thought of as a distribution of moving acoustic dipoles of
strength lidS.
The last two terms represent the sound radiation due to volume displacement effects of the
moving surface. This is basically equivalent to a monopole radiation, however, in equation
(5.1) the volume displacement effects are expressed in terms of a dipole source proportional
to the acceleration bj of the surface and a quadrupole source of strength POWiWj .
In the following chapters the aerodynamic noise generation processes relevant to turbomachinery are categorized with respect to the three aeroacoustic source mechanisms described above: Section 5.2 is concerned with the sound due to volume displacement effects
80
(monople radiation), sections 5.3 and 5.4 deal with the tonal and random noise generation
due to steady and unsteady forces exerted by the flow on the blades, vanes and casing of
the machine (dipole noise), and in section 5.5 the sound generation due to volume sources
(quadrupole noise) is described.
5.2
The sound generated by volume displacement effects is often called blade thickness noise.
Early papers dealing with this subject were published by Ernsthausen (1936), Deming (1937),
Deming (1938), Ernsthausen (1938), and Gutin (1942). The moving impeller blades displace
fluid mass and generate periodic pressure fluctuations in the adjacent field. The azimuthal
phase velocity of these pressure fluctuations is of course equal to the rotor shaft speed.
For subsonic circumferential phase speeds of the source region, the radiated sound field is
relatively weak, because of the local cancellation of outofphase regions within the acoustic
near field of the source. In case of low Mach number fans used for ventilation or industrial
applications, the azimuthal phase velocity of the blade thickness source pressures is well
below sonic, and so their acoustic radiation efficiency is low. This is why blade thickness
noise does not play an important role in industrial fan noise. However, blade thickness noise
is an important noise generation mechanism for high speed turbomachines, like propellers
and helicopter rotors. The intensity of the blade thickness noise is dependent on the blade
crosssectional area.
5.3
5.3.1
The forces experienced by the blades of a fan rotor operating in a spatially uniform steady
flow field (see the schematic in Figure 5.1) are steady. An observer in a fixed frame of
I
/
..
"
"
\
\
{~
' .
\
I
r/
" '
./
Figure 5.1: Schematic of a rotor interacting with a uniform steady flow field
reference, however, will sense periodic pressure fluctuations at the blade passage frequency
81
and harmonics thereof. This rotor noise component is named after Gutin (1936), who was
the first to study the sound radiation due to the steady aerodynamic forces on the blades of
a propeller. In his theory the blade forces were represented by a source distribution of dipole
order. Like in case of blade thickness noise, the azimuthal phase velocity of the bladelocked
forces is equal to the rotor shaft speed, and thus their acoustic radiation efficiency is low in
case of low Mach number turbomachines. Consequently, the 'Gutinnoise' is an insignificant
noise source in case of the low to medium speed fans used for ventilation and industrial
applications while it represents an important noise generation mechanism for high speed
turbomachines, in particular highspeed propellers and helicopter tail rotors (Chou (1990)).
The steady blade forces are necessary to do the flow work and, therefore, cannot be
avoided. As Wright (1976) pointed out, however, the acoustic radiation produced by these
forces can be controlled by increasing the blade number.
At supersonic rotor speeds, the rotor locked blade forces become efficient radiators of
sound. Since the flow relative to the blade is supersonic, there is a shock wave ahead of
each blade, similar to the shock wave of an aircraft travelling at supersonic speed. If the
shock waves from all blades would be identical, the tone heard by a distant observer would
be equal to the blade passage frequency and its harmonics. In practice, however, there are
small differences in the blade geometry, spacing or aeroelastic behavior which result in sound
radiation at the shaft order frequencies. This effect occuring at supersonic speeds, which is
discussed in a little more detail in the section on nonuniform rotor geometry, is often called
"buzzsaw" noise or "multiple tone" noise.
5.3.2
When a rotor operates in a steady but spatially nonuniform flow environment, see the
schematic in Figure 5.2, each blade will experience unsteady fluid forces, since both mag
/'
,...."
1
I'
k
/
\
\
" .;
/
/
Figure 5.2: Schematic of a rotor interacting with a spatially nonuniform steady flow field.
nitude and angle of attack of the oncoming flow as seen by the moving blade change with
the angular position. The spectrum of the noise due to a rotor interacting with a nonuniform steady flow field is discrete with spectral lines at the blade passage frequency and
its harmonics. It was shown by Tyler & Sofrin (1962) that the azimuthal phase velocity of
these unsteady blade forces is commonly much higher than the rotor speed itself and can
even become supersonic. This is demonstrated in Figure 5.3. In Figure 5.3a is shown the
82
oj
~ \~.
0
I.~ ~_
bl~\~) ~
4
.t.
5
$
7
Figure 5.3: Interaction between rotor and stator (after Tyler & Sofrin (1962)).
case of a rotor with B = 8 blades interacting with a stator of V = 6 vanes. The rotor is
represented by the 8 spokes in the center, one of which is marked with a dot, and the stator
vanes lie concentric around the rotor. This schematic corresponds directly with a centrifugal
fan or compressor with outlet guide vanes, but the principle shown also applies to axial
flow machines. In position "0", the marked rotor blade coincides with the right hand vane,
and a similar event occurs on the opposite side. The coincidences are identified by arrows.
Successive positions show that the coincidence pattern rotates faster that the rotor itself;
in the particular case considered here, at four times the rotor speed. Similarly, in Figure
5.3b is shown a rotor with Z = 8 blades interacting with stator with V = 9 vanes. Here,
coincidence of blades and vanes occurs only once along the periphery, and the coincidence
pattern rotates in a direction opposite to that of the rotor, at a speed of eight times the
rotor speed.
As a result of their high azimuthal phase velocity, the acoustic radiation efficiency of
the unsteady blade forces is high, and, hence, they represent the most important cause of
turbomachinery noise.
In case of turbomachines with a free inlet, nonuniform velocity profiles of the intake flow
are caused by inhomogeneities of the air entering the fan or by the asymmetric position of
the intake with respect to adjacent floors, ceilings, or walls. Cross flows near the inlet which
may be produced by other machines would have a similar effect.
In Figure 5.4 the noise mechanism due to the ingestion of large scale atmospheric turbulence of a helicopter main rotor is shown. The acceleration of the main flow stretches the
eddies, and the repeated passing of these eddies by the rotor blades gives rise to the tonal
83
ATMOSPHERIC,
TURBULENCE
EDDY
/~
, /
~\
><===>
C~_ _
ELONGATED
EDDY
I
I
"....... .
/
~ RECIRCULATION
OF WAKES
a) MAIN ROTOR TURBULENCE INGESTION
NOISE MECHANISM
a..
en
o
3:
BROADBAND
NOISE
CD
oa:
a:
z
10 2
10 3
FREOUENCY.Hz
b) EXAMPLE OF TURBULENCE INGESTION
NOISE SPECTRUM
Figure 5.4: Turbulence ingestion noise mechanism and resultant nOIse spectrum (after
Schlinker & Brooks (1982)).
spectrum shown.
If a turbomachine with an inlet duct is considered, nonuniform steady flow fields are
produced by upstream obstructions like struts or inlet guide vanes duct bends, corners and
other duct elements. Effective sound radiation will occur when the circumferential wavelength of the flow distortion is of the order of the blade spacing. The effect of a stator on
the flow into the rotor is shown schematically in Figure 5.5.
Irregularities of the duct boundary layer are very effective sound generators because of the
interaction with the blade tips which are the fastest parts of a turbomachine. Asymmetries
of the casing or an eccentric position of the impeller have a similar effect, because unsteady
forces are set up on the blade tips.
Helicopter rotor blades encounter unsteady aerodynamic forces due to their complex
flow environment: The linearly varying circumferential velocity along the blade span is
superimposed by the forward flight speed of the aircraft. Additionally, the blades are twisted
along their lengths and the stagger angle is controlled to vary around the circumference. This
situation is summarized in Figure 5.6.
84
Stator
Rotor
,,
I
I
I
I
I
I
~I.::~
Vo
1
V
I
I
I
I
I
I
(a)
,\\,
o"o~
,,"
,,0
~,,'" //
~<i6 //
",'\
oG
oc<
//
Rotor blade
velocity V
_~i) f3/ /
6~ AbsOl ut
e veloCity
v:o
( b)
Figure 5.5: Effect of inlet stator vanes on the flow into the rotor (after Sharland (1964)).
REVERSED
FLOW
REGION
INFLOW VELOCITY
VECTOR TO BLADE
&"*'1: BLAD E
ELEMENT
ADV~
BLADE SIDE
Figure 5.6: Angleofattack contour plot of the main rotor for a helicopter in an arbitrary
forward flight condition (after Schlinker & Brooks (1982)).
5.3.3
In case of multistage turbomachines, the presence of adjacent blade or vane rows results in
nonuniform mean flow conditions. Even if a frictionless medium is considered, the potential fields of blade and vanes generate unsteady interaction forces. These potential fields,
however, decay rapidly with distance, and thus the sound to potential field interaction is important only when the axial distance between rotor and stator is very small. The asymmetric
85
casing of centrifugal fans or compressors has a similar effect in that it sets up unsteady blade
forces on the upstream rotor.
The wakes from upstream blades or vanes do not wash out as quickly as the potential
fields, and thus the unsteady interaction forces set up on the rotating blades as well as
on the stationary vanes are the most important generation mechanism of the tonal noise
component. Likewise, the unsteady flow leaving the rotating impeller channels of centrifugal
fans or compressors generates pressure fluctuations of high amplitude at the cutoff of the
casing which is regarded the main source region of the blade tone noise for these machines
when no outlet guide vanes are present as is common in centrifugal fans.
Moreover, a vorticity field is attached to the wakes of upstream rows that are caused by
spanwise variation in blade or vane loading which generate random noise components which
will be discussed in a following section.
In Figures 5.7 and 5.8 experimental results of the early study of Sharland (1964) are
120
110
:<
.,
100
Qj
>
'OJ
:<
~Single
oa..
0
::J
'"
70
"2
OJ
<3
60
Figure 5.7: Increase in axial fan interaction noise as stages are added (after Sharland (1964)).
reproduced which show how the noise of a rotor is increased when a stator row and subsequently more fan stages, are added.
5.3.4
IMPULSIVE NOISE
One source of impulsive noise is the interaction between the rotor and the blade tip vortex
of a preceding blade. In helicopter aeroacoustics this is called blade slap. Impulsive noise
is a special type of rotational noise. The spectrum is characterized by peaks at the blade
passage frequency and harmonics, see the spectra shown in Figure 5.9. Blade slap is, when
it occurs, the most prominent helicopter noise component.
5.3.5
ROTATING STALL
This flow phenomenon occurs when a fan is operated at low flow rates and some local
disturbance stalls the flow on one of the blades. A schematic of this flow condition is shown
86
90
.,
>
80
70
0.
"
., >.
0 u
Vl
..D
C
::J
.,
o c
::J
cr
o .,
50 
~!:
., v
E en
o 0
0
40
C
::J
'"
0'"
u..o.
1,000
Frequency ('/s)
Figure 5.8: Effect of added stages on the sound pressure spectrum of an axial flow machine;
dashed line: single rotor; dotted line: single rotorstator stage; solid line: 3 rotorstator
stages plus IGV; (after Sharland (1964)).
in Figure 5.10. As a result, the flow in the passage on the suction side of this blade is partially
blocked, and the mean flow is diverted away from this passage. This in turn increases the
angle of attack on the blade nearest to the suction side of the originally stalled blade so that
the flow is stalled there. On the pressure side of the originally stalled blade, the flow angle is
reduced, and the flow is 'unstalled'. In this way the stall cell moves around the blade row at
about 0.3 to 0.5 times the shaft frequency, opposite to the impeller rotation. In consequence
unsteady blade forces are set up, and low frequency pressure pulsations, noise and vibrations
are generated. Rotating stall occurs in axial as well as centrifugal fans, see for example the
papers by Tanaka & Murata (1975) and Suzuki, Ugai & Harada (1978). When more than
one cell of stalled flow exist in a turbomachine, which is often the case in the centrifugal flow
type, the resultant peak in the spectrum lies above the rotor shaft frequency. As an example
the spectra of a centrifugal fan are shown in Figure 5.11 for four operating conditions and
a speed of n = 1500/min= 25/s. At zero flow rate and 47% best efficiency flow rate strong
peaks at about 36Hz exist which are not present in the spectrum at the optimum flow rate
and free flow. This result indicates that two stalled flow cells exist which rotate at 72% shaft
speed.
5.3.6
Subsonic Rotors
For an asymmetric rotor, the blade passage sequence repeats once per revolution rather than
Z times as for the symmetrical rotor with Z blades. Hence, an asymmetric rotor can radiate
sound at the rotor shaft frequency and its harmonics even when the inlet flow is steady. The
blade tone will be frequency modulated about the blade passage frequency, i.e., additional
spectral lines will appear to both sides of the blade passage frequency at a distance of the
rotor shaft frequency. The total sound energy radiated by the rotor alone will be the same
as for a fan with evenly spaced blades.
87
90 ~.
70
_~II;.Jl~iil!{ll,i1i1l2jH~zlti~lt~.r~'~j~'5~Hizit~i'j~lr~lIi~'Jltiltii1iJ1I1~il~~~tH~
(a)
+::;:::t,:
.,: ._
60
....r:::r..::.
50
70
CD
"
10
100
Frequency (Hz)
Figure 5.9: Narrowband analyses of blade slap of a helicopter main rotor (after Leverton
(1989)).
It was already mentioned in section 5.3.1 that shock waves are attached to the blade of a
rotor operating at supersonic tip speeds. This situation is sketched in Figure 5.12. In the
upper sketch, the case of a perfectly periodic rotor is shown where the shock waves from all
blades are identical. The time history picked up by a microphone at a distance from the
88
~
.....J
0...
V)
8 0 ~++R!jr'r't"'''o''r+tHf++J
8 0 I++#,~H
70
60
sg 8 0 1++!j'T'i"'''trr,,++HiIlIlt:t1
0:: 7 0 f1h:V+fV)
60
8 0 1+'+!Ic7'T'"'t":"::;'r+tHil>N.HrI
0:: '7 0
V)
60
20
30 40 60
80 100
200 Hz 300
500
Figure 5.11: Spectra of a centrifugal fan at different fan operating conditions (D = 450mm;
Z = 12; n = 1500/min; after Suzuki, Ugai & Harada (1978)).
rotor would be strictly periodic giving rise to sound radiation at the blade passage frequency
and its harmonics.
In practice small irregularities in blade geometry, spacing, and stagger angle exist which
cause a perturbation of the regular shock pattern, i.e., strength and spacing of the individual
shock waves are not identical anymore. Since due to nonlinear effects higher amplitude
shocks propagate faster than lower amplitude ones, the resultant waveform of the fluctuating
pressure changes with distance from the rotor, see the papers by Hawkings (1971), and
Kurosaka (1971). Close to the rotor, the time history is determined by the blade passage
sequence, but as the distance from the rotor is increased, the spacing and the strength of the
shock waves become less and less regular. As a result, a very uneven system of shocks which
is no longer periodic with the blade passage evolves, and thus multiple tones are generated
at the rotor shaft order frequencies. A typical "buzzsaw" noise spectrum of a supersonic
rotor is shown in Figure 5.13. The relative blade flow on the downstream side of the rotor
is generally either subsonic or very low subsonic so that the unsteady pressure field decays.
This is the reason why multiple tone noise is observed only on the inlet side of the rotor.
Multiple tone noise is generated in modern aircrafts at high power, in particular during
take off, when the fan blade tips are running at supersonic speed (Smith (1989)).
5.3.7
When the flow distortions entering the fan are unsteady, the sound spectrum changes from
discrete to continuous. Lowfrequency variations result in a bandspreading of the spectrum
about the blade passage frequency, and the stochastic velocity fluctuations of turbulent
intake flow generate random blade forces and, hence, a broadband sound spectrum. The
next chapter is devoted to the generation of random noise components.
89
Airflow
,r
IJ\f\JW
)./ Microphone
Discretenoise
waveform
y~~
~,
'
'",
"
, Mach wa"
'
u
F an blade tips
a) Idealized wave pattern
Airflow
/~~crop~o~= __ I ~
,~Mach
Multiplepuretone
noise waveform
wave
Figure 5.12: Schematic of multiple tone (buzzsaw) noise generation in superSOnIC rotors
caused by nonuniform rotor geometry (after Goldstein (1976)).
4E
100.
10E
Fan blade
passing
frequency (33E)
20E
30E
2E
90.
80.
SPL
70.
(dB)
60.
50.
40.
0.0
2.0
4.0
6.0
8.0
10.0
Frequency (kHz)
Figure 5.13: Typical multiple tone noise spectrum of a supersonic rotor (after Smith (1989)).
90
5.4
5.4.1
v\,?
~
Incident Turbulence
Figure 5.14: Broadband noise sources of a twodimensional stationary airfoil (from Chou
(1990) ).
twodimensional stationary airfoil are summarized. Additional threedimensional effects are
present at the blade tip.
The generation of random noise was studied theoretically as well as experimentally by
Sharland (1964). His theoretical model is derived here by beginning with equation (4.12) of
chapter (4) on the influence of solid surfaces on the aerodynamic sound: When the integrals
involving the source terms qi and qij are neglected, the intensity of the far field sound radiated
from a stationary surface can be expressed in the form
1 p,2 _
XiXj
 poao  16 7r 2 poa 3 x 4
o
11 8fi ( ( , _ ~)8ji(.
!)t
8' 8
u~t' y~,t
ao u
Y1"t
_~) dS'dS
ao
(5.2)
If it is assumed that the time derivatives of the force fluctuations are correlated only within
a limited flow region which is small compared with the sound wavelength, the differences in
retarded time t  t' in equation (5.2) may be neglected in this region, and the covariance
integral can be expressed in terms of a correlation area Se :
~)]2
ao
(5.3)
Introducing the fluctuating lift force per unit span, L = cji, inserting equation (5.3) into
equation (5.2), and integrating the intensity over a sphere gives the radiated sound power
2
P _
1
dL l l
 127r poa~ [ dt ]
e
91
(5.4)
where I is the blade span, and Ie is the spanwise correlation length. This result implies that
the sound radiated from the whole blade is the sum of the radiation from l/ Ie uncorrelated
regions, each of spanwise extent Ie. Thus, not the whole blade span must be small compared
with the sound wavelength to be considered a compact source, but only the coherent regions
must be compact. The requirements for this are Ie A and c A.
Equation (5.4) may be written in spectral form as
dP
dw
(5.5)
Here GL(w) is the power spectral density of the blade loading fluctuations. Equation (5.5)
allows to calculate the sound power spectrum radiated from a blade, provided the blade loading spectrum GL(w) is known. The above theoretical considerations were used by Sharland
(1964) to develop early prediction fornmlas for the random noise due to the pressure fluctuations in a turbulent boundary layer, due to inlet turbulence and due to vortex shedding
nOIse.
5.4.2
Even in a completely undisturbed steady flow field, the fan blades will experience random
force fluctuations that are due to the pressure fluctuations in the turbulent blade boundary layer. However, it is not the direct radiation from the quadrupole sources within the
turbulent boundary layer, away from the leading and trailing edge, that is responsible for
the noise but, rather, the fact that the blade turbulent boundary layer is passing the blade
trailing edge. The edge serves to scatter the basically nonpropagating nearfield pressures
into a propagating sound field. Early papers on this problem were published by Powell
(1959) and Ffowcs Williams & Hall (1970). A model for the sound radiation from boundary
layer/trailing edge interaction was proposed by Mugridge (1971) in which the unsteady blade
forces were estimated from wall pressure measurements. The results showed that the overall
levels due to turbulent boundary layer radiation are some 10dB lower than the vortex shedding noise predicted by Sharland (1964). This result indicates that in practice the turbulent
boundary layer noise is most often dominated by the noise due to incidence turbulence and
vortex shedding so that boundary layer noise can be considered the lower limit of random
turbomachinery noise, see the following two sections.
More detailed discussions of the noise generation by the turbulent boundary layer/trailing
edge interaction are given in the survey papers on rotor broadband sound by SchEnker &
Brooks (1982) and Chou (1990).
5.4.3
It was mentioned before that velocity fluctuations of the intake flow with a stochastic time
history generate random blade forces and, hence, a broadband sound spectrum. The importance of turbulence ingestion noise was first demonstrated by Sharland (1964). In his
experiments a small flat plate was first placed in the nominally smooth flow of a jet close to
the exit nozzle and then further downstream into the mixing region. The experimental results are shown in Figure 5.15. In the first case the noise was dominated by vortex shedding
(see the following section) and in the second case by incident turbulence.
92
i'J
110
o
0..
/""'~B.L. Radiation
/
60/
50
I
//,
I
I
I
~1070~2~070~~3~0~04~0~0~'6~070~8~00~1,0~0~0~
Velocity at plate centre Ut/secl
Figure 5.15: Noise radiated from a small flat plate (after Sharland (1964)).
5.4.4
When a vortex separates from a blade, the flow circulation around the blade is changed,
and thus force fluctuations on the blades are induced. A classical example for this is the
"Karman vortex street" occurring behind circular cylinders which are responsible for the
aeolian tones. Two types of vortex shedding are important for turbomachines, the laminar
boundary layer vortex shedding and the turbulent vortex shedding from blunt blade trailing
edges.
Fukano, Kodama & Senoo (1977) argued that in case of no inlet flow turbulence present,
vortex shedding dominates the broadband noise of low pressure axial fans, and presented an
analytical model for this noise component which is based on the model by Sharland (1964).
Laminar Boundary Layer Vortex Shedding Noise
It was pointed out by Archibald (1975), Wright (1976), Longhouse (1977), and Bridelance
(1986) that the origin of laminar vortex shedding noise are instability waves in the laminar
boundary layer on the blade suction side that travel downstream and generate a pressure
disturbance when passing the blade trailing edge. This pressure is radiated also upstream to
reinforce the instability wave, and thus an aerodynamic/acoustic feed back loop is formed.
A schematic presentation of this mechanism is given in Figure 5.16. The spectrum of the
vortex shedding noise or 'instability noise' was observed to be fairly narrow band in the
studies by Archibald (1975), Wright (1976), Longhouse (1977) and distributed over a wider
frequency range by Bridelance (1986). As Schlinker & Brooks (1982) explained, "the nearly
periodic surface pressure fluctuations due to shedding renders a quite peaked but continuous
spectrum shape. When this phenomenon occurs for a rotor, the variation of local blade
velocity results in the generation of a range of narrowband frequencies. Thus the spectrum
sensed by a far field observer has the appearance of broadband noise." Laminar boundary
layer noise can be avoided by tripping the flow to cause the laminar boundary layer to
93
Acoustic waves
POint of
Instability
I
Origin of
acoustic WO ves
Turbulent boundary layer
(no vortex shedding occurs)
Figure 5.16: Schematic of laminar vortex shedding noise generation (after Longhouse (1977)).
become turbulent, as will be shown in more detail in chapter 12 on noise reduction methods
for axial flow fans
5.4.5
Figure 5.18 shows two cases of flow separation occurring on an airfoil at high incidence
angle. In the upper sketch the onset of stall is visible in the small separation region near the
trailing edge. Note that there is also a small separation bubble near the leading edge of the
blade. If the angle of attack is further increased, the flow separation point moves upstream
to the leading edge to form a large leading edge vortex. The effect of the separated flow
is to set up random force fluctuations on the blades which are known to produce random
noise. Comparative spectra for stalled and unstalled rotors are shown in Figure 5.19. While
the spectrum of the unstalled rotor is characterized by the blade tone fundamental and
harmonics, the stalled rotor spectrum is almost entirely broadband.
94
   .
SHARP TE
BLUNTED
BLUNTED
BLUNTED
(I = 0)
TE (t = 2.5 mm)
TE (I = 1.9 mm)
TE (I = 1.1 mm)
50
=0
1/0'
J:
40
tlo'
=064
..........
SINO
(/) a.
.
30
I , ......
~.
.:...."
0.3
10
FREQUENCY (kHz)
Figure 5.17: Effect of trailing edge bluntness noise for various freestream velocities and
degrees of the trailing edge (TE) bluntness (after Schlinker & Brooks (1982)).
(a) LIGHT STALL
TRAILlNGEDGE SEPARATION
SEPARATION EDGE OF
BUBBLE
VISCOUS LAYER
iL
Uoo
LAMINAR
FLOW
STRONG INTERACTION
VISCOUS LAYER = (9(AIRFOIL THICKNESS)
(b) DEEP STALL
VORTEX DOMINATED
VISCOUS LAYER = (9(AIRFOIL CHORD)
Figure 5.18: Flow separation on an airfoil at large angles of attack (after McCroskey (1982)).
95
100
Fan speed 10,000 r.p.m.
45 to fan axis
.3 ft from intake centre
90
d)
.::; 80
v>
..!!
l' 70
::>
~
~
:..
l'<l0
"
'"\
60
::>
,,
V)
50
40
10
100
10,000
1,000
Frequency (cis)
Figure 5.19: Spectra from stalled and unstalled rotor cascades (after Sharland (1964)).
5.4.6
Unducted Rotors
Unsteady blade forces are also generated by secondary flows like the flow around the blade
tip which is driven by the pressure difference between pressure and suction side of the rotor
blades. The resulting flow, see the schematic depicted in Figure 5.20, consists of a vortex on
SEPARATION
LINES
REGION OF
OUTWARD FLOW
UNDER VORTEX
PLAN VIEW
ATIACHMENT
POINT
SEPARATION
POINTS
WING
EDGE VIEW
Figure 5.20: Schematic of the blade tip flow of an unducted rotor (after Schlinker & Brooks
(1982)).
the blade suction side with a thick viscous core (Schlinker & Brooks (1982)). The formation
96
Figure 5.21: Tip vortex at the impeller blades of ducted axial flow machines (from Vavra
(1969)).
outlet side of the impeller disk of an axial fan or compressor drives a secondary flow through
the tip clearance gap which extends over the entire circumference, see the flow sketch in
Figure 5.22. The size of the radial gap and the pressure built up by the fan are decisive for
the strength of the secondary flow which in turn influences the mean flow field in the tip
region including the flow separation processes at the blade tips.
Tip clearance noise can be a significant source of noise in axial fans, when the tip clearance
exceeds a certain limit. While Longhouse (1978) and Fukano, Takamatsu & Kodama (1986)
reported increased broadband noise levels as a result of an enlarged tip clearance, Kameier,
Nawrot & Neise (1992) and Kameier (1994) found, on top of that, drastic level increases
within limited almost narrowband frequency regions. Figure 5.23 shows the effect of the tip
clearance ratio T = s / D on the sound pressure spectrum of an axial fan. In the high frequency
region (upper diagram), the random level is increased uniformly as the tip clearance grows.
A particularly steep level increase of almost 20dB occurs at 370Hz, below the blade tone
fundamental, when the tip clearance s is enlarged from T = s / D = 0.0013 to 0.0053 (lower
diagram).
Measurements of the pressure fluctuations on the casing wall close to the impeller tips
and on the rotating blades revealed that a flow instability mechanism occurs at the blade tip
which is associated with the tip clearance noise. In Figure 5.24 a series of pressure spectra
measured on the impeller blades is shown for tip clearance ratios of T = s / D = 0.00066
and 0.0053. From the top to the bottom, the flow rate is reduced. In each diagram, four
spectra are shown which were measured at the following blade positions: No.1: r / R = 0.88,
x/c = 0.3, suction side; No.2: r/R
0.99, ;r/c = 0.3, suction side; No.3: r/R = 0.99,
97
1:=0
___
.P~~~~
___________
.p       J l.._     ~~~~
Figure 5.22: Schematic view of the secondary flow in the blade tip area driven by the pressure
difference across the impeller disk of a ducted axial fan or compressor (from Kameier (1994)).
x/c = 0.6, suction side; No.4: r/R = 0.99, x/c = 0.44, pressure side; (r = radial distance
from axis, R = impeller Radius, x = axial distance from leading edge, c = blade chord).
Beginning at the best efficiency point, a broad hump with superimposed spikes (RI) occurs
when the tip clearance is large. This spectral signature represents a rotating source or
vortex mechanism which, similar to the rotating stall phenomenon, moves in the azimuthal
direction, relative to the blade row. Hot wire measurements in the tip region showed that
this rotating instability component only occurs if a reversed flow condition exists in the
tip clearance gap. The effect of the reversed flow is to increase the thickness of the casing
wall boundary layer upstream of the impeller such that the axial mean flow component
is eliminated. As a result, vortex separation takes place at the tip in a circumferential
direction. If the azimuthal wavelength of the shed vortices is equal to the blade spacing, a
strong interaction of the vortex separation on individual blades occurs which results in the
drastic increases of the radiated noise.
When the flow rate is further reduced, the blade flow is separated over the entire blade
span over part of the impeller circumference, which is the well known phenomenon of rotating
stall.
5.5
5.5.1
QUADRUPOLE NOISE
RANDOM NOISE
It was pointed out at the beginning of this chapter that the first term on the right hand side
of equation (5.1) involving the double derivative of the volume integral over the shear stress
tensor is equivalent to a sound radiation from a distribution of moving acoustic quadrupoles.
The direct quadrupole type sound radiation from the turbulent velocity fluctuations is gen
98
100
90
80
70
r1
60
0)
U
LJ
Q.
1
50
40
30
1=
20
0.0013
T =0.00066
10
0
10000
8000
6000
4000
2000
[Hz]
f
100
90
,=0.0053
80
,=0.0027
/ ..\
70
r1
0)
LJ
0..
1
60
50
40
30
20
10
0
0
100
200
300
400
500
600
700
800
900
1000
f [Hz]
Figure 5.23: Sound pressure spectra in the outlet duct of an axial fan with outlet guide vanes
0.22; after
as a function of the tip clearance (D = 450mm; Z = 24; n = 1400/min; 'Popt
Kameier (1994)).
erally considered negligible compared to that from the dipole radiation due to the fluctuating
99
140
<D,.,0.250
\=0.285
1:=0.0053
RF
n
110
u
Q 100
..J
90
JJo~~i~tM~$t~L~~
,
,/'i:'_~:;'~~:~~/('::~] ~::':')~::;N:j:
lt+j++1+,I;t;1'I=ri:'~r'~""""
(1)=0.200
130
\1'=0.433
80
70
~L'~'L'~'~'~'_L~~L~~~
100
Lt'O
300
400
500
GOO
700
800
900
1000
f [Hz]
100
200
300
400
500
(,00
700
800
900
1000
( [Hz]
Figure 5.24: Impeller blade pressure spectra of an axial fan with outlet guide vanes (D
450mm; Z = 24; n = ILiaD/min; after Kameier (1994)).
forces. Ffowcs Williams & Hawkings (1969b) pointed out that a dimensional analysis of fluc100
tuating forces radiation yields a sound intensity that varies as the sixth power of a typical
velocity, while an eighth power relationship is valid for the quadrupole radiation. Since the
sixth power dependence was observed experimentally for the random part of turbomachinery
noise, it is fair to conclude that the effect of quadrupole radiation may be neglected against
the sound produced by fluctuating aerodynamic forces.
5.5.2
DISCRETE NOISE
It was recognized by Ffowcs Williams & Hawkings (1969b) that the dipolelike force fluctuations on the rotor blades are not the only source mechanism in turbomachinery responsible
for discrete sound. Another  quadrupole type  source is given by the interaction of inlet
flow distortions with the potential flow fields of the moving blades. The resultant sound
spectrum contains the blade passage frequency and harmonics. Cumpsty (1977) concluded
from theoretical considerations by Morfey (1971) and Goldstein, Rosenbaum & Albers (1974)
that the quadrupole radiation becomes important only at blade tip Mach numbers over 0.8.
Therefore this mechanism is important for high speed compressors, aircraft engines, propellers, and helicopter rotors, but irrelevant to the noise of industrial and ventilation fans.
5.6
CONCLUSIONS
From the discussion in the preceding section it appears that monopole radiation (blade
thickness noise) and quadrupole radiation (random noise due to turbulent shear stresses and
discrete noise due to the interaction of inflow distortions with the blade tip potential field)
can be important in some cases where the blade tip mach number is high. However, for
most applications the primary cause of turbomachinery noise are the steady and unsteady
forces that are exerted by the turbulent flow on the blades, vanes and the fan casing. The
blade forces can be periodic as well as random in nature, and consequently the resulting
sound field has discrete and broadband components. In Figure 5.25 the basic aerodynamic
generation mechanisms relevant to turbomachinery noise are summarized. They are grouped
with respect to the acoustic source types they are equivalent to. Also shown are various ways
for setting up steady and unsteady aerodynamic forces.
An alternative way of categorizing the various noise generation mechanisms is to differentiate between rotor self noise (produced by the rotor in uniform flow alone) and interaction
noise, see Blake (1986):
Sources of self noise:
blade thickness noise (discrete)
noise due to steady blade forces (Gutin noise; discrete)
turbulent boundary layer noise (random)
vortex shedding noise (laminar and/or turbulent; narrow band and/or random)
tip vortex flow (narrow band or random)
stalled blade flow (random)
Sources of interaction noise:
101
Turbomachinery Noise
I
I
Monopole
Dipole
Quadrupole
blade forces
discrete + broadband
turbulence (broadband)
inflow/rotor (discrete)
I
I
(GUTINnoise)
discrete
discrete
discrete
broadband
I
Uniform
stationary
flow
Nonuniform
stationary
flow
Nonuniform
unsteady
flow
discrete
continuous
broadband
Secondary
flows
Vortex
shedding
discrete
narraw band
broadband
broadband
Turbulent
boundary
layer
broadband
102
5.7
BIBLIOGRAPHY OF CHAPTER 5
ARCHIBALD, F. S., 1975. The laminar boundary layer instability excitation of an acoustic
resonance. Journal of Sound and Vibration 38, 387402.
BLAKE, W. K., 1986. Mechanics of FluidInduced Sound and Vibration. Academic Press
Inc. New York, USA.
BRIDELANCE, J. P., 1986. Aeroacoustic study of axial fans with small diameter. Analysis
and suppression of instability noise. In Proceedings Internoise '86 (Cambridge, USA),
pp. 141146.
BROOKS, T. F. & HODGSON, T. H., 1981. Trailing edge noise prediction using measured
surface pressures. Journal of Sound and Vibration 78, 69117.
CHOU, S. R., 1990. A study of rotor broadband noise mechanisms and helicopter tail rotor
noise. NASACR 177565, National Aeronautics and Space Administration, USA.
CUMPSTY, N. A., 1977. Review  A critical review of turbomachinery noise.
Transactions, Journal of Fluids Engineering 99, 278293.
ASME
DEMING, A. F., 1937. Noise from propellers with symmetrical sections at zero blade angle.
N ACA Technical Memorandum 605, N ACA.
DEMING, A. F., 1938. Noise from propellers with symmetrical sections at zero blade angle,
II. NACA Technical Memorandum 679, NACA.
ERNSTHAUSEN, W., 1936. Die Quelle des Propellergerausches. LuJtfahrtforschung 8,433440. Translated as "On the Origin of Propeller Noise" NACA TM 825, 1937.
ERNSTHAUSEN, W., 1938. Untersuchungen tiber das Luftschraubengerausch. Akustische
ZeitschriJt 3,141146.
FFOWCS WILLIAMS, J. E. & HALL, L. H., 1970. Aerodynamic sound generation by
turbulent flow in the vicinity of a scattering halfplane. Journal of Fluid Mechanics 40,
657670.
FFOWCS WILLIAMS, J. E. & HAWKINGS, D. L., 1969a. Sound generated by turbulence and
surfaces in arbitrary motion. Philosophical Transactions of the Royal Society (London)
A 264, 321342.
FFOWCS WILLIAMS, J. E. & HAWKINGS, D. L., 1969b. Theory relating to the noise of
rotating machinery. Journal of Sound and Vibration 10, 102l.
FUKANO, T., KODAMA, Y. & SENOO, Y., 1977. Noise generated by low pressure axial
flow fans. Journal of Sound and Vibration 50, 6374.
FUKANO, T., TAKAMATSU, Y. & KODAMA, Y., 1986. The effects of tip clearance on the
noise of low pressure axial and mixed flow fans. Journal of Sound and Vibration 105,
291308.
GOLDSTEIN, M. E., ROSENBAUM, B. M. & ALBERS, L. U., 1974. Sound radiation from
a high speed axialflow due to Che inlet turbulence quadrupole interaction. N ASATN
D7667, National Aeronautics and Space Administration, USA.
103
GOLDSTEIN, M. E., 1974. Unified approach to aerodynamic sound generation in the presence of solid boundaries. Journal of the Acoustical Society of America 56, 497509.
GOLDSTEIN, M., 1976.
York.
GUTIN, L., 1936. On the sound field of a rotating airscrew (English translation of Russian
title). Zhurnal Technicheskoi Fiziki 6,889909. English translation of Russian original
published as N ACA TM 1195, 1948.
GUTIN, L., 1942. On the rotational sound of an airscrew (English translation of Russian
title). Zhurnal Technicheskoi Fiziki 12, 7683. Translated as British National Lending
Library for Science and Technology RTS 7543.
HAWKINGS, D., 1971. Multiple tone generation by transonic compressors. Journal of Sound
and Vibration 17, 241250.
KAMEIER, F., NAWROT, T. & NEISE, W., 1992. Experimental investigation of tip clearance noise in axial flow machines. In Proceedings DGLR/AIAA 14th Aeroacoustics
Conference (Aachen, Germany), Deutsche Gesellschaft fur Luft und Raumfahrt, Bonn,
Germany, pp. 250259.
KAMEIER, F., 1994. Experimentelle U ntersuchung zur Entstehung und Minderung des
BlattspitzenwirbeWirms axialer Stromungsmaschinen. FortschrittBericht VDI Reihe 7,
Nr. 243, Verein Deutscher Ingenieure, VDIVerlag GmbH, Dusseldorf, Germany.
KUROSAKA, M., 1971. A note on multiple pure tone noise. Journal of Sound and Vibration
19, 453462.
LEVERTON, J. W., 1989. Twentyfive years of rotorcraft aeroacuostics: Historical prospective and important issues. Journal of Sound and Vibration 133, 261287.
LONGHOUSE, R., 1977. Vortex shedding noise of low tip speed, axial flow fans. Journal of
Sound and Vibration 53, 2546.
LONGHOUSE, R., 1978. Control of tip clearance noise of axial flow fans by rotating shrouds.
Journal of Sound and Vibration 58, 201214.
MCCROSKEY, W. J., 1982. Unsteady airfoils.
285311.
MORFEY, C., 1971. Tone radiation from an isolated subsonic rotor. Journal of the Acoustical
Society of America 49, 16901692.
MUGRIDGE, B., 1971. Acoustic radiation from aerofoils with turbulent boundary layers.
J07Lrnal of Sound and Vibration 16, 593614.
POWELL, A., 1959. On the aerodynamic noise of a rigid flat plate moving at zero incidence.
Journal of the Acoustical Society of America 31, 16491653.
SCHLINKER, R. H. & BROOKS, T. F., 1982. Progress in rotor broadband noise research.
Preprint A823851D, 38th Forum of the American Helicopter Society, Anaheim, California, USA.
104
I. J., 1964. Sources of noise in axial flow fans. Joumal of Sound and Vibration
1, 302322.
SHARLAND,
SMITH, M.
J. T., 1989. Aircraft Noise. Cambridge University Press, New York, USA.
& MURATA, S., 1975. On the partial flow rate performance of axialflow
compressor and rotating stall. Bulletin of the Japanese Society of Mechanical Engineers
18, 256271.
TANAKA, S.
J. M. & SOFRIN, T. G., 1962. Axial flow compressor noise studies. Transactions
of the Society of Automotive Engineers 70, 309332.
TYLER,
VAVRA,
E., 1976. The acoustic spectrum of axial flow machines. Journal of Sound and
Vibration 45, 165223.
WRIGHT, S.
105
106
Chapter 6
DUCT ACOUSTICS
Contents of Chapter 6
6 DUCT ACOUSTICS
6.1 INTRODUCTION
6.2 WAVE EQUATION FOR FLOW DUCTS WITH FLOW AND THERMAL
BOUNDARY LAYERS. . . . . . . . . . . . . . . . . . . . . . . . . . . . ..
6.3 SOUND PROPAGATION IN RIGIDWALLED RECTANGULAR DUCTS
WITH NO FlOW AND NO TEMPERATURE GRADIENTS . . . . . . . .
6.3.1 GENERAL SOLUTION OF THE HOMOGENEOUS WAVE EQUATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.3.2 BOUNDARY CONDITIONS AT THE RIGID DUCT WALLS .
6.3.3 BOUNDARY CONDITIONS AT THE DUCT TERMINATION
6.3.4 BOUNDARY CONDITIONS AT THE SOUND SOURCE . . .
6.4 SOUND PROPAGATION IN HARDWALLED CYLINDRICAL OR ANNULAR DUCTS IN THE ABSENCE OF TEMPERATURE GRADIENTS AND
MEAN FLOW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.5 SOUND PROPAGATION IN RECTANGULAR DUCTS WITH UNIFORM
FLOW. . . . . . . . . . . . . . . . .
6.6 CONCLUSIONS . . . . . . . . . . .
6.7 BIBLIOGRAPHY OF CHAPTER 6
6.1
107
107
108
110
110
112
114
115
116
121
122
123
INTRODUCTION
Many turbomachines like the fans, compressors and turbines of aircraft engines, the fans
of ventilating or air conditioning systems, or most industrial and power plant fans are connected to ducts. Optimum control, prediction and measurement of turbomachinery noise
in such installation conditions can be achieved only through a basic understanding of both
the generation mechanisms of the noise sources and the way the sound propagates in and
interacts with the environment. Hence, the fundamental knowledge of the sound propagation characteristics of rectangular and circular ducts is the prerequisite for sound prediction
schemes, for the development of sound power measurement procedures inside ducts, and for
107
the design of noise control procedures, including primary methods at the source, conventional
silencers as well as active noise control methods.
Turbomachines are fluidhandling machines, and for this reason the general analysis of
the duct sound field must also include the convective, refractive and scattering effects of
mean flow, velocity, and temperature gradients on the sound propagation and attenuation.
A complete treatment of this topic, however, would go beyond the scope of this chapter, and
therefore only the most important aspects of sound propagation in hardwalled rectangular
and circular ducts with no flow and with uniform mean flow are discussed in the following.
6.2
In a large number of papers (PridmoreBrown (1958), Tack & Lambert (1965), Mungur &
Gladwell (1969), Mungur & Plumblee J1'. (1969), Shankar (1971), Savkar (1971), Mariano
(1971), Unruh & Eversman (1972), Ko (1972), Kapur & Mungur (1972)), wave equations
governing the sound propagation in ducts of various crosssectional shape have been derived
with regard to various conditions of mean flow and transverse gradients of mean flow and
temperature. The basic equations underlying all these derivations are the NavierStokes
equation for inviscid media (conservation of momentum), the equation of continuity (conservation of mass) and energy equation (conservation of entropy). The acoustic fluctuations are
treated as small perturbations of velocity, pressure, density and temperature superimposed
on the steady state values. Combining the linearized expressions of the above equations
leads to the wave equation. If a rectangular duct is considered, the wave equation for the
sound pressure is written in cartesian coordinates (see Figure 6.1) as
dy
0,
108
(6.1)
where U, a o and po are the steady state values of flow velocity, sound speed and density
which are functions of both transverse coordinates, but constant along the duct axis; the
dependence of po and a o on y and z is due to the assumed temperature profile. u, v and w
are the fluctuating components of the particle velocity.
For a circular or annular duct, the wave equation is expressed in cylindrical coordinates
(see Figure 6.1) as
o.
The steady state values, U, a o and po are functions of the radial coordinate only;
acoustic velocity in the radial direction.
(6.2)
Vr
is the
The terms in the first parentheses on the left hand side (1.h.s.) of both equations (6.1) and
(6.2) can be written using the Laplace operator: \l2p . The terms in the second parentheses
represent the material derivative of the sound pressure which includes the convective effect
of superimposed mean flow. The last two terms represent the refraction of sound owing to
the presence of transverse gradients of mean flow and temperature.
The effect of the mean velocity gradients in a duct is that the resulting propagation
velocity of the sound waves varies across the duct. For downstream propagation, the wave
travels faster on the duct axis than near the duct walls, and as a result the sound wave is
bent towards the wall. Accordingly, the wave is bent towards the center of the duct when
the sound propagates against the flow.
The refractive effect of transverse temperature gradients can be explained in a similar
manner. In case of a cooled flow the sound wave is bent towards the wall due to the higher
sound speed on the duct axis, and if the temperature of the wall is higher than that of the
fluid the energy of sound is concentrated in the center of the duct.
The simplest case to be considered is that of an isothermal medium at rest: po
const,
a o = const, and U = O. In this case equations (6.1) and (6.2) reduce to the familiar form of
the homogeneous wave equation:
(6.3)
T'he sound field in the duct is not completely determined by the wave equation alone; the
acoustic pressure must also satisfy the boundary conditions at the duct walls, at the duct
termination and at the sound source.
109
6.3
6.3.1
The sound pressure field in a rectangular duct is governed by the wave equation which is
conveniently expressed in the cartesian coordinate system depicted in Figure 6.1. For an
isothermal medium at rest the wave equation reads:
[j2p
[J2p
[J2p
1 cPp
++=0
ox 2 oy2 OZ2 a6 ot 2
(6.4)
Equation (6.4) can be solved by the method of separation of variables which is outlined
below. Let p, which is a function of the three space coordinates x, y, z, and time t, be
represented by the product of four functions, each depending on only one variable:
p(x,y,z,t)
(6.5)
(6.6)
The expression on the r.h.s. of equation (6.6) is a function of t alone, and the only way it
can equal the 1.h.s., which is a function of x, y, and z only, is for both sides to be equal to
the same constant. For convenience, let
2
1 dT _
a6 T
dt 2
k2

(6.7)
T(t)
ejkaot
+B
ejkaot
(6.8)
Each term on the 1.h.s. of equation (6.6) is a function of one variable only, and by an argumentation similar to that above one concludes that each term must be equal to a constant:
(6.9)
Z dz 2
The solutions of these ordinary differential equations are analogous to that in equation
(6.8), and one can write for the solution of the wave equation (6.4):
110
p(x,y,z,t)
ejkxX)(A2 e jkyy
X (A ejkaot + B ejkaot)
(AI e jkxx
+ BI
+ B2
ejkYY)(A3 e jkzz
+ B3 e jkzz )
(6.10)
Each of the four terms in parentheses on the r.h.s. of equation (6.10) is the mathematical
description of a wave motion, depending on only one variable. The variation of the acoustic
pressure with time is determined by the last term, where A and B are arbitrary constants,
and kao = w is related to the angular frequency of the source fluctuation. The phase of the
fluctuation can be chosen such that B = O. The remaining three terms represent the pressure
field in the duct as a triple product of waves going in the positive and negative X, y, and
zdirection, the corresponding wave numbers being kx, ky, and kz, respectively. Inserting
equation (6.10) into the wave equation yields the socalled dispersion relation between the
wave numbers:
(6.11)
where kx, ky, and kz appear as components of the wave number vector k. Hence, the waves
going in the X, y, and zdirections form resultant waves of wave number k, propagating at
an angle to the duct axis. The reflections of these waves from the duct walls generate an
interference pattern over the cross section of the duct ("standing waves"). These interference
patterns are termed the modes of the duct.
The acoustic velocity components u, v, and ware obtained from the linearized condition
of momentum conservation in a homogeneous medium at rest (Eulerequation):
au
1 ap
.
=    = Jkaou
at
po ax
av
1 ap
.
=    =Jkaov
at
po ay
ow
at
1 ap
(6.12)
=    = Jkaow
po az
Together with equation (6.10) one obtains the acoustic velocity components in the
and zdirection:
X,
y,
u(x, y, z, t) = kk x (AI e jkxx  BI ejkxX)(A2 e jkyy + B2 ejkYY)(A3 e jkzz + B3 e jkzz ) A ejkaot
poao
(6.13)
y
v(:r, y, z, t) = kk (AI e jkxx + BI ejkxX)(A2 e jkyy  B2 ejkYY)(A3 e jkzz + B3 e jkzz ) A ejkaot
poao
(6.14)
w(x,y,z,t) = kk z (AI ejkx''''+B I ejkxX)(A2 ejkYY+B2 ejkYY)(A3 ejkzz_B3 e jkzz ) A ejkaot
poao
(6.15)
111
6.3.2
If the walls of the ducts are rigid, the velocity fluctuations normal to the wall must vanish,
and the consequences of this boundary condition are
v
v
w
W
at
at
at
at
0
0
0
0
o : A2 = B2
d y : sin kyd y
0: A3 = B3
d z : sin kzd z
(6.16)
(6.17)
(6.18)
(6.19)
and
dy
(6.20)
where m and n are integers including zero. The expressions for the pressure and velocity
field from equation (6.10) and (6.13) now become
Pmn
cos
k xmn cos
kpoao
[Amn ej(wt+kxmn X )
+ Bmn
ej(wtkxmnx)]
(6.21)
[Amnej(wHkxmnx) _ Bmn ej(wtkxmnx)]
dz
Here Amn = AAIB2B3 and Bmn = AB1 B 2 B 3; the axial wave number kx is now multi valued
and denoted by k xmn . From equation (6.11) and (6.20) one obtains
(6.22)
For each integer value of m and n, equation (6.21) represents one possible solution of the
wave equation (6.4) together with the boundary conditions in equations (6.16) to (6.19) or,
in other words, one possible mode of sound propagation in the duct. The complete solution
is given by the superposition of all the linearly independent solutions, i.e., the sum over all
possible modes:
00
P=
00
2:
2:Pmn
m=On=O
(6.23)
Wave propagation in the axial direction is possible as long as the axial wave number k xmn
is real or, likewise, k;mn > O. According to equation (6.22), this is true for
2
7r
W> ao ( md )
y
+ (n 7r )2 =
dz
wmn
cutoff
c
= wmn
(6.24)
Below this 'cutoff' frequency w~n' the axial wave number k xmn becomes imaginary, and
the propagation factors in equation (6.21) turn into exp(jwt  Ikxmnxl), which means that
the amplitudes of these modes decay with axial distance from the source, they are 'cutoff'
(this critical frequency is often also called "cuton frequency" , because a new mode becomes
propagational when the frequency is raised beyond this limit).
The discussion above has shown that the effect of the boundary conditions at the wall
is to permit only certain types of fluctuations, that are excited by the source, to propagate
along the duct. Each of these modes is characterized by its amplitude profile over the duct
cross section, depending only on the duct dimensions. A schematic presentation of the first
twelve mode patterns in a rectangular duct with aspect ratio of 3:4 is given in Figure 6.2.
Figure 6.2: Mode patterns in a rectangular duct with aspect ratio 3:4.
The outer rectangle represents the wall of the duct, where the acoustic pressure has a local
maximum. The inner lines indicate the positions of zero sound pressure, and the shaded and
unshaded areas represent the zones of positive and negative amplitude at a fixed instant of
time. The instantaneous profiles of the sound pressure amplitudes are drawn to the sides of
the duct walls.
Any given mode described by fixed values m and n can only propagate above its cutoff
frequency, and thus only a limited number of modes is propagational at any given frequency;
113
the other modes which may be excited by the source decay with axial distance from the
source. The plane wave mode is characterized by m = n = 0, and its cutoff frequency is
w80  O. Hence, plane wave propagation is possible at all frequencies.
Two constants are still unkown in equation (6.21): Amn and Bmn. It will be shown in
the following two subsections that they are determined by the boundary conditions at the
sound source and at the duct end.
6.3.3
As was mentioned before, the left and the right exponential terms in equation (6.21) represent
waves going in the negative and positive x directions, respectively. If the sound source is
located, say, in the y  z plane at x = 0, then for x > 0 the left term represents a
reflected term. If there were no axial change in the medium property or the duct geometry,
there should be no reflected wave, and hence Amn = O. However, if the duct is finite in
length, reflections will occur at the duct end. For a given duct termination impedance,
Zmn = Pmn/umn, one may express Amn in terms of Bmn. In the absence of reflections (i.e.,
Amn = 0), the termination impedance is equal to poaok / k xmn ; this is usually called the modal
impedance which is not a function of x, y or z. For the plane wave mode k xmn = k, and the
modal impedance becomes equal to the characteristic impedance poao.
It is necessary to know the reflection and transmission properties at a duct termination,
not only to describe the sound field in the duct but also to determine the acoustic field
outside the duct radiated from the termination. In general, the reflected waves will even
affect the acoustic pressure and velocity at the sound source, and hence the sound power
output of the source.
If a pressure wave in the form of a single acoustic duct mode is incident on the duct
termination, the reflected and transmitted pressures and acoustic velocities are not confined
to that mode alone. In general, other modes will be "generated" at the termination. Hence,
complete specification of the reflection and transmission process requires that the complete
expressions for the sound fields inside and outside the duct are matched at the termination
by applying usual boundary conditions of continuity of pressure and particle velocity (or
displacement) .
Figure 6.3 shows the pressure reflection coefficient Tp = IAoo/ Boo I of flanged and unflanged
circular pipe ends for the fundamental mode (m, n = 0) as functions of the nondimensional
wave number kR as quoted from Rayleigh (1945) and Levine & Schwinger (1948). R is the
pipe radius. At the low frequency end, the majority of sound energy is reflected back into
the tube. As the wavelength becomes smaller, more and more energy is radiated from the
tube end. The effect of the flange is that less energy is reflected back into the tube than for
the unflanged tube at the same frequency, because of the better impedance match between
inside and outside when the pipe end radiates into a halfsphere only (infinite flange) rather
than into free space (unflanged).
In the past, important papers have been published which are concerned with the calculation of the reflection and transmission properties of duct terminations, not only for an
isothermal fluid at rest but also for flow ducts with transverse gradients of flow and temperature (see Tyler & Sofrin (1962), Morfey (1969), Lin & Martensen (1969), Doak (1973),
Homicz & Lordi (1975), Wright (1972), Lansing, Drischler & Pusey (1970), Lansing & Zorumski (1972), Mungur, Plumblee & Doak (1974)). An interesting result is that any well
cuton mode suffers very little reflection; it carries on past the termination as if it were continuing down an infinitely long duct. Another important feature is that modes that would
114
10


'...
'...
0.8
'.
"'..
OJ
'.
'. "
=
G:i
Unflonged
'.
0.6
" '.
'...
l?u
0.4
e2
'...
'...
Flonged~ _
OJ
~
::J
U1
U1
0.2
OJ
L.
 
D
0.5
1.5
1.0
kR
2.0

Figure 6.3: Pressure reflection coefficients of flanged and unflanged circular pipe ends for
the plane wave mode (after Rayleigh (1945) and Levine & Schwinger (1948)).
be cutoff in an infinite duct can carry appreciable acoustic power down to frequencies an
octave or so below their cutoff frequency when the axial distance between source and duct
end is only of the order of a representative duct width.
6.3.4
A sound source located at, say, x = 0 may be expressed in terms of a fluctuating velocity
distribution or a fluctuating force. If a velocity source is considered, the boundary condition
to be satisfied at x = 0 is that the summation of the axial velocities in the duct over all
modes must be equal to the xcomponent of the source velocity us. When reflections from
the duct end are disregarded, Amn = 0, it follows from equation (6.21) that
_
Us 
f(y,z )
~ ~ 1 k xmn
 ~o~poaokBmncos
jwt _
(m7rY)
d,
cos
(n7rz)
y ejwt
(6.25)
To determine the mode amplitude B mn , both sides of equation (6.15) are multiplied by
cos (m'7ry/d y ) cos (n'7ry/d z ) and integrated over the duct crosssection:
dy
1 k xmn
L
L
k Jo Jo
m=O n=O poao
00
00
rdz
dy
Bmn cos
(m7rY)
dy
(6.26)
cos
(m'7r
dY) cos (n7rz)
d
y
cos
(n'7rz)
d dydz
'Z
Due to the orthogonality of the cosine functions, only those of the integrals on the r.h.s. of
equation (6.26) are nonzero for which m' = m and n' = n, and a simple expression for the
mode amplitude Bmn is obtained:
~~~: k x: n fady fa
dz
f(y,
115
dydz,
(6.27)
where Dij is the Kronecker delta, equal to unity when i = .i and zero otherwise. Owing to
the factor kjkxmn' which according to equations (6.11) and (6.24) is equal to
(6.28)
the mode amplitude Bmn is highly frequency dependent. That mode whose cutoff frequency
is closest or equal to the frequency of the source w will tend to dominate the pressure
field in the duct. Or in other words, if for a given source amplitude f(y, z) the source
frequency w comes near the cutoff frequency w~n of one particular mode, the pressure in
the duct will increase sharply. This is a typical resonance effect. As mentioned earlier in this
chapter (compare equation (6.24)) all the modes which are excited by the source distribution
f(y, z) but whose cutoff frequencies Wmn lie above the source frequency w, cannot propagate
but decay with axial distance from the source.
For more detailed discussions of sound sources and source distributions in hardwalled and
lined ducts, see the papers by Doak (in: Richards & Mead (1968)), Doak (1973), Swinbanks
(1975), and Mariano (1975).
w~n
6.4
The wave equation for an isothermal medium at rest, expressed in cylindrical coordinates
reads
fJ2p
fJ2p
1 op
1 02p
1 02p
(6.29)
(6.30)
~d X
X dx 2
dr2
r dr)
r2 8 d0 2
~1d T
ag T
(6.31 )
dt 2
+ Bl
ejkxX)(A2 e jke (}
+ B2
ejkeO)(A ejwt
+B
e jwt )
(6.32)
(6.33)
Due to the periodicity in the coordinate 0 and the condition that the sound field must be
singlevalued at any position 0, kg is restricted to integers in a cylindrical or annular duct
without radial splitters:
116
kg = 0, 1, 2, 3, ... , m
(6.34)
In that case, the solution of equation (6.33) is given by Bessel and Neumann functions of
order m, and the complete solution of the wave equation (6.29) may be written in the form
where J m and N m are the Bessel and Neumann functions of integer order m. (In case of a
duct with lined radial splitters, the azimuthal wave number kg = m can be fractional and
complex.) The wave numbers are related by
2
k =
(:r
=k;+k;.
(6.36)
The first term on the r.h.s. of equation (6.35) is the wellknown representation of waves
travelling in the positive and negative xdirections. Accordingly, the second term describes
waves going in both circumferential directions. The superposition of both wave types represents waves propagating along the duct and spinning at the same time; according to the
third term on the r.h.s. of equation (6.35), the wave amplitudes are functions of the radial
coordinate.
Such spiralling modes can be excited by the rotating blade of ducted fans or compressors;
they can be related to the modes of a rectangular duct that, as was pointed out by Morfey
(1964), would be formed by cutting the annulus along a radius and "unrolling" it until it
becomes of rectangular crosssection.
As in case of the rectangular duct, the various constants Ai, B i , and ki are determined
by the boundary conditions at the duct walls, the duct termination and the sound source.
For a circular duct, the condition of finite amplitude of the pressure on the axis requires that
B3 = O.
The radial component of the fluctuating velocity can be derived from equation (6.35) by
using the momentum continuity equation
OV r
ot
op
po
or
= JWV r =    .
(6.37)
The boundary condition of zero radial acoustic velocity at the duct wall yields the following
relation,
op
 = 0 at
or
l'
=R
'
(6.38)
117
the radial
number of
number is
the eigenaxial wave
(6.40)
Like in the rectangular duct case, wave propagation in the axial direction is possible as
long as the axial wave number is real, compare equation (6.35). According to equation (6.40)
this is true for
 > krmn
(6.41)
ao
or
w
cutoff
= W~n
(6.42)
The meaning of this cutoff frequency is as previously discussed for the rectangular duct
case: Below the cutoff frequency, the axial wave number k xmn becomes imaginary, and the
propagation factors turn into exp(jwt  Ikxmnxl), i.e., the amplitudes of these modes decay
exponentially with axial distance from the sound source, the modes are "cut off". Once
again the effect of the boundary condition at the duct wall is to permit only certain types
of fluctuations to propagate along the duct. In the circular duct, each of these duct modes
is characterized by its radial amplitude profile, depending only on the duct size. Like in the
rectangular duct case, any given mode described by fixed values m, n can only propagate at
frequencies above its cutoff frequency, and at a given frequency only a limited number of
modes are propagational; the other modes  even though they may be excited by the source
 decay with axial distance from the source.
Like in the rectangular duct case, the axial phase velocity is given by
w
k'
xmn
The azimuthal angular phase velocity can be written accordingly as
a xmn =
w
w
aemn =  = ke
m
(6.43)
(6.44)
Again, the plane wave mode is characterized by m = n = 0, and its cutoff frequency is
equal to zero. This means that plane wave propagation is possible at all frequencies. The
first higherorder mode is given by m = 1 and n = 0, and its cutoff frequency in a circular
duct of diameter d according to equation (6.45) is f{o = 0.586 ao/d.
The various constants A 1 ,2,3 and B 1 ,2 in equation (6.35) are determined by the boundary
conditions at the source and at the duct end. For a more detailed discussion of higherorder
mode sound propagation see the book by Morse & Ingard (1968).
In Figure 6.4 the patterns of four modes in a circular duct are shown in a similar fashion
as the mode shapes in rectangular ducts (compare Figure 6.2). The dashed diametral and
circumferential lines indicate the positions of zero acoustic pressure, and the shaded and
unshaded areas represent the zones of positive and negative pressure at a fixed time instant.
Also shown are the radial and circumferential profiles of the instantaneous sound pressures.
Each mode (m, n) has m nodal planes, extending radially outwards from the axis, and n
cylindrical nodal surfaces concentric with the axis. The spinning motion of a (3,0)mode in
a circular duct is shown schematically in Figure 6.5 .
In Figure 6.6 , the crosssectional patterns of the first twelve modes in a circular duct are
shown together with the corresponding eigenvalues j'(m, n) = krmnR. The cutoff frequencies of the modes (m, n) can be determined from the following relation:
118
(2,1) Mode
c
Jmn
"(
= J
) ao
(6.45)
m,n 27fR
Upon inserting the condition for mode propagation (equation (6.42)) into equation (6.44),
one obtains the following relation for the minimum azimuthal phase velocity of a propagational mode:
c
aO mn
w~n
aokrmn
= ;;;: =
m
mn 
aemnR
ao
aoj'(m, n)
mR
Ma~n
(6.46)
j'(m, n)
m
(6.47)
Comparing the eigenvalues j'(m,n) given in Figure 6.6, one finds that Ma':nn > 1, which
means that the azimuthal duct modes (m, n) in a circular duct must rotate at a peripheral
speed higher than the speed of sound to propagate as sound waves. The lower the mode
order m, the higher the critical peripheral Mach number Ma':nn' This result also applies to
119
Figure 6.5: Schematic presentation of the (3,O)mode propagating down a circular duct (after
Ghiladi (1981)).
n =1
n =0
jim,") = 0
jim, n)
5.3311.
6.70613
8.5363
9.9691. 7
=+
L..20 119
=7.0156
jim, n)
(]) @ ~
E8 @) ~
3.051.0
= 3,8317
0
1.81.118
n =2
11.31,59
8.01521.
~ ~

...
Figure 6.6: Amplitude patterns and eigenvalues of the first twelve modes in a circular duct
(after Stahl (1987)).
annular ducts; it was shown by Tyler & Sofrin (1962) that the larger the ratio of inner to
outer duct diameter, the smaller the critical Mach number Ma':nn (see also chapter 7).
120
6.5
The wave equation governing the pressure field in the presence of transverse gradients of
mean flow and temperature has been presented in equation (6.1). If uniform mean flow and
an isothermal medium are considered, oU/oy = oU/oz = op%y = op%z
0, and from
equation (6.1) follows:
( 6.48)
Upon inserting the normal propagation solution for the positive xdirection,
p(x, y, z, t)
= (A2
e jkyy
+ B2
+ B3
e jkzz ) ej(wtkxx) ,
(6.49)
into equation (6.48), one obtains for the axial wave number
(6.50)
where Ma = U/ ao is the flow Mach number. As in the noflow case, the linearized momentum
equation leads to a relation between the fluctuating velocities and the acoustic pressure
OV
Ox
(6.51)
Ow
o;r;
The boundary conditions at the rigid duct wall are the same as in the noflow case and
lead to the same eigenequation for the wave numbers, ky = m7r / dy and kz = n7r / dz , compare
equations (6.16) to (6.20). With these, equation (6.50) can be rewritten in terms of the axial
wave number in the positive and negative xdirection, kx+ and k x , respectively:
k"mn
~ 1 _ ~a' { k Ma
+ C~)']}
(6.52)
The axial wave number is real, when the root in equation (6.52) is real, or in other words
+ C~)
']
( 6.53)
With the cutoff condition for the noflow case in equation (6.24), the following relation can
be derived for the cutoft' condition of the uniformflow case:
(~)2
ao
>
I(W~,n)21
ao
(6.54)
Aifa
121
Note however that sound propagation in the positive xdirection, i.e., in the direction of
flow, requires that the axial wave number is positive; this is true when
(6.55)
>
(W~n)'
(6.56)
Ma
where w~n is the socalled blocking frequency (see Stahl (1987)). In conclusion we note
that the effect of uniform flow in a duct is to reduce the cutoff frequency w~n above which
undamped sound propagation is possible (equation (6.54)), however, up to another frequency
limit, the blocking frequency w~n' the sound wave:: can propagate only in the direction
opposite to the mean flow. For downstream sound propagation, the frequency of a mode
m, n has to be larger than the blocking frequency w~n which is equal to the cutoff frequency
for the noflow case. The above statements as well as equations (6.54) and (6.56) were derived
for the rectangular duct, but they also apply to circular ducts.
6.6
CONCLUSIONS
will tend to dominate the pressure field in the duct (resonance). The plane wave mode
is characterized by m = n = 0, and its cutoff frequency is Woo = O. Hence, plane wave
propagation is possible at all frequencies.
6.7
BIBLIOGRAPHY OF CHAPTER 6
MORSE, P. & INGARD, K., 1968. Theoretical Acoustics. McGrawHill Book Company, New
York.
MUNGUR, P. & GLADWELL, G., 1969. Acoustic wave propagation in a sheared fluid
contained in a duct. Journal of Sound and Vibration 9, 2848.
MUNGUR, P. & PLUMBLEE JR., H., 1969. Propagation and attenuation of sound in a softwalled annular duct containing a sheared flow. Tech. Rep. SP207, National Aeronautics
and Space Administration, USA.
MUNGUR, P., PLUMBLEE, H. & DOAK, P. E., 1974. Analysis of acoustic radiation in a
jet flow environment. Journal of Sound and Vibration 36, 2152.
PRIDMOREBROWN, D. C., 1958. Sound propagation in a fluid flowing through an attenuating duct. Journal of Fluid Mechanics 4, 393.
RAYLEIGH, J. W. S., 1945. Theory of Sound, vol. 2. Dover Publications New York.
RICHARDS, E. J. & MEAD, D. J., 1968. Noise and Acoustic Fatigue in Aeronautics. John
Wiley and Sons Ltd., London.
SAVKAR, S. D., 1971. Propagation of sound in ducts with shear flow. Journal of Sound and
Vibration 19, 355372.
SHANKAR, P. N., 1971. On acoustic refraction by duct shear layers.
Mechanics 47, 8191.
Journal of Fluid
STAHL, B., 1987. Experimentelle Untersuchung zur Schallerzeugung durch die Turbulenz
in einer Rohrstromung hinter einer unstetigen Querschnittserweiterung. Acustica 63,
4259.
SWINBANKS, M. A., 1975. The sound field generated by a source distribution in a long duct
carrying sheared flow. Journal of Sound and Vibration 40, 5176.
TACK, D. H. & LAMBERT, R. R., 1965. Influence of shear flow on sound attenuation in
lined ducts. Journal of the Acoustical Society of America 38, 655.
TYLER, J. M. & SOFRIN, T. G., 1962. Axial flow compressor noise studies. Transactions
of the Society of Automotive Engineers 70, 309332.
UNRUH, J. F. & EVERSMAN, W., 1972. The utility ofthe galerkin method for the acoustic
transmission in an attenuating duct. Journal of Sound and Vibration 23, 187197.
VON HEESEN, W. & REISER, P., 1991. Einsatz eines Kanalprufstandes nach DIN 45635,
Teil 9, bei der Parameteroptimierung von Axialventilatoren. In Proceedings Ventilatoren
im industriellen Einsatz (Dusseldorf, Germany), vol. 872 of VDIBerichte, VDIVerlag,
Dusseldorf, pp. 275290.
WRIGHT, S. E., 1972. Waveguides and rotating sources. Journal of Sound and Vibration
25, 163178.
124
Chapter 7
GENERATION OF DUCT MODES
BY TURBOMACHINES AND
THEIR EXPERIMENTAL
ANALYSIS
Contents of Chapter 7
7 GENERATION OF DUCT MODES BY TURBOMACHINES AND
THEIR EXPERIMENTAL ANALYSIS
125
7.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 125
7.2 MODES GENERATED BY A ROTOR ALONE. . . . . . . . . . . . . . .. 126
7.3 DECAY OF NONPROPAGATIONAL MODES. . . . . . . . . . . . . . .. 127
7.4 MODES GENERATED BY ROTOR/STATOR INTERACTION . . . . . . 127
7.5 MODES GENERATED BY THE INTERACTION OF TWO COUNTERROTATING ROTORS OF EQUAL BLADE NUMBER AND SPEED. . .. 131
7.6 ANALYSIS OF DUCT MODES TO DETERMINE THE DOMINANT
AEROACOUSTIC SOURCE MECHANISMS IN A PROPFAN MODEL .. 133
7.6.1 GENERAL REMARKS .. . . . . . . . . . . . . . . . . . . . . . .. 133
7.6.2 TEST FACILITIES. . . . . . . . . . . . . . . . . . . . . . . . . . .. 135
7.6.3 ANALYSIS OF AZIMUTHAL AND RADIAL MODES . . . . . . .. 136
7.6.4 EXPERIMENTAL RESULTS . . . . . . . . . . . . . . . . . . . . .. 136
7.6.5 PREDICTION OF THE FARFIELD SOUND BASED ON NEARFIELD DATA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 142
7.7 CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.8 BIBLIOGRAPHY OF CHAPTER 7 . . . . . . . . . . . . . . . . . . . . . . 142
7.1
INTRODUCTION
Sound propagation in ducts of rectangular and circular cross section in the presence of
uniform mean flow is discussed in chapter 6. To recall, the sound propagation in ducts is
125
governed by the homogeneous wave equation which may include the effects of mean flow,
velocity and temperature gradients. The effect of the boundary conditions at the rigid duct
walls is to permit only certain types of fluctuations that are excited by the source to propagate
along the duct. Each of these acoustic duct modes is characterized by its amplitude profile
across the duct, which depends on the duct geometry and size. In circular ducts, the acoustic
modes are characterized by a spinning motion around the duct axis and a radial amplitude
profile described by Bessel functions. In addition, the sound pressure amplitude may vary
with the azimuthal angle due to the superposition of modes.
In this chapter it is shown how these duct modes are generated by aerodynamic source
mechanisms in turbomachines: By a rotor alone, by the interaction between the rotor blades
and stator vanes, and by two counterrotating rotors. Conversely, it is demonstrated how a
measured sound pressure distribution in a duct cross section can be resolved into acoustic
duct modes to conclude back what the dominant source mechanisms are.
The description of mode generation due to the action of a single rotor and due to rotor/stator interaction follows very closely the original presentation given by Tyler & Sofrin
(1962).
7.2
Consider a rotor with Z evenly spaced blades rotating in a perfectly concentric duct containing uniform steady mean flow. In this situation, an observer in the fixed frame of reference
close to the rotor would sense fluctuations that are associated with the local displacement of
fluid as the result of the moving blades (blade thickness noise) and with the force field due
to steady aerodynamic blade loading (Gutinnoise), compare chapter 5.
Both types of fluctuations are locked to the rotor and, therefore, rotate with the rotor
angular velocity ,0 = 27rn relative to the fixed coordinate system (n = rotor speed). The
fluctuating pressure at angular position B and time t can be expressed in the form of a
travelling wave in the circumferential direction:
p(B, t)
p(B  nt)
(7.1 )
This expression reflects the fact that the pressure at a particular place and time arrives at
another locations B radians forward at a time B/,0 later.
Since the bladelocked pressures are periodic with the rotor geometry, i.e., with 27r /Z,
the circumferential pressure distribution can be expressed as a Fourier series
p(B, t)
= Re
{~ah
ej[hZ(OOi)+iI>hl}
(7.2)
h=l
where ah and <Ph are the amplitude and phase angle of the Fourier components (blade
harmonics). For brevity, the notation of the real part Re{} is omitted in the following. The
angular speed of all of these harmonics is equal to the rotor speed n. Nothing has been
said yet about the radial distribution of the fluctuating pressure. This variation may be
incorporated by allowing the amplitudes and phases of each harmonic to be a function of
the radial coordinate. With this equation (7.2) becomes
p(B, t) =
+00
L ah(r)ej[hZ(BOi)+iI>h(r)l
h=l
126
(7.3)
7.3
The rate at which these modes decay is determined by the propagation term in equation
(6.35) which can be rewritten with equations (6.40) to (6.42) to become
(7.5)
For frequencies below cutoff (w
of the form
<
w~n)'
from which is obvious that the modes decay with distance x from the source (evanescent
modes). Tyler & Sofrin (1962) calculated the decay rates for hollow circular and annular
ducts for various azimuthal mode orders m as functions of the "wall Mach number" agR/ aD,
see the results depicted in Figure 7.1. A general observation for all duct geometries is that
the higher the azimuthal mode order m is, the more rapidly the mode amplitudes decay.
On the other hand, the higher m, the smaller the critical Mach number Ma':nn above which
undamped propagation is possible. Also visible in Figure 7.1 is what was stated before: the
larger the ratio of inner to outer duct diameter, the smaller the critical Mach number Ma':nn'
7.4
The physical mechanism of the rotor/stator interaction was described in chapter 5 on aerodynamic sound generation mechanisms. To evaluate the azimuthal modes generated by the
rotor/stator interaction, we consider a rotor with Z evenly spaced blades and a stator with
V evenly spaced vanes. The stator may be placed upstream or downstream of the rotor.
Figure 7.2 shows in a schematic form the case of a rotor with Z = 8 blades interacting with
stators of a) V
6 and b) V = 9 vanes. The rotor is represented by the 8 spokes in the
center, one of which is marked with a dot, and the stator vanes lie concentric around the
rotor.
127
1000
J:
..
le>
1000
J:
I
'"
.J
"
100
I.
12
.J
I.
"
100
"':;;
:::> ....
"':;;
::> ....
<~
5<t~"
0"
0::,<::
0:::::
0:.0
c:.o
w..,
W..,
cc~
CC~
m'l
10
10
.J
:;;
.J
c..
'"
1~~~7~~~~~~~~~~~~~~~~
.3.1.!I
,6
.7.8
.9
lO
1,1
I~
I.l
1.04
I.~
1.61.7
Ie
',l
."'"
.~
1000
J:
to
1.1
1.21.3
I.~
1.4
16
'7
J:
e>
z
w
le>
~l
.J
100
16
"':;,
2,
0"
;;!;g
.....
0::.4
~
.,
1000
l
w..,
cc<l
.7,8
"
.,
.J
"

v>';;
:::>,
5 ..
<t<l
0:,
0:.0
w'O
10
CC<l
m'l
10;_ _ __
~o
~
.J
c..
<Jl
1~~~.4~
.~
. .~7~.8~.9~10~~lI~I~2~1~~~'4~,~~~,76~'.7~'"
I.~)~A~
. ~:.1~~.'.~':''''''0LL~'L.'2:L:,:,):,,:~,~~;';,,:;I";.7:"
(J
= 0.75
Figure 7.1: Decay rates of higherorder modes in circular and annular ducts in decibels per
duct radius length (after Tyler & Sofrin (1962)).
The pressure fluctuations generated by the rotor at a point near or on a stator vane
repeat every time a blade passes by and, hence, are periodic with time, the fundamental
frequency being equal to the rotor angular velocity multiplied by the blade number Z. The
periodic pressure fluctuations can be expressed in terms of a Fourier series,
(7.7)
where ah(O) and <h(O) are amplitude and phase of the blade tone harmonic h which may
vary with angular position. Expanding the complex amplitude of the blade tone harmonic
{lh(O) = ah(O) exp[j<Ph(O)] into a series of azimuthal modes with modal amplitudes amh and
phases <Pmh yields
(7.8)
m==oo
128
I.e
& _ l{~.oj
.
~\~\
..
I.~ ~ ~
b}~\~)
~
_
Figure 7.2: Schematic presentation of the interaction between rotor and stator (after Tyler
& Sofrin (1962)).
Inserting equation (7.8) into equation (7.7) leads to:
p(O, t) =
+00
+00
L L
amh ej(m(lhZOt+iI>mh)
(7.9)
h=l m=oo
Equation (7.9) can be interpreted as the pressure time history at position 0 and time t which
is produced by the interaction of the rotor blades with one stator vane, say vane #1; such
events occur at each vane. The effect of the rotor interacting with the entire vane row is
found by superposing the effects on the single vanes, in the correct space and time sequence.
The time required for a particular rotor blade to turn from vane #1 to vane #2 (t6.0 = 27f IV)
is given by t6.t
t6.0 In = 27f Ivn. The pressure generated at vane #2 is the same as at
vane #1, however at the angular position 0 + t6.0 and time t + t6.t. To evaluate what the
contribution of the interaction at vane #2 is to the pressure at position 0 and time t, one
has to replace 0 by 0  t6.0 and t by t  t6.t in equation (7.9). Consequently, the contribution
of the interaction of the rotor with the (l 1)th vane is obtained by replacing 0 by 0  lt6.0
and t byt lt6.t in equation (7.9). Summing over the contributions of all interaction events,
i.e., l = 1 to V, leads to
p(O, t) =
+00 +00
L L
h=l m=oo
amh
ej[m(elb.O)hZO(tlb.t)+iI>mhl
1=1
129
(7.10)
p(B,t) =
+00 +00
I:: I::
amh ej(mIJhOZt+iI>mh)
lint
(7.11 )
h=l m=oo
. I mt
~e
j[21fI( hZ /V m/V)]
(7.12)
1=1
m/v)
V for s = hZ/V
o for s # hZ/V _ m/V ,s = .. , 1,0,1, ..
(7.13)
p(B, t) = V
+00 +00
I:: I::
amh ej(mIJhOZt+iI>mh)
(7.14)
h=l m=oo
where due to the interference term the azimuthal mode order is restricted to the following
values:
kg = m = hZ + sV
(7.15)
and the azimuthal phase speed is given by
hZO
ao = hZ + sV
(7.16)
Equations (7.15) and (7.16) are now applied to the examples shown in Figure 7.2: In
position "0" of Figure 7.2a, the marked rotor blade coincides with the right hand vane,
and a similar event occurs on the opposite side. The coincidences are identified by arrows.
Successive positions show that the coincidence pattern rotates faster that the rotor itself; in
the particular case considered here at four times the rotor speed. For h = 1 (blade passage
frequency) and s = 1, equation (7.15) predicts m = 8  6 = 2, which is interpreted as the
number of lobes (coincidences) along the periphery, and equation (7.16) gives aIJ = 40 for
the angular phase speed.
Similarly, Figure 7.2b represents a rotor with 8 blades interacting with a stator of 9
vanes. Here, coincidence of blades and vanes occurs only once along the periphery, and the
coincidence pattern rotates in a direction opposite to that of the rotor, at a speed of eight
times the rotor speed. According to equations (7.15) and (7.16), m = 1 and ae = 80, for
h = 1 and s = l.
The azimuthal mode orders and angular speed of modes generated by the interaction of
a rotor with nonuniform How can be obtained from equations (7.15) and (7.16) by letting
V=l.
The case of a rotor alone in uniform How is also governed by equations (7.15) and (7.16)
by letting V = O. For this case the azimuthal mode order is m = hZ, and the angular speed
is equal to that of the rotor, ae
0, compare section 7.2.
The effect of a higherorder duct mode generated just below and just above its cutoff frequency was shown experimentally by von Heesen & Reiser (1991); in Figure 7.3 are
130
x = 3.25 m
120
+
X
= 2518/min
. n = 2387/min
100
80
5000
f. Hz
Figure 7.3: Sound pressure spectra in the outlet duct of an axial fan (D=630mm, Z=8, V=9;
after von HeetJen & Reiser (1991)).
shown the sound pressure spectra measured in the anechoic outlet duct of an axial fan with
D = 630mm impeller diameter, Z = 8 impeller blades and V = 9 stator vanes. A slight
change of the fan speed from 2387/min to 2518/min results in a drastic increase of the blade
passage frequency level. According to equation (7.15), the lowest azimuthal mode order of
the blade passage frequency component, which is generated by the rotor/stator interaction,
is m = Z  V = 1. The cutoff frequency of this mode in a 630mm diameter duct is
319Hz. The blade passage frequency of the fan is 318.3Hz at n = 2387/min and 335.7Hz
at n = 2518/min, i.e., just below cutoff at the lower speed and above cutoff at the higher
speed.
The corresponding variation of the blade passage frequency level with the impeller speed
is shown in Figure 7.4 for two axial distances x between impeller and measurement position.
The vertical line marks the fan speed at which the (I,O)mode of the blade passage frequency
component becomes propagational. The cutoff or cuton effect is clearly visible in the
acoustic far field (x = 3.25m) while close to the fan (x = LIm) the tone level is high even
below the cutoff frequency. Below cutoff, the (I,O)mode decays with distance from the
source which is the reason for the level difference found at the two measurement positions.
7.5
This case was treated by Holste & Neise (1992b) in a way similar to the considerations by
Tyler & Sofrin (1962) regarding the interaction between rotor and stator: Since both rotors
have the same number of blades, the pressure fluctuations due to the interaction between
rotor 1 and rotor 2 are periodic with time, the fundamental frequency being equal to the
rotor angular velocity n multiplied by the blade number Z, and thus can be expressed as in
131
x = 1.10 m
110
100
90
1"'\
.
f
+'++IV"."+
110
100
90
2000
2400
n,1/min
2800
Figure 7.4: Variation of the blade passage frequency level of an axial fan with fan speed for
two axial distances from the impeller (D=630mm, Z=8, V =9; after von Heesen & Reiser
(1991)).
the case ofthe rotor/stator interaction, i.e., by equation (7.7). Again the complex amplitude
of the blade tone harmonic !J.h(O) = ah(O) exp[j<h(O)] is expanded into a series of azimuthal
modes according to equation (7.8), and one finally arrives at equation (7.9) for the pressure
fluctuations generated by one of the rotors.
In the present case of rotor/rotor interaction, equation (7.9) is interpreted as the pressure
time history at position 0 and time i which is produced by the interaction of the blades of
rotor 2 with one blade of rotor 1. Since both rotors have the same number of blades and the
same speed in opposite directions, the pressure pattern due to the rotor/rotor interaction
is repeated 2Z times over one full revolution. Hence, if rotor 2 moves an angular distance
~O = 271" /2Z during a time interval ~t = ~O /0" the pressure generated by the interaction
with the next blade of rotor 1 is the same as described by equation (7.9), but at a location
0' = 0 + ~O and at the time i' = t +~t. To evaluate what the contribution of this interaction
to the pressure at position 0 and time t is, one has to replace 0 by 0  ~O and t by t  ~t.
Consequently, the contribution of the interaction of rotor 2 with the (I  1)th blade of rotor
1 is obtained by replacing 0 by 0  I~O and t by t  l~i in equation (7.9). Summing over
the contributions of all interaction events, i.e" I 1 to 2Z, leads to
p(O, t)
+00 +00
=
L L
2Z
amh
h=1 m=oo
ej[m(OIt.B)hZO(tl~t)+<l>mhl
(7.17)
1=1
p(O, t)
+00 +00
=
L L
h=1 m=oo
132
(7.18)
. _
] mt

2Z
~
L...,
1=1
ej [21rI(h/2m/2Z))
(7.19)
2Z for
o for
h/2
m/2Z)
=
,8
.. ,
1,0,1, ..
(7.20)
From this, the azimuthal mode orders are derived which are generated by the interaction
between the counterrotating rotors
m = Z(h  28)
(7.21)
Using a different approach, Hanson (1985) arrived at the same result for the interaction
between unducted counterrotating propellers.
7.6
7.6.1
In this section an experimental investigation by Holste & Neise (1992b) is described in which
a modal decomposition technique was employed to determine the dominant generation mechanisms of the tonal noise components of the propfan model CRISP. CRISP stands for Counter
Rotating Integrated Shrouded Propfan; it is a novel concept for future aircraft engines which
was introduced by MTU (Motoren und TurbinenUnion Munchen) and partners for use in
commercial aircrafts. The envisaged bypass ratio is up to 1:20 with a predicted savings in
specific fuel consumption of nearly 20%. The fan producing the bypass flow comprises two
counter rotating rotors driven by a core engine (gas turbine).
A scaleddown model was constructed by MTU for aerodynamic and acoustic testing; a
schematic of this propfan model is depicted in Figure 7.5. The two counter rotating blade
rows are 400mm in diameter and have 10 blades each. The rotors are driven by a compressed
air turbine via a spider gear. The shroud is supported by 7 struts placed downstream of
the second rotor. The inner diameter of the shroud in the outlet plane is O.4m, and the
hubtotip ratio is 0.5.
The sound field in the exit plane of this model propfan was measured by using conventional 1/ 4inch microphones with nose cone. Initial tests proved that the presence of the
microphone probes in the exit flow does not interfere with the sound pressure distribution
to be measured. The spatial distribution of the complex acoustic pressure is then resolved
into azimuthal acoustic modes with the following objectives:
There are four source mechanisms responsible for the radiation of the blade tone fundamental and its harmonics: The interaction between the inlet flow and the first rotor
133
Microphone
paths (lltp = 30)
Measurement
plane
7 Struts
Figure 7.5: Schematic of the CRISPO.4mmodel (after Holste & Neise (1992b)).
blade row, the interaction between the two rotors, the interaction between the second
rotor and the struts, and, finally, the interaction between the wake flow from the first
rotor, which is convected through the second row of blades, and the struts. Each interaction process generates a characteristic set of spinning modes, and if the resulting
mode distribution in the exit plane of the propfan is known, one is able to conclude
what the dominant source mechanisms for the various tone components are.
Knowledge of the dominant acoustic modes is necessary to optimally tune the acoustic
treatment (lining) of the shroud to achieve maximum noise reduction.
Establish an experimental technique for assessing the tonal noise characteristics of a
propfan in the absence of acoustic freefield conditions.
Each of the various source mechanisms responsible for the blade passage frequency component and its harmonics generates a set of azimuthal modes, the orders of which follow
from sections 7.2, 7.4, and 7.5:
uniform inlet flow/rotor 1: m
hZ
+ sV
m = hZ + sV
rotor 2/struts: m = hZ
rotor 1/struts:
7.6.2
TEST FACILITIES
One test configuration of the CRISPO.4m model was measured in the open test section of
the GermanDutch Wind Tunnel (DeutschNiederlandischer Windkanal, DNW). For these
experiments, a microphone traversing mechanism was used, which was mounted directly on
the hub of the propfan model, see the photograph depicted in Figure 7.6. There are two
Figure 7.6: View of the microphone traverse system mounted in the exit plane of the CRISPO.4mmodel (after Holste & Neise (1992b)).
supporting rods placed at 180 angular distance, each carrying three 1/ 4inch microphones
with nose cones. The radial distances of the microphones from the axis are r / R = 0.96,
0.785, and 0.61 on both sides (R = shroud radius). The two microphone rakes are mounted
on a drum to be traversed simultaneously over an angular span of 180 with an angular
distance of 3 between measurement points.
To perform a spinning mode analysis, the circumferential distributions of both magnitude
and phase of the fluctuating pressure need to be measured. This was accomplished by using
up to three twochannel FFT analyzers hp 3562A. A oneperrevolution trigger signal was
used to obtain the complex pressure spectra averaged in the time domain and relative to a
fixed rotor position. This way all nonrotational signal components were suppressed which is
important in view of the turbulent pressure fluctuations of the jet flow. To eliminate effects
of speed variations, the FFTanalyzers were operated in an external sampling mode, again
using a signal derived from the shaft rotation.
135
7.6.3
The circumferential sound pressure distribution in the exit plane can be described in terms
of azimuthal acoustic modes (compare chapter 6):
E((), r, t) =
+=
2.:
Am(r)eimOeiwt
(7.22)
m==
Here the complex value p( (), T, t) describes the amplitude and phase, and Am(T) is the complex
amplitude of the azimuthal mode m. Am(T) can be resolved into radial modes n by using
the Bessel and Neumann functions:
+=
Am(r) =
(7.23)
n=o
P..w(r, 0)
(7.24)
m=(X)
Experimentally, the modal amplitudes Am(r) can be determined from the measured circumferential distribution of the complex pressure by means of a Fourier transform with
respect to the circumferential angle 0; this requires equidistant measurement positions. Alternatively, a linear system of equations can be established by approximating the measured
distribution by the least square fit method.
In the present study measurements were made at 3 intervals which allows analysis of
azimuthal modes up to orders of m = 59. For some conditions, both methods described
above were used to determine 119 azimuthal modes, which led to the same results.
Radial modes n (A:nn in equation (7.23)) were determined by using the azimuthal mode
distributions obtained for different radial distances from the axis by establishing a linear
system of equations by the last square fit method for each azimuthal mode m.
7.6.4
EXPERIMENTAL RESULTS
A spectrum of the unsteady pressures measured in the exit plane of the propfan model
is depicted in Figure 7.7. Clearly, the spectrum is dominated by the spikes at the blade
passage frequency and its harmonics. In Figure 7.8 the measured pressure amplitude and
phase of the third blade tone harmonic are plotted versus the azimuthal angle. The two
curves shown represent the cases of low and increased turbulence intensity of the wind
tunnel flow. There are hardly any differences between the two cases which proves that the
wind tunnel turbulence did not affect the tonal noise generated by the propfan model. As
a byproduct, the data document the excellent measurement accuracy and reproducibility of
the test equipment used.
The azimuthal distributions of the complex pressure amplitude are resolved into a distribution of acoustic modes. A typical result is shown in Figure 7.9 for the first six blade tone
harmonics. The blade stagger angle is the same for both rotors, f3
_6 /  6, the wind
tunnel flow Mach number is Ma = 0.22 at zero angle of attack with respect to the propfan
axis, a = 0, and the relative pressure ratio of the fan 7r r cl = 0.8; 7r r cl is representative for
the engine thrust. The mode distributions measured on different radial distances from the
axis are labelled by different line symbols, while the various point symbols indicate which
136
} 10 dB
OJ
"U
GJ
>
Measurement
GJ
GJ
'
::J
U1
U1
GJ
'
Cl..
"U
C
::J
(j)
20
Figure 7.7: Spectrum of the pressure fluctuations in the exit plane of the CRISPO.4mmodel.
interaction mechanism is responsible for the particular azimuthal mode. Two vertical arrows
in each graph mark the range of propagational modes in the annulus between shroud and
hub.
For all harmonics, only the propagational modes occur in the exit plane, the other modes
are of much lower amplitude. This indicates that the shroud helps to reduce the noise in
comparison with unducted propfans.
For all blade tone harmonics, the mode distributions obtained for the three radial distances from the axis exhibit the same characteristic behavior, i.e., the dominant modes are
the same at all three radii. The amplitude of a particular azimuthal mode depends on the
radial position which indicates that higher radial modes are generated as well. The existence of radial modes can be explained by the fact that the flow profile of the blade wakes
is nonuniform in the radial direction, in both magnitude and flow angle.
At some blade tone harmonics, modes occur which are neither generated by the interaction between the two rotors nor by the interaction between a rotor and the struts. As was
shown by Zandbergen, Laan & Zeegmans (1983), these modes are formed by a transformation of modes at the struts. For example, a mode that is generated by the rotor l/rotor 2
interaction is propagated downstream until it impinges on the struts. Due to its rotation, it
interacts with the struts just like a rotor wake. As a result, this mode is spread into a series
of modes the order numbers of which are spaced by the strut number.
While most of the blade tone harmonics are generated by the interaction between the
counter rotating blade rows, this is not true for the blade passage frequency component. At
this frequency, the rotor l/rotor 2 interaction produces only evanescent modes, the strongest
of which are m =
10 and m = 10. The dominant modes at the blade passage frequency
are m
3 and m
4; the most likely cause of these is the interaction between rotor 2
and struts, however they can also be generated by the transformation of mode m = 10 at
the struts. In any case, enlarging the distance between rotor 2 and struts would lower the
blade passage frequency level substantially, because it would weaken both source mechanisms
involved: the wakes from the second blade row would wash out over a longer distance, and
at the same time the nonpropagational mode m
10 would decay more.
137
CD
D
>
C
0)
::J
CT
0)
'
4
OJ
=
ro
(J)
(J)
ro
Cl.
0)
ro
.D
>I<
(Tl
4
0)
>
0)
(J)
0)
0)
0
180
90
[!
[!
[!
[!
=
0)
120
[!
::J
CT
0)
60)
'
ro
([)
([)
ro
Cl.
"Cl
ro
60
.D
>I<
(Tl
4
120
0)
(J)
ro
.180
0
4
tl
J l
J'
[!
[!
[!
[! [
, [! I
[ t ~
1[ !/
:\1 ! i
i[ !,
II !i
j[ !j
it !j
it !j
j[ !j
1 ,
tI
1
i
i
i
;
;
I
I
[
1~ !
1
I
i! ~\ !
~I
1, ,J
r
j
:'i
~~
i
r
it
I;
'j!
.,
,.:1
o'
.,
1
1
1
i \1,
;
/1
, !
i
i
I
I
.:
/,'1
[!
.~
0)
;~
[!
[!
0)
r.
[!
[!
[!
[!
[!
[!
0)
360
270
180
.."
'
>c
[
[
1 :.
ill "
[
1
1
[
[
:'11
~\i
i 1
i 1
\jl
. ,[
:~\
I:
I:
I:
C
[
[
[
[".,
I"
i!
:l
90
1
I
270
180
Azimuthal angle, degrees
360
Figure 7.8: Azimuthal distribution of amplitude and phase of the third blade tone harmonic
measured in the exit plane of the CRISPO.4mmodel (after Holste & Neise (1992b)).
In the mode distributions of the second harmonic, the mode m = 0 generated by the
interaction between the two rotors dominates the sound field and is mainly responsible for
the noise radiation at this tone component. The second most important modes m = 8
and m = 6 are produced either by the interaction rotor 2/struts or, less likely, by the
transformation of the mode m = 20 at the struts.
The mode distributions for the blade tone harmonics 3 to 6 show that modes which were
produced by a mode transformation at the struts can have amplitudes almost as high or even
higher than the primary mode. For example at the third blade tone harmonic, the modes
m  17,11,18 originate from the mode m = 10. Similarly for the fourth harmonic, the
modes m = 13,15,22 out of m = 20 and m = 22, 13 from m
20.
It is obvious from the above discussion that the levels of the blade tone harmonics 2 to
6 can be reduced most emciently by increasing the distance between the two rotors. Also,
unequal numbers of blades should be used so that interaction modes with high mode orders
138
60
so
~40
~ZO
30
10
3 x blade passing
frequency
10
2Q
30
40
50
60
"0
I
"0
"0
"
"
,1
0
:1
e
<
60
o
o
x
A
so
10 dB
"0
40
30
10
10
20
30
40
SO
60
60
rotor l/rotor 2
rotor 2/struts
rotor l/sfruts
rofor l/rofor 2 and
rotor l/sfruts or rotor 2/struts
other modes
r=l92mm
r=157mm
r= 122mm
SO
40
30
20
10
10
20
30
40
p = 6/6
rotor l/rofor 2
o rotor 2/sfruts
o rotor l/sfrufs
x rotor l/rofor 2 and
rofor l/sfrufs or rofor 2/sfrvts
A of her modes
,.,=0.8
Ma =0.22
,,=0
It
r= 192mm
r=157mm
r= 122mm
50
60
II =6/6
,.,=0.8
Mo=0.22
=0
It
Figure 7.9: Azimuthal mode distributions for three radial distances from the aXIS for the
blade tone harmonics 1 to 6 (after Holste & Neise (1992b)).
are generated which have a lower acoustic radiation efficiency than the low order ones.
In Figure 7.10 is presented the mode distribution obtained for the maximum thrust
condition or, equivalently for the maximum relative pressure ratio 1rrel = 1.0. Wind tunnel
flow Mach number, blade stagger angle, and angle of incidence are the same as in Figure
7.9. As expected, the levels of the dominant modes are increased as a result of the higher
rotor speed, this is particularly true for the harmonics 4 and 6. Also observed is a shift of
the dominant modes to other order numbers. In case of the second harmonic, e.g., the mode
139
I I
V
'f
20
I'j'"
]0
['i"rrTT,
'f
<
60
SO
(0
]0
20
10
0
10
2'0
30
~O
SO
j l l l / ' I I " ' l l ; , , ) ) ) , , / ' " ' 1 ' /"1 ) 1 ) , / ) ) ) , /
3 x blade passing
frequency.
60
so
40
3D
20
60
60
10
10
20
rotor l/rotor 2
rotor 2/struts
rotor l/struts
rotor 1/rotor 2 and
rotor l/struts or rotor 2/struts
other modes
r= 192mm
 r= 157mm
r= 122mm
40
30
20
10
10
20
30
':0
30
<0
so
60
60
50
40
30
20
10
10
2:0
30
40
50
60
o
o
x
so
p = 6/6
,01 = 1.0
Ma =0.22
a=O
o
o
x
1t
rotor l/rotor 2
rotor 2/struts
rotor l/struts
rotor 1/rotor 2 and
rotor l/struts or rotor 2/strvts
other modes
r= 192mm
___ r=157mm
r= 122mm
p = 6/6
,01=1.0
Ma =0.22
a=O
1t
Figure 7.10: Azimuthal mode distributions for the blade tone harmonics 1 to 6 at maximum
thrust (after Holste & Neise (1992b)).
0 is dominant at both thrust conditions (compare Figures 7.9 and 7.10), while the mode
m
8 is reduced by more than 10dB as the thrust is raised; the modes m = 6 and m
13
are increased by about the same amount. The third blade tone harmonic exhibits the same
dominant modes at both thrust conditions with an increase in amplitude with rotor speed
at the higher thrust. From the fourth harmonic on, a larger number of contributing modes
is generated by the rotor l/rotor 2 interaction as the pressure ratio goes up, and some of
these modes are subject to a mode transformation at the struts. Note that the m = 0 mode
m =
140
60
50
40
3D
20
blade pOS:;ing
frequency
i
II
II
60
50
I,
1Al~~t
40
40
'I '
"
50
I
fiG
1
0.
<
1\
11
~~~~~~~rt~~~4~f,M~~~~~~:~
60
so
40
30
20
10
10
20
30
40
SO
60
rotor l/rotor 2
o rotor 2/51ruts
o rotor llstruts
x rotor l/rotor 2 and
rotor llstruts or rotor 2/struts
A other modes
r=l92mm
r=157mm
 r=l22mm
p = 6/6
It tol;::;
0.8
Ma = 0.22
,,= 20
0:
20 0 angle of
an angle of incidence of 0: = 20. Otherwise, the operating conditions are the same as in
Figure 7.9. There is very little difference between the mode distributions obtained for 0: = 0
and 0: = 20, especially for the first three harmonics. Only at the higher harmonics, minor
141
changes in the dominant modes are observed, but they are all generated by the interaction
between the two rotors or by a mode transformation at the struts. The conclusion to be
drawn from this result is that significant flow separation occurs neither at the lip of the
shroud nor on the blades. This is true for the cruise condition (0'
0) as well as for the
start case (0' = 20).
7.6.5
Holste (1994) presented a method to calculate the sound radiated by an aircraft engine into
the acoustic far field based on the mode distribution measured in the acoustic near field,
more specifically, in the intake or exit plane. The sound radiation is simulated by considering
a distribution of ring sources in the engine interior which reproduces the measured near
field pressures and at the same time satisfies the acoustic boundary conditions at the solid
surfaces of the machine (zero acoustic particle displacement). A comparison of the measured
and predicted sound pressure field is shown in Figure 7.12. The sound pressure is plotted
versus the circumferential angle "theta" and the polar angle "phi" which is relative to the
rotational axis of the engine. The agreement between the experimental and predicted sound
pressure contours is excellent.
7.7
CONCLUSIONS
The discussion of the papers by Tyler & Sofrin (1962) and Holste & Neise (1992b) gave insight
into the mode generation process due to the interaction of a rotor with a spatially uniform
steady flow as well as with a nonuniform steady flow, due to the interaction between a rotor
and a upstream or downstream stator, and due to the interaction between two counterrotating rotors with identical blade numbers and speed.
A modal decomposition technique after Holste & Neise (1992b) was described in which
the complex sound pressure field measured in the inlet or outlet plane of a turbomachine is
resolved into a distribution of acoustic duct modes. Since the various interaction mechanisms
responsible for the blade tone spectrum generate different sets of acoustic modes, one can
conclude from the measured mode distributions what the dominant noise generation mechanisms are. Experimental evidence for the applicability of this method was given for the case
of a propfan model. It was shown by Holste (1994) that the mode distributions measured in
the acoustic near field (inlet or exit plane) of an aircraft engine can be used to calculate the
sound field radiated into the acoustic far field. With these methods it is possible to assess
the tonal noise characteristics of an aircraft engine at an early stage of the development of a
new aircraft engine, because the prediction method is based on experimental data which can
be determined in the absence of acoustic freefield conditions, for example in a compressor
test stand environment.
7.8
BIBLIOGRAPHY OF CHAPTER 7
609~617.
HOLSTE, F. & NEISE, W., 1992. Experimental determination of the main noise sources
in a propfan model by analysis of the acoustic spinnung modes in the exit plane. In
142
Q)
....
:::)
Vl
Vl
....Q)
0.
47
theta, deg.
100
110
120 130
140
phi, deg.
...
Q)
:::)
Vl
Vl
....Q)
0.
47
theta, deg.
phi, deg.
70
Figure 7.12: Comparison of the measured and predicted farfield sound pressure distribution
of the third blade tone harmonic of the CRISPO.4mmodel; jet flow Mach number=0.43;
windtunnel flow mach number=O.22; (after Holste (1994)).
143
144
Chapter 8
OPEN ROTORS
Contents of Chapter 8
8 OPEN ROTORS
145
8.1
INTRODUCTION
145
8.2
146
8.3
148
8.4
BIBLIOGRAPHY OF CHAPTER 8
153
8.1
INTRODUCTION
The sound generation by open rotors like propellers, wind turbines, and helicopter rotors
will be discussed next. Some theory will be discussed in this chapter while the applications
on propeller noise, helicopter noise, and wind turbine noise are dealt with in the following
two chapters.
Like any turbomachinery noise, the noise of open rotors consists of tonal and broadband
contributions. The earliest acoustic theories for propeller noise treated the tonal noise contribution, only. Gutin (1936) studied the noise related to the rotating blade forces (now
termed loading noise). Ernsthausen (1936) and Deming (1938) considered independently
the noise created by the displacement of the fluid due to the rotating blades (now termed
thickness noise).
The procedure of Gutin (1936) was to assume that the blade forces can be concentrated in
a point on each blade. Even if the forces on each blade remain constant during one revolution
of the propeller an observer in the far field will realize a periodic pressure fluctuation. This
can be expanded into a Fourier series in the frequency domain. The method of Gutin
(1936) can be extended to include blade load variations as shown by Wright (1969). This is
important for the sound generation of helicopter rotors.
Many observed effects can be explained with the point source method. However, the
precision of predictions must necessarily remain limited. Modern rotor noise studies are
based on the solution of the Ffowcs WilliamsHawkings equation (4.9). Both methods will
be discussed in the following.
145
8.2
The acoustic analogy of Lighthill (1952) with the extension by Ffowcs Williams & Hawkings
(1969a) permits the calculation of the sound generation by any source distribution and this
includes the sound emission of open rotors. The integral solution for the sound pressure in
the far field in free space with the ambient fluid in arbitrary uniform motion was presented
in chapter 4 as equation (4.9). The result of a transformation into the rotating coordinate
system of the rotor (after Goldstein (1974)) was presented for zero ambient flow speed in
form of equation (4.13). This equation can also be extended to include an ambient mean
flow speed parallel to the rotor axis as shown by Hanson (1980). However, this parallelity is
approximately valid only for the cases of propellers but not for helicopter rotors.
The rotorfixed coordinate system has the advantage of enabling closed form solutions in
the far field requiring only some minor simplifications within the source fields. On the other
hand, the spacefixed coordinate system enables the calculation of any problem in the time
domain
The far field solution (4.9) for the space fixed coordinate system must generally be used
for open rotors. The equation is repeated here:
(8.1 )
where
q = pou 2
q
TO
(1 + L) _(1 _
poa6
po) p'.
P
(8.2)
ro ' fro, Un, and C = 1  UTa / ao are defined in chapter 4. The first integral in equation
(8.1) describes the sound due to those velocity fluctuations in the source region that are
directed toward the observer position (with emission angle eo). The second integral is the
result of fluctuations of the surface force component directed toward the observer (loading
noise) and the third integral is the sound due to volume displacement by the moving blades
(thickness noise).
The quadrupole strength qq has to be known in the vicinity of the blades as function of
space and time. The moving dipole strength fi and the normal velocity component Un have
to be known on the surface as a function of time.
Two strategies are used to solve the integrals. Solutions in the time domain are straightforward and require the numerical solution of equation (8.1). The state of this art is described
by Farassat (1986). Solutions in the frequency domain were performed by Hanson (1980)
who derived integrals for the steady loading noise and the blade thickness noise. Parry &
Crighton (1989) presented an asymptotic solution of the Hanson integrals for a symmetric
loading in a uniform stream for subsonic rotors with a large number of blades which was
extended to supersonic rotors by Crighton & Parry (1991) and Crighton & Parry (1992).
U
146
Blade passing
frequency
(forward row)
_ _ _ Measured
'\_ Predicted
steady loading
10dB
'.
60
30
90
'.
120
150
Angle to flight axis
Figure 8.1: Comparison between asymptotic solution of Parry & Crighton (1989) and measurements (Fairey Gannet) for the bladepassage frequency.
2nd harmonic
of blade passing
frequency
/
.\_ _ _ _ Measured
Predicted
steady loading
'<_ _'';\__ Predicted
' , '
thickness
/
j/. .......
10dB
/,
'\
.I ,
/
, I
I
I
"
\
\
\
\
I
\ I
\
\
1/
1/
1/
30
60
90
120
150
Figure 8.2: Comparison between asymptotic solution of Parry & Crighton (1989) and measurements (Fairey Gannet) for the second harmonic of the bladepassage frequency.
The earlier propeller noise theories also produced closed form solutions, but were based on
point source approximations ofthe loading noise integral in equation (8.1) (Gutin (1936) and
Stuff (1982)) and for the thickness noise integral (Ernsthausen (1936) and Deming (1938)).
Figures 8.1 and 8.2 are examples for the good agreement between results of the asymptotic
solution and noise measurements of the Fairey Gannet. Loading noise is seen to dominate
in the rear while thickness noise is directed more forwardly.
For undisturbed inflow conditions there is no sound emission on the rotor axis when
the rotor has at least two equal blades separated equidistantly in the azimuthal direction.
However, a single blade rotor like the one that was developed for a wind turbine will radiate
147
uf
..
rotor
Figure 8.3: Stream tube passing through the propeller plane is accelerated from flight speed
Uf to Uj downstream of the propeller.
toward the rotor axis as a consequence of the rotation of the circumferential blade forces.
The relative importance of loading noise versus thickness noise largely depends on the
helical bladetip Mach number which is defined as the ratio of helical flow speed, Ch, over
the ambient speed of sound, ao. This is normally defined for simplicity by
(8.3)
where u is the circumferential tip speed of the propeller and Uf is the flight speed. Loading
noise dominates for typical rotors (blade thickness of 0.07 times chordlength) and for typical
aerodynamic loadings up to helical bladetip Mach numbers of Mah = 0.6 .. 0.7. Thickness
noise dominates for higher values of Mah.
The value determined with equation (8.3) is slightly smaller than the real helical Mach
number of the blade profile because the axial component U1x of the velocity in the propeller
area is slightly higher than Uf . This is illustrated in Figure 8.3 where the stream tube
passing through the propeller plane is shown. An approximate equation for U1x is
(8.4)
The value of Uj is determined by the thrust of the propeller. U1x is in the order of 8 % larger
than Uf for a general aviation propeller in cruise.
8.3
However, an impression of how rotors generate sound can be obtained by discussing the
generation of discrete tones by the loading noise component described by the second integral
in equation (8.1) with the procedure of Wright (1969).
The first step of the analysis is to consider the sound radiated from a continuous force
distribution along the circumference in a rotor plane defined by
Fi (r, B, i) =
i\ (r ) exp [j (wi 
kg B  1
(8.5)
(0)
XI
Yi
D=
=
Pi
(b)
Pi =
Pi
COS
(wt: kg 8 ~)
/\
IFI sin f3
/\
If I cos f3
Xi
Pi =
R!P! (COS a COS1/Isin ~ sin 8  COS a sin1f!sin ~ COs 8 + sin a cos ~).
Figure 8.4: Coordinate system and force geometry used for equation (8.5). (Adapted from
Wright (1969).)
used and the force geometry.
The sound pressure in the acoustic far field at an observer point Xi, radiated by the force
Fi in the source position Yi(r, B) is given by
p(Xi' i)
JW
47r Rao
x exp
{j [W(i 
koB]}.
(8.6)
Here it was assumed that the observer is positioned in the 1,3 plane ('ljJ
0). 0,0 is the speed
of sound, R is the distance of the observer from the center of the rotor, and IP';(r) I is the
amplitude of the resultant force. See Figure 8.4 for the remaining notation.
The first and the second term within the braces on the first line of the equation represent
the contributions from the drag and the thrust components of the force, respectively. If the
force is distributed over an azimuthal surface element dS
r flr dB,
(8.7)
then the radiation from the whole circumference is obtained by integrating equation (8.6)
over the circumference, and one has
p(Xi' t) =
{cos fJ sin 0
sin
fJ~/(J
}
wr 0,0
(8.8)
This is an expression for the sound pressure in the far field radiated from a single rotating
mode where J ke is the Bessel function of the first kind and order k(J.
The next step of the analysis is to integrate the loading on each blade to obtain a net
loading L acting at an effective radius re of the blade (see Figure 8.5):
D/2
L=
J L(r)
(8.9)
dr.
0/2
L =J L (r) dr
B blades
L(B)
Lo
+L
Ls cos(sB  <Ps),
s=l
150
(8.10)
Figure 8.6: Variation of loading around rotor. (Adapted from Wright (1969).)
00
L(8)
s
Figure 8.7: Azimuthal Fourier components of blade loading. (Adapted from Wright (1969).)
where Lo is the steady blade loading. The variation of the loading around the circumference
is shown in Figure 8.6 and the blade loading harmonics Ls are shown in Figure 8.7.
Each blade loading harmonic generates a set of n rotating pressure modes. Each of these
modes is characterized by the generating blade loading harmonic s and by the mode order
number n:
Pns
+ cos {nEDt
(nB
+ s)() + <Ps}} ,
(8.11)
where
as
Ls/ Lo .
D is the angular speed of the rotor and B is the number of blades of the rotor. The mode
pair in equation (8.11) is of the same form as the rotating force distribution in equation
(8.5) when the imaginary part is suppressed there. The angular frequency is now w nED,
the azimuthal wave number ke = nB s, and the azimuthal phase speed ae = w/k e =
nBD/(nB s). Note that the modes generated by the steady blade loading (s = 0) always
rotate with the angular velocity of the rotor, which is the reason for the low radiation
efficiency of these modes at subsonic operation. Phase speeds higher than the angular
velocity of the rotor are generated only by the unsteady blade loading (s i 0).
151
It was shown by Wright (1969) that of the two cosine terms in equation (8.11) only the
left term (ke = nB  5) radiates sound effectively for 5 ::/ 0, hence the right term is considered
in the following only for 5 = O.
The sound pressure radiated by the n, 5mode can then be calculated from equation (8.8)
for 5 ::/ 0:
p(Xi,t)
Here,
D/2
LT
J L(r)
dr = 27rrJ5(re),6.rja s
(8.13)
(8.14)
(8.15)
(8.16)
(8.17)
TT
FT
Mae
ke
LT cos f3
LT cos f3
wre
ao
nB  5
For the radiation from steady load, equation (8.9) has to be multiplied by 2 to consider
the second term in equation (8.11) (Wright (1969)).
The radiation from all modes n generated by all blade loading harmonics 5 is obtained by
summing equation (8.12) for all nand 5. The radiation efficiency and the polar directivity
of the modes in equation (8.12) are determined by the term
(8.18)
in equation (8.12). This term is plotted in Figure 8.8 as a function of nB for different blade
loading harmonics 5. The higher the blade loading harmonic number 5, the more modes are
generated in the far field. The upper and lower cutoff frequencies ie of the mode spectrum
for which '"'ike vanishes depends on the blade loading harmonic number 5:
ie =
is
or
(8.19)
(8.20)
Equations (8.19) and (8.20) show that the radiation pass band is centered about the
source frequency is or the source mode number nB = 5.
The sound pressure level radiated from the various modes nB that are generated by the
steady load component (5 = 0) is depicted in Figure 8.9 for an observation angle 0" = 30 deg
and various effective Mach numbers Mae. For subsonic speeds the radiated spectrum due to
steady loading falls off rapidly with increasing nB which implies that rotors with large blade
numbers radiate less steady loading noise than rotors with fewer blades. For an effective
Mach number of 0.5, the difference in sound pressure level between rotors with 4 and 20
blades for the first harmonic of the blade passage frequency (n = 1) is 70 dB. However,
152
20 log nB
20
10
m
.:g
I
1
1
,I
Is=12
r>.:x V
I
\I
\1It
\ I
1\
\ I
;\./
X,
"'.
I
I
s=48
/(~:r":~'~l(X/jxrx~\!IX \ '\\~n{j(\(\{"\(l:'~~
1
X, x/x
("\ I X \
s=24
XXx.......
:;q.
'L .
!"X, lX
I "
'
.
x,
/s=48
'X
."\"\
!
i
Ii
'x
x,
i
30
40
504LL~8~~12~~16~L2LOL~2~4l2~8~~~~~~~L~J~4~8~~5~2~
nB
Figure 8.8: Radiation efficiency of radiated modes as function of mode number n and blade
number B for different blade loading harmonics s. (Adapted from Wright (1971).)
higher blade numbers are less favorable near sonic blade speeds, because the nBspectrum
is flat.
The radiation from an unsteady lift component, blade loading harmonic s = 12, is shown
in Figure 8.10 where the radiation pass band is very distinct, in particular for low Mach
numbers. At low values of nB the contribution from unsteady blade loading is small which
together with Figure 8.7 proves that the lowfrequency part of the sound spectrum is governed
by the radiation from the steady loading of the blades. At supersonic speed, the steady and
unsteady lift components radiate equally well.
The point force concept described above seems justified when the wave length is long
compared with both the blade chord and the span of the blade. Better approximations
consider the distribution of the forces along a radial line rather than a single radius as was
pointed out by Goldstein (1976).
8.4
BIBLIOGRAPHY OF CHAPTER 8
D. G. & PARRY, A. B., 1991. Asymptotic theory of propeller nose part II:
Supersonic singlerotation propeller. AIAA Journal 29, 12, 20312037.
CRICHTON,
CRICHTON,
153
120
100
80
~
tIl
.::s
....J
60
0..
(f)
40
._e_e._.._....
..........
_e...._._.,
.
M e =I25
..
.~...........................,
:.\
..........
.'.'.
......... ....................
~~75
"""
............
Figure 8.9: Effect of the effective Mach number Mae upon the sound radiation due to steady
blade loading (s = 0). Observation angle (Y = 30deg. (Adapted from Wright (1969).)
DEMING, A. F., 1938. Noise from propellers with symmetrical sections at zero blade angle,
II. NACA Technical Memorandum 679, NACA.
ERNSTHAUSEN, W., 1936. Die Quelle des Propellergerausches. Luftfahrtforschung 8,433440. Translated as "The Source of Propeller Noise" NACA TM 825, 1937.
FARASSAT, F., 1986. Production of advanced propeller noise in the time domain. AIAA
Journal 24, 4, 578584.
FFOWCS WILLIAMS, J. E. & HAWKINGS, D. L., 1969. Sound generated by turbulence and
surfaces in arbitrary motion. Philosophical Transactions of the Royal Society (London)
A 264, 321342.
GOLDSTEIN, M. E., 1974. Unified approach to aerodynamic sound generation in the presence of solid boundaries. Journal of the Acoustical Society of America 56, 497509.
GOLDSTEIN, M., 1976.
York.
GUTIN, L., 1936. On the sound field of a rotating airscrew (English translation of russian
title). Zhurnal Technicheskoi Fiziki 4, 889909. English translation of Russian original
published as NACATM 1192, 1948.
HANSON, D. B., 1980. Helicoidal surface theory for harmonic noise of propellers in the far
field. A1AA Journal 18, ] 0, 12131220.
154
.~
/"
100
...
/
........
80
CD
"0
.J
(L
Iii
I'
1'/
60
(j)
\\
\
II!I r/
\\
. ,
1
. , 1\
\
.I ,I
I ,I
\
,,
,
I
//
... /\
40
20
M=O'5
e
\
\
M e =O'25
80
Figure 8.10: Effect of the effective Mach number Mae on the sound radiation from unsteady
blade loading (3 = 12). The observer angle is (J = 30 deg. (Adapted from Wright (1969).)
LIGHTHILL, M. J., 1952. On sound generated aerodynamically. 1. General theory. Proc.
Royal Soc. London A 211,564587.
PARRY, A. B. & CRIGHTON, D. G., 1989. Asymptotic theory of propeller noise part I:
Singlerotation propeller. AIAA Journal 27, 9,11841190.
STUFF, R., 1982. Propellerliirm bei Unterschallblattspitzenmachzahlen, Umfangskraft und
Axialkraft. Mitteilung DFVLRMitt. 8217, Deutsche Forschungs und Versuchsanstalt
fiir Luft und Raumfahrt e.V., Koln/Porz, Germany.
WRIGHT, S. E., 1969. Sound radiation from a lifting rotor generated by asymmetric disk
loading. Journal of Sound and Vibration 9, 223240.
155
156
Chapter 9
PROPELLERS
Contents of Chapter 9
9
PROPELLERS
157
9.1
INTRODUCTION .....................
157
9.2
158
9.3
159
9.4
161
9.5
EFFECTS OF COUNTERROTATION
165
9.6
165
9.7
BIBLIOGRAPHY OF CHAPTER 9
167
9.1
INTRODUCTION
Propeller noise obtained renewed interest in recent years with the extension of aircraft noise
legislation to propeller driven generalaviation and ultralight aircraft. The noise emission of
these aircraft is small in comparison to their larger siblings, but these aircraft often operate
on small airports near residential areas and in resort areas with a noise sensitive population.
Additional interest for propeller noise came from manufacturers of new propeller driven
commuter aircraft. Here, the main objective was a reduction of the annoying cabin noise
to improve passenger acceptance of their aircraft in comparison to jet aircraft. Propeller
noise is also involved in discussions about new large propulsion systems with highspeed
propellers (propfans) which were developed as demonstrators for use in large commercial
aircraft. Propfans are seen as an alternative to large fan engines when fuel prices should
dramatically increase in the future.
Propeller noise has also obtained considerable interest in the field of ship propellers for
the reasons of passenger comfort in civilian ships and of detect ability of military vessels.
A voiding cavitation is one important objective, here. Ship propeller noise is not discussed
here.
Experimental data shall mainly be discussed in this chapter.
157
110 _
(a)
100
90
80
70
.
a
a..
::L 60
'"
"
50
Q)
.;
40
>
'"~
120
(b)
on
"
110
::l
100
90
80
. !III
~~I
VvU
70
~!~
60
50
0
10
20
30
Figure 9.1: Measured and predicted propeller noise spectra for two helical bladetip Mach
numbers (a) Mah = 0.71, (b) Mah = 0.86 (from Farassat & Succi (1980)).
9.2
Typical narrowband spectra of propeller noise are shown in Figure 9.1. The peaks in the
spectra are multiples of the bladepassage frequency. The figure is also an example for
the possible quality of computational results based on the integrals of the farfield solution
in the time domain, given by equation (8.1). The relative amplitude of the harmonics of
the bladepassage frequency is a strong function of helical bladetip Mach number Mah as
shown in Figure 9.2 (already shown in Figure 1.29 on page 30). A dramatic increase of
higher harmonics can be seen when the helical bladetip Mach number is increased above
Mah = 0.77. The typical bladepassage frequency of a twobladed propeller of a light general
aviation aircraft during takeoff is 90 Hz. The range between the tenth and the fortieth
higher harmonics is the range of best ear sensitivity. The reason of this increase of higher
harmonic content is the Doppler factor C in the denominator of equation (8.1) which causes
a strong Doppler amplification of the sound amplitude and a compression of the signal in the
time history when the component of the blade velocity vector directed toward the emission
direction 00 (see Figure 3.2 on page 48 for a definition of the emission direction in flight)
approaches Mach number 1.
The influence of helical bladetip Mach number on the decay of the higher harmonics
of bladepassage frequency were also studied by Parry & Crighton (1989). They could
158
120~~~~~
g:
CT =0.06
dB
r /0= 2.0
Cl.>
.l
Cl.>
L...
100
::J
(f)
(f)
Cl.>
L...
D'"0
C
80
::J
U)
60~~~~~
10
20
30
40
Harmonic Order
Figure 9.2: Influence of harmonic order of bladepassage frequency on the sound pressure
level for different helical bladetip Mach numbers Mah. Note the dramatic increase of higher
harmonics, when Mah is increased from Mah = 0.77 to Mah = 0.90 (from Heckl & Miiller
(1994)).
demonstrate with their asymptotic solution that the decay is weak for high and rapid for
low helical Mach numbers.
9.3
The noise of any rotor is increased when the inflow is not uniform, because the blade forces
are no longer stationary in the rotating system.
Unsteady blade loads are generated even without turbulence when the propeller axis
is not aligned with the inflow velocity. This occurs when the airspeeds of propeller driven
aircraft are low and large angles of attack are required to generate the necessary lift. Another
origin of misalignment is the influence of the lift generating circulation around a wing on
the airflow. This causes an upwash in front and a downwash behind the wing which are
especially large when the aircraft operates with high lift coefficients. The situation is shown
in Figure 9.3. Both effects (large angle of attack and upwash) add up for propellers in front
of a wing and are especially strong during takeoff and landing when the airspeeds are low.
The angle of attack and the helical bladetip Mach number are a function of angular blade
position in these cases yielding cyclic propeller loads. The load and helical Mach number
of the downward moving blades are higher resulting in a larger noise emission toward the
ground. Experimental results of the influence of an angle between propeller axis and flow
speed were reported by Dobrzynski (1986).
The disturbances of the mean velocity profile are higher than in the just mentioned
case when the propeller is mounted downstream of obstructions like struts or wings. These
propeller locations were repeatedly selected for improved passenger comfort, e.g., for the
unobstructed view and reduced cabin noise when the engine is positioned behind the passenger cabin of a general aviation (or ultra light) aircraft and the propeller is rear mounted.
Other reasons for this design are a reduced aerodynamic drag of the fuselage or the wing
159

~
Figure 9.3: Illustration of misalignment between propeller axis and mean flow velocity vector
due to angle of attack and wing upwash.
when pusher propellers are installed. The cause of the increased sound generation in these
cases is that the blades have to pass through narrow wakes which causes sudden changes of
blade loads and a subsequent increased contribution of higher harmonics of bladepassage
frequency.
Additional broadband sound is also generated in these cases, because the wakes are always
turbulent and turbulence in the inflow causes increased broadband noise levels. This was
shown by Bottcher & Gehlhar (1993) who studied the influence of nearly isotropic turbulence
on the sound generation of a fourblade propeller (about 1/8th scale of a general aviation
propeller) in a wind tunnel. One result is shown in Figure 9.4 for a helical bladetip Mach
number Mah = 0.76 and three different turbulence levels. It can be clearly seen that the
broadband noise level as well as the levels of the higher harmonics of the bladepassage
frequency are increased. The figure shows the results for an observation angle of 120 degrees
relative to the upstream direction. The observed effects are larger in the rear arc than in
the forward arc. The tonal noise components are also increased despite of the orientation of
the propeller axis in the direction of the mean flow velocity.
A similar result was observed by Dahlen, Dobrzynski & Heller (1988) for a propeller
mounted in the turbulent wake of an obstruction simulating the pilot in front of the propeller
as found in many ultralight aircraft. One result is shown in Figure 9.5. The level of the
bladepassage frequency is unchanged, but the higher harmonics are increased by up to 35
dB for a bandwidth of 3.1 Hz. The broadband level is increased by 20 dB.
While no sound is emitted in the direction of the propeller axis (for a propeller with
a regular blade distribution and identical blades) when the uniform inflow is coaxial, a
considerable noise is emitted in the axial direction when the inflow is distorted by the wake
of an obstruction. This is shown in Figure 9.6. The peak emission angle (observation
angle is shown in the figure) is moved further into the rear arc. The influence of the same
configuration on the azimuthal directivity is shown in 9.7. A negative influence on the sound
emission to the ground during flyover can be avoided, when the rotor direction is chosen such
that the blade crosses the wake in the upward direction. This requires opposite directions
of the propeller rotation for twinengined aircraft when the engines are mounted on struts.
Because of the already high contribution of higher harmonics of the bladepassage frequency
when the inflow is distorted, the effect of an increased helical bladetip Mach number is less
160
110
~
= 120 deg
100
(])
0
90
.S;
0:;
:
'"
e
80
'
:::J
U)
U)
d:'"
70
0
C
:::J
0
CI)
60
50
40
0
10
Frequency in kHz
15
20
Figure 9.4: Influence of grid turbulence on tonal and broadband sound generation of a model
scale propeller with 4 blades (Bottcher & Gehlhar (1993)). Helical bladetip Mach number
Mah = 0.76, propeller diameter D = 0.273 m, tunnel flow speed V f = 15 mis, observation
angle of outofflow microphone () = 120 deg. reI. upstream direction. Turbulence levels are
0.3 %, 7.5 %, and 13.5 % for the three shown spectra.
pronounced than with an undistorted inflow.
It can be concluded that disturbed inflow conditions must be avoided in aircraft designs.
The classic designs with tractor propellers are the quietest.
9.4
The method of a nonuniform blade distribution was already discussed in subsection 5.3.6
on page 87. The sound power at the bladepassage frequency is partly redistributed to
harmonics of the rotational shaft frequency. The frequency spectra contain a more even
distribution of tonal components. It was demonstrated in an experimental study of Landskron & Neise (1989) in a closed wind tunnel for a four and a fivebladed propeller that
a nonuniform distribution of the blades does not reduce the sound power but reduces the
sound's annoyance. One result for the four bladed propeller is shown in Figure 9.8. For rotor
balance purposes, the blade lengths of the nonuniform propeller were also nonuniform. The
propeller shaft speed was slightly higher for the nonuniform propeller to achieve the same
thrust. It can be seen, that the spectral peaks for multiples of the bladepassage frequency
are slightly decreased when the blade distribution is nonuniform at the expense of additional
161
110
Q.J
>
Q.J
....J
6f = 3.1 Hz
dB
E
...... 90
U
:J
L..
Q.J
0.
(j)
Q.J
80
lJ)
z 70
L..
Q.J
OJ 60
0.
0
L..
0...
50
0.8
0.4
1.2
kHz
1.6
Frequency
Figure 9.5: Influence of mean velocity profile and turbulence of an obstruction on the sound
pressure spectrum of a propeller of an ultralight aircraft (Heckl & Miiller (1994)). Harmonics
of the bladepassage frequency as well as broadband noise are drastically increased.
120,,,,,,,
dB
0 ,1,86m
110  MH' 0,7 _~
~
X' 270'
gesliirle
Zuslriimung
CD
0.
'"
2 100  
::c::!
0
.r::
Vl
90
ungesliirle
Zuslriimung
~~~~ ~~~~~~
<:)
Vl
OJ
L!:>
~~I :~;~~:J
I
70'
I
90'
() I
110'
130'
150'
Figure 9.6: Influence of distorted inflow on the polar directivity of the sound emitted by a
propeller (Heckl & Miiller (1994)).
tonal components.
Dobrzynski (1993) studied, how advantage could be taken of the redistribution of acoustic energy to lower frequencies. He predicted with computations in the time domain, and
substantiated this with fullscale windtunnel tests, that the Aweighted sound pressure level
is reduced for nonuniform spacings. The spacing between subsequent blades was optimized
162
110
dB
~rI
o ' 1.86m
II
Azimuloler Abslrohlwinkel X
Figure 9.7: Influence of inflow distortion on the azimuthal directivity of the sound emitted
by a propeller (Heckl & Miiller (1994)). Sound emission is highest in the direction of the
blade motion when it passes the wake.
2000
1000
n
1/min
sym.
 asym.

CD
0
l800
1980
S
N
167
167
Lp
tot
dB
113.2
114.0
LPA
dB
107.9
108.0
. ".
D..
I"
.J
I"
"
11
,t
"
I "
""
80
1000
f, Hz
2000
Figure 9.8: Comparison of frequency spectra of the sound power of a fourbladed propeller
with uniformly (solid line) and ununiformly (dotted line) distributed blades (after Landskron
& Neise (1989)). Helical bladetip Mach number is Mah = 0.6.
for minimum Aweighted sound emission in the propeller plane. This interference effect is a
function of helical bladetip Mach number Mah. Figure 9.9 shows the chosen propeller setup.
Figure 9.10 is a comparison between the theoretical (left side) and experimental (right side)
results. The achieved reduction is 3 dB which decreases in the upstream and downstream
163
(\
./


,,
~Hn~r
\J

I; \
I'
J'
W""..J
I \
I \ 'j
I
\j
" ,
,
,
'~,
~
:
;" 1
r.~
\~
~X2~
102
(a)
(b)
/"~
100
r..
co
''
>
v
U
;:J
98
,../
\
\
/
I
\
\
\
\
\
90
\
\
\
\
I
I
88
50
""\
/
/
92
"\
94
,.. /~
96
Vl
100
50
100
Figure 9.10: Polar Asound level directivities for a propeller with six blades. Dashed line is for
symmetrical blade distribution, solid line for unsymmetrical distribution. (from Dobrzynski
(1993)). Helical bladetip Mach number is Mah = 0.68, propeller diameter D = 1.7 m, blade
spacing angle 20 deg.
directions. For angles 0 > 120 0 (downstream direction) the interference benefits are lost and
the Alevel is even increased.
In a computational study of Lohmann (1993) it was shown that similar interference effects
can also be achieved by asymmetric blade sweep of otherwise uniformly distributed blades.
164
9.5
EFFECTS OF COUNTERROTATION
Counterrotating propellers are attractive for aerodynamic reasons because the angular momentum of the flow that is generated by the first rotor is taken out by the second rotor. Such
an arrangement improves the propulsive efficiency considerably, especially at high subsonic
flight Mach numbers when the advance ratio of the propeller is large and a considerable angular momentum is generated by a single propeller. Therefore, counterrotating propellers
are proposed for propfans and were previously used in highspeed propellerdriven aircraft.
Examples are the Fairey Gannet (carrier aircraft of the late 1950s with a single engine) and
the Tupolev Tu 114 (fourengine long range transport aircraft). Both rotors had identical
blade numbers in these cases and the helical bladetip Mach numbers were supersonic in
highspeed flight, yet the blades were straight.
The sound emission of these counter rotating propellers was very high and had a high
and annoying pitch. Similar results must be expected for future projects although they will
incorporate blade sweep to reduce the effective helical blade Mach number and different
blade numbers on both rotors. The reason is that the blades of the second rotor have to
pass through the blade wakes of the first rotor. The mechanism for the generation of the
additional sound is very similar to the case of inlet distortion. However, the effect is more
pronounced, because the wakes of the first rotor impinge on the blades of the second rotor
with the relative speed between the rotor blades. This has the double value of the stationary
wake leading to higher velocity gradients in the system of the rotating second rotor. This
increases particularly the higher harmonics of the bladepassage frequency and causes the
high pitch.
9.6
L_
90
dB
.J
..
.L   3
.L ._
85
>
.J
t7l
300;=:l
,.r
a
Q
80
0;
w
l) 75
70
0.60
0.65
0.70
0.75
0.80
. 0.85
Figure 9.11: Flyover noise data in level flight versus helical bladetip Mach number. Bladetip Mach number is changed through change of propeller diameter. Decreased diameter is
offset by increased blade number. (Dobrzynski & Gehlhar (1993))
The results of Figure 9.11 show the noise reduction potential of lowering the helical bladetip Mach number. The reduction of the linear sound pressure level is most eminent when the
helical bladetip Mach number is decreased from Mah = 0.83 to 0.75 while the blade number
is doubled from two to four. This is mainly due to a reduction of the higher harmonics of
the shaft rotational frequency. A reduction of the Mach number below Mah = 0.7 is less
effective because the sound pressure level is increasingly dominated by the fundamental of
the bladepassage frequency. A second reason for the levellingoff of the linear sound level
is the constant contribution of engine exhaust noise. Because of the performance loss of
the small diameter propellers, the authors conclude that fourbladed propellers are best for
retrofitting today's aircraft. The noise reduction potential is in the order of 7 dB for an
initial helical bladetip Mach number Mah = 0.83 of the twobladed propeller.
llSPl(dB)
 Asymptotic


~======:::=::::=
L__;m;=11
0..5
0.6
0..7
0..8
Figure 9.12: Reduction of the sound pressure level of the harmonics of the bladepassage
frequency for a propeller with 12 blades and a 50deg sweep at the tip (from Parry & Crighton
(1989)). Emission angle eo = 90 deg.
The beneficial influence of blade sweep can be studied with Figure 9.12. The results of
Parry & Crighton (1989) for a propeller with 12 blades and a 50deg sweep are shown. The
solid lines are numerical solutions of the Hanson integrals, the broken line is the asymptotic
solution which is independent on the product mE 1, where m is the harmonic mode
number and E is the number of blades. It can be seen that blade sweep particularly reduces
the higher harmonics of the bladepassage frequency. The effects are largest for, but not
restricted to, high helical bladetip Mach numbers.
9.7
BIBLIOGRAPHY OF CHAPTER 9
BOTTCHER, J. & GEHLHAR, B., 1993. Experimental investigation of propeller noise under
highly turbulent inflow condition. DLRIB 12993/21, Deutsche Forschungsanstalt fur
Luft und Raumfahrt e.V.
DAHLEN, H., DOBRZYNSKI, W. & HELLER, H., 1988. Aeroakustische Untersuchungen
zum Liirm von Ultraleichtflugzeugen. Forschungsbericht DFVLRFB 8803, Deutsche
Forschungsanstalt fur Luft und Raumfahrt e.V., Koln/Porz, Germany.
167
DOBRZYNSKI, W. & GEHLHAR, B., 1993. Propeller blade number: A parameter for flyover
noise reduction. In International Noise and Vibration Control Conference, (eds. M. J.
Crocker & N. I. Ivanov), vol. 1, pp. 6570.
DOBRZYNSKI, W., 1986. The effect on radiated noise of nonzero propeller rotational plane
attitude. AIAA Paper 861926. AIAA Aeroacoustics Conference, Seattle, Washington.
DOBRZYNSKI, W., 1993. Propeller noise reduction by means of unsymmetrical bladespacing. Journal of Sound and Vibration 163,123136.
FARASSAT, F. & SUCCI, G. P., 1980. A review of propeller discrete frequency noise prediction technology with emphasis on two current methods for time domain calculations.
Journal of Sound and Viobration 71, 399419.
HANSON, D. B., 1979. The influence of propeller design parameters on farfield harmonic
noise in forward flight. AIAA Paper 790609.
HECKL, M. & MULLER, H. A., 1994.
Verlag, Berlin, Germany.
LANDSKRON, R. & NEISE, W., 1989. Gerauschmessungen an Propellern mit ungleichformiger Blatteilung und variabler Blattlange. In Fortschritte der AkustikJ DAGA J89,
DPGGmbH, Bad Honnef, Germany, pp. 587590.
LOHMANN, D., 1993. Numerical optimization of propeller aeroacoustics  Using evolution
strategy. In Proceedings Noise J9B (St. Petersburg, Russia), Interpublish Ltd, St. Petersburg, Russia, pp. 103114.
PARRY, A. B. & CRIGHTON, D. G., 1989. Asymptotic theory of propeller noise part I:
Singlerotation propeller. AIAA Journal 27,9, 11841190. ,
168
Chapter 10
10.1
INTRODUCTION
The main and tail rotors are the dominating sound sources of helicopters. The flow throw a
helicopter rotor is very complicated and only recent research has improved our understanding
of helicopter sound generation. This was achieved by numerical work as well as by substantial
experimental investigations in which the GermanDutch Wind Tunnel, DNW (an acoustic
wind tunnel with an open 8 m x 6 m test section) played a significant role.
Wind turbines have become common in many areas with high average wind speeds. Despite the environmentally welcome contribution to the supply of electric energy without
carbon dioxide emissions, the sound emission of these turbines often prohibits their installation near housing areas despite of the noise levels being small on an absolute scale. The
noise mechanisms of horizontal axis wind turbines are very similar to those of propellers.
Vertical axis wind turbines and exotic designs are not covered here.
169
60 m/s
120 kts
..
tip vortex
of preceding
blade
vertical speed
~...".:..:~~
tip speed
10.2
HELICOPTERS
A survey of the state of the art and of perspectives in helicopter rotor noise research was
presented by Heller, Splettstoesser & Schultz (1993), the current status and future direction
of helicopter noise prediction was described by Brentner & Farassat (1992). An overview is
also included in the book of Heckl & Miiller (1994).
10.2.1
100~~~~~~~~''''''',
dB
Gesoml sc ho II drue kp ege\
901+
0
~ 80~~HH__+~~nrr_P.~_T~r~
:r:
601+ji
Hormonlsche des Heckrotors
501~01LL~LL,L~4L~8~10')'~~~5'H~z~1O4
Frequenz
Figure 10.2: Frequency spectrum of the sound pressure of an UH1A helicopter (from Heckl
10.2.2
The maximum flight Mach number of the helicopter of Figure 10.2 is Maj = 0.16, the
maximum circumferential blade tip Mach number is Ma j = 0.73, which results in a maximum
Mach number of the advancing blade relative to the flight direction of Mab = 0.89 while it is
only Mab = 0.57 on the retreating blade. The helical bladetip Mach numbers are even higher.
Due to the Doppler factor C in the denominators of equation (8.1), Doppler amplification
will be largest toward the flight direction and will affect all noise sources that move with the
blades.
Several sound generation mechanisms are identified for the main rotor. Thickness noise
is generated by the advancing blade, it is of tonal nature, only, and contributes to the bladepassage frequency and its harmonics. Loading noise is also important. The rotating blade
forces can be described in a Fourier series and will also generate tonal components. The
turbulent inflow and the presence of the tip vortices near the blades are responsible for
fluctuating blade forces which contribute to broadband noise.
Two impulsive noise contributions are especially noteworthy, because they are responsible
for the annoyance of helicopter noise. Bladevortex interaction noise is generated when a
blade passes close by the vortex of the preceding blade or even strikes it. This is especially
likely to happen during descend. The most annoying contribution is described as highspeed
171
1000 ,y,,~,,!,.
Hoc hgesc hwindigk eil simpulslo rm
OJ
Cl
.~ 500
OJ
Blollwirbelinlerferenz'impul slorm
I
en
I
1
o~~~~~~P~~+r~!~~D7Ber~eic~h~des~
impulslorms
OJ
~ 500
l223
douerhof!es
Auftreten von
impulstorm
("""I
..
moximoter
impulstor m
.S
en
1000 OL_J20_~LO~60:8~0:1~00::~12~O;1/'~O~~15~O
Ftuggeschwindigkeif in Knolen
Figure 10.3: Flight regimes of bladevortex interaction noise (flight speeds < 100 knots) and
of highspeed impulsive noise (flight speeds> 105 knots) of UH1A (from Heckl & Muller
(1994)) .
impulsive noise. Current understanding is that thickness noise is only partially responsible
for it and that quadrupole noise plays a considerable role. This may be a consequence of the
farreaching influence of a blade in transonic motion on the flow field around it.
The flight regimes in which the two impulsive noise mechanisms appear are shown in
Figure 10.3. Highspeed impulsive noise can be avoided according to Heckl & Miiller (1994)
by limiting the bladetip Mach number to Mab < 0.85 and by use of optimized bladetip
planforms.
In a recent study, an actuator system was designed that permits a higher harmonic control
of the blade pitch. Figure 10.4 demonstrates, how the angleofattack can be arbitrarily
changed with this design. This can be used to improve ride comfort on the helicopter but
also to reduce noise. Results at the German Dutch Windtunnel show that this cannot be
achieved at the same time. Experiments at NASA Langley indicate that the noise level of
the main rotor can be reduced by about 4 dB.
10.2.3
The noise mechanisms for tail rotor noise are the same as for the main rotor. The inflow
for the tail rotor is much more disturbed and the frequency range of the bladepassage
harmonics is higher. The directivity is also directed more to the ground due to the vertical
orientation of the rotor. The tail rotor noise can dominate the total sound emission under
certain conditions, e.g., during climb when the main rotor is relatively quiet.
The latest helicopter designs avoid the problem of tail rotor noise, because the traditional
tail rotor was eliminated. Eurocopter (formerly Aerospatiale) replaced the tail rotor by the
"Fenestron" which is a multibladed axial fan of very light construction. Its noise is masked
by the main rotor. One of the produced models is shown in Figure 10.6 taken from Lambert
(1990). McDonnell Douglas (formerly Hughes) installed a cylindershaped tail boom with
two fixed horizontal slots running the length of its right side. Low pressure air is blown
out and remains attached to the cylindrical boom through the Coanda effect. The force on
172
conventional helicopters
'7
....::::..:::::""" /'
retreating blade
advancing blade
Figure 10.4: Effect of higher harmonic control on the angleofattack of a main rotor blade
during one rotor revolution
Figure 10.5: Helicopter with a Fenestron which replaces the tail rotor.
(1990)).
(from Lambert
the tail boom is sufficient to control the helicopter. The necessary air is supplied by a fan
with a pressure of about 14 kPa (compared to an atmospheric pressure of 101 kPa). The
manufacturer claims a noise benefit of 10 dB (EPNdB ?) in comparison to the competitors.
A produced model is shown in Figure 10.6.
173
Figure 10.6: Helicopter using the Coanda effect for directional control (MD 530N). (from
Lambert (1990)).
10.3
WIND TURBINES
10.3.1
Wind turbines may have many different designs, a selection is shown in Figure 10.7. Most
wind turbines have horizontal axis rotors but some have a vertical rotor axis (Darrieus type)
or even an oblique axis. An advantage of the latter two types is that the machine house can
be placed on the ground. Yet, these types are unimportant and will not further be discussed
here.
The aerodynamic sound generated by horizontalaxis wind turbines is similar to propeller
noise. Tip speeds are generally smaller for wind turbines. Rotor advance ratios (wind speed
over tip speed) are much smaller. This reduces the distance between the blades and the
tip vortices of the preceding blades and enhances bladevortex interaction noise. Boundary
layers are thicker on windturbine blades than on propeller blades. Therefore, the noise due
to the turbulent boundary layers passing the trailing edge of the blades is an important
contribution.
Wind turbines operate in the atmospheric boundary layer. The mean wind speed is a
function of height above ground which creates a cyclic load on the blades. In addition,
the velocity vector of the wind may change direction with time which cannot be followed
immediately by the rotor axis. This yawed operation also introduces cyclic loads. The
turbulence level is high in the atmospheric boundary layer increasing tonal and broadband
noise. Some windturbine rotors operate behind the tower which causes the blade loads to
change abruptly each time a blade passes through the wake. The result is a thumping noise
which may propagate over large distances.
174
C J
'///////
upwind
rotor
Darrieus
downwind
rotor
Western
single bladed
oblique axis
10.3.2
The aim of a wind turbine is to gain mechanical power out of the wind by reducing the wind
speed to a smaller level downstream of the rotor. This shall be discussed with a stream tube
of air passing the rotor of a wind turbine as shown in Figure 10.8. The wind speed upwind
of the rotor is C(XJ which is reduced to a smaller value C3x downwind of the rotor. The index
x indicates the axial component parallel to the rotor axis. A constant wind speed over the
rotor plane is assumed in the following discussion.
The axial component of the wind speed in the rotor plane can be approximated by
Cl x
C2x
(10.1)
The wind power Pw available in a cross section equal to the rotor area A is given by
C
Pw = pC(XJA ; ,
(10.2)
...

   
Figure 10.8: Stream tube of air passing through the rotor of a horizontal axis wind turbine.
The ratio Pm / Pw is given by
(10.4)
This ratio is zero for C3x/ Coo = 1 when the wind turbine is no operating and would have a value
of Pm / Pw = 0.5 for the theoretical limit C3x/Coo = o. A maximum value of Pm / Pw = 0.59 is
achieved for C3x/ Coo = 1/3 when the wind speed is reduced to one third of its original value.
The axial component of the flow speed in the rotor plane is then accoring to equation (10.1)
Cl x / Coo = 2/3.
The mechanical power extracted by the wind turbine can also be described by Euler's
turbine equation,
(10.5)
where u is the circumferential speed of the rotor blade and C2u is the circumferential component of the flow speed behind the rotor. Cl u = 0 at the inlet. It is assumed that the product
C2uU is constant over the rotor plane, a condition that has to be satisfied at least in the outer
region of the blades which contribute most to the rotor's power. The velocity triangles are
shown in Figure 10.9.
Combining equation (10.3) and (10.5) yields
C2u
(10.6)
C2x
(10.7)
C2x
for best power extraction from the wind. The ratio coo/u is the advance ratio of the blade.
The ratio of turbine power over wind power was studied by Molly (1990). The result
is shown in Figure 10.10 as a function of tip speed ratio A u/ Coo, (which is the inverse
176
'<lJ
~ 0.5
0..
modern
turbines
'"0
0.4
'
<lJ
first
generation
3:
o
~ 0.3
.
..a
'
E=~
2 0.2
II
drag
a.
U
0.1
10
12
14
16
Figure 10.10: Wind turbine power over wind power as a function of tip speed ratio, lifttodrag ratio of profile and number of blades. (after Molly (1990))
of the advance ratio), lifttodrag ratio E of the profile and number Z of blades. It can be
seen that the theoretical maximum of Pm / Pw = 0.59 can only be approached by high tip
speed ratios and very good profiles. It can also be seen that the number of blades is rather
unimportant for high tip speeds and this is why a wind turbine with only a single blade
was constructed. Most wind turbines have twobladed rotors. Three blades are preferable
from the standpoints of rotor dynamics and acoustics and this has recently become the most
common design. The number of blades becomes important when the aerodynamic blade
performance is poor. The Western Mill design with tip speed ratios of about A = 4 for
driving piston pumps is a typical example for this kind of turbine. The first generation
modern wind turbines for electric power generation have tipspeed ratios of about A = 8
while last generation modern turbines reach values of about A = 14. Higher values are
177
E
6
c
45
::J
~
30
O"l
QJ
>
0
..a
.:: 15
C
O"l
'(jj
C
0
0
0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
Figure 10.11: Atmosheric boundary layer profiles for different surface roughness parameters
in comparison with a typical medium size wind turbine. The wind speed is plotted in terms
of the wind speed at the rotor axis (after Molly (1990)).
possible but not feasible because of the accompanying noise generation.
A second problem of wind turbines is illustrated in Figure 10.11. Typical atmospheric
boundary layer profiles for different surface roughness parameters are plotted. The wind
speed is normalized with the wind speed at the rotor hub. It can be seen that the mean
wind speed changes between 20% and 30% during one revolution of the rotor. In addition,
the turbulence level of the wind is in the order of 10% to 20%.
The next Figure 10.12 shows the mechanical poswer of a wind turbine as a function of
rotor speed with wind speed as parameter. It can be seen that an operation with constant
rotor speed requires a minimum value for the wind speed. An optimum performance can
only be achieved with variable rotor speed which requires expensive electrical equipment.
The line of constant momentum is followed by the Western Mills when they are used for
pumping water with piston pumps.
10.3.3
Figure 10.13 presents onethirdoctave level spectra typical for large wind turbines that were
measured by Hubbard, Shepherd & Grosveld (1981). This wind turbine has an upwind rotor
with a diameter of 90 m and a nominal power of 2.5 MW. The power output was between
0.9 MW and 2.0 MW while the measurements were performed.
Two broad peaks are prominent in the spectra. The one at low frequencies f < 50 Hz
is related to the steady blade forces, to the deterministic force Iiuctuations created by the
mean profile of the atmospheric boundary layer, and to random force Iiuctuations caused by
wind turbulence and blade vortex interaction. The second peak in the 8001300 Hz range
is mainly related to the interaction between the turbulent boundary layers and the trailing
edges of the blades.
A linear spectrum of a turbine with a downwind rotor is shown in Figure 10.14. The
trailind edge noise peak is hardly visible, here. A different aerodynamic source is much more
178
150
.,.,,~
peak momentum
125
constant
momentum
<
OJ
3:
o
75
12m/s
0..
50
25
rotor speed
..
Figure 10.12: Mechanical power of a wind turbine as a function of rotor speed for various
wind speeds. Operation at the aerodynamic design point of the blades requires a change
of rotor speed with wind speed. Most turbines operate with constant rotor speed which
requires a minimum wind speed for operation (after Molly (1990)).
Measuring point \
70
 v60......
(Q
'0
Qj
:>
qu*,+
~_
!x+{! '
60
...'"
;:J
til
til
...'"
0.
'0
50
g
U)
40
30~
2.1)
____~____~__~~____~____~____
100
J __ _ _ _ _ _~_ __ J_ _ _ _ _ _ _ __ _
1;<)00
10000
Figure 10.13: Onethirdoctave level spectra of a large wind turbine in three upwind positons
and background level (dotted) (after Hubbard, Shepherd & Grosveld (1981)).
prominent  the cooling fan of the generator. A narrowband spectrum of the low frequency
range f < 100 Hz is shown in Figure 10.15.
The next Figure 10.16 demonstrates the influence of tip speed and of rotor size. The
Aweighted sound pressure level is plotted as a function of distance from the wind turbine.
The measure levels are quite considerable and are well above the sound levels pemitted at
night in residential areas. The downwind machine WTS4 having the highest tip speed is
179
70
Wlnd
>
Measuring pOin7
60
50
Sound
pressure 40
level,
dB
30
20
400
800
1200
Frcqu~ncy,
1600
2000
Hz.
Figure 10.14: Linear spectra of a large wind turbine (WTS4) with downwind rotor in two
downwind locations. (after Shepherd & Hubbard (1983)).
20
60
40
Frcquc~cy I J{1.
Figure 10.15: Narrowband spectra of the lowfrequency range of a wind turbine with a rotor
downwind of the mast in two downwind positions. The harmonics of the rotor speed are
clearly seen. (after Shepherd & Hubbard (1983)).
seen to be the noisiest turbine at large distances.
The influence of radiation direction is shown in Figure 10.17 It can be seen that the
onethird octave band levels are lower in the plane of rotation than on the axis of rotation.
The main difference is in the low frequency range where multiples of the shaft frequency and
of the blade passage frequency play an important role for this downwind machine.
The loading noise can be approximated with the method of Wright (1969) described
180
Tip ?seed,
m sec
Blade arca,
2
WTS4
122.6
208
MOD2
83.8
197
MODOA
80.0
32
Machine
80
70
Aweighted
sound
prcssure
level,
dB
60
50
1.1 0D  2
40
WTS4
300L2~505~007~5~0~1~000
Distance from wind turbine generator, m
Figure 10.16: Aweighted sound pressure level of three different wind turbines as a function
of distance from the rotor. The influence of tip speed and wind turbine size can be clearly
seen. (after Shepherd & Hubbard (1983)).
90
1Bl
On axis of rotation
I7Z2Zl
In plane 01 rotation
Sound
pressure
level,
dB
50
40
301LOl~00:::;;1OO::O:~I;;;;O000
One~third
Figure 10.17: Onethird octave band spectra of the noise from a large wind turbine (WTS4)
in a distance of 90 m on axis (higher levels) and in plane of rotation (lower levels) (after
Shepherd & Hubbard (1983)).
in chapter 8 or with the method of Stuff (1982). Both methods account for load changes
during one revolution. The force fluctuations are expanded in a Fourier series. Each source
component of the Fourier series generates a large number of farfield modes.
The peak frequency of trailing edge noise in a onethird octave band spectrum
181
IS
gIVen
(10.8)
where Uh is the helical speed and II is the streamwise correlation length at the trailing edge.
If we use the helical speed at 90% of the rotor radius,
(10.9)
where C1x = (2/3)c oo was used, and use Uh
in Figure 10.13 for II
0.07 m.
10.3.4
1 kHz
NOISE REDUCTION
The noise of wind turbines is most annoying for low wind speeds when the background noise
level is low. The simplest noise reduction measure is to operate the wind turbine with a
rotor speed proportional to wind speed. This would also allow a consistent operation at the
optimum efficiency of the rotor. An electric frequency converter between the generator and
the main power line is required for the operation of the turbine.
Beside this, quiet wind turbines are characterized by:
upwind rotors (almost cost neutral)
at least 3 blades (cost increase)
small tip speeds, requiring long profile chord lengths (cost increase)
slightly larger rotors to compensate reduction of tip speed ratio (see Figure 10.10) (cost
increase)
Unfortunately, these are the design parameters of the wind mills in the last century which
are economically not feasible.
10.4
BIBLIOGRAPHY OF CHAPTER 10
BRENTNER, K. S. & FARASSAT, F., 1992. Helicopter noise prediction: The current status
and future direction. In DGLR/AIAA) 14th Aeroacoustics Conference, pp. 724735.
DGLR/ AIAA Paper 9202122, DGLRBericht 9203.
HECKL, M. & MULLER, H. A., 1994.
Verlag, Berlin, Germany.
HELLER, H., SPLETTSTOESSER, W. & SCHULTZ, K.J., 1993. Helicopter rotor noise
research in aeroacoustic wind tunnels State of the art and perspectives. In NOIBE93, vol. 4, pp. 3960. Proceedings of the International Noise & Vibration Control
Conference, St. Petersburg, Russia.
HOWE, M. S., 1978. A review of the theory of trailing edge noise. Journal of Bound and
Vibration 61, 437465.
182
Chapter 11
EFFECTS OF ACOUSTIC
LOADING ON FAN NOISE
Contents of Chapter 11
11 EFFECTS OF ACOUSTIC LOADING ON FAN NOISE
185
11.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . .
185
11.2 PASSIVE ELECTRICAL ONEPORTS AND TWOPORTS
187
11.3 PASSIVE ACOUSTIC ELEMENTS . . . . . . . . . . . . . .
189
190
11.4 FANS MODELLED AS ACTIVE ACOUSTIC ELEMENTS
11.5 EXPERIMENTAL DETERMINATION OF THE PASSIVE TWOPORT
PARAMETERS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 191
11.6 MODELLING OF FAN NOISE BASED ON ACOUSTIC TWOPORT DATA 194
196
11.7 CONCLUSIONS . . . . . . . . . . .
11.8 BIBLIOGRAPHY OF CHAPTER 11 . . . . . . . . . . . . . . . . . . . . . . 196
11.1
INTRODUCTION
In the discussion of sound generation by and sound propagation from different turbomachines in the previous chapters, no allowance was made for the possible effects the acoustic
environment may have on the noise generation process. In other words, the sound power
emitted by a turbomachine under fixed aerodynamic operating conditions was regarded as
a given quantity, independent of the acoustic environment. This assumption holds true for
a wave motion with a wavelength much smaller than the lateral dimensions of the fan, its
casing or ducts. This is the case for thermal wavelengths but not for the low frequency
sound of turbomachines. If open rotors are considered, there are hardly any changes in the
acoustic environment and neglect of the environmental effects seem reasonable. However, it
was shown by Cremer (1971) and his coworkers Baade (1977) and Wollherr (1973) that the
sound power generated by a fan is not only a function of its impeller speed and operating
condition, but it also depends on the acoustic impedances of the duct systems connected to
its inlet and outlet.
It is generally well understood that the sound pressure to be measured in a fixed location
relative to the turbomachine is altered when there are for example sound reflections from a
185
nearby wall in the case of open rotors or from a duct discontinuity in case of ducted machines.
However, the effect discussed in this chapter is that, apart from the change in the farfield
sound pressure, the strength of the acoustic source, i.e., the total sound power emitted, may
be altered by the influence of the environment on the source.
In the acoustic theory the effects of the acoustic environment are of course included,
implicitly, because the source terms in for example equation (4.3) would change as a result
of environmental effects.
DUCT LENGTH, L
TELESCOPING
AIR
FLOW
&0
ZBLADED
PROPELLER
FIXED MICROPHONE
WITH NOSE CONE
FAN
BLADE PASSAGE
l
W
>
W
l
i 1"
BAND
500
1000
2000
<t
(I)
>
<t
I
o
a
>
I
<t
l
W
a:
63
125
250
4000
8000
FREQUENCY IN Hz
Figure 11.1: Effect of the inlet duct length on the sound output of a small axial fan (from
Baade (1977)).
pressure level in the entrance plane of the duct. When the effective length of the inlet duct
186
is made to be equal to onehalf of the wavelength of the blade passage frequency component,
a very noticeable increase in the tonal fan noise occurs. Another result from Baade (1977)
is shown in Figure 1l.2. Here the sound pressure spectrum of an axial fan is shown which
lr,ERTIA
BLOCI~
RUBBER
MOUNT
" X
ROTATIONS/SEC
5 X
ROTATIONS
/SEC
R X
ROTATIONS/SEC
 tjf'rj/+
2000
FREQUENCY, HZ
3000
3500
Figure 1l.2: Narrowband sound pressure spectrum of an axial fan (from Baade (1977)).
was placed with its impeller in the middle of a hardwalled cylindrical duct. Clearly, the
random noise level peaks whenever the length of each half of the duct is equal to a multiple
of onehalf of the sound wavelength.
Cremer (1971) employed the "theory of black boxes", which is commonly used in communication engineering to describe the characteristics of electric or electronic elements circuits,
to account for the impedance effects on fan noise. Cremer's approach is outlined in the three
following sections. Experimental methods to determine the acoustic parameters necessary
to characterize the fan as an acoustic "black box" source are presented in section 11.5. In
section 1l.6 is shown how the results of these measurements can be used to further the
understanding of fan sound generation.
11.2
The simplest electric element has two poles (electric terminals) which form one set of terminals, or one "port", see the schematic depicted in Figure 11.3. If this box does not contain
sources, no output voltage or current can be measured and the element is called a passive
one. A passive oneport may be described by an impedance Z 1 or admittance Y 1:
187
..
1.1
12
1.1
~1t:j
J~1
~ t~2
(b)
(0)
Figure 11.3: Schematic presentation of passive electric elements; a) oneport; b) owo port
(after Cremer (1971)).
"'i 1
z
!1
u
_
1
71
1
1
(11.1 )
Here 1.1 and 11 are the complex voltage (amplitude and phase) across the two terminals and
the complex current going into the element. If the element has two pairs of terminals, for
example input and output, it is called a twoport (Figure 11.3) and its behavior is described
by one of the following three relations which involve the voltages and currents on both sides
of the element in different ways. The two voltages may be expressed in terms of the two
currents:
1.1 = Z1111
1.2 = Z 2111
+ Z1212,
+ Z 2212,
(11.2)
+ Y121.2,
+ Y 22 1.2,
(11.3)
(11.4)
~1
12
(11.5)
Y 12 ) ( 1.1 )
Y 22
1.2
(11.6)
Y 21
(1111
1121
1112 ) ( 1.2 )
1122
12
(11.7)
) = ( Y 11
( 1.1)
11

~1
Z12 )
Z22
12
Y 12
Z21,
Y 21 ,
11111122  1112112l = 1.
188
(11.8)
11.3
The acoustic behavior of a fan or another type of machinery can be described in a similar
fashion, provided the acoustic field at the input or output is sufficiently described by two
variables. In the frequency region where the sound wavelength is large compared with the
lateral dimensions of the fan, the sound field at the fan inlet and outlet is characterized by
the sound pressure p and the volume velocity q. A schematic representation of a fan acting
as an acoustic oneport and twoport is given Tn Figure 11.4; here an axial impeller is used
1,
I
i
I
I
II
:6P~ __
...........1 q 1
I ~~~I
I
L _______ 1
(11.9)
(11.10)
(11.11)
The socalled transfer matrix form in equation (11.11) is particularly useful for duct
systems where the various elements are coupled in series, and each element is described in
terms of an input/output relationship.
189
11.4
An active twoport system can be derived by first considering a passive threeport system
with the matrix representation
(~)
(11.12)
and then by treating the third port as internal one where the primary aerodynamic sources
act, see the sketch in Figure 11.5. Only the first two lines in equation (11.12) are of interest
= YuE l
= Y 21 E1
+ Y 12E2 + Y 13&,
+ Y 22E2 + Y 23&
(11.13)
The terms
I
= Y 13 PL3 = 100
q ,
q I
= Y 23 PL3 = 200
q ,
2 Pl =P2 =0
q
1 Pl=P2=0
(11.14)
are interpreted as the volume velocity that would be present at the fan inlet and outlet
when the acoustic pressures at these two locations are zero. Inserting equation (11.14) into
equations (11.12) and (11.13) yields
'II
'1 2
'II )
'12
= (
+ Y uEl + Y 12E2,
'1200 + Y 21El + Y 22E2,
'1100
'1100 )
'1200
+ ( Yu
Y 12
Y 21 Y 22
(11.15)
'1 )
E2
(11.16)
Equations (11.15) and (11.16) are valid for sinusoidal quantities only, i.e., for the tonal fan
noise components. A proper discussion of the corresponding relations for the random noise
components would go beyond the scope of this chapter, and the reader is referred to the
presentation given by Wollherr (1973).
190
11.5
EXPERIMENTAL DETERMINATION OF
THE PASSIVE TWOPORT PARAMETERS
The main difficulty in measuring the passive acoustic twopole parameters of a fan is that
there is no reliable experimental technique for the determination of the acoustic volume
velocity. Wollherr (1973) developed an experimental procedure in which this problem was
avoided by operating the fan under different acoustic loading situations where it was sufficient
to measure acoustic pressures and acoustic impedances. The following relationships can be
derived from equation (11.11):
(11.17)
fl12 =
flll E21
'
(11.18)
:2 P1=O
(11.19)
pI'
'=l.
51.1
fl21
P2=O
= t22 (flll 
12
Ell
E2 Q1=O
).
(11.20)
The required acoustic loading conditions of zero acoustic pressure at the fan inlet or
outlet (PI = 0 or P2 = 0) were approximated by mounting ducts with a length of multiples
of onehalf wavelength (L = n)"/2, n = 1,2, .. ), and the conditions of zero acoustic volume
velocity (ql = 0 or q2 = 0) by ducts with a length of odd multiples of onequarters wavelength
(L = (2n + 1)"/4, n = 0,1,2, .. ). A schematic of the four duct configurations measured by
Wollherr (1973) is shown in Figure 11.6. The impedances PI! q1 and pd q2 were measured by
using the conventional Kundttube method.
The experimental setup used by Wollherr to determine the active source parameters
qlQO and q200 for the blade passage frequency component is depicted in Figure 11. 7. The
~onditionof zero acoustic pressure at the fan inlet or outlet (compare equation (11.14)) was
established by superposing a loudspeaker signal on the fan pressure signal. Two microphones
in the inlet and outlet plane were used to monitor the resultant pressure. The volume velocity
was measured by microphones located a quarter wavelength from the inlet or outlet plane.
Wollherr's approach is applicable to small model fans but not to fans of realistic size and
make because of the experimental effort necessary to establish the required acoustic loading
conditions. Terao & Sekine (1989) devised a method for determining the active and passive
twoport parameters of a fan in which it is not required to change the duct configuration.
The method was developed further by Abom, Boden & Lavrentjev (1992). Only the principle
features of the technique are outlined here, for more details see the original papers quoted
above. A principal sketch of the experimental setup of the method is given in Figure 11.8.
Test ducts are connected to the fan inlet and outlet; no anechoic terminations are necessary.
Loudspeakers are mounted in each duct to excite a sound field independent of the fan noise.
Microphone pairs at the fan inlet and outlet are used to measure the amplitudes and phases of
the sound waves going in the positive and negative axial direction (twomicrophone method,
see for example Chung & Blaser (1980)), i.e., the pressures emitted by the fan (subscript
"+") and the pressures incident on the fan (subscript "").
191
..:LA
4
L,
,9l 6192
1J
J
Y
,P 9l
.LA
I"
u
I
11

"D
til
J
11
P'I
P2 'l!.:.O_
l.A
u
I
~ r9
I
p
_2
~
~

L
II
P21
.2, Q2=O
1::)"
~
.LA
2
UI
I'
P2i
.92 p,=O
"I
E'I
.9, P2=O
Figure 11.6: Experimental setups used by Wollherr (1973) for measurement of the passive
twoport parameters of a fan.
Abom, Boden & Lavrentjev (1992) used a "scattering matrix" representation of active
acoustic twoports which is yet another form of describing these elements; in this form the
pressure amplitudes of the emitted sound waves (p+) at the fan inlet (cross section "b") and
outlet (cross section" a") are expressed as functions of the source pressures ps, the incident
pressures (p_), and the scattering matrix S which involves the sound reflection coefficients
P and transmission coefficients T of the fan:
) = ( P;+ )
( Pa+
PH
PH
+ (Pa
Ta
Tb) ( Pa ) .
Pb
Pb
(11.21)
Pa+
p~+
PH = Pb+
+ PaPa + TbPb,
+ TaPa + PbPb
(11.22)
In the first step of the experimental procedure, the passive acoustic parameters (scattering
matrix S) are determined and in the second step the source vector Ps. Both quantities are
considered to be functions of the mean flow velocity.
For the determination of the passive parameters, two test states are created: (A) loudspeaker A on, loudspeaker B off, and (B) A off, B on; this can be done with or without
superimposed airflow. The scattering matrix components are then determined from various
192
r~H_r___c=}____t__r_~
Phas e
Onethird
octave
analyzer
Counter
Amplitude
'' f c + f s
Microphone Narrowband
amplifier
filter
Figure 11.7: Experimental setup used by Wollherr (1973) for determination of the active
twoport parameters of a model fan.
Pa3
Pa2
L
Pbl
ff
s,
a'
Pal
L
f, f
Sb
,I
~Pa :....'\1', s
...,.AI\. p
Pb3
Pb2
I
I
I
I
I
I
I
~'Pb+~
P~:Pb~
,
a+, Pa+
za
______ 1
zb
Figure 11.8: Experimental setup for determination of the acoustic twoport parameters of a
fan (from Abom, Boden & Lavrentjev (1992)).
transfer function measurements between the electric signals driving the loudspeakers and
the pressure signals in the duct for the two test conditions A and B. One important advantage of this twosource method proposed by Terao & Sekine (1989) is that the test duct
configuration remains unchanged throughout the measurements.
Once the passive twoport parameters are determined, the source vector Ps is determined
by measuring the unknown quantities in equation (11.21) or (11.22) with both loudspeakers
A and B turned off. To suppress the influence of the flow noise pressures, two additional
microphone positions" a l " and "b l " are introduced. The unwanted flow noise signal is then
removed from the acoustic signal by using cross correlation techniques. To ensure that
the flow noise pressures at the locations" a" and "b" are uncorrelated from those at the
positions "al " and "b l " , respectively, the axial spacing between these locations has to be
made sufficiently large.
193
The method described by Abom, Boden & Lavrentjev (1992) is valid for discrete as well
as random noise components.
11.6
Cremer (1971), Wollherr (1973) and Baade (1971) showed how the results obtained from
the twoport experiments can be used to further the understanding of the sound generation
processes in a fan. In a purely acoustical experiment, Baade placed a small loudspeaker box
in the middle of a cylindrical tube and measured the sound pressure near one of its ends, see
the sketch in Figure 11.9. There were two loudspeakers mounted backtoback inside the box.
Monopole
source
q1=q2
5 dB
II
I:
I:
I I
II
I~:
,I
I I
I \
I
I \ \ ' '
I
I
\,/1 I 0 IpO e
I\
I\ /
II
f~I~I#I
\
I I
II
II
II
II
il
1
,I
\~
V
I
I' source
\1
q1 =q2 
\\II\IJ:II\\\l111
500
1000
1500
Frequency [Hz]
Figure 11.9: Excitation of a long cylindrical tube by a monopole source and a dipole source
(after Baade (1971); adapted from Cremer (1971)).
When the loudspeakers were excited in phase, they represented an acoustic monopole; in this
case the resonance peaks in the spectrum occurred at frequencies where the length of each
duct half equals odd multiples of onequarter wavelength (L = (2n + 1),/4; n = 0,1,2, .. ).
Exciting the loudspeakers in antiphase simulates a dipole source and yields resonant peaks
of the sound pressure spectrum at frequencies where the length of each duct half is equal to
multiples of onehalf of the sound wavelength (L = nA/2; n = 1,2,3, .. ). Comparing this
result with the sound pressure spectrum of an axial fan mounted with its impeller in the
middle of the duct, see the spectrum shown in Figure 11.2, reveals that the resonant peaks
excited by the fan and by the dipole source occur at the same frequencies which is a clear
indication of the nature of the aerodynamic fan noise, not only for the tonal but also for the
random components.
194
At this point it is worth mentioning that experimental proof of the dipole character
of axial fan noise was also provided by Margetts (1987) who measured directly the phase
difference across an axial fan impeller of 600mm diameter in a ducted inlet, ducted outlet
configuration. The phasing between the inlet noise and the outlet noise was found to be
180 0 , for both tonal and random noise components, at frequencies from almost zero to just
below the cutoff frequency of the first higherorder duct mode.
Knowing that the axialflow fan is a dipole source, Baade (1977) developed a model
consisting of a source sheet with a pressure drop Ilpo and a passive impedance lao], see the
schematic presentation of the fan depicted in Figure 11.10 with ducts connected to both sides.
rj

)D~
MOTOR
IIf'MPELLER
.
~.
AIR FLOW
~~~~
;m
lD~
+F
I
II Po
qa
Zrad
Figure 11.10: Dipole model of an axial fan with inlet and outlet duct (after Baade (1977)).
Note that the ducts and the radiation characteristics of the duct ends are also described in
terms of acoustic twoports.
A centrifugal fan is a much more complicated geometric system and therefore much more
difficult to model. Wollherr (1973) measured the passive twoport parameters of a small
model fan, see the sketch in Figure 11.11, using the test arrangements shown in Figure 11.6
and from his experimental results he concluded that the air contained in the inlet and outlet
sections of the casing acted as masses with the volume between them as a resilience. Further,
the volume of the casing together with the hole through which the impeller shaft enters the
casing act as a Helmholtz resonator. Wollherr was also able to show that the aerodynamic
source of the blade passage frequency sound is of dipole origin. In Figure 11.12 is plotted the
difference in blade passage frequency level between outlet duct and inlet duct as function of
frequency. The dashed curve represents the case of the aerodynamic source; it was obtained
by varying the impeller speed. The solid curve was measured with a dipole source placed
near the cutoff simulating the aerodynamic source. The agreement between the two cases
shown is very good which supports the dipole model for the tonal noise of centrifugal fans.
The results of Baade (1971), Baade (1977), and Wollherr (1973) show that the knowledge
of the acoustic twoport parameters of a fan allows some useful conclusions with regard to
the inside of the "black box" , i.e., the interior sound generation processes.
195
11.7
CONCLUSIONS
In this chapter methods were discussed to describe and experimentally determine the acoustic
source and transmission characteristics of fans which playa significant role in fan noise when
the wavelengths of the sound waves generated are large compared to the cross dimensions of
the fan; this is true for the low frequency noise of small fans used for household appliances
or computer equipment. In these methods, the fan is treated as an active acoustic twoport.
The methods were described in connection with fans but are also applicable to other passive
or active elements. Once the active and passive twoport parameters of a fan are known
from experiments, the sound power emitted under arbitrary acoustic loading conditions, i.e.,
when connected to arbitrary duct systems, can be predicted, provided the acoustic twoport
parameters of the acoustic load (duct system) are also known.
In case of very large fans, acoustic loading effects are less important, because they take
place predominantly in the frequency range of plane wave propagation with respect to the
cross dimensions of the fan inlet or outlet, i.e., at very low frequencies.
Acoustic loading effects on fan noise will be discussed further in chapter 14 on installation
effects on fan noise. There it will be shown that acoustic loading effects can be found not
only in the frequency range of plane wave sound propagation in the fan ducts but also at
higherorder duct mode frequencies.
11.8
BIBLIOGRAPHY OF CHAPTER 11
196
30
20
II I
I I
I \
I I /'
I
I I I I
I \ / \
1/ I
I
IJ
\
I
\
/
\
I
\
I
I
\
\
\
\
CD
.::s
'l
"'\ \
10
"'l
I
I
I
I
I
I
I
I
I
Io'cIo~0::::;0:~2000
{(Hz)
Figure 11.12: Difference in the blade passage frequency levels measured in the fan inlet and
outlet of a model centrifugal fan. Dashed line: measured with aerodynamic source (rotating
impeller); solid line: measured with artificial dipole source at the cutoff (after Wollherr
(1973); adapted from Cremer (1971)).
BAADE, P. K., 1971. Die Behandlung des Axialventilators als akustisches Zweitor. Doctoral
dissertation, Technische Universitat Berlin, Germany.
BAADE, P. K., 1977. Effects of acoustic loading on axial flow fan noise generation. Noise
Control Engineering 8, 515.
CHUNG, J. Y. & BLASER, D. A., 1980. Transfer function method of measuring induct
acoustic properties (part 1 and part 2). Journal of the Acoustical Society of America
68, 907921.
CREMER, L., 1971. The second annual fairey lecture: The treatment of fans as black boxes.
Journal of Sound and Vibration 16, 115.
MARGETTS, E. J., 1987. A demonstration that an axial fan in a ducted inlet ducted outlet
configuration generates predominantly dipole noise. Journal of Sound and Vibration
117,399406.
TERAO, M. & SEKINE, S., 1989. Induct pressure measurements to determine sound
generation, characteristic reflection and transmission factors of an air moving device in
airflow. In Proceedings Internoise'89 (Newport Beach, Ca., USA), pp. 143146.
197
WOLLHERR,
198
Chapter 12
199
201
.......................
201
203
203
204
206
....................
207
211
.................................
215
216
199
12.1
INTRODUCTION
Only primary noise control methods are discussed here, i.e., methods that alter the strength
of the aeroacoustic source mechanism. Noise reduction achieved by silencers or active control
means are not treated here, except when they are particular to turbomachinery. Noise control
at the source in general means avoiding or reducing as many of the various noise generation
mechanisms as possible, i.e., diminishing the strength of the fluctuating forces or minimizing
their acoustical effect. In the following sections experimental evidence from various sources
is given for noise reduction in axial turbomachines by means of suitable design.
12.2
One important generating mechanism for the tonal noise of axial turbomachines is the interaction between moving impeller blades and stationary guide vanes, see chapter 5. Increasing
the distance between the impeller blades and the stator vanes reduces both the interaction
between their potential fields and the effect of the blade wakes impinging on the stator vanes.
A summary of experimental results published in 1968 is shown in Figure 12.1. When the
en
0
~
C
0
0..
<lJ
U
C
<lJ
10
L
<lJ
'+
<lJ
L
<lJ
>
0
<lJ
L
<lJ
5
C
<lJ
:::: 10
0
~I~~~~~~~~~~~~rr
<lJ
>
<lJ
1
0.1
1.0
10
x/c
Figure 12.1: Noise reduction by increasing the rotor/stator axial spacmg (after Lowson
(1968)).
initial distance between rotor and stator is less than a blade chord, doubling the distance
yields a noise reduction of 4dB, while the improvement is only 2dB per doubling at axial
distances of more than a blade chord. In Figure 12.2 experimental data of Benzakein (1972)
are shown who found a reduction of 9dB at the blade passing frequency and some 5dB in
broadband noise when the axial distance 5 between the rotor and stator of an aircraft engine
fan stage was enlarged from 0.15 to 2 blade chords (c). This result was obtained with stator
vanes which were inclined by 30 against the radial direction, see the sketch in Figure 12.2.
Also, the experiments were made with clean inflow conditions. It is to be anticipated that
the noise reduction potential of an enlarged rotor/stator axial spacing may not be achieved
200
90
11
80
60
mC]rat\
6/c
0.15
1
2
Figure 12.2: Effect of enlarging the axial distance between rotor and stator (after Benzakein
(1972)).
in the case with disturbed inflow conditions where the interaction between incoming flow
and rotor blades may be the dominant source mechanism for the blade passing tone of the
turbomachine. Unfortunately Benzakein did not report whether or not the aerodynamic
performance of the fan stage was affected by the increased rotor/stator axial spacing.
12.3
Another possibility to weaken the overall strength of the aeroacoustic source given by the
rotor/stator interaction is to introduce an axial, radial, or circumferential phase shift into
the unsteady force field. This can be achieved by various means which are described below.
12.3.1
Leaned vanes are characterized by leading edges that are inclined against the plane of the
rotor to produce a radial phasing of the rotor wake interaction with the downstream stator.
As a consequence, local cancellation of the source fluctuations takes place, and the blade
passage tone radiated into the far field is reduced. The effect of such a procedure is shown in
Figure 12.3 for the case of an aircraft engine fan stage (after Benzakein (1972)). Even when
the rotor/stator axial spacing was fairly large, which diminished the strength of the local
interaction forces, reductions of 12dB at the blade passing frequency and 8dB in broadband
noise were obtained with a lean angle of 30. The effect on the aerodynamic performance
was not reported.
12.3.2
Tilted vanes are inclined in the circumferential direction while the axial distance from the
rotor blades is constant over the radial extent of the vane. The purpose of this method is
to generate an azimuthal phase shift in the interaction forces between the rotor blades and
201
110
Blade passing frequency
100
90
80
Lp
dB
70 
60 50
0
234
567
10
f,kHz.....
Figure 12.3: Effect of leaning the stator vanes against the impeller disk (after Benzakein
(1972)).
the stator vanes. Nemec (1967) studied the effectiveness of this method on a small axial fan
of D = 320mm impeller diameter with inlet guide vanes. In Figure 12.4 the reductions in
Z = 20
20
 
V = 20
105 5= 0
dB
100
95
20
20
30
0'66
""1
90
cImpeller blade
85
c.
.J
1 Z ~2:~'i
::;(
~
7
j' T LJ ? ;,;, !~
70
50 59 67 50 59 67 50 59 67 50 59 67 50 59 67 50 59 67 1Is
n,1/s
Figure 12.4: Reduction in blade tone levels obtained by tilting the inlet guide vanes of an
axial fan (after Nemec (1967)).
the blade passage frequency level (St = (fDjU)(njZ) = 1; f = frequency, U = impeller tip
speed) and its harmonics (St = 2, 3, ... ) are shown for a constant number of impeller blades
(Z
20) and various vane numbers V. First of all, the blade tone levels are influenced
strongly by the number of stator vanes, for this see section 12.4. The effect of tilting the
inlet guide vanes is most pronounced in those cases where the tone levels to start with, i.e"
202
with the radial vanes, are highest due to an unsuitable ratio of blade to vane numbers. Note
that the difference in the blade passage frequency level between the worst case (V
22,
radial vanes) and the best case (V = 30, tilted vanes) is about 25dB.
Suzuki & Kanemitsu (1971) performed a similar study on an industrial type fan with
outlet guide vanes. They were able to reduce the blade passage frequency level by 5dB by
tilting the outlet guide vanes by 45. However, at the same time the maximum fan efficiency
was impaired from 0.77 to 0.71.
12.3.3
Duncan & Dawson (1975) used stator vanes that were irregularly spaced in the axial, the
circumferential, or both the axial and the circumferential direction, to generate a phase shift
in the fluctuating forces set up on adjacent vanes. In Figure 12.5 the sound spectra of a
Rotational harmonic number n
I
90
dB
80
70
60
50
t
l
Q
if)
90
(b)
80
70
60
50
5
10
kHz
20
Frequency  ____
Figure 12.5: Noise reduction of an axial fan by irregular axial spacing of the outlet guide
vanes; a) evenly spaced vanes; b) every other vane displaced axially by 4mm (after Duncan
& Dawson (1975)).
24, V
22, impeller speed n = 1l000/min) with even
small axial fan (D = 104mm, Z
and uneven vane spacing are compared. In the latter case, every other vane was positioned
4mm further downstream than the other vanes. Reductions of up to 7.5dB in the blade tone
fundamental and 23dB in the first harmonic were observed.
12.3.4
Schaub & Krishnappa (1977) suggested stepped stator vanes as a means of noise reduction for
aircraft engines. In this design the chord length of the vanes varies over its radial extent, see
203
the schematic in Figure 12.6. Again the purpose of this method is to produce a radial phasing
SHADED SECTION FOR
30'f. ENDS OF STEPPED
STATOR BLADE AND FOR
REFERENCE STATOR BLADE
R/2
,,I
30"10 H
H 40%H
I I
JJ~
ROTOR
FIRST
EXIT 0.3, LEADING EDGE
PLANE
OF STATOR BLADE
Figure 12.6: Schematic of the stepped stator concept suggested by Schaub & Krishnappa
(1977).
of the source fluctuations. Experimental results for a model fan are shown in Figure 12.7for
impeller speeds of 6000/min and 12000/min; the impeller diameter was 305mm. The blade
passage frequency level was diminished over a wide range of the directivity characteristic, for
both the fan intake and the fan outlet. On the other hand, the aerodynamic fan performance,
pressure and efficiency, were reduced noticeably.
12.4
Tyler & Sofrin (1962) have shown that the interaction between a rotor with Z blades and a
stator with V vanes generates acoustic modes of azimuthal mode order m which is given by
the following expression, compare chapter 7:
m = hZsV
(12.1 )
where h = 1, 2, 3 for the blade tone fundamental and its harmonics, and s = 0, 1, 2,
3, .... Tyler & Sofrin also showed that substantial tone level reductions can be achieved by
appropriate choice of the number of impeller blades and stator vanes such that the dominant
blade tone harmonics are excited as nonpropagational acoustic modes, i.e., as modes that
cannot propagate down the duct but decay exponentially with axial distance from the source
(compare chapter 6).
This concept was employed by Duncan, Dawson & Hawes (1975) to diminish the blade
passing frequency level of a small axial fan (D = 105mm, n = 21000/min). In Figure 12.8
the inlet side and outlet side sound pressure spectra are shown for the cases Z /V
8/7 and
204
_~__~_
Reference stator
6000 r. p.m.
St epped stator
12000 r.p.m.
10
20
30
40 50 60 70
Angle to the fan axis 
80
90
10dB
\
Stepped stator
ou tl e t sid e ====~==:::C==~====~==::J:=:===:c===j
10dB
10
20
30
40
50
60
70
80
90

Figure 12.7: Effect of the stepped stator on the free field characteristic of the blade tone
fundamental (after Schaub & Krishnappa (1977)).
Z/V = 10/7. In the latter case the blade passage frequency level is reduced by almost 20dB
compared to the former case. The explanation for this result is as follows:
In the case Z /V
8/7 the blade passage frequency is 2800Hz. According to equation
12.1 the lowest azimuthal mode order for this component (h = 1) is m
8  7 = 1, the
cutoff frequency of which in a 105mm duct is 1932Hz (neglecting the influence of the mean
flow, compare Figure 6.6 and equation 6.45). Hence the blade passage frequency component
is generated as a propagational mode. In the case Z/V = 10/7, the blade passage frequency
is 3500Hz, and the lowest azimuthal mode order m
3. Since the cutoff frequency of this
mode is 4410Hz, it cannot propagate down the fan duct which is the reason for the low tone
level in this case. The reduction of the second harmonic (2 x BP F, h = 2) can be explained
205
f
70
) 1
(:''f;
60
r: t: t
I," "
I:
I~
Ii
200
Ii
; \:
\ i j ,'r'
5000 Hz
20000
f
DABClin
'"
Figure 12.8: Noise reduction of an axial fan obtained by suitable choice of blade and vane
numbers (after Duncan, Dawson & Hawes (1975)).
with the same arguments.
It was pointed out by Duncan, Dawson & Hawes (1975) that the maximum possible tone
level reduction may not be achieved in practical applications due to manufacturing tolerances
in the spacing of blades and vanes.
12.5
12.5.1
Irregular blade spacing of vaneless fans results in a spreading of the tonal sound energy over a
wider frequency band which makes the radiated noise less annoying. The total sound power
emitted, however, is not changed. Mellin & Sovran (1970) demonstrated the effectiveness of
this method on an axial fan with 5 blades, see the results shown in Figure 12.9.
Duncan & Dawson (1974) used irregularly spaced impeller blades on fans with outlet
guide vanes. Here the idea is that the asymmetric rotor wake distribution excites fluctuating
forces on adjacent stator vanes that are shifted in phase and therefore radiate sound less
efficiently. In this concept the total sound energy emitted is reduced.
206
5 Bladed Fan
5Bladcd Fan
2
'"'
,~
<j
25
cc
;::c
'"'
;~~
.:
4 
02
0
~C2
0,/)$
v. 
6
LOptimuf11
c"'
co
Cetimum
3
32
40
48
64
56
I
72
5 Bladed Rotor
'0
102 
)Bladed Fan
<L
2:: ]00 ~,
v.
q
s
:0
P.S
Optimum
<L
96
40
48
MINIMUM
72
0QUAL SPACING
56
A,~GLE.
DEG.
OPTII,\UM BALANCED
Figure 12.9: Comparison of the reduction in tonal annoyance produced by balanced and
unbalanced blading of a small vaneless axial fan (after Mellin & Sovran (1970)).
Figure 12.10 shows experimental results for an axial fan with d = 104mm impeller diameter, n = 1l000/min, and V = 22 stator vanes. In the upper diagram the reference spectrum
measured with Z = 24 evenly spaced blades is shown. In the first step of the approach, 'a
range of blade numbers Zmin to Zmax is sought all of which generate only nonpropagational
modes, compare section 12.4. The blade number in the middle of that range is adopted for
the modified impeller; in the present example this is Z = 15. The angular position Oq of the
blades is then adjusted according to the following equation
(12.2)
where Oq is the angular distance of the q  th blade from an arbitrary reference blade (q = 0)
and the amplitude of the displacement variation 00 is chosen such that the minimum angular
displacement corresponds to the blade number Zmax, i.e., Omin 2: 27r /Zmax, and accordingly
the maximum displacement Omax ::; 27r / Zmin.
In the middle diagram of Figure 12.10 is shown the sound pressure spectrum for the
case with Z
15 evenly spaced blades. The blade tone fundamental is shifted to a lower
frequency and reduced in level by about lldB. The spectrum obtained with irregular blade
spacing according to Oq = (27r /15)[q + 0.3 sin 27rq/15)] (lower diagram) shows a spreading
of the blade tone components and a reduction of the overall level by 2.5dB.;
12.5.2
Suzuki & Kanemitsu (1971) studied the effect of rotor blades that were leaned forward in
the axial direction. In this design, the unsteady blade forces due to inlet flow distortion
are thought to be phased in the spanwise direction, and consequently the noise is reduced.
207
10
i!
15
I
20
30
40
11rriI:tlil;rTTTr!
80
C)
~
...J
7C
CL
vi 60
90~
80 c
10
20
Frequency Hz)
Figure 12.10: Effect of blade number and irregular blade spacing on the sound pressure
spectrum of an axial fan (after Duncan & Dawson (1974)).
With the rotor/stator axial spacing kept constant over the blade span, i.e., the vanes were
leaned in the same direction, they found reductions in the specific noise level of up to 6 dB,
depending on the fan operating condition, mainly as a result of lower broadband noise, with
almost no loss in the aerodynamic performance, see the results depicted in Figure 12.11.
Note that leaned blades may be combined advantageously with straight vanes to reduce the
interaction noise; this combination, however, was not tested by Suzuki & Kanemitsu.
12.5.3
The main purpose of swept impeller blades is the same as that of leaned blades: to introduce a spanwise phasing of the unsteady blade forces generated by the interaction with
a nonuniform inlet flow. Brown (1977) investigated impeller blades swept in the forward
circumferential direction on a truck cooling fan (D = 710mm, n = 2400/min, Z
15).
A reduction of 7dB(A) was measured under insitu conditions at constant fan speed. To
maintain the same aerodynamic duty, the modified impeller had to be run at a 10% higher
speed; still a reduction of 7dB(A) was measured.
While the first axial fan designs employing circumferentially swept blades suffered aerodynamic losses, advanced design methods have overcome these difficulties. In Figure 12.12 is
shown a comparison of the aerodynamic and acoustic performance curves of a vaneless axial
7, n = 2000/min). The design point lies at <p = 0.15S.
fan (Stutz (1991); D = 414mm, Z
The effect of the blade sweep on the noise is most pronounced at offdesign operation. It
was shown by Ohtsuta & Akishita (1990) and StUtz (1991) that impeller blades that are
20S
80
70
'"
I\~~
60
' / r(~
". ~
,,~
Lws
d8(A)
50
~,,~
40
"""',,:
~.:::::"/
I)
~\\
~,,~./W
V
I)
.;;~ r15
~ .
"".
0.6
,/1
'\~:1
'\,0 15
~~
30
~30
0.5
0,4
0.3
~'~
8 r = 00 ""15 0
);;;
0.4
0.2
8r = 0
1 I;' 30 0
~.
0.8
0.6
Lws
0.2
V
0,1
0.2
0.3
<p
0,4
0.5
0
Figure 12.11: Effect of the blade lean angle on the aerodynamic and acoustic performance
of an axial fan (D = 564mm, n = 1500/min; after Suzuki & Kanemitsu (1971)).
swept in the forward circumferential direction are superior to backward swept blades from
both the aerodynamic and the acoustic point of view. The reason given is that with forward
sweep, the low momentum fluid particles in the blade boundary layer have to travel a shorter
distance before they reach the trailing edge and, therefore, induce smaller losses.
Miiller (1986) optimized the shape of the blade tips of a vaneless axial fan by using the
evolution strategy and obtained a 3dB(A) reduction compared to a standard design. The
evolution strategy was developed by Rechenberg (1973) for finding optimum solutions for
technical problems; the general approach of this method is briefly described in chapter 13.
For acoustical optimization, the evolution strategy was first used by Hillebrand, Neise &
Barsikow (1980) to find the optimum shape of the volute of a centrifugal fan.
The evolution strategy was also used by Lohmann (1993) in connection with numerical
codes for the aerodynamics and acoustics of fan blades to find an optimum spanwise geometry for the blades of a vaneless automotive cooling fan. A typical result for a threebladed
impeller with rotating shroud is shown in Figure 12.13. For the Sshaped blade design, a reduction in noise of 3dB(A) was predicted numerically and verified experimentally, compared
to a fan with straight radial blades.
12.5.4
In a recent paper, Carolus (1992) studied the effect of the radial distribution of the aerodynamic loading of the blades on the noise and performance of low pressure axial fans without
guide vanes. The optimum design, both from the aerodynamic and the acoustic point of view,
is characterized by a free vortex blade design (condition of radial equilibrium: T' Cu 2 =const;
209
0.5
0.4
0.3
9
0.2
0.1
00
0.04
0.08
0.12
0.16
0.2
100
95
90
ill
'1;)
85
:,
.J"
80
Radial blades
..00.(.
6~
0"......0
6/6___.
75
6___
70
65
0.04
0.08
\
0_
ooG\1
66\'If\::"6 6 
0.12
0.16
0.2
Figure 12.12: Effect of blade sweep on the aerodynamic and acoustic performance curves of
a vaneless axial fan (after Stutz (1991)).
Figure 12.13: Schematic of the automotive cooling fan developed by Lohmann (1993).
0.8 ,            ,           ,
0.6
0
~ 0.1.
CL
<l
;f'
0.2
'//
'1,/
//
(
o "_ _'___'_ _
L_ _.l
0.2
0.1.
80
0.8
 rc
=canst, Z = 6 } load increases
rc u~ ~ canst, Z = 5 linearly with r
  rc \ = canst, Z = 11. constant radial
u_
load distribution
:(
OJ
'0,
0.6
m'/s
Q,
"
60
<!
CL
..J
r'
s.s 
80.6
iI 73.3
~',
I ,
72.6
LJ
1.0
125
250
500
1000
2000
f, Hz
1.000
8000
Figure 12.14: Comparison of three rotor designs for a vaneless axial fan (D = 305mm; after
Carolus (1992)).
12.5.5
80
dB
'">
'"
'"
0
~
~
'"0.
D
50
Vl
40
0
Frequency 
Figure 12.15: Reduction of vortex shedding noise by means of blade serrations mounted at
the blade leading edges of an axial fan at optimum efficiency operation (after Longhouse
(1977)).
Bridelance (1986) placed a trip wire on the suction side of the blade at 30% chord distance
from the leading edge of a small axial fan. This method was found efficient in reducing the
noise at flow rates larger than the design rate. The spectra in Figure 12.16 show that with
the fan operating near free delivery, the high frequency noise was substantially reduced. An
Fig 10  Suppression of instability noise. Fan diameter 85mm, 3900 rpm
Flow rate near free delivery
25
.
~15
,,
Obstacle
(l
With obstacle
r.n
5
50~710~O~~~~~3~O~070~4~O~O~O~r5~O'0
Frequellcy (Hz)
Figure 12.16: Reduction of instability noise obtained by placing a trip wire on the blade
suction side of a small vaneless axial fan (D = 85mm, n = 3900/min; after Bridelance
(1986)).
overall reduction of 5dB(A) with no loss in the fan pressure or efficiency was reported.
In both cases substantial reductions in the high frequency broadband noise were achieved,
in particular at flow rates higher than the optimum.
different blade profiles were compared experimentally on a fan with outlet guide vanes. The
recommendation given was to employ NACA 65 series profiles and a free vortex blade design
(rc u 2 = const).
In another study Suzuki, Ugai & Komatsu (1985) found that the specific noise level of
a vaneless axial fan can be reduced by about 4dB by doubling the chord length. Using the
same fan type, they also studied the effect of blade camber and obtained a noise minimum
for a camber of 9% of the blade chord. Finally, they looked at the effect of the pressure
distribution along the blade and came to the conclusion that a blade profile with a 'gentle'
pressure distribution on the suction side (modified Gottinger profiles) gave lower random
noise in the high frequency region.
Sigel (1985) reported that the blade tone noise of axial fans can be reduced by placing
the maximum blade camber midway between the blade leading and trailing edge. This result
was obtained using N ACA blade profiles.
Hay, Mather & Metcalfe (1987) presented a blade design scheme which is directed towards
lower sensitivity to inflow distortion. The basic underlying idea is to design the fan well away
from the stall line to allow a wide range of operating conditions in which the fan can operate
without encountering the high noise levels associated with stall.
To obtain low sensitivity to inflow distortions, Mugridge (1975) suggested to use a blade
flow angle, at which the gradient of the liftincidence distribution is small, because the noise
due to inlet flow distortions is proportional to this gradient.
12.6
Reducing the clearance between blade tip and fan duct is beneficial to both performance
and noise of axial fans, as was shown by Marcinowski (1953), Longhouse (1978) and Fukano,
Takamatsu & Kodama (1986). Figure 12.17 shows an experimental result of Longhouse
for a vaneless axial fan (D = 356mm, Z = 8, n = 3900/min). In this example, even
a small tip clearance gap of only s = 0.76mm (s/ D = 0.0043, s/c = 0.0133; c = blade
chord) results in much higher spectral levels than in the case s = 0.05mm (s/ D = 0.00028,
s/c = 0.0009); the maximum difference in the overall level was 15dB(A). It was pointed out
by Fukano, Takamatsu & Kodama (1986) that the smaller the clearance is made, the more
accurately the fan rotor has to be centered with the duct axis, so that the blade tip sees
a uniform boundary layer. Otherwise more noise may be generated than with a larger tip
clearance. In cases where small clearances are impractical, e.g., in automotive applications,
a rotating impeller shroud is helpful in diminishing tip clearance noise, see schematic and the
experimental results obtained by Longhouse (1978) which are reproduced in Figure 12.18.
A rotating shroud cannot be used for ducted axial fans. For this application, Kameier
& Neise (1993) achieved quite substantial reductions in the tip clearance noise of an axial
fan with outlet guide vanes (D = 452mm, Z = 24, n = 1400/min) by inserting a turbulence generating device into the 2.4mm wide tip clearance gap (T = s/d = 0.0053), see the
schematic presentation in Figure 12.19. In this example, a Velcro tape was used for this
purpose. A comparison of the aerodynamic and acoustic performance curves with and without the turbulence generator is given in Figure 12.20. A few data measured with smaller
tip clearances (T = s/ D = 0.0027; 0.0013; and 0.00066) are also included for comparison.
The reduction in noise is accompanied by improvements in the fan pressure and efficiency.
In Figure 12.21, finally, the sound pressure spectra in the fan outlet duct are compared.
Note that the narraowband tip clearance noise component at about 315Hz is diminished by
almost than 40 dB.
213
100
90
LpA
dB(A)
80
70
60~
0.3
__
L _ _L _ _~_ _~~
0.5
0.7
0.9
11
1.3
O/ONom 
75
55
45
f, kHz    
Figure 12.17: Effect of the tip clearance gap on the noise of a vaneless axial fan (after
Longhouse (1978).
12.7
CASING MODIFICATIONS
Temporal and spatial variations in the fan duct boundary layer produce fluctuating blade
forces at the tip which are efficient sound sources. Moore (1975) obtained level reductions
of up to 8dB, mainly at the higher harmonics of the blade passing frequency, by boundary
layer suction. The bleed flow rate required was substantial, i.e., 5% of the main flow. He
pointed out that great care must be taken to remove the boundary layer uniformly so that
no extra distortions are created that would produce more noise rather than less.
Bard & Klee (1986) presented a scheme for a vaneless axial fan that exhibits no stall
region. It is claimed that a 'stabilization ring', which is basically a short duct expansion
directly in front of the impeller entry plane, completely eliminates stall problems.
A similar design was suggested by Wieland et al. (1989) which is depicted in Figure
12.22. Here the annular cavity in the fan casing is coupled to the interior by means of a
perforated structure. Experimental results were not conveyed.
In a European patent application by Sellmann & Koch (1984) it is suggested to provide
a short but steep diffuser section or even an abrupt duct enlargement directly behind the
impeller of a vaneless fan as a noise control means. Together with rounded off impeller tips
this is to achieve a reduction in the specific noise level of 3dB(A).
214
HI?:.0.sJ
..... 
8ellmouth
;01"
[;tolioo",)
~
"'Recirculating airflow
t
100.,...,,
   stationary bellmouth
  Rotating shroud
. /14'4% Clearance
dB
QJ
>
.!!!
QJ
:J
if>
if>
90
QJ
'"
... ~"
.?'
>;  ,.....
......
0.
"

133%
C
::l
o
if>
U
QJ
\..
::l
80
if>
QJ
3600 rpm
70~
____
______
____
______
1'5
Flow coefficient,
____
______
25
3x10 1
rp ___
Figure 12.18: Noise reduction of an automotive cooling fan by means of a rotating shroud
(after Longhouse (1978)).
12.8
CONCLUSIONS
Noise control at the source in general means avoiding or reducing as many of the various
noise generating mechanisms involved as possible. In turbomachinery, in particular in low
to medium speed machines, this is equivalent to reducing the fluctuating forces on the
blades vanes and the casing. Based on a previous discussion of the generation of steady and
unsteady forces, various noise control methods for axialflow fans were discussed, and their
effectiveness was demonstrated by using published experimental data.
It is important to note that the noise reduction data, that were quoted as the result of
a certain noise control procedure, also depend, of course, on the initial configuration of the
215
,
2 I.
\,
\
/
'
Figure 12.19: Schematic of the turbulence generator mounted into the tip clearance gap of
an axial flow fan (after Kameier & Neise (1993)).
fan tested, and therefore one should not expect to obtain exactly the same level reduction
when applying this method.
Also, the combined effect of a number of noise control methods is in general not equal
to the sum of the individual reductions. This means that after a substantial noise reduction
has been achieved by one fan modification the additional application of a second or third
noise control method will most likely be not quite as rewarding.
12.9
BIBLIOGRAPHY OF CHAPTER 12
BARD, H. & KLEE, D., 1986. Axialventilatoren mit kontinuierlicher Kennlinie ohne Pumpgrenze fur VLVSysteme in der Elektronikindustrie. In Ventilatoren im industriellen Einsatz (Dusseldorf, Germany), vol. 594 of VDIBerichte, VDIVerlag, Dusseldorf, pp. 247255.
BENZAKEIN, M. J., 1972. Research on fan noise generation. Journal of the Acoustical
Society of America 51, 14271438.
BRIDELANCE, J. P., 1986. Aeroacoustic study of axial fans with small diameter. Analysis
and suppression of instability noise. In Proceedings Internoise '86 (Cambridge, USA),
pp. 141146.
BROWN, N. A., 1977. The use of skewed blades for ship propellers and truck fans. In Proceedings 98th ASME Winter Annual Meeting (Noise and Fluids Engineering) (Atlanta,
Georgia, USA), pp. 201207.
CAROLUS, T., 1992. Acoustic performance of low pressure axial fan rotors with different blade chord lengths and radial load distributions. In Proceedings DGLRIAIAA
14th Aeroacoustics Conference (Aachen, Germany), Deutsche Gesellschaft flir Luft und
Raumfahrt, Bonn, Germany, pp. 809815.
CHANAUD, R. C., KONG, N. & SITTERDING, R. B., 1976. Experiments on porous blades
as a means of reducing fan noise. Journal of the Acoustical Society of America 59,
564575.
216
0.4
0.3
0.8 
0.7
t=
0,5
0.5
/"
0.4
0.3
60
fIfIf!t'ff"~~jll
, 0 '
L
r
* 1<
.. .
50
~,~~
.. ~.~
'
30
.. .
~
__~L~L~~~~~
0.15
0.18
0.2
0.22
0.24
0.26
20~~L~LLL
0.1
0.12
0.14
Figure 12.20: Effect of the turbulence generator in the tip clearance gap on the aerodynamic
and acoustic performance curves of an axial fan (after Kameier & Neise (1993)).
DUNCAN, P. E. & DAWSON, B., 1974. Reduction of interaction tones from axial flow fans
by suitable design of rotor configuration. Journal of Sound and Vibration 33, 143154.
DUNCAN, P. E. & DAWSON, B., 1975. Reduction of interaction tones from axial flow fans
by nonuniform distribution of the stator vanes. Journal of Sound and Vibration 38,
357371.
DUNCAN, P. E., DAWSON, B. & HAWES, S. P., 1975. Design techniques for the reduction
of interaction tonal noise from axial flow fans. In Proceedings Conference on Vibrations
and Noise in Pump, Fan, and Compressor Installations (University of Southampton,
U.K.), pp. 143161.
FUKANO, T., TAI{AMATSU, Y. & KODAMA, Y., 1986. The effects of tip clearance on the
noise of low pressure axial and mixed flow fans. Journal of Sound and Vibration 105,
291308.
217
120
Outlet duct
110
100
OJ
lJ
90
Empty gap
II
0..
~
't =0.0053
( ,
80
70
,.r \
'1",,,'
"\ j'
60
50
40
100
200
300
400
500
600
700
800
900
1000
f [Hz]
Figure 12.21: Sound pressure spectra in the fan outlet duct measured with (cp = 0.20,?,Ot
0.419) and without (cp = 0.20,?'ot = 0.368) turbulence generator (after Kameier & Neise
(1993)).
Fig. 1
16
13
22
31,30
2
18
8
(
.L
~
22"
13
21.
///
/
Figure 12.22: Vaneless axial fan design presented by Wieland et al. (1989).
HAY, N., MATHER, J. S. B. & METCALFE, R., 1987. Fan blade selection for low noise.
In Proceedings Industrial Fans
Aerodynamic Design (London, UK), Institution of
Mechanical Engineers, London, UK., pp. 5157.
HILLEBRAND, N., NEISE, W. & BARSIKOW, B., 1980. AnwendungderEvolutionsstrategie
auf die Formgebung eines Radialventilatorgehauses im Hinblick auf geringe Schallerzeu
218
gung. In Fortschritte de, Akustik) DAGA '80 (Miinchen, Germany), VDEVerlag Berlin,
Germany, pp. 471474.
KAMEIER, F. & NEISE, W., 1993. Verfahren zur Reduzierung der Schallemission
sowie zur Verbesserung der Luftleistung und des Wirkungsgrades bei einer axialen
Stromungsmaschine und Stromungsmaschine. Deutsche Patentanmeldung P 43 10 104.6,
Deutsches Patenamt Berlin, Germany.
LOHMANN, D., 1993. Numerical optimization of propeller aeroacoustics  using evolution
strategy. In Proceedings Noise '93 (St. Petersburg, Russia), Interpublish Ltd, St. Petersburg, Russia, pp. 103114.
LONGHOUSE, R., 1977. Vortex shedding noise of low tip speed, axial flow fans. Journal of
Sound and Vibration 53, 2546.
LONGHOUSE, R., 1978. Control of tip clearance noise of axial flow fans by rotating shrouds.
Journal of Sound and Vibration 58, 201214.
LOWSON, M. V., 1968. Reduction of compressor noise radiation. Journal of the Acoustical
Society of America 43, 3750.
MARCINOWSKI, H., 1953. Einfluss des Laufradspaltes und der Luftfiihrung bei emem
Kiihlgebliise axialer Bauart. Motortechnische Zeitschrift (MTZ) 14, 259262.
MELLIN, R. & SOVRAN, G., 1970. Controlling the tonal characteristics of the aerodynamic
noise generated by fan rotors. ASMETransactions) Journal of Basic Engineering 92,
143154.
MOORE, D., 1975. Reduction of fan noise by annulus boundary layer removal. Journal of
Sound and Vibration 43, 671681.
MUGRIDGE, B. L., 1975. Axial flow fan noise caused by inlet flow distortion. Journal of
Sound and Vibration 40, 497512.
MULLER, K.D., 1986. Optimieren mit der Evolutionsstrategie in der Industrie anhand von
Beispielen. Doctoral dissertation, Technische Universitiit Berlin, Germany.
NEMEC, J., 1967. Noise of axial fans and compressors: Study of its radiation and reduction.
Journal of Sound and Vibration 6, 230236.
OHTSUTA, K. & AKISHITA, S., 1990. Noise reduction of shortly ducted fans by using
forward swept and inclined blades. AIAAPaper 903986.
RECHENBERG, 1., 1973. Evolutionsstrategie. Frommann Verlag Stuttgart.
SCHAUB, U. W. & KRISHNAPPA, G., 1977. The stepped stator concept: Aerodynamic
and acoustic evaluation of a thrust fan. Paper presented at 4th AIAAAeroacoustics
Conference, Atlanta, Georgia, USA 771362, American Institution of Aeronautics and
Astronautics.
SELLMANN, M. & KOCH, W., 1984. Leitradloser Axialventilator, insbesondere zm
Beliiftung von Wiirmetauschern. European Patent Application EP 0 143 235 AI, European Patent Office, Munich, Germany.
219
220
Chapter 13
13.5
13.6
13.7
13.8
13.1
240
241
241
241
242
242
243
245
247
247
INTRODUCTION
Several papers have been published in the past which describe various methods available
for reducing the noise of centrifugal fans. Most of these studies were concentrated on the
control of the blade passage tone. The reason for the interest in this particular component
of the overall noise spectrum is obvious. From a subjective point of view, the tone is the
most annoying component and needs to be reduced. The production of the blade tone
is known to originate from the interaction of the mean air flow leaving the impeller with
the asymmetrical casing, in particular the cutoff which is closest to the rotating impeller.
Hence, the generation mechanism of the discrete frequency sound seems to be understood
quite satisfactorily, and its main source region is rather compact and can be easily modified.
This may be the reason why most of the past studies in centrifugal fan noise reduction were
focused on the harmonic part of the spectrum only.
It seems much more difficult to reduce the random noise component of centrifugal fans
which is thought to be generated by the turbulent flow acting on the solid surfaces of the
impeller as well as the fan casing. The sound radiation from the turbulent fluctuating
quantities in the flow itself is considered to be of minor importance, at least when the mean
flow Mach numbers involved are small, compare chapter 5. Reducing the random noise would
require controlling the turbulent flow in both the impeller and in the casing. Consequently,
only a few studies aimed at a lower broadband noise level have been published in the past.
13.2
13.2.1
Very close to the impeller periphery, the circumferential velocity profile of the mean flow
leaving the impeller exhibits sharp minima and maxima due to the blade wakes and the
nonuniform velocity distribution in the rotating blade channels. A cutoff placed in this
222
region would experience strong pressure fluctuations which in turn result in an effective
sound radiation into the acoustic far field. Due to the periodicity of the flow field, the sound
spectrum is discrete with spectral lines at the blade passage frequency and its harmonics.
When the distance from the impeller is enlarged, the circumferential velocity profile
becomes smoother, and the amplitudes of the pressure fluctuations excited at the cutoff are
diminished. A second effect is that a cutoff placed close to the impeller blocks the flow
between two blades periodically. The unsteady flow sets up unsteady blade forces which are
known as a dominant cause of rotor noise compare chapter 5. It is obvious that the flow
blockage is less severe when the cutoff is placed further away from the impeller.
Increasing the cutoff clearance was a very early recommendation for fan noise control;
see the handbooks by Madison (1949) and Harris (1957). The first experimental results on
the noise reduction achieved by this means were reported by Regenscheit in the book by Eck
(1962), Eck (1973).
Leidel (1969) studied the effect of the cutoff configuration on both the aerodynamic and
the acoustic performance of a centrifugal fan. A schematic of the test fan is given in Figure
13.1. The volute of the fan casing was adjustable so that the cutoff radius as well as the
Logarithmic
spiral
~
Impeller "
Inlel duel
125
3:
N
100
~.
:::
95
~
::>
90
0
~0
85~''~~~~~~~~rr~
"0
.!?
.0
Figure 13.2: Influence of the cutoff clearance on the level of the blade passage frequency
(after Leidel (1969)).
95r,,,,,~
90
85
3rr
'"o'"
~
o
0.
'"
.2
"0
.a
o
'">
..!!!
]~Ij
o
0'02
004
O 06
008
010
012
0'14
0.16
'P
Figure 13.3: Influence of the cutoff radius on the level of the blade passage frequency level
(after Leidel (1969)).
reduction is about 6dB for each of the three cutoff clearances tested (8/ R
0.25).
The lowest blade passage frequency level was achieved by combining the cutoff clearance
= 0.25 with the cutoff radius r / R = 0.2, compare the bottom graph in Figure 13.3
with Figure 13.2. This minimum in the level of the blade tone fundamental was associated
8/ R
224
with the maximum in fan efficiency. Thus, there was no performance penalty to be paid for
gaining a lower noise level. The only drawback of Leidel's results can be seen in the enlarged
fan casing.
The second and third harmonic of the blade tone were reduced even more than the
fundamental. The reason for this is clear, because the circumferential velocity profile close
to the impeller is very peaky and therefore produces impulselike pressures at the cutoff. The
resultant far field sound spectrum has many harmonics of high amplitudes. With increasing
distance from the impeller periphery, the flow profile is washed out and the pressures at the
cutoff become less spiky. Thus the higher harmonics are attenuated more than the blade
tone fundamental.
In Figure 13.4 are shown the results of various experimenters for the noise reduction
10
.',
(0)
',/,
Reference point
5~
:::
>
,/
"
;~
/,w"
fi~
,
10
.~
co
'illl
~ 'V
"
Smith e/ 0/,
"'<
'''.''''''/
Embleton
.~
III
Leidel
____
15
:i;
"
0
w 20
.:=
0,01
0'
0'02
0,03 0,04
0,06 0,080,1
0,2
0'3
0,4
0'6
0'8
6r1R. olR
"
'0
10
15
( b)
J!
5
10
15
 20 '      :    !     L 4!s6!J:7c8~9!:c'!:0'::2:'0;}3'::0f40~S:,0~60::;!;:.80:::"OO
6(,0 (mm)
Figure 13.4: Effect of enlarging the cutoff clearance or rib clearance, respectively, on the
blade passage frequency level of centrifugal fans (after Neise (1976)).
achieved by enlarging the cutoff clearance. In the upper graph, the difference in blade
passage frequency level is plotted versus the cutoff clearance 8 normalized by the impeller
radius R. The reference point was chosen to be 8/ R
0.06 because this distance lies within
the range covered by all the studies involved. In the lower graph, the data are plotted as
a function of the absolute value of 8 to include the results from Regenscheit (see the book
by Ed: (1962), Eck (1973)) where the impeller diameter was not reported. Except for the
results quoted from Embleton (1963), the blade passage frequency level is found to decrease
monotonously as the cutoff clearance is increased. Embleton's data show a sharp minimum
225
at a certain cutoff clearance, however it is felt here that this result is is a rather atypical one
which was supposedly caused by some unknown secondary effect.
The results given by Leidel (1969) and Smith, O'Malley & Phelps (1974) agree very well
with each other; in both studies, centrifugal fans with backward curved blades operating
in conventional volutes with a logarithmic spiral were used. Leidel, however increased the
entire casing as the cutoff clearance was enlarged, while in the work of Smith, O'Malley &
Phelps (1974) only the position of the cutoff in one and the same casing was altered. The
change in blade passage frequency level is the same in both cases, but Leidel achieved a
reduction of the broadband noise of 3 to 5dB, presumably owing to the lower flow velocities
in the larger casing, while no change was reported by Smith et al.
In both the studies of Leidel (1969) and Smith, O'Malley & Phelps (1974) the fan efficiency was not reduced by the increase in the cutoff clearance. Leidel found that the minimum in fan noise was associated with a maximum in fan efficiency ("7 = 0.82 at optimum
operation; I)IR = 0.25 and rlR = 0.2).
The two curves quoted from Ploner & Herz (1969), who studied the noise generation by
ribs of various cross sections placed near periphery of a centrifugal impeller, show that the
efficiency of an enlarged cutoff depends on the shape of the rib. It is believed that this result
holds also true for the shape of the cutoff of a conventional fan casing.
Recently Suzuki (1991) investigated the effect of a cutoff clearance which varies over the
axial width of the impeller. His results showed that it is the average distance between cutoff
and impeller that reduces the noise and not so much the axial variation of this distance.
In conclusion we note that shape and position of the cutoff are two important design
features with respect to centrifugal fan noise. The only drawback of an enlarged cutoff
appears to be that it requires a larger fan casing, at least when the initial cutoff clearance
is small. Thus the method cannot be applied in cases where severe space limitations exist.
13.2.2
When the conventional spiral is replaced with a rectangular one and the cutoff is omitted
altogether, the level of the blade passage tone is reduced, according to Honmann (1973). No
experimental data on a fan with a rectangular casing were published. However, due to the
insufficient guidance of the flow, the aerodynamic performance of a fan with a rectangular
casing is thought to be lower than in the case of a spiral casing.
13.2.3
Krishnappa (1979) replaced the conventional scroll of a 11.2kW blower by a circular casing.
By doing that, the cutoff clearance which was initially 25mm (I) I R = 0.075) was increased
to a uniform radial distance of 77 mm (I) I R = 0.23) between impeller periphery and casing
wall. Figure 13.5 shows the sound pressure spectra measured in an anechoically terminated
outlet duct for three operating conditions labelled A (free delivery), B (best efficiency point),
and C (choked) which are marked on the fan performance curve in Figure 13.6 .
While the harmonics of the blade passage frequency were completely eliminated at all
three flow rates tested, the fundamental itself was reduced only at off design operation.
The fan performance curve was not changed by the casing modification, but one has to
be careful in generalizing this result. The fan tested in Krishnappa's work with its straight
radial blades and small hubtotip ratio is of low aerodynamic quality. In such a fan, most
of the flow losses occur in the impeller, and thus the flow in the casing does not play an
important role for the fan performance. In case of aerodynamically designed impellers, the
226
. VOLVT{

(J<cu.M
CA!. ... jO
'ASING
"
"
90
"
.0
10
a)
200
400
"0
.
11
~.
vou.n(
CR'CU,..vt
(.ASWG
C.A.StlfG
,00
~
~
'0
.0
a
10
b)
'0
200
HIO
600
eoo
1000
rR(OlHCY
IN
tWO
1400
1600
Ie()()
H.c
'20
ORCULAR
CASItfG
0 VQU1(
CASNG
,.
.'"
.,""
,'.
'."
'0 ,
~o
10
c}
800
FR(OO(~'I'
1200
1"1
'''00
1&00
1&00
2000
HI
Figure 13.5: Influence of a circular fan casing on the sound power spectra in the anechoic
outlet duct of a centrifugal fan (D = 667mm; Z = 10 (flat, radial); n = 3500jmin; after
Krishnappa (1979)).
shape of the casing becomes more important, and one would expect a noticeable change in
the fan performance and efficiency when the volute is replaced by a circular casing.
13.2.4
ACOUSTICAL OPTIMIZATION OF A
CENTRIFUGAL FAN CASING
In most of the studies discussed so far, it was tried to diminish the intensity of the blade tone
by modifying only the cutoff region of the casing. Hillebrand, Neise & Barsikow (1980) varied
the shape of the entire scroll to find an optimum design with respect to the noise generated.
As an experimental tool, the evolutionary strategy proposed by Rechenberg (1973) was used.
This strategy is basically an experimental procedure for finding the optimum solution of a
technical problem which is governed by a large number of variables. Since this method can
227
I~OO
VOLUTE
+ CmCu.M
CASI"IG
CASING
1000
10
0.'
..
Figure 13.6: Mass flowpressure rise characteristic of the experimental fan used by Krishnappa (1979).
be applied to all sorts of technical problems, using experimental or numerical "tests", the
general approach is briefly described here.
The evolution strategy adopts the principle of the biological evolution of plants and
animals: The original form of a certain species of, say, an animal, produces a large number
of offsprings which differ slightly in their characteristic properties. From the members of
this first generation of offsprings, those survive and produce a second generation which are
adapted best to their environment ("survival of the fittest"). The same selection process takes
place among the members of the second generation, and a third generation is formed, etc.
This process is repeated until the set of characteristic properties (variables) is found which
provides optimum adaption of the species to its environmental conditions. Characteristic of
the biological evolution is that the variables are varied randomly from specimen to specimen,
i.e., offspring to offspring, rather than systematically.
It was shown by Rechenberg (1973) that the principle of the biological evolution can be
used effectively for the optimization of technical problems in that the optimum solution is
obtained faster than by systematic variation of the variables. Very often the evolutionary
strategy leads to "unexpected" solutions which would have been difficult to anticipate based
on engineering intuition. Also, this technique does not necessarily require a detailed knowledge or understanding of the physical processes involved and is therefore applicable also to
problems which are not yet fully understood. Yet, a deeper insight into the physics and
the important parameters involved may be gained by exploring the reasons which led to the
solution found.
In the technical problem of a quiet centrifugal fan, Hillebrand, Neise & Barsikow (1980)
described the scroll by ten variables PI to P10 which are the radial distances of the spiral
from the impeller axis at various angular positions, see the schematic of the experimental
fan with adjustable casing depicted in Figure 13.7. The scroll was made of 0.25mm thick
flexible sheet metal; a layer of 4mm thick leadrubbercompound was bonded to the wall
to eliminate vibration problems. The cutoff radius 'J' and the width of the casing were
kept constant throughout the experiments. Sound measurements were made in anechoically
terminated inlet and outlet ducts.
The starting point of the study was a logarithmic spiral casing with a small cutoff clearance of 15/ R = 0.036. This noisy configuration was used purposely to see how rapidly the
evolutionary strategy would lead to a quiet configuration. From the initial form, a set of 9
offsprings was generated by random variation of the variables PI to P10. Each configuration
228
Spacer 5
'r~.~~
$b "0150
0, "01.10
<DbG "0380
"0.050
P7
Adjusl,n9
mechanism
for PI
Oullel
e
$
Figure 13.7: Schematic of the experimental fan used by Hillebrand, Neise & Barsikow (1980).
was tested experimentally at a constant rotor speed (n = 3000/min) and constant flow rate
(best efficiency point rp = Q/7fbDU = 0.11; b = impeller width). The best of the 9 offsprings
formed the new "parent" from which the second generation was derived, again by random
variation of the variables PI to P10. During the first 6 generations, the cutoff clearance was
increased rapidly, During generations 7 to 11, the rest of the scroll was shaped. From there
on the differences in noise level among the members of one family of offsprings were smaller
than OAdB(A). After 18 generations finally, the level difference between the best offsprings
of two subsequent generations were equal to the reproducibility margin of the experimental
apparatus.
The result of the evolutionary strategy is again a logarithmic spiral, however with a large
cutoff clearance of 8/ R = 0.31. In Figure 13.8 the aerodynamic and acoustic fan performance
curves for the original and the optimum form are plotted as functions of the flow coefficient.
In Figure 13.9 the result of the evolution strategy is compared to a casing with a moderate
cutoff clearance of 8/ R = 0.09, which is with the range of industrial designs, and to a casing
designed according to the findings of Leidel (1969) (8/R = 0.25).
In a second experiment, Hillebrand, Neise & Barsikow (1980) used the evolutionary
strategy to optimize the geometry of only the cutoff region in the range of I = 15 to
90 0 , compare Figure 13.7. Ten variables were used again to describe the radius and the
position of the cutoff and the contour of the casing. An unexpected result of this series
of experiments was that small changes in the geometry of the cutoff have a substantial
effect on the blade passage frequency level, even when the cutoff clearance is quite large
(8/R = 0.31). Hillebrand, Neise & Barsikow (1980) reported that a cutoff contour which
bulged out toward the impeller, as compared to the logarithmic spiral, gave a 6 to 8dB
lower tone level. This was found at the fan operating conditions specified before. Further
experiments, however, showed that this optimum contour did not prove successful at other
fan speeds and/or throttling conditions. Thus, in this particular example the evolutionary
strategy has led to an optimum specific to the fan operating conditions of the test.
In general, the evolutionary strategy has proven to be a powerful tool for determining the
optimum casing geometry for a given impeller type. It is also capable of finding solutions
which would be difficult to anticipate on the basis of theoretical considerations or engineering
intuition.
229
130r~~~~~~~~~~~~~~
dB
dB(A)
110
100
dB
dB(A)
110
A wei gh ted
~ __ xx
100
90 Inlet
1.2
ljJ(~
Xx __ x__
1.0
0.8
1.0
0.6
x'
0.8
xx
O/R =0.31
0.6
0.4
0.2
o. 4 O'~O:.0~2'00~4'00:':6:'00":8:':0.~10'0."'12'0.~14'0.'16''0.18
\fl
Figure 13.8: Comparison of the aerodynamic and acoustic performance curves of the original
(8/R = 0.036) and the optimized fan casing (8/R = 0.31); D = 280mm, n = 3000/min,
'Popt = 0.11; (after Hillebrand, Neise & Barsikow (1980)).
13.2.5
Evteev & Padalkin (1992) suggested to improve the aerodynamic performance of centrifugal
blowers with forward curved blades by reducing the cutoff clearance; the resultant increase
in blade tone noise was to be counteracted by placing a flow obstacle in the interior of the
impeller, see the schematic depicted in Figure 13.10. The purpose of the rib, the width of
which may vary over its axial length, is to block the flow in the blade channels which would
discharge the flow to the cutoff region and thereby generate unsteady pressures responsible
for the tonal noise. The experimental results obtained with this device for the fan efficiency
and noise are shown in Figure 13.11. It must be noted at this point that the effect of reducing
the cutoff clearance is in contradiction to the findings of an experimental optimization study
where an optimum value of 8/ R = 0.3 was determined for centrifugal blowers with forward
curved blades, compare section 13.5.
230
130
dB
dB(A)
Lp
110
LpA 100
90
linear
dB
r
Lp
dB(A)
110
LpA 100
90
Inlet
i!++++~++++++++++l1.
1.2
0.8
1.0
tV
0.8
11
0.6
04
0 0
xx
___
0
0.02
O.OL.
":...
'" "
/
0.06
0.6
~
'"
6/R = 0.09~'~ '';,'x
6IR = 0.25
~
0,4
.~
6/R = 0.31
.;,\
".~ 0.2
0.08
0.10
\.P

0.12
0.14
0.16
0.18
Figure 13.9: Comparison of the aerodynamic and acoustic performance curves of the
optimized fan casing (8jR = 0.31) with casing geometries according to Leidel (1969)
(8j R = 0.25) and a industrial design (8j R = 0.09); D = 280mm, n = 3000jmin, i.popt = 0.11;
(after Hillebrand, Neise & Barsikow (1980)).
13.3
13.3.1
In conventional centrifugal fan design, the cutoff edge and the trailing edge of the impeller
blades are parallel to one another, and the pressures excited by the flow along the cutoff
edge are in phase. When the cutoff or the blades are inclined, a phase shift of the source
pressures is produced which results in local cancellation and lower sound radiation into the
acoustic far field. This effect is similar to that of inclined stator vanes or in the case of axial
fans, compare chapter 12.
231
Jl Jl
Figure 13.10: Schematic of centrifugal blower with rib placed in the impeller interior (after
Evteev & Padalkin (1992)).
0.5
0.4
0.3
~/
/"
.~
.7/
.
x.
~~~
:/"
A 6/R = 0.2 no rib
80
~~
2000
6000
10000
B 6/R
=0.12
C 6/R
=0.12
no rib
with rib
14000
Q m
/h
Figure 13.11: Influence of an increased cutoff clearance and a ribtype flow obstacle in the
impeller interior on the efficiency and noise of a centrifugal blower with forward curved blades
(D = 500mm, Z = 42, n = 1180/min; after Evteev & Padalkin (1992)).
Sloping the blades has the drawback of being difficult to manufacture. Embleton (1963)
studied the effect of bevelled blades on a multivaned centrifugal blower with forward curved
blades as well as on impellers with backward curved blades. In the first case, a reduction of
12dB in the blade passage frequency level was found for the case of constant rotor speed; the
modified impeller, however, had to run faster to deliver the same aerodynamic duty which
232
in turn increased the random noise over that of the reference impeller, and very likely the
overall level as well. A comparison between the standard design and the modified impeller
for a constant volume flow was not made experimentally.
In the case of the impeller with backward curved blades, there was only a small reduction
of 2 to 3dB in the blade passage frequency level when the cutoff was sloped by 12, at the
expense of a slightly reduced pressure head.
An inclined cutoff is easier to produce than inclined impeller blades, and it was also found
to be more effective in reducing noise. The cutoff has to be sloped so as to span at least a
full blade pitch. The noise reduction effect obtainable by this means strongly depends on the
distance between impeller and cutoff as was shown by Ploner & Herz (1969) who simulated
the acoustic effect of the fan casing by a rib placed close to the impeller periphery. In Figure
13.12, the blade passage frequency level measured with this arrangement is plotted against
1 1 0 ,    ,      .      , _ _ _.._,
100
"'C7
to
~
90
"C7
.J
a : 50
80
Without rib
70~~~~L~
0,02
0'04
006
0,08
0'10
Figure 13.12: Blade passage frequency level of a centrifugal fan impeller as a function of
the distance between its periphery and a rib (D = 430mm, n = 2500/min, r.p = 0.16; after
Ploner & Herz (1969)).
the rib clearance for various cross sectional shapes of the rib and for an inclined rib. For one
and the same rib shape, the effect of bevelling the rib depends on the rib clearance or vice
versa.
Lyons & Platter (1963) measured a 10dB reduction at the blade passage frequency by
skewing a sharp cutoff placed close to a multivane impeller over two bladetoblade spacings,
without a loss in efficiency. This was measured at the best efficiency point; at other operating
conditions, the inclined cutoff was found less effective.
The slope angle necessary to span a full blade spacing is small in case of the multivane
impellers, but it becomes large in the case of impellers with backward curved or radial blades
where 6 to 12 blades are common. This results in a rather complex fan outlet geometry which
causes flow losses. Khoroshev & Petrov (1965) measured a 6% loss in fan pressure and a
decrease in efficiency from 0.77 to 0.72 when the cutoff edge was inclined by 70, see the
results presented in Figure 13.13. On the other hand, a drastic reduction of the blade passage
frequency level and the random noise was achieved. In contrast with Lyons & Platter's results
but in agreement with the data reported by Ploner & Herz, the inclined cutoff reduced the
blade tone level most strongly at offdesign operation.
In order for the modified fan to generate the same pressure head as the original fan, the
233
180
1L.0
  ......
LlP
2
kp/m 100
0.8
.;
" '"
"" ""
"" "
"""~
1]
./
0.6

,...,
"'..;:
/./
L
p
80
dB
70
60
50

L.O
100
80
1000
f. Hz .;
10000
dB
70
Lp
60
....
__ 
.. .
   Straight cutoff
    Cutoff inclined by 70
../50L~~__~~~~~
300
400
500
600
O.m 3/ h 
Figure 13.13: Influence of an inclined cutoff edge on the aerodynamic fan performance, on the
blade passage frequency level and the sound spectrum of a centrifugal fan (after Khoroshev
& Petrov (1965)).
impeller speed has to be raised by approximately 3%, compare the fan performance curve in
Figure 13.13.
13.3.2
Double inlet blowers often consist of two halves which are manufactured separately and
mounted together afterwards. If the blades of one half are placed halfway between the
blades of the other half, the pressures produced at the cutoff are out of phase and the blade
tone is reduced. By this means Lyons & Platter (1963) lowered the blade passage frequency
level by 8 to lOdB. It is believed here that a similar effect can be achieved by staggering the
blades of double row impellers, but no results on this possibility have been published yet.
Since the noise reduction effect of staggered blades is based on the local cancellation
of the cutoff pressures, one has to assume that, like in the case of an inclined cutoff, the
obtainable noise reduction depends on the cutoff clearance.
Staggering the blades is an easy modification in case of small impellers constructed as
described above. In case of large welded fan units, the asymmetric impeller structure causes
problems resulting from thermal stresses and deformation of the impeller back plate.
13.3.3
This method was already discussed in chapter 12 on noise reduction methods for axial
fans. The total sound energy generated is not reduced by uneven blade spacing but spread
234
over a wider frequency band; this is desirable from a subjective point of view. Irregular
blade spacing for centrifugal fans was first suggested by Hubner (1961) for the radial bladed
impellers used for cooling electric motors. Later on Krishnappa (1980) showed that at certain
operating conditions the irregularly spaced blades may generate side band tones with higher
levels than the blade tone fundamental of an impeller with evenly spaced blades.
13.3.4
Wille (1972) suggested to mount a wedgeshaped guide belt around the impeller to reduce the
tone intensity in a manner similar to that of an inclined cutoff. No experimental evidence was
given in the patent description of the noise reduction effect of this method and its influence
on the aerodynamic fan performance. It is assumed here that due to the additional friction
losses caused by the flow along the guide belt, the fan pressure and efficiency is impaired to
some extent.
13.4
IMPELLER MODIFICATIONS
13.4.1
This noise control procedure was proposed by Petrov, Khoroshev & Novoshilov (1970),
Khoroshev & Petrov (1971), Petrov & Khoroshev (1971), and Petrov & Khoroshev (1992)
for impellers with radial blades and for impellers with forward curved blades. Figure 13.14
shows a centrifugal impeller with mesh screens mounted along both the inner and the outer
Transition meshes
Figure 13.14: Schematic of a centrifugal fan impeller with transition meshes at the leading
and trailing edges of the blades (after Petrov, Khoroshev & Novoshilov (1970)).
circumference of the blade row. The mesh at the leading edge of the blades is meant to form
235
a smallscale turbulent flow behind the mesh and a turbulent blade boundary layer in order
to shift the point of flow separation further downstream. The outer transition mesh is to
smooth the outlet velocity field and to further reduce the turbulence scale.
The meshes were found most effective when used on fans with poor mean flow conditions
in the rotating impeller ducts. In Figure 13.15, the measured noise reduction is shown for
CC'
I
If)
co
l"
CC'
co
q
r0
co
q
0
~
q
<;]"
...,<J"
tD
0)
<J"
tD
(j\
tD
~
I
0)
tD
tD
(j\
qq
<2:'
If)
N
I
u;,
~
.:::
N
I"N
<J"
...,
0)
<J"
<J"
If)
f (Hz)
6.::.000
Mesh type
V (mJ/h)
LlP (kp/ml)
'7
13 x 13 x 0250101
54 x 54 x 104 mm
4100
4100
4100
146
138
135
0626
0545
0536
Figure 13.15: Effect of transition meshes on the noise of a centrifugal fan; D = 325mm,
n = 2870/min, Z = 24, forward curved blades with a radial tip; (after Petrov, Khoroshev &
Novoshilov (1970)).
two different mesh types. The fan used in the experiments had blades with a forward curved
heel and radial blade tip. The fine mesh (1.3 x 1.3 x 0.25mm) reduced the sound pressure
in the frequency range from 100 to 10kHz by 4 to 9dB. The penalty to pay was a 6% loss
in the pressure head and a drop in fan efficiency from 62.6 to 54.5%. The use of the course
mesh (5.4 x 5.4 x 1.4mm) yielded significantly lower noise levels between 100 and 2000Hz,
whereas the high frequency components were intensified. The losses in pressure head and
fan efficiency were slightly greater than before.
If one wants to compare the noise data with and without transition meshes at the same
flow work, one has to consider slightly higher speeds for the impeller with meshes to compensate for the loss in pressure head. For the fine mesh, the fan speed is to be increased
by 3%, and the resulting frequency spectrum lies 0.6dB higher than shown in Figure 13.15,
based on the assumption of a USdependence for the fan sound power. For the course mesh,
the corresponding figures are 4% higher fan speed and 0.8dB higher level.
From Figure 13.15, it is obvious that the mesh dimensions, i.e., rate of open area and
the diameter, are important parameters with respect to noise reduction. No exact data were
given by the authors which would allow selection of the optimum mesh type for a given fan
without experiments.
In nearly all cases reported, the fan performance was impaired when meshes were used.
Although the fan efficiencies to start with, i.e., without meshes, were not very high. One
has to anticipate that there would be an even larger drop in fan efficiency in the case of
impellers of higher aerodynamic quality.
236
The use of meshes is problematic, if not altogether impossible, in cases where contaminated air or gases are to be handled because the meshes will become blocked after a relatively
short time of operation.
13.4.2
The inlet nozzle of centrifugal fans is normally somewhat smaller in diameter than the inlet
aperture of the impeller shroud so that it can reach into the impeller mouth with an axial
overlap of a few millimetres, depending on the impeller size. Since the static pressure in the
casing is higher than in the fan inlet, there is a relatively strong secondary flow through the
annular gap between the nozzle and the rotating impeller shroud which injects energy into
the main flow and thereby helps the flow to turn from the axial into the radial direction.
Suzuki & U gai (1977) have found that the size of the annular gap plays an important role
not only for the aerodynamic performance but also for the noise. In Figure 13.16 are shown
80
T = 10 7
75
.!. J.~
SPL A
SPL A 70
dB (A)
65
?'
'
':".::;.:,.""?~

f~
80
~
kp/m 2
~0
I'(~
,<. ,,~
t?::
~~
"~'"
",
", <>\l"'2::
" "', 3
PT "
".
"
0.8
0.6
/;::;:= 
"/
0.4
/1
40
20
;",..,>.
1]
//
107 5
60
'" :::,,,,,
>~ "'.1
1='
1]
3
 v>'
:;~ ~. ~. .:~~
~
T = 110
75
0.2
o1/
o
S=4
20
40
60
80 m3/min 120
Figure 13.16: Influence of the radial clearance T between impeller shroud and intake nozzle
of a centrifugal fan on the aerodynamic and acoustic performance (after Suzuki & U gai
(1977)).
experimental data for a fan of D = 604mm with backward curved blades. The radial
clearance between intake nozzle and shroud was varied between T = 1 to 10mm while the
axial overlap S was kept constant. The smallest radial clearance (T = 1mm) yielded the
best results. It is obvious that the optimum clearance is not the same for all fan types and
sizes, but Suzuki & Ugai did not make any suggestion as to how to scale the ring clearance
size with the fan dimensions.
237
13.4.3
Bommes (1980) studied the effect of the blade number of high pressure centrifugal impellers
on the radiated noise. For a constant flow work, the blade loading is reduced with increasing
number of blades. As a result both fan efficiency and noise are improved. Blade numbers of
12 to 15 were recommended for these low specific speed fans.
13.4.4
ROTATING DIFFUSER
Henke, Schlender & Schuster (1990) tested a rotating radial diffuser attached to the outer
impeller circumference as a noise reduction means. A schematic presentation of this arrangement is depicted in Figure 13.17. Both the backplate and the shroud are extended to
t t t
II
\
;
I
0
..II

'$.
r;...
'So
N
..
I

Figure 13.17: Sketch of the centrifugal impeller equipped with a rotating diffuser after Henke,
Schlender & Schuster (1990).
a diameter of 1.2D and are inclined by 15 against the radial direction. A straight radial
center plate subdivides the diffuser into two halves to form a "multidiffuser". Figure 13.18
shows a comparison of the sound pressure spectra of an impeller operating without a casing
for the cases with and without the rotating diffuser. The impeller diameter was not given.
Clearly, the broadband noise level was lowered over the entire frequency range of interest.
The reduction in noise was accompanied by an increase of 7% in the flow rate, 18% in fan
pressure, and 11 % in fan efficiency, at the operating point of maximum efficiency.
13.4.5
AIRFOIL BLADES
Airfoil shaped blades are very often thought to have an advantage over blades manufactured
from flat sheet metal, both in the aerodynamic performance and in the noise. A direct
comparison is shown in Figure 13.19 for a centrifugal fan with backward curved blades. The
data were taken from a paper by Neise (1989). Two impellers with airfoil blades and flat
blades but otherwise identical geometries were run in one and the same casing under identical
test conditions. In the upper graph of Figure 13.19, the pressure coefficient 'ljJ = 26.pt! POU 2
is plotted as a function of the flow coefficient 'P
4Q (rr D2 U, and in the lower diagram are
shown the total sound power spectra of the two fans (sum of inlet power and outlet power)
238
65
CD
0
0.
.J
55
45 ..
35
50
100 200
500
1K
2K
5K
10K
f. Hz
Figure 13.18: Influence of the rotating diffuser on the sound spectrum of a casingless centrifugal impeller (after Henke, Schlender & Schuster (1990)) .
.8
3
.6
.4
.2
Ccn~r'i
00
.2
100
90
.4
lp
80
LJ
3:
_I
70
50 ljlopf
0.21
50]r~8:1~~12~O~5llo"~2Ik~~k"~8kt] in
f, Hz
Figure 13.19: Comparison of airfoil shaped blades and flat sheet metal blades for a centrifugal
fan with backward curved blades (D = 510mm, Z = 12, ipopt = 0.21, n = 1600/min).
at the best efficiency point. There is only a small difference in the aerodynamic performance,
239
and the airfoil blades gave a mere 3 to 5dB reduction in the high frequency random noise
which results in a marginally lower total and Aweighted noise level.
13.4.6
Curvers & Dupre (1992) performed an experimental study to compare the noise of four
different impeller designs under the following conditions: Constant aerodynamic duty in a
given duct system (Q = 5.1m 3 /s, ClPt = 1800Pa), constant fan speed (n
985/min), and
usage of one and the same casing. The principal designs and dimensions of the four test
impellers are shown in Figure 13.20 and the sound pressure spectra measured on the fan
Z =20
FAN W1
~ Z=13
ttt
Figure 13.20: Schematics of the test impellers used by Curvers & Dupre (1992).
outlet sides are depicted in Figure 13.21 . The Aweighted levels averaged over a number of
measurement points on the outlet side are also included in Figure 13.21. Fan No. 2 with
radial blades and forward curved leading edges was found to give the lowest Aweighted
sound pressure level and a spectrum with very little tone components. On the other hand,
the low frequency portion of the spectrum of this impeller design exposes much higher levels
than the impellers with backward curved blades.
240
90.,
10~,
eo ..
..................... .......... .
:::,i;X ..... .
70
60
................. "\ ..
60
;.;;;:::.;:;tt+~~!i:t\s.'~}f+Y!+f:hfl<~~~.;,
10
100
600
FAN
:::: .':\:::: ~:::::::::::~:::: ....... ........... " .... ...... :............; ........... ~.
'
'.~.
.:':;
::~~~3J,R~J,V,i;;
100
100
)00
FAN
.co
lOa
,00
. . . . . . . J,
.......... I
.
<:::: .~ .......... ;. f'
90
h~
60
..
.00
600
100
100
100
.... "
.. " ,I
. , J.
.::::::::::::::::.:::::;:::::......
.......
.. . . . . . . ~ ..
.. ........ .
1 1 .
::ITr'~'iJht~~+B;SJr;fj~?~~
SO
1,.0
::::::::::::::::::::::::1::::::::::.
..... ,~ ........ , ~ .... , ..... ; ... , . , .... ~ ..... ~ ...... ; ............~ ....
. . . . . . . . . . . . . . . . . . . . . . . J , ..1 , J , 1 . . . . . . . . . . . . . . . . . . . . . . . ~
100
)00
FAN
9 0.\
, ....................
,
SO .:
100
)0
100
:: :::: :::::
~::::::::::: t:::: :::::::;: :::: ::: ::: i::::: :::::: ~:::::::::::~::::::: ::::~: ::::::~: :100
100
N3 L :: 89.0 dB(A)
)00
.00
FAN
pA
lOa
600
100
100
W4 L :: 88.6 dB(A)
pA
Figure 13.21: Comparison of the sound pressure spectra radiated by the test fans used by
Curvers & Dupre (1992).
13.5
Konieczny & Bolton (1990) studied the effect of various geometrical parameters on the
noise and performance of centrifugal blowers with forward curved blades. The parameters
varied were blade type, radial and axial inlet clearance, scroll development angle, scroll
developmenmt length, and cutoff clearance. The test impellers were D = 160mm in diameter
and b = 100mm wide. Two different commercially available bladetypes were used: "tablock"
and "bladestrip". The optimum values of the design parameters studied are summarized in
Figure 13.22 . The effect of each single design parameter on the noise and the performance
was not described.
13.6
ACOUSTICAL MEASURES
13.6.1
It was shown in chapter 11 that the sound power generated by a fan is not only a function
of its aerodynamic operation conditions but also dependent on the acoustic load impedance
of the duct system connected. More precisely, the sound power radiated, say, into the outlet
duct is a function of the fan's acoustic source impedance, of the acoustic impedance of
the outlet duct, and of the impedance of the inlet duct. Thus it is possible to influence the
sound power radiated into either one of the ducts by manipulating the geometry and thereby
impedance of the same duct or that of the other, or both. Examples for the application of this
technique were given in chapter 14. It must not be forgotten however, that mismatching the
impedances of fan and duct system cannot be a standard procedure for a given fan, because
241
Blower
Housing
Linear Expansion:
=: r (l + py)
r
J
j
D1
t
cc
r?'
1
'"\
C.1
Parameters Varied
Expansion angle (P) [deg]
Development angle (y J [deg]
Cutoff Clearance (cc) [mm]
Radial Inlet Clearance ( DJDj)
Axial Inlet Clearance ( C0 [mm]
Investigated
Range
6  12
300  320
6  24
0.94  1.08
2.5  10
Optimum Values
Tablock
8
300
24
1
2.5
Bladestrip
8
300
24
1
7.5
Figure 13.22: Design parameters varied by Konieczny & Bolton (1990) and their optimum
values.
the success of any such modification depends on the acoustic installation conditions. Also,
the desired noise reduction effect occurs only at fixed frequencies which limits the application
of this method to fans of constant speed.
13.6.2
Applying acoustical liners to the interior of the fan casing effectively reduces both the tone
intensity and the random noise, see Bartenwerfer et al. (1977). In Figure 13.23 is depicted
the model fan used by Bartenwerfer et al. (1977), where the volute is made of perforated
sheet metal covered by a layer of nonwoven fabric. The volute itself is mounted in a closed
box filled with glasfiber. Experimental data obtained with this fan are presented in Figure
13.24.
Compared to a fan of the same geometry of the interior of the casing but with a solid
volute, both the tone intensity and the random noise level in the outlet duct are reduced. This
was found at two different operating conditions; the value of the flow coefficient cp = 0.078
corresponds to optimum fan efficiency. Depending on the fan speed, the Aweighted sound
pressure level is diminished by as much as 12dB. The noise level in the fan inlet duct is
changed only insignificantly.
13.6.3
It was shown in chapter 16 that the acoustic resonance and radiation characteristics of
centrifugal fans play an important role for the noise emission, in particular in its spectral
distribution. These resonance/radiation effects are the reason why, for example, the level
of the blade passage frequency component does not increase monotonically with fan speed
242
o
o
If)
Rockwool
Figure 13.23: Schematic of the test fan with acoustically lined volute (dimensions in mm;
after Bartenwerfer et al. (1977)).
but, rather, exhibits sharp maxima and minima at certain speeds. Bommes (1980) used this
phenomenon to reduce the level of the blade passage frequency component of a centrifugal
fan substantially, see chapter 14 on installation effects on fan noise.
The method proposed by Bommes can be employed only for fans of constant speed.
Also, the acoustic resonance and radiation characteristics of a fan are determined not only
by the geometry of the fan itself but also by the acoustic impedance of the duct system
connected. That means that the positions of the minima of the blade passage frequency
level are affected by the duct system and, hence, that in general the modification of the fan
required to minimize the noise has to be tailored for each special application.
13.6.4
Neise & Koopmann (1982), Neise & Koopmann (1984) demonstrated experimentally that
the blade tone noise of centrifugal fans can be effectively reduced by introducing acoustic
resonators into the cutoff of the fan casing. A principal sketch of this method is shown in
Figure 13.25 . The quarterwavelength resonators are excited by the unsteady flow leaving
the impeller via the perforated cutoff plate. With appropriate amplitudes and phases of the
resonator oscillations, the pressure fluctuations at the cutoff are reduced. However, there
is only one mechanical parameter for each of the two resonators to adjust amplitude and
phase, i.e., its length.
In laboratory experiments using different types of model fans, reductions of the blade
passage frequency of up to 25 dB were achieved. A sample result is shown in Figure 13.26.
For both the inlet and outlet duct, the level of the blade passage frequency component is
plotted as function of the impeller speed, for the case with a conventional solid cutoff and
for the case with the resonators mounted. The results for three different settings are shown,
each of which was obtained by optimum tuning of the two resonator lengths.
In case of a 3 MW induced draft power plant fan of 2930 mm impeller diameter, the
octave band level containing the blade passage frequency was lowered by as much as 13dB
near the fan casing (Neise (1983), Koopmann, Neise & Hoover (1985)). Lipkens & Tournoy
243
dB
~~
5eft
";f
n '3000 rev/min
'l',0078
400~~4~OO~~~8~00~~~12~00~~1~6~00~~HZ~20oo
130r~~~,
dB
. .  .                 1
1 110
_.__.. _. _._...........j
90
70
n , 3000 rev/min
.~:.~
,_I
'l' ,0162
~L~~~~~~~~~~~
400
800
1200
1600
Hz 2000
fFlg.8(a)
130
130,
dB
dBA
hy/'"
p:.'\
.........
110.
hard~
soft
LpA
,,
~
.. /
"
/ / . "5011
90
90,/ / ....
\j!=0078
70 \j!=OQ78
70L__
__________
1000 1500 2000 rev/min 4000 1000
1500 2000 rev/mn 4000
~
n
Figure 13.24: Effect of the acoustically lined volute on (a) the sound pressure spectra and
(b) the total and Aweighted sound pressure level in the fan outlet duct (after Bartenwerfer
et al. (1977)).
(1990) and Tournoy & Bruyere (1992) reported that a 10dB(A) reduction was measured at
the mouth of the chimney downstream of a 5.8MW power plant fan by means of a single
quarterwavelength resonator.
244
Cutoff with
double resonator
~Outlet
duct
...
Perforate
resonator
mouth
Inlet duct
Impeller
Figure 13.25: Schematic of the model test fan with two quarterwavelength resonators
mounted in the cutoff (after Neise & Koopmann (1982)).
From their field experiments Neise & Koopmann concluded that the variety of flow conditions, cutoff geometries and fan inlet configurations encountered in practice often made
the appropriate flow excitation and tuning of the resonators difficult and sometimes impossible. For this reason, the fluid driven resonators were replaced by electromechanical sound
sources. With this device it is possible to diminish not only the blade passage frequency
component but also its harmonics, see the following section.
13.6.5
The arrangement for introducing the secondary sources into the fan casing is shown in Figure
13.27 . The cutoff is made from a perforated plate, and behind it are two cavities to which
conventional loudspeakers are connected. The loudspeakers are driven with signals that
are synchronized to the rotation of the fan impeller. Each loudspeaker signal comprises a
component at the blade passage frequency and one at one of its harmonics. To achieve
minimum blade tone levels, both amplitude and phase of each of the loudspeaker signal
components have to be properly adjusted using appropriate electronic circuitry.
As a typical result for the active cancellation, Figure 13.28 shows the reduction of the
blade tone fundamental plus its second harmonic in the anechoic fan ducts obtained at
optimum fan operation (i.popt = Q/7rbDU = 0.0944). Experiments were also made at two
other loading conditions, i.e., i.pmin = 0.039 and i.pmax = 0.134. For these three operating
conditions, sound pressure level reductions for the fundamental range from 1518dB in the
inlet and 1723dB in the outlet duct. Similarly for the second harmonic, reductions range
from 1420dB in the inlet and 1015dB in the outlet duct.
A similar set of results for active cancellation was obtained for the blade tone plus its
third harmonic which are not shown here. In all of the experiments described above, the
blade tone reductions were achieved at frequencies where only plane sound waves in the fan
ducts exist. From the experimental results one concludes that, with appropriate electronic
circuitry, it should in principle be possible to reduce the fundamental plus all of its more
dominant higher harmonics, as long as they propagate as plane waves.
The next set of experiments was performed to determine the effectiveness of the active
source cancellation method at frequencies of high order duct mode propagation (Neise &
Koopmann (1991)). For this test, a fan of industrial design and make was used where
245
120
dB
Inlet duct
110
'
'
100
without
resonator
Ll
\
\
'
/
J
'
\ I
1/
90
80
L1
(mm)
fres 1
(Hz)
L2 (mm)
fres 2
(Hz)
3[,[,
2[,7
300
283
27[,
310
271
313
265
320
231
367
120.,.
dB
Outlet duct
......
110
~~.
....
~dhout
100
resonator
\ ~  ,::</~/ \
\
\ /
/ . \ .. /
'. '\ /
\!
0= 280
.
\.
"'Y'
\ I
90
,..
........... .
\. )
\i
IIII
'I'
mm
80  Z = 6. backw.
lfJ opt
70L____________~~~~~~~.
2000
3000
1/min
[,000
n
200

300
Hz
400
Figure 13.26: Blade passage frequency level of a centrifugal model fan as function of the
impeller speed with and without resonators mounted in the cutoff (after Neise & Koopmann
(1982)).
the blade tone fundamental was excited as the first highorder duct mode (D = 508mm,
b = 152mm, Z = 12, n = 2300/min). The reduction of the blade passage frequency level
obtained for this case was about 6dB in the fan outlet duct for both almost free delivery
and heavily throttled fan operation. The decreases in the outlet duct were accompanied by
increases of about 2dB in the inlet duct, which resulted in overall reductions of about only
2dB and 4dB for the two operating conditions mentioned above.
The reductions found in the frequency regime of higherorder mode sound propagation are
small compared to the results that were obtained in the plane wave region. Neise & Koopmann (1991) provided experimental evidence that the aerodynamically generated source
pressure field around the cutoff is strongly nonuniform, both in amplitude and phase, and
hence is too complex to be successfully counterimaged by only two active sources introduced
in this region. For a more efficient application of this active noise control technique in the
higherorder mode frequency regime, it was found necessary to employ a more sophisticated
246
Volute
Outlet duct
Perforate cutoff
Figure 13.27: Schematic of a model centrifugal fan with two active sources mounted into the
cutoff of the casing (D = 280mm, b = 42mm, Z = 6; after Koopmann, Fox & Neise (1988)).
active source, i.e., a larger number of individually driven loudspeakers and an automated
optimization scheme for finding the most efficient source excitation.
13.7
CONCLUSIONS
In this chapter several methods of reducing the noise of centrifugal fans were reviewed.
Most of the studies discussed were aimed at a lower blade passage frequency level because
tonal noise components are of particular annoyance. It has been established in a number
of studies that shape and position of the cutoff of the fan casing play an important role
for the generation of the blade tone, while they do not seem to influence the aerodynamic
performance significantly. Other modifications suggested have a detrimental effect on the
fan pressure and efficiency. Finally it appears that some of the methods proposed need to
be confirmed by additional tests in other laboratories.
Like in the previous chapter on noise reduction in axial turbomachines, it is important
to note that the noise reduction data obtained by the various experimenters depend on the
initial configuration of the test fans, and therefore one should not expect to obtain the same
level reduction when applying a certain method to a different fan.
The combined effect of two or more noise control methods is in general not equal to the
sum of the individual reductions. This means that after a substantial noise reduction has
been achieved by one fan modification the additional application of a second or third noise
control method will most likely be not quite as rewarding.
13.8
BIBLIOGRAPHY OF CHAPTER 13
BARTENWERFER, M., GIKADI, T., NEISE, W. & AGNON, R., 1977. Noise reduction in
centrifugal fans by means of an acoustically lined casing. Noise Control Engineering 8,
100107.
247
120
BPF
110
103.9
99.4
100
2 BPF, n = 3100/min,
lfiopt = 0.0944
Active control off
 Active control on
+
IT]
90
86.0
81.3
1
80
70
400
800
120,~~~~~~~~~.
110
106.3
100
IT]
=
 90
=
1
80
89.7
70
400
800
1200
1600
2000
f, Hz
Figure 13.28: Reduction of the blade passage frequency and the second harmonic of a model
fan (D = 280mm, b = 42mm, Z = 6) by active sources in the cutoff (after Koopmann, Fox
& Neise (1988)).
L., 1980. Larmminderung bei einem Radialventilator kleiner Schnellaufigkeit
unter Beriicksichtigung von Zungenform, Zungenabstand und Schaufelzahl. Heizung
Liiftung Haustechnik 31,173179 and 210218.
BOMMES,
CURVERS,
ECI(, B., 1962. Ventilatoren, 4th ed. Springer Verlag, Berlin, Germany.
ECK,
B., 1973. Fans, 1st English ed. Pergamon Press Ltd., Oxford.
EMBLETON,
248
EVTEEV, I. & P ADALKIN, A., 1992. The new active method of decreasing the siren noise of
radial fans. In Proceedings Fan Noise (Senlis, France), Centre Technique des Industries
Mecaniques (CETIM), pp. 267270.
HARRIS, C. M., 1957. Handbook of Noise Control)) 5th Edition. McGraw HillBook Company, Inc., New York.
HENKE, K., SCHLENDER, F. & SCHUSTER, C., 1990. Rotierende Radialdiffusoren an
breiten Laufriidern  eine Moglichkeit des Energieriickgewinns und zur Liirmminderung.
KI KlimaKiilteHeizung 9, 386389.
HILLEBRAND, N., NEISE, W. & BARSIKOW, B., 1980. AnwendungderEvolutionsstrategie
auf die Formgebung eines Radialventilatorgehiiuses im Hinblick auf geringe Schallerzeugung. In Fortschritte der Akustik) DAGA )80 (Miinchen, Germany), VDEVerlag Berlin,
Germany, pp. 471474.
HONMANN, W., 1973. Radialventilator. Deutsche Patentanmeldung 2210271, Deutsches
Patenamt Berlin, Germany.
HUBNER, G., 1961. Geriiuschminderung bei gro en elektrischen Maschinen.
Zeitschrijt 35,111121.
Siemens
KHOROSHEV, G. & PETROV, Y., 1965. Decreasing the noise level of centrifugal ventilators.
Sudostroenie 3, 1517.
KHOROSHEV, G. & PETROV, Y., 1971. Some new methods of fan noise reduction. In Proceedings of the 7th International Congress on Acoustics (Budapest), no. Paper 19N16.
KONIECZNY, P. & BOLTON, J. S., 1990. The influence of design parameters on broadband
noise generation by small centrifugal blowers. In Proceedings Internoise 190 (Gothenburg, Sweden), pp. 717720.
KOOPMANN, G. H., Fox, D. J. & NEISE, W., 1988. Active source cancellation ofthe blade
tone fundamental and harmonics in centrifugal fans. Journal of Sound and Vibration
126, 209220.
KOOPMANN, G. H., NEISE, W. & HOOVER, R. M., 1985. Active cancellation of blade
tones in power plant fans. Final Report Project 16497 EPRI 3260, Electric Power
Research Institute, Palo Alto, California.
KRISHNAPPA, G., 1979. Some experimental studies on centrifugal blower noise.
Control Engineering 12, 8290.
Noise
KRISHNAPPA, G., 1980. Effect of modulated blade spacing on centrifugal fan noise. In
Proceedings Internoise 1980) December 810 (Miami, Fla.), pp. 215218.
LEIDEL, W., 1969. Einfluf3 von Zungenabstand und Zungenradius auf Kennlinie und Gera,usch eines Radialventilators. Forschungsbericht DLRFB 6916, Deutsche Forschungsanstalt fiir Luft und Raumfahrt, K6ln/Porz, Germany.
LIPKENS, B. & TOURNOY, D., 1990. The use of a quarter wavelength resonator in large
centrifugal fans. In Proceedings Internoise )90 (Gothenburg, Sweden), pp. 497500.
LYONS, L. & PLATTER, S., 1963. Effect of cutoff configuration on pure tones generated by
small centrifugal blowers. Journal of the Acoustical Society of America 35, 14551456.
249
MADISON, R., 1949. Fan Engineering (Handbook), 5th Edition. Buffalo Forge Company,
Buffalo, New York.
NEISE, W. & KOOPMANN, G., 1982. Anwendung von A/4Resonatoren zur Larmminderung
bei Radialventilatoren. In Fortschritte der Akustik  FASE/DAGA '82 (Gottingen,
Germany), pp. 801 ~804.
NEISE, W. & KOOPMANN, G., 1984. Noise reduction on the centrifugal suction fan of a
Berlin street sweeper truck. Noise Control Engineering 23, 78~88.
NEISE, W. & KOOPMANN, G., 1991. Active sources in the cutoff of centrifugal fans to
reduce the blade tones at higherorder duct mode frequencies. ASMETmnsactions)
Journal of Vibmtion and Acoustics 113, 123~ 131.
NEISE, W., 1976. Noise reduction methods in centrifugal fans
of Sound and Vibmtion 45, 375~403.
NEISE, W., 1983. Turbulenz und Gerauschminderung bei Ventilatoren. VGB Kmftwerkstechnik 63, 957~969.
NEISE, W., 1989. Noise rating of fans on the basis of the specific sound power level. In
Proceedings Tenth A ustralasian Fluid Mechanics Conference (Melbourne, Australia),
pp. 1.41~1.44.
PETROV, Y. 1. & KHOROSHEV, G. A., 1971. Improving the noise level of centrifugal fans.
Russian Engineering Journal 51, 42~44.
PETROV, Y. 1. & KnOROSHEV, G. A., 1992. Fan noise reduction by affecting the blade
air flow structure. In Proceedings Fan Noise (Senlis, France), Centre Technique des
Industries Mecaniques (CETIM), pp. 247~252.
PETROV, Y. 1., KHOROSHEV, G. A. & NOVOSHILOV, S. Y., 1970. Reduction of the noise
level in centrifugal fans by means of transition meshes. Sudostroenie 5, 22~25.
PLONER, B. & HERZ, F., 1969. New design measures to reduce siren tones caused by
centrifugal fans in rotating machines. Brown Boveri Revue 56, 280~287.
RECHENBERG, 1., 1973. Evolutionsstrategie. Frommann Verlag Stuttgart.
SMITH, W., O'MALLEY, J. & PHELPS, A., 1974. Reducing blade passages noise in centrifugal fans. ASHRAETmnsactions, Part II 80, 45~51.
SUZUKI, S. & UGAI, Y., 1977. Study on high specific speed airfoil fans. Bulletin of the
Japanese Society of Mechanical Engineers 20, 575~583.
SUZUKI, S., 1991. Study on the noise reduction in a centrifugal fan. In Proceedings Internoise'91 (Sydney, Australia), pp. 71~74.
TOURNOY, D. & BRUYERE, E., 1992. Reduction of the noise generated by industrial
centrifugal fans: Use of a quarter wavelength resonator. In Proceedings Fan Noise
(Senlis, France), Centre Technique des Industries Mecaniques (CETIM), pp. 279~285.
VVILLE, R., 1972. Gehause fur eine radiale TurboArbeitsmaschine. Deutsche Patentanmeldung 1503624, Deutsches Patenamt Berlin, Germany.
250
Chapter 14
INSTALLATION EFFECTS ON
FAN NOISE
Contents of Chapter 14
14 INSTALLATION EFFECTS ON FAN NOISE
14.1 INTRODUCTION . . . . . . . . . . . . . . . .
14.2 EFFECTS OF INFLOW CONDITIONS ON FAN NOISE . . . .
14.3 EFFECTS OF FAN OPERATION CONTROL ON FAN NOISE.
14.4 ACOUSTIC LOADING EFFECTS ON FAN NOISE
14.5 CONCLUSIONS . . . . . . . . . . .
14.6 BIBLIOGRAPHY OF CHAPTER 14 . . . . . . . . .
14.1
251
251
252
259
261
268
269
INTRODUCTION
The primary cause of the aerodynamic noise from low to medium speed industrial fans are
the steady and unsteady forces that are exerted by the flow on the fan blades and vanes.
Unsteady forces are generated when a fan rotor operates in a spatially nonuniform flow field,
because magnitude and angle of attack of the flow as seen by the blades change with angular
position. Steady flow distortions lead to periodic blade forces and tonal noise components,
and stochastic variations of the flow field produce random blade forces and in turn random
noise components.
Fan noise control at the source means avoiding or reducing as many of the various noise
generation mechanisms as possible, i.e., diminishing the fluctuating forces on blades and
vanes. Many studies have been published in the past in which methods were developed to
design quiet fans; for a review of these see Neise (1988a), Neise (1992) and chapter 12 and
13 of these course notes. To achieve the lowest possible noise level in a fan installation, it is
also necessary to select the optimum fan type for the given aerodynamic task. It is shown in
chapter 18 that the different fan types vary significantly in their specific sound power level
which is the fan sound power level reduced in a certain way by the fan aerodynamic duty (see
also Neise (1988b) and Neise (1989)). For example, centrifugal fans with backward curved
blades prove quieter in this comparison than axial fans.
251
In practical fan installations often unexpected fan noise problems arise although the fan
is acoustically well designed and appropriately selected. In this chapter, the problem of
excess fan noise is discussed which is generated by the fan installation. Three different types
of installation effects on the noise of different fan types are discussed: Inlet flow conditions,
acoustic loading and fan operation control.
14.2
Inlet flow distortions and incident turbulence generate unsteady forces on the rotating impeller blades and, hence, generate noise. Such nonuniform flow conditions are caused by
various duct elements like inlet boxes, bends, corners, diffusers, throttling devices, guide
vanes, etc.
Schmidt (1976) investigated the effect of a 90 0 duct bend mounted at various axial distances upstream of an axial fan on the radiated sound pressure spectrum (Figure 14.1). As
x/o
00
0.16
0.48
...... 1.1
'Vv 1.8
0'0
4 .0
0  0 8.1
X~x
,*\
80
+_._+
Q)
'D
6~t1
=
'
00
70
60 50
100
500
f. Hz
1000
5000
10000
Figure 14.1: Sound pressure spectra of an axial fan with a 90 0 duct bend mounted at various
distances x from the fan inlet (after Schmidt (1976)).
the result of secondary flows and flow separation in the duct bend, the velocity profile behind
the bend is not axisymmetric any more, and the turbulence intensity is increased. At short
distances between duct bend and axial fan, the low frequency random noise components
are increased significantly, and the same is true for the blade passage frequency component
which in this case lies in the 400Hz onethird octave band. The most critical range for the
noise emission of the axial fan is the axial distance x / D = 1.5 to 4 (D = impeller diameter).
From his experimental data, Schmidt concluded that a minimum distance of eight duct diameters is necessary between the duct bend and the fan to avoid excess noise generation. It
is noted here that this minimum distance is also a function of the radius of curvature of the
duct bend which was not reported by Schmidt.
In many practical centrifugal fan installations, an inlet box is mounted at the fan intake
which involves a 90 0 turn of the flow directly before entering the impeller. This situation
252
is similar to a 90 duct bend upstream of the fan inlet. Barsikow & Neise (1978) showed
that the presence of the inlet box increases the noise output of a fan by up to 4.5dB(A)
as compared to an axisymmetric inlet nozzle, see Figure 14.2, where the linear and the A
9080 70
20
110
" ,
100
,,
30
U, m/s
100
n = 1200/min
90
co
'D
" 80
1
Centrifugal fan
70
o = 720 mm,
60
   intake nozzle
 inlet box
50~
10
lJlopt
Figure 14.2: Influence of an inlet box on the noise level of a centrifugal fan (after Barsikow
& Neise (1978)).
weighted sound pressure level in the fan outlet duct are plotted versus the impeller tip speed
U. The sound pressure spectra measured in the anechoically terminated fan inlet duct show
that due to the distorted velocity profile at the entrance plane of the impeller, the tonal as
well as the random noise components are increased, however, this effect is not as strong as
in the case of the axial fan, compare the spectra in Figure 14.1. The experiments described
above were made with a set of inlet guide vanes at the fan intake, for both the axisymmetric
inlet nozzle and the inlet box, which were fully open during the measurements. Thus, the
inlet guide vanes were not used to control the fan operation condition, but they did have
the effect of stabilizing the intake flow, as will become obvious when discussing the following
results.
In a study by N eise & Hoppe (1986) a centrifugal fan with an inlet box but with no
inlet guide vanes (Figure 14.3) was used to study the effect of the inflow conditions on
253
Inlet box
Im~eller
/\
diameter, 710
"''~
\ "'"
\V=r::P~_'""=L==~tr',,\,c\,f
Figure 14.3: Schematic of the test fan with an inlet box used by Neise & Hoppe (1986).
the noise generation. This type of fan is normally used as a suction fan for street sweeper
trucks. It was found that the presence of the inlet box in the absence of inlet guide vanes
leads to unsteady fan operation due to steady and unsteady swirl components generated
in the inlet box. This swirl was successfully eliminated by means of a short splitter plate
that extends from the impeller shaft to the bottom of the inlet box. As a result, operation
of the fan became smooth and steady, and the aerodynamic and acoustic fan performance
characteristics were improved, see Figure 14.4.
The nondimensional parameters used are defined as follows: Flow coefficient 'P =
Q/7rbDU, pressure coefficient 'ljJ = 26pt/P OU 2 , and fan efficiency 'TJ = Q6pt/Pel . Here Q
and 6pt are volume flow and total pressure head of the fan. Width, diameter and tip speed
of the impeller are denoted by b, D and U, respectively. The density of air is po. Note that
since only the electric power input Pel to the drive motor was measured, the fan efficiency
as defined above represents not only the flow losses in the test fan but also the losses in the
DCmotor and pulley drive.
Sound measurements were made in an anechoic outlet duct and at one freefield position
close to the intake of the inlet box. The presence of the splitter plate improves the fan
characteristics mainly at high volume flows. The maximum efficiency is raised from 0.63
to 0.68, the corresponding pressure coefficient by about 4%. At the same time the noise
radiated is decreased over a wide range of fan operating conditions with the largest reductions
occurring in the outlet duct. The level of the blade passage frequency which represents the
most annoying fan noise component is diminished by as much as 14 dB. Figure 14.4 also shows
sound pressure spectra measured at n = 3000/min impeller speed and at a fan operation
condition close to the design point ('P = 0.145). The addition of the splitter plate lowers the
frequency spectra on both inlet and outlet over the entire frequency range.
To explain the experimental observations described above, hot wire measurements were
254
0= 710 mm
n = 3000/min
\jl = 0.145
~
110
n.
'
70L~~~~
f. Hz
150r~~1~1~1~~
Outlet duct
~
iia.
140 I
130 l
120 f
~~~
.....J
110'100~
Intake
______L1______L _ _ _ _
~1
______
______
11.0~~.~~~
Outlet duct
130
<:
a.
.....J
120
Intake
110
~.~~~~~rir~~~
100 L . _ _ _ __
140r~,~~~~
co
130
it
120
0
co
a.
.....J
110
100
90~
____
_ J_ _ _ _ _ __ L_ _ _ _ _ _~_ _ _ _ _ _L __ _ _ _~
1.l.
1.2
1.0
~
3"
0.8
0.6
0.4
o = 710 mm.
0.2
n = 3000/min
fan with inlet box
x   x Splitter in inlet box
00
Ori~inal
L _ _ _ _ J_ _ _ _ _ _ L______
0.05
0.15
______L __ _ _ _
0.2
0.25
Figure 14.4: Effect of a splitter plate in the inlet box of a centrifugal fan on the acoustic and
aerodynamic fan performance curves and on the sound pressure spectrum. n = 3000/min
(after Neise & Hoppe (1986)).
made at vanous locations
In
the inlet box. The results are depicted in Figure 14.5. The
255
lP max
lJl max
lJlmin
lJlmed
I
I
I
k
+I
/01/
.O~ .05/
. Vf#;.
.+
I
~
"'~
I
I
I
I
Y.+.
4
splitter
plate
diameter
of intake nozzle
In
Figure 14.5: Flow velocity vectors and turbulence levels in the inlet box of a centrifugal fan
with and without splitter plate (after Neise & Hoppe (1986)).
arrows represent the vectors of the local flow velocities, and the figures give the local turbulence intensity, i.e., the rmsvalue of the velocity fluctuations divided by the mean flow
256
14.3
In practical fan installations it is often required to vary the volume flow over a fairly wide
range. The most commonly used ways of volume flow control in industry are inlet guide vane
control, damper control (throttling), and fan speed control. The aerodynamic performance
curves of a centrifugal fan of D = 0.72m impeller diameter with Z = 8 backward curved
blades are plotted in Figure 14.9 for various angles of the inlet guide vanes. For each angle,
1.L.
1.2
1.0
0.8
tV
06
0.4
0.2
0
0
0.05
0.1
0.15
0.2
Figure 14.9: Dimensionless performance curves of a centrifugal fan for different angles of the
inlet guide vanes (after B arsikow & Neise (1979)).
the point of maximum fan efficiency is indicated by the value of the fan efficiency in percent.
The parabola 'IjJ ex cp2 indicates the change in fan pressure and volume flow when the fan
speed is varied. It was shown by Piltz (1974), Barsikow & Neise (1978), and Wieland (1986)
that it is advisable to control the flow rate by changing the fan speed rather than by throttling
or by inlet guide vane control, because that is advantageous not only with respect to energy
consumption but also from an acoustic point of view. In Figure 14.10 the linear and the
Aweighted sound pressure levels in the anechoic inlet duct of a centrifugal fan are plotted
as functions of the flow rate for the three ways of volume flow control mentioned above.
While the sound pressure level radiated increases with decreasing flow rate in the cases of
throttling and inlet guide vane control, it is sharply reduced in the case of impeller speed
control. If a volume flow of half the nominal rate is considered, the reductions of the linear
and Aweighted noise levels are 810dB and 20dB(A), respectively.
The effect of the inlet vane control on the noise radiated by a centrifugal fan into the outlet
duct is shown in Figure 14.11. The onethird octave band spectra shown were measured at
the respective point of maximum efficiency for each setting angle of the inlet guide vanes. The
blade passing frequency component and its harmonics become more and more pronounced as
the setting angle is increased while the random noise components remain almost unaffected
up to angles of 60. As a result, the linear and Aweighted sound pressure levels in the duct
are amplified, despite the fact that the aerodynamic fan power is reduced. The increase of
259
8,,,'~~~
dB
o ttL"",/'it=<o.~...
Throttle
I
I
'I'
12
Overall SPL
16
02
0.4
0.6
t
<1:
4
Cl
...J
<l
8 12
16
10
II'...'Q:x
.~
I
I
08
Throttle
''~'''
liAr,'''''
/'l/
'rrli
I~+~ 
,,/'"
 4 f+l~I~+. /~
I
Fan speed control
8
dB(A)
fifl1:~;
I.
Y
<J
0.2
Fan speed
control
II
/1
.L
~~Ispd/
I
Aweighted
o
...I...
SPL / /
0.4
06
0.8
1.0
Figure 14.10: Effect of volume flow control of a centrifugal fan on the sound pressure level
(after Barsikow & Neise (1979)).
fm

Figure 14.11: Onethird octave band spectra in the fan outlet duct of a centrifugal fan for
different angles of the inlet guide vanes (after Barsikow & Neise (1979)).
the blade tone levels can be explained by the interaction of the impeller blades with the
wake flow downstream of the inlet guide vanes which is shown in Figure 14.12 for a vane
angle of 45 and for three axial distances x / d (d = inlet duct diameter) from the vanes.
In many computer applications, the cooling fan is commonly selected to meet the requirements of the worst case situation. In case of constant speed drive motors, unnecessary
large flow rates and, hence, noise levels are generated by the fan under normal operating
requirements. As was pointed out by Kaiser & Wohrle (1989), substantial reductions in fan
noise as well as electric power can be achieved by using variable speed drives.
The flow rate and pressure rise of axial fans can also be controlled by variation of the
impeller blade stagger angle, in some constructions even while the fan is running. In Figure
260
0'"
90
180
/
Pitch angle of the inlet
x/d
/
/
2~~=~~/~~~~~~~~~/~====~
/
/
/
d
x Position of Hot wire Probe
Figure 14.12: Wake flow profiles downstream of the inlet guide vanes of a centrifugal fan
generated in the fan inlet duct (after Barsikow & Neise (1979)).
14.13 the aerodynamic and acoustic performance curves of an axial fan of D = 450mm
impeller diameter are plotted for five blade stagger angles 6.j3s. The noise levels are presented
in terms of the specific sound power level Lw spec = Lw  10 19( Q / Qo)  20 19 (6.pt/ 6.ptO)
(Qo = 1m3 /s; 6.ptO = 1Pa). Clearly, variation of the flow rate by means of blade stagger
angle control involves much lower sound levels than by means of throttling at a constant
stagger angle.
14.4
It was shown by Cremer (1971), Baade (1977), and Wollherr (1973) that the sound power
generated by a fan is not only a function of its impeller speed and operating condition,
but it also depends on the acoustic impedances of the duct systems connected to its inlet
and outlet. Therefore, fan and duct system should be matched not only for aerodynamic
reasons but also because of acoustic considerations. Care has to be taken that tonal fan
noise components do not coincide with resonance frequencies of the duct system.
Baade (1977) showed experimentally that a very noticeable increase in blade tone level
occurs whenever the total effective length of a fan duct is made to be equal to half the sound
wavelength or multiples thereof (i.e., when l = n . )../2; n
1, 2, 3, ... ). In Figure 14.14
is shown the variation of the sound pressure level radiated by a small axial fan with the
impeller speed, or rather the frequency of the blade tone fundamental, for the case of a)
ducted inlet and b) ducted outlet. Clearly, the sound pressure level exposes distinct peaks
which occur at frequencies where the above resonance condition is established.
On the other hand, a lower noise level can be obtained if a tone falls on an antiresonance
of the system. Bommes (1980) reported a case where by appropriate choice of fan diameter,
261
rr
70
60
so
30
20
0.8
0.1
0.2
0.3
0.4
Axialflaw fan
o cD.I.'irn. n c 3DOO/rnin. Z = 2/.
0.7
0.5
0.4
0.3
0.7
If'
0.8
'r~
"
+\
+
21.
\+
\
12
0.2
f:.,
~s =
0.1
0.1
_120
30
00
0.6
00"
0\"\
0.6
3
0.5
0.2
0.3
0.4
0.5
0.6
0.7
lp
0.8
Figure 14.13: Aerodynamic and acoustic performance curves of an axial fan with variable
impeller blade stagger angle.
speed and blade number, the blade tone was placed on a frequency where the radiation
efficiency of the fan was a minimum. In Figure 14.15 the blade passing frequency level
of a centrifugal fan is plotted versus the fan speed. In the original configuration (impeller
diameter D = 510mm), the fan is run at a fan speed of 5000/min (point B1), where the blade
tone level reaches a local maximum. By reducing the impeller diameter and increasing the
fan speed, to maintain the fan aerodynamic duty, the blade tone level is reduced substantially
for the following reasons: firstly, the cutoff clearance was increased, because the fan casing
was kept the same, and secondly the fan was now operated at a speed where the blade
passing frequency level has a local minimum. The comparison of the frequency spectra of
the original and the modified configuration in Figure 14.16 shows that the resultant tone
level reduction was 25dB.
Wollherr (1973) showed that a substantial change in the blade passing frequency level
can be obtained by mismatching the acoustic impedances of the fan (source) and the duct
system (load). In his example, a 17 dB drop in the tone level in the fan inlet d nct was achieved
by manipulating the length of the outlet duct, compare the experimental data depicted in
262
1147mm
....
I 1
+
....
H =tJ
f.
I I
I
I I
I I
I
!
!
I
!
I
I
I I,
I
I
I
I
,,' t
I HI
'.1
r
=t~~
'
+;~
1__ i~1 I  
,
L.i
I~I:
5(B
f.
1 +

,
\
I
I
11
"
.L..,
t
\11
Impeller at
Discharge
Opening
t
I'
II
,i)
11,f f.
U )4
   10 log (A
\1
II
+1
I~
f.f. 1 ,
f.I+
'I
1 ~j
l+~
_.

~!
j ,ri
~
1'
:
I
 +1I
II 11.11
rt
III
1,'=
1I _.1
'Pj
J I 1
f
I
,
:'
1
~rl
'J
1
IH
;.
I'
 
IY
r
r'lj
h",r bD
i
1/
t
f
"
II
i~
I
R"
Il'
J..
f. H~
i"1
Vi
:1'
\ I,
' I
f'"
.1
If'r=
I
1\
11\
' ~

ir++
I
I
I!
1,
Impeller 47 mm
from Inlet
Opening
I'
250
I 1_1..1J I I I
\j
1111111
500
750
1000
II
I
I 1
1250 Hz
250
I I I
If
I I I
500
Frequency
I I I
.lJJ
750
I I I I
I I
I
1000
1250 Hz
Frequency
Figure 14.14: Sound pressure level radiated by an axial fan (D = 67 mm) as a function
of the impeller speed; a) ducted fan inlet, microphone placed 50 mm off inlet opening; b)
ducted fan outlet, microphone located 50 mm off discharge opening (after Baade (1977)).
11.0,,,,,
6000
1.500
n ,1/min
Figure 14.15: Blade passing frequency level of a centrifugal fan as a function of the impeller
speed (after Bommes (1980)).
Figure 14.17.
Acoustic loading effects became visible in an experimental study by Neise, Holste &
Hoppe (1988) and Holste & Neise (1992a) in which the sound power of six fans of different
263
140,r~,,,~~~,.
dB
510mm Z=12
n = 5000/min
130
tp=tpopt=0026
/0 = 450 mm
Z= 12; tp =tpopt=0.034
n = 5555/min
t
0.
110
100 
....J
90
80
6. f = 5 Hz
70
1000
2000
3000
4000 Hz 5000
f to
Figure 14.16: Comparison of the sound pressure spectra of the original and the improved
version of a centrifugal fan at the same aerodynamic duty (after Bommes (1980)).
l}
p~,>:~~
I
Anechoic
termination
~ :j': ,I
~f[' 'io
110
I :'f':
90m
) , ~ 11.5
I
fTl
II
i,".r
I
,0tr
.~'I
I
~'il\
o I
100
Centrifugal fan
lviicrophone
,~+ I''''r'~'
r'<."
90
Inlet due t
100
10J
,.
Y\ y
'\~/'f!
'i;9~~
't
lOO
400
~
500
i, Hz
Figure 14.17: Inlet duct sound pressure spectra of a centrifugal fan for two different lengths
of the outlet duct section (after Wollherr (1973)).
designs was measured using different standardized measurement procedures. Figure 14.18
shows the results for one of the test fans, a centrifugal fan with backward curved blades, and
for two of the measurement procedures employed, i.e., the freefield method ISO 3744 (1981)
and the induct method ISO 5136 (1990). The first method is to be applied to determine the
264
sound power radiated from the fan inlet or outlet into free space, and the induct method
yields the sound power radiated into a duct.
1OO"rr~''~~~,
I~ flO 0
Inlet duct
,
oj,
1
Inlet side
90
m
co
!'~'
lOOr"
500 mm
90
80
BO
flO
500 mm
~.~!~rl
!I~'
\'\~
6\~
70

""
oj,
70
50
60
'~o
90
Outlet side
BO
70
50
50
50
50
10ci1
53
125
250
500
1k
2k
4k
63
Total
12~
250
500
lk
2k
Bk A lin
4k
Totol
90
m
co
:V~.
BO
~~,
70
50
80
Centrifugal fan
b.c. airfoil blades
~""
inlet/ducted outlet
~Ducted inlet/free outlet
5031 53 125 250 500~ 2'k '4'k
Centrifugal fan
b.c. airfoil blades
~~:~
70
",'
~.
s'k A lin
50]16312s'
250
'500'
1'k '
2'k'
',>~
\.~
'4kB'kAtln
f, Hz
f. Hz
Figure 14.18: Effects of acoustic loading on the sound power level of a centrifugal fan with
backward curved blades. a) for different outlet side configurations; b) for different inlet side
configurations (after Holste & Neise (1992a)).
The top diagram in the left column of Figure 14.18 depicts the sound power spectra
measured in the anechoic fan inlet duct, the diagram in the middle shows the spectra measured on the outlet side using the two measurement methods named above, and the bottom
one gives the total sound power emitted, i.e., the sum of inlet and outlet noise. The right
column of Figure 14.18 is analogous to the left, only that the roles of inlet and outlet are
interchanged; the comparison of induct levels and freefield levels is now shown in the top
diagram. The vertical lines mark the cuton frequency of the first higherorder acoustic mode
265
Inlet duct
100
m
u
d = 500rnm
i,o@
100
d',SOO/31Smm
90
90
80
80
70
JO
11~1
63
125
Outlet duct
m
u
100
100
90
90
80
80
70
70
31
110
63
125
250
4k
8k
~~~~~~~~~~~~L~~
in
31
110
"0
80
2k
100
90
lk
Total
aJ
.!f
125
250
500
lkI
2k
'I
LikI.,rrrr
8k A 1in
Total
100
90
~~
~~M6~~,lO~=~~90/~~~, Q=3.3m'3/s ~66
80
~Oucted
inlet/ducted outlet
l<*Oucted inlet/free outlet
Jill
31
63
125
250
~~
500 1k
f, Hz
2k
~k
,~Ll
8k A 1in
7ill"7''~~,=,",=~~~L~1L>''~~_~
j1
63
125
250
500
1k
2k
4k
8k A lin
f, Hz
Figure 14.19: Effects of acoustic loading on the sound power level of an axial flow fan with
outlet guide vanes. a) for different outlet side configurations; b) for different inlet side
configurations (after Holste & Neise (1992a)).
acoustic mode.
If the total sound power is considered, the axial fan shows a different behavior than the
centrifugal ones, compare the bottom diagrams in Figures 14.18 and 14.19: the total sound
power of the axial fan is highest with the outlet unducted.
In case of very large fans, acoustic loading effects are less important, because they take
place only in the frequency range of plane wave propagation with respect to the cross dimensions of fan inlet or outlet, i.e., at very low frequencies.
In the frequency region of higherorder duct mode sound propagation, the sound power
levels measured using the freefield method are higher than those obtained from the induct
method, compare Figures 14.18 and 14.19, but this is probably due to imperfections of the
measurement methods involved, see Holste & Neise (1992a).
267
110
Lp
315 He
= 109.7 dB
100
90 OJ
80
blade pasSing
70
Axial flow fon
40~
Hz
Figure 14.20: Narrowband sound pressure spectra on the outlet side of an axial fan (after
Holste & Neise (1992a)).
14.5
CONCLUSIONS
Experimental results from published literature are presented to show the excess fan noise
produced by installation effects. Three different types of installation effects are discussed:
Inlet flow distortions, fan operation control and acoustic loading.
Examples are given for the noise of a centrifugal fan with and without inlet box, the
effect of unsteady swirl flow in the entrance plane of a centrifugal impeller, the influence of
an upstream duct bend upon the noise generated by an axial fan, the effect of an intake nozzle
or a streamlined cowl on axial fan noise, the influence of flow diffusers mounted upstream
of centrifugal as well as axial fans, and for the excess noise as a result of fan volume flow
control by means of throttling and inlet guide vanes.
The experimental results reveal that axial fans are more sensitive to disturbed inflow
conditions with regard to the radiated noise than centrifugal fans.
Experimental data are also presented for the change in total fan sound power with the
acoustic loading of the fan when operating in type B, C, and D installations, i.e., with
free inletjducted outlet, with ducted inlet/free outlet, and with ducted inlet/ducted outlet,
respectively. For centrifugal fans, quietest fan operation is achieved when the fan inlet is not
ducted. The total sound power of the axial fan is highest with the outlet unducted.
Acoustic loading effects are not only found in the regime of plane wave sound fields in
the duct but also at higherorder duct mode frequencies. For example, increased free field
levels as compared to the induct levels were observed on the outlet side of an axial fan when
higherorder mode tone components where generated which were cut on in the fan annulus
but not yet propagational in the circular fan duct.
268
co
l:J
\\
,/'\
>C
\
\
I
\
100
\1
*.
I
I
Axialflow fan
3
0:: SOOmm, n:: 2970/min, 0:: 3.3m /s
ducted inlet/free outlet
    measured
  approximated 1.0mode
in 31S/S00mm annulus
9SL______~~~~
90
180
270
360
<P, Grad
Figure 14.21: Azimuthal sound pressure level distribution of three tone components in the
315 Hz onethird octave band in the unducted exit plane of the axial fan (after Holste &
Neise (1992a)).
14.6
BIBLIOGRAPHY OF CHAPTER 14
BAADE, P. K., 1977. Effects of acoustic loading on axial flow fan noise generation. Noise
Control Engineering 8, 515.
BARSIKOW, B. & NEISE, W., 1978. Der Einfluf3 ungleichmaf3iger Zustromung auf das
Gerausch von Radialventilatoren. In Fortschritte der Ak7Lstik) DAGA )78 (Bochum,
Germany), VDEVerlag Berlin, Germany, pp. 411414.
BARSIKOW, B. & NEISE, W., 1979. The influence of nonuniform inflow conditions on
centrifugal fan noise. In Proceedings Internoise )79 (Warsaw, Poland), pp. 8993.
BOMMES, L., 1980. Liirmminderung bei einem Radia.lventilator kleiner Schnellaufigkeit
unter Beriicksichtigung von Zungenform, Zungenabstand und Schaufelzahl. Hcizung
L74tung Haustcchnik 31, 173179 and 210218.
CREMER, L., 1971. The second annual fairey lecture: The treatment of fans as black boxes.
Journal of Sound and Vibration 16, 115.
FUCHS, H. V., MANTE, J., NEISE, W. & STAHL, B., 1978. Untersuchungen des L~1.rms
von Axialgeblasen in Kraftfahrzeugen. In Fortschritte der Akustik) DAGA )78 (Bochum,
Germany), VDEVerlagBerlin, Germany, pp. 403406.
HOLSTE, F. & NEISE, W., 1992. Experimental comparison of standardized sound power
measurement procedures for fans. Journal of S07J,nd and Vibration 152, 126.
269
HOPPE, G. & NEISE, W., 1987. Vergleich verschiedener GerauschmeBverfahren fur Ventilatoren. Forschungsbericht FLT 3/1/31/87, Forschungsvereinigung fur Luft und Trocknungstechnik e.V., Frankfurt/Main, Germany.
ISO 3744, 1981. Acoustics  Determination of sound power levels of noise sources  Engineering methods for freefield conditions over a reflecting plane. International Standard,
International Organisation for Standardization.
ISO 5136,1990. Acoustics Determination of sound power radiated into a duct by fans  Induct method. International Standard, International Organisation for Standardization.
KAISER, L. & WOHRLE, K., 1989. Variable speed blowers  a solution to control the noise
of computers. In Proceedings Internoise)89 (Newport Beach, Ca., USA), pp. 157162.
NEISE, W. & HOPPE, G., 1986. Effect of inflow conditions on centrifugal fan noise. In Proceedings First European Symposium on Air Conditioning and Refrigeration (Brussels,
Belgium), pp. 165172.
NEISE, W., HOLSTE, F. & HOPPE, G., 1988. Experimental comparison of standardized
sound power measurement procedures for fans. In Proceedings Internoise )88 (Avignon,
France), pp. 10971100.
NEISE, W., HOPPE, G. & HERRMANN, I., 1986. Intercomparison of fan noise measurements using the induct method ISO/DIS 5136. Interner Forschungsbericht DFVLRIB 2221486/B1, Deutsche Forschungs und Versuchsanstalt fur Luft und Raumfahrt,
Koln/Porz, Germany.
NEISE, W., HOPPE, G. & HERRMANN, I., 1987. Gerauschmessungen an Ventilatoren
 Vergleichende Untersuchungen nach dem KanalVerfahren DIN 45 635 Teil 9 bzw.
ISO/DIS 5136. Heizung Liiftung Haustechnik 38, 343351.
NEISE, W., 1988a. Fan noise  Generation mechanisms and control methods. In Proceedings
Internoise )88 (Avignon, France), pp. 767776.
NEISE, W., 1988b. Gerauschvergleich von Ventilatoren. Die spezifische Schalleistung
zur Beurteilung des Gerausches unterschiedlicher Ventilatortypen. Heizung Liiftung
Haustechnik 39, 392399.
NEISE, W., 1989. Noise rating of fans on the basis of the specific sound power level. In
Proceedings Tenth A ustralasian Fluid Mechanics Conference (Melbourne, Australia),
pp. 1.411.44.
NEISE, W., 1992. Review of fan noise generation mechanisms and control methods. In
Proceedings Fan Noise (Senlis, France), Centre Technique des Industries Mecaniques
(CETIM), pp. 4556.
PILTZ, E., 1974. Enel'giebedarf und Schallel'zeugung bei verschiedenen Methoden del' Volumenstromvariation in Ventilatol'anlagen. Heizung Liiftung Haustechnik 25, 207214.
SCHMIDT, L., 1976. Der EinfluB eines 90Grad Rohrkrummers auf die Schallemission eines
Axialventilators. Luft und Kiilietechnik, 287290.
WIELAND, H., 1986. Vergleich verschiedener Systeme zum Verandern der Forderleistung
bei RadialvenLilatoren. In Proceedings Ventilatoren im ind71,siriellen Einsaiz (Diisseldorf,
Germany), vol. 594 of VDI.. 13erichte, VDIVerlag, Dusseldorf, pp. 267281.
270
WOLLHERR,
271
272
Chapter 15
15.1
INTRODUCTION
This chapter is concerned with the determination of the sound power radiated from a
fan's inlet, outlet, casing, or combinations thereof. The three basic measurement methods based on sound pressure measurements available for this task are mentioned above. Of
the reverberationroom methods, ISO 3741 (1975) and ISO 3742 (1975) specify precision
methods for broadband sources and narrowband sources, respectively. For all practical
purposes, the standards ISO 37431 (1994) and ISO 37432 (1994) with engineering grade
accuracy are sufficient for fan noise testing. These two standards as applied to fans will be
discussed in more detail in section 15.2.
For the freefield method, three international standards with different grades of accuracy
are available: the precision method ISO 3745 (1977), the engineering method ISO 3744
(1981), and the survey method ISO 3746 (1992). Of these, the most relevant one for fan
noise testing is ISO 3744 (1981) which will be described in section 15.3.
The induct method ISO 5136 (1990) is specified only for engineering grade accuracy
measurements and is presented in section 15.4.
273
At the present time, an international test code for fan noise measurements is being
prepared by ISO /TC 117/WG 2/N8 (1991) which describes the special requirements for
applying the basic standards to fans and the specifications for mounting and operating the
fans during the test. More information about this document is given in section 1.5.5. Similar
national standards are specified in DIN 45 635 Teil 38 (1986) (Germany) and BS 848 Part
2 (1985) (Great Britain).
In section 15.6, results of experimental studies are presented in which the sound power
radiated from the inlet or outlet of different fan types was measured using the three basic
acoustical measurement methods.
15.2
REVERBERATIONROOM METHOD
In ISO 37431 (1994) a relatively simple engineering method is specified for determining the
sound power levels of small movable noise sources within test rooms with hard and sound
reflective walls. A comparison method is used to determine the sound power. The spatial
average of the octave band sound pressure levels produced in the test room by the source
under test < L p  st > is cO,mpared to the spatial average of the octave band sound pressure
levels < L p  rss > produced by a reference sound source of known sound power output. The
difference in sound pressure levels is equal to the difference in sound power levels if conditions
are the same for both sets of measurements (sound pressure levels corrected for background
noise ):
rss
>
+< L
p  st
>
(15.1)
The Aweighted sound power level is calculated from the octave band levels. The frequency range of interest includes the octave bands from 125 to 8000 Hz. At least three
microphone positions are required. If discrete tones or narrowband noise can be observed
by aural examination, the number of microphone positions is increased to six. If the standard
deviation of the sound pressure over the six measurement positions exceeds a value of 2.5
dB in an octave band, tests with two source locations are required; if the standard deviation
is larger than 4.0 dB, two additional source locations are required in another test room with
different dimensions. The minimum room volume required in ISO 37431 (1994) is 40m 3 ,
and the largest dimension of the source is 1.0m in a 40m 3 room and 2.0m in a 100m3 room.
Hence, this method is restricted to small and medium size fans.
In ISO 37432 (1994) a direct method for determining the sound power level of a noise
source is specified which eliminates the need for a reference sound source and permits the
determination of the Aweighted sound power level from Aweighted sound pressure measurements, but requires a special reverberation test room having a specified reverberation
time over the frequency range of interest. The volume of the test room shall be at least 70m 3
and preferably greater if the 125 Hz octave band is within the frequency range of interest. If
the 4 and 8 kHz bands are of interest too, the volume shall not exceed 300m 3 . The volume
of the source under test should not exceed 1 % of the room volume. With the above limits
for the test room volume, the recommended maximum size of the noise source is 0.7 to 3m 3 ,
which again limits the method to small to medium size fans.
The reverberation time T of the test room shall fall within limiting curves specified in
ISO 37432 (1994) which are defined by
T
T
f
f >
6300Hz and
6300Hz, where
(15.2)
fo
The minimum number of measurement points necessary to obtain the specified measurement accuracy has to be determined following an initial procedure which involves six
microphone positions. From the standard deviation of the six initially measured sound pressure levels, the number of measurement positions and source locations in the test room is
derived. Usage of a moving microphone traversing path in the test room instead of fixed
positions is also permitted. The sound power level of the source follows from the spatial
average of the sound pressure < L p  st >, the nominal reverberation time of the test room
Tnom(To Is), and the test room volume V :
l'nom
+ 10lg Va
13dB
(15.3)
The constant 13 dB instead of 14 dB, which appears in other ISO documents, and the
variation of the reverberation time with frequency account approximately for the increase of
sound energy density near the surfaces of the special reverberation test room and near the
source.
Note that both parts of the engineering method ISO 37431 (1994) and ISO 37432 (1994)
are limited to the determination of octave band and Aweighted sound power spectra. If
onethird octave band spectra are required, the precision methods ISO 3741 (1975) or ISO
3742 (1975) have to be used.
15.3
FREEFIELD METHOD
ISO 3744 (1981) specifies an engineering method for determining the sound power emitted by
a machine by measuring the sound pressure levels on a measurement surface enveloping the
source, under essentially freefield conditions near one or more reflecting planes. Free field
conditions are usually not encountered in typical machine rooms where sources are normally
installed. If measurements are made in such installations, corrections may be required for
background noise or undesired reflections from the room walls. These corrections reduce the
accuracy of the sound power determination.
To construct the measurement surface on which the microphone positions are located, a
hypothetical reference box is defined which is the smallest rectangular parallelepiped that
just encloses the source and terminates on the reflecting plane or planes. Elements protruding from the source which are not significant noise contributors may be disregarded. Two
different shapes of the measurement surface are prescribed in ISO 3744 (1981): a hemispherical surface (or partial hemispherical surface) and a rectangular parallelepiped whose sides
are parallel to those of the reference box. Figure 15.1 shows an example of the reference box,
the box shaped measurement surface, and the arrangement of measurement points for a tall
machine with a small base area. The preferred measurement distance d between reference
box and measurement surface is 1 m and shall be at least 0.25m.
The sound power level of the source under test is given by the following equation
(15.4)
275
Path 5
Reference
~x
Path 3
\J
+
r1
II
Path 1
Figure 15.1: Example of a measurement surface and microphone positions (paths) for a tall
machine (II :s; d; [2 :s; d; 2 d < [3 :s; 5 d) according to ISO 3744 (1981).
where < Lpf > is the sound pressure level averaged over the measurement surface A, ]{1
the background correction and ]{2 the environmental correction. ]{2 is to be determined
following one of the procedures specified in ISO 3744 (1981). For engineering grade accuracy,
the following limits apply: ]{1 :s; 1.3dB and ]{2 :s; 2dB. The frequency range covered by ISO
3744 (1981) ranges from the onethird octave bands 50 to 10 OOOHz.
If the sound power radiated from a fan inlet or fan outlet is to be determined, special
considerations are necessary to specify the dimensions of the reference box. In DIN 45 635
Teil 38 (1986) the reference box for such a case is defined in such a way that it encloses a
portion of the space adjacent to the fan aperture, see the example shown in Figure 15.2.
The dimensions of the reference box are functions of the fan dimensions: Length and width
are equal to the largest cross dimension D of the circular or rectangular fan inlet or outlet,
and the measurement distance d shall be equal to D for D ~ 0.5m and equal to 0.5m for
D < 0.5m. Different measurement locations are specified for the part of the measurement
surface facing the inlet and the outlet of the fan to ensure that the microphone is not
affected by flow noise pressures. The rectangular parallelepiped is the preferred shape for
the measurement surface according to DIN 45 635 Teil 38 (1986). In the example shown in
Figure 15.2, the number of measurement points is 16 for the fan inlet and 17 for the fan
outlet. The ducts connected to the nonmeasured side of the fan have to be terminated
anechoically, however, the requirements for the pressure reflection coefficient of the duct
terminator are much less stringent than for the test ducts according to the induct method
ISO 5136 (1990).
In the British fan noise testing standard BS 848 Part 2 (1985) a hemispherical measurement surface is described for freefield sound power measurements on fans with free inlet
276
u
UI
Figure 15.2: Reference box and measurement surface for a fan inlet or outlet aperture after
DIN 45 635 Teil 38 (1986); 1.5 d < h < 4.5 d; X additional point for fan inlet, 0 additional
points for fan outlet.
/ducted outlet (type B installation) or ducted inlet/free outlet (type C installation) which
is schematically shown in Figure 15.3. The hemispherical surface of radius R is centered
around the fan inlet or outlet aperture. The number of measurement positions is 13, and
the limits for the radius R are specified by R ;:: 2( h + D /2) or R ;:: 4h, whichever is larger;
h is the height of the fan inlet or outlet above the floor. The ducts connected to the test fan
are to be terminated anechoic ally, with the same requirements for reflection coefficient of
the duct terminator as for the test ducts according to the induct method ISO 5136 (1990).
15.4
INDUCT METHOD
The sound power radiated by a source into a duct depends on the type of duct, characterized by its acoustical impedance. Therefore the duct has to be specified for a measurement
method. In ISO 5136 (1990), the test duct is of circular cross section and terminated nearly
anechoically. The sound power determined under these special conditions is a representative value for actual applications, because the anechoic termination provides an acoustical
impedance which lies about midway between the higher and lower impedances encountered
in practice and is nearly independent of frequency. The termination also eliminates the problem of axial standing waves in the duct, i. e., the sound pressure is constant along the axial
direction. The duct terminations are specified in ISO 5136 (1990) in terms of a maximum
reflection coefficient as a function of frequency.
A schematic layout of the test arrangement for sound measurements on the fan inlet side
277
o
S
Fan
Duct
Microphone position
'"
ru
a
0.
ru
ru
ru
...ru
L.
a::
'
1"
dLf~~
Reference plane
r = radius of hemisphere
Figure 15.~): Microphone positions for a hemispherical measurement surface for fan installation types Band C according to BS 848 Part 2 (1985).
and the fan outlet side according to ISO 5136 (1990) is presented in Figure
of test duct diameters covered by ISO 5136 (1990) is from 0.15 to 2.0m.
can be used for a limited, specified range of fan sizes by employing conical
which have to meet certain requirements. Intermediate ducts have to be
278
   ___~f 
  
 

++,..LtlTi
He;::urement
plane
Test duct
Anechoic
termination
<,/~ 
~j~''r:..Lj
!,:,::mediO',
 duct
conical transition
Terminating
duct
Throttle
Anechoic
termination
Anechoic
termination
Anechoic
termination
Figure 15.4: Test arrangement for the induct method ISO 5136 (1990).
fan inlet and outlet to ensure undisturbed flow conditions. If the nonmeasured side of the
fan is normally ducted in the practical application, a terminating duct with an anechoic
termination has to be mounted on this side. If the nonmeasured side of the fan is normally
unducted, no terminating duct is required. Sound measurements can be made on the inlet
and outlet side simultaneously, of course, by providing anechoic test ducts on both sides.
The microphone placed in the duct is subject not only to the acoustic pressures but also
to the pressure fluctuations associated with the turbulent duct flow. To suppress the effect of these turbulent pressures on the microphone, the use of a long cylindrical windscreen
(,turbulence screen', 'sampling tube', 'slittube', see Neise (1975b)) is prescribed. The turbulence screen according to ISO 5136 (1990) shall suppress the turbulent pressure fluctuations
by at least 10dB in the frequency range of interest, and its directivity characteristic has
to be within specified limits. Two procedures are described in ISO 5136 (1990) to determine whether or not there is a sufficient signaltonoise ratio of sound to turbulence. The
maximum flow velocity in the test duct is 30m/s, and the maximum swirl angle is 15.
The microphone with the windscreen is mounted at a specified radial position such that
the measured sound pressure is acceptably well related to the sound power by the plane wave
formula, even in the frequency range of higherorder duct modes. The radial measurement
position T' is a function of the test duct diameter d, i. e., T' = 0.8d/2 for 0.15m:S d < 0.5m and
T' = 0.65d/2 for 0.5m:S d :S 2.0m. A circumferential average has to be obtained by measuring
at least three evenly spaced azimuthal positions or by a continuous circumferential traverse.
The sound power level in the test duct of cross section A (Ao = 1m2) is determined by
279
f (Hz)
3744
5136
50
63
80
5.0
5.0
5.0
3.5
3.0
2.5
100
125
163
5.0
3.0
3.0
3.0
2.5
2.0
2.0
3.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
200
250
315
37431
3.0
2.0
37432
400
500
630
1.5
2.0
1.5
1.5
1.5
3150
4000
5000
1.5
2.0
1.5
1.5
1.5
2.0
2.0
2.5
6300
8000
10000
2.5
3.0
2.5
2.5
2.5
3.0
3.5
4.0
Table 15.1: Estimated standard deviations of reproducibility of sound power levels according
to various ISOstandards
the circumferentially averaged sound pressure level < Lp >using the following relation:
pa
pa
(15.5)
Here C 1 is the free field microphone response correction, C2 the frequency response correction
of the turbulence screen, C3 the flow velocity correction which accounts for the change in
the frequency response of the turbulence screen as a result of the superimposed flow, and C 4
the socalled modal frequency correction which accounts for the directivity characteristics of
the microphone with the turbulence screen in view of the propagation angle of the higherorder duct modes. The headon frequency response C 2 has to be calibrated under acoustic
freefield conditions. Data for C3 as a function of frequency and flow velocity, and for C 4 as
function of frequency and duct diameter are tabulated in ISO 5136 (1990). The onethird
octave band centre frequency range covered in the standard is from 50 to 10000Hz.
In Table 15.1 the estimated values of the standard deviations of reproducibility of sound
power levels of the induct method are compared to those of the reverberationroom method
and the freefield method. ISO 37431 (1994) and ISO 37432 (1994) are applicable only
280
for octave bands between 125 and 8000Hz, while ISO 3744 (1981) and ISO 5136 (1990) are
for onethird octave bands in the range 50 to 10000Hz. Up to 315Hz, the induct method
yields the best accuracy and is only slightly worse than the freefield method up to 4000Hz.
The data for the induct method were derived from a European roundrobin test (Bolton
(1992b)) which was carried out by six laboratories using an axial fan and a centrifugal fan.
Each laboratory performed the measurements firstly with a reference test rig, which was
shipped together with the fans, and secondly with a rig of its own design. The results of this
test were used to establish the standard deviation data for ISO 5136 (1990) which are listed
in Table 15.1.
The results of the roundrobin test also indicated that problems arise with the turbulence
screen when there is a substantial amount of swirl in the test duct; this is frequently the case
in the outlet duct of axial fans. It was shown by Farzami & Guedel (1992) that these problems
can be overcome by placing a flow straightener between the fan and the measurement plane.
This, however, requires revision of ISO 5136 (1990).
The induct method ISO 5136 (1990) has been adopted in full or in part by several
national standards institutions, see DIN 45 635 Teil 9 (1989), BS 848 Part 2 (1985), and
AMCA 330/ ASHRAE 68R SPC (1994).
15.5
The measurement procedures specified in the draft standard prepared by ISO /TC 117/WG
2/N8 (1991) cover the following fan installations: Free inlet/free outlet (type A), free inlet/ducted outlet (type B), ducted inlet/free outlet (type C), and ducted inlet/ducted outlet (type D). To provide a specified acoustic load impedance of the fan during the sound
measurements, all ducts connected to the fan under test have to be fitted with anechoic
terminations.
In general, the sound power radiated from the fan inlet into free space and into a duct are
different and each is affected by what is connected to the outlet of the fan. Similarly the sound
power radiated from the fan outlet into free space and into a duct are different and depend
on the installation condition on the fan inlet side. Considering all possible combinations
for the above installation conditions, the following twelve different sound power levels are
defined in ISO/TC 117/WG 2/N8 (1991):
1. Lw( Ain): freeinlet sound power level, type A installation.
+ c):
0.40
0.35
0.30
0.8
0.7
0.6
100
125
> 163
0.25
0.15
0.15
0.5
0.3
0.2
Table 15.2: Maximum pressure reflection coefficients of the test ducts in ISO 5136 (1990)
and the terminating ducts in ISO /TC 117/WG 2/N8 (1991)
Polyesterfilm
all areas, including bottom,
except mounting panel
and exit port
o
o
..
Piezometer
pressure ring
behind panel
Mounting
panelassy, Testfan
Figure 15.5: Test plenum for sound power measurements on small air moving devices (after
ISO/DIS 10302 (1993)).
fan, which is mounted on one side wall of the plenum, can be adjusted by means of a slider
mechanism in the opposite wall. With this device, the fan sound power can be determined
by using either the reverberation room method ISO 37431 (1994) or ISO 37432 (1994), or
the free field method ISO 3744 (1981).
15.6
Holste & Neise (1992a) performed an experimental comparison of the following standardized
sound power measurement procedures as applied to fans: the reverberationroom method
ISO 3741 (1975) and ISO 3742 (1975), the freefield method over a reflecting plane ISO
3744 (1981), and the induct method ISO 5136 (1990). In spite of the fact that the induct
method is aimed at a principally different portion of the total fan noise than the other two,
there is considerable practical interest to compare the results of the different methods and
eventually to convert induct sound power levels to freespace sound power levels and vice
versa.
The following six fans with impeller diameters between 450 and 510mm were used for
the experiments: centrifugal fan with backward curved airfoil blades; centrifugal fan with
backward curved sheet metal blades; centrifugal fan with straight radial blades; centrifugal
fan with forward curved blades (scirocco blower); halfradial fan (mixedflow impeller running in a centrifugal fan casing); axialflow fan with outlet guide vanes. The sound power
283
measurements were made at the respective optimum fan operation conditions. More details
are given in the original paper by Holste & Neise (1992a).
Figure 15.6 shows schematically the measurement setups for the three sound power mea
Reverberation room
Indue tmethod
0
0
0
>
lrl
Intake
xJ
lrl
e
0
0

~*
Microphone positions
lrl
x/
0
0
If)
e
o
0
If)
0
0
0
N
i;
= 200 m 3
S
0
0
lrl
co
N
e0
0
If)
0
0
0
N
S
0
0
lrl
320x635
500<1>
500<1> 500<1>
500'1>
1~.~~I.~~_Jl~.=~_~I.
~50~O<l>_ _.~I~_1
1200
2000
280 2000
2500
1050
Figure 15.6: Measurement setups used by Holste & Neise (1992a).
surement procedures for the inlet side of a centrifugal fan. The test setups for the outlet
sides are accordingly. During the freefield measurements on the inlet side, simultaneous
sound measurements were made in the anechoic test duct remaining on the outlet side, as is
shown in Figure 15.6, and vice versa.
The reverberation room used for the tests had a volume of 200m 3 . The direct method
specified in ISO 3741 (1975), ISO 3742 (1975), i.e., measurement of the reverberation time,
was used to determine the fan sound power. The sound field was measured at six microphone
positions.
The freefield tests were made in a laboratory room of about 500m 3 volume. It was necessary to increase the room sound absorption by mounting about 200m 2 of 50mm thick rubber
foam mats at varying distance from the walls to meet the requirement for the environmental
correction f{2 :::;: 2 dB in ISO 3744 (1981). The number of measurement points was 15 to 17,
depending on how many points were occupied by parts of the equipment.
The induct facility meets the requirements of ISO 5136 (1990). The test ducts are
terminated anechoically. They are fabricated from 1.25mm thick sheet metal and wrapped
284
with a 10mm thick layer of rubber foam plus a 1mm thick lead sheet to eliminate wall
vibrations and to increase the sound insulation. The microphones fitted with turbulence
screens are mounted in short rotatable duct sections and can be continuously traversed over
a full circumference while taking the time average.
With both inlet and outlet duct connected to the test fan, the fan operation condition
was controlled by means of the throttle at the end of the outlet duct, and the volume flow
was measured via the static pressure in the inlet nozzle, compare Figure 15.6. In case of the
freefield and the reverberationroom tests on the inlet side, the volume flow was determined
by measuring the velocity profile in the outlet duct. Finally, with the fan outlet unducted,
the fan operation was controlled by inserting screens between inlet nozzle and anechoic
termination, to ensure uniform inlet flow conditions for the test fan; the volume flow was
monitored via a velocity traverse in the inlet duct.
The experimental data presented by Holste & Neise (1992a) showed very good agreement
between the results of the freefield method and reverberationroom method. Both methods
seem equally well suited for the determination of the sound power radiated from an unducted
fan inlet or outlet, however the reverberation room method is applicable only to smaller fans
because the size of the test object should preferably be less than 1% of the room volume.
In the plane wave region, the induct method yielded higher levels than the freefield and
the reverberationroom method which is due to the reflection of the sound waves at the fan
inlet or outlet, when the duct is removed. In Figure 15.7 the differences between freefield
inlet side
m
u
10
20 
outlet side
OJ
'l
u
.~
:<
10
0
,u
110~
:<
20
"
o 0 0___<l:.~. ::;':
* ..
s...0"i(~ .~'
8:
'r' 
Q 0
!QR~g~~s80
r_ _ _ _ _
o"'~_';'~~_o_
'
'
0 __0_0_ _ _ _ .. _
/'
Roq(mm)
inlet
182
182
128
205
250
Gullet
254
251.
251.
251.
250
Figure 15.7: Difference between freefield and induct sound power levels (after Holste &
Neise (1992a)).
spectra and induct spectra are plotted for the inlet and outlet side of all test fans. The
285
two additional curves show for comparison the effect of the sound reflection at an unflanged
circular pipe and a duct ending in a wall. The measured differences between induct and free
space sound power level can be described approximately by using the reflection characteristics
of a flanged duct end. Another effect which results in differences between induct and free
space levels is the change in acoustic loading when the duct is removed.
In the frequency range of higherorder mode sound propagation, the induct sound power
levels are lower than the freefield levels. The level difference is frequency dependent with
average values of about 3dB on the inlet side and about 5dB on the outlet side. Similar results
were reported by Bolton (1992a). In this frequency regime, where the wavelength is small
compared to the cross dimensions of the duct, effects of sound reflection and acoustic loading
are unlikely to playa significant role, and one would expect all test methods to deliver the
same result. Nevertheless, the induct levels are somewhat lower than the levels measured in
the free field or reverberation room. Since the latter two yield consistent results, it appears
that the reason for this discrepancy lies in the induct method. As a possible cause, Holste
& Neise (1992a) named the two frequency correction terms C3 and C4 in ISO 5136 (1990)
which are based on theoretical and extrapolated experimental data. The data for the flow
velocity correction C3 are the result of theoretical investigations by Neise (1975b) which were
not verified experimentally at that time. However, recent experiments by Neise (1991) show
that there is reasonable agreement between measured and calculated C3 data. The modal
correction C4 accounts for the directivity of the microphone with the turbulence screen with
respect to the propagation angle of the higherorder duct modes. The data given in ISO
5136 (1990) are based on theoretical and experimental studies by Bollder & Crocker (1972)
and Bolleter, Cohen & Wang (1973) which were extrapolated to higher frequencies and duct
diameters. For the theoretical part, an empirical model for the directivity of the microphone
probe was employed. Recent theoretical considerations by Michalke (1992) show that higher
values for the modal correction C4 are in fact obtained when the theoretical slittube model
described by Neise (1975b) is used to calculate the microphone probe directivity. From the
above discussion it appears that the modal correction data C4 presented in ISO 5136 (1990)
need to be reexamined.
In a recent European round robin test, freefield sound power measurements were performed on an axialflow fan and a centrifugal fan, both of 600mm impeller diameter. Each
fan was measured in two installation configurations: with an open inlet and an anechoic
outlet duct (Type B installation according to ISO 5801 (1994)) and with open outlet and
an anechoic inlet duct (type C installation). In each installation condition, simultaneous
sound power measurements were made in the ducts according to the induct method ISO
5136 (1990). The objective of the intercomparison was to investigate differences in measured
sound power levels in the free field at the open inlet or open outlet of the test fans when
different measurement methods or, rather, different measurement surfaces are applied and
used by different laboratories. This information was intended to provide the basis for the
selection of an optimum procedure which can be incorporated into a unified European fan
noise measurement standard.
Five European laboratories took part in the round robin test. The two test fans were
circulated between the participants, one a beltdriven centrifugal fan and the other an axial
fan with no guide vanes with closecoupled motor mounted in the hub. The fans were
mounted in test rigs assembled from a standard set of ducting which was circulated with the
fans. This was to ensure identical installation conditions for every measurement.
Three different surfaces were employed for the freefield tests:
a) A hemispherical measurement surface with nominally 13 measurement points according
to BS 848 Part 2 (1985). For the test fans given (D
600mm impeller diameter) and the
height of the fan aperture above the floor (h 1357 to 1705mm), the radius of the hemisphere
286
according to BS 848 Part 2 (1985) was R = 5428 to 6820mm. A measurement surface that
large was not practicable for the test environments available in the various laboratories.
Therefore the limits on the radius of the measurement hemisphere specified in BS 848 Part
2 (1985) were relaxed so that the radius used by each of the participating laboratories was
the largest that could be accommodated in their test facility.
b) A rectangular box (parallelepiped) as described in DIN 45 635 Teil 38 (1986). For the
given fan sizes and height h of the fan inlet or outlet above the sound refiecting surface (fioor),
the nominal number of measurement points was 16 on the fan inlet side and 17 on the fan
outlet side, compare Figure 15.2. The dimensions of the reference box are d x d x (h + 1.5 d),
where the characteristic dimension of the reference box d is equal to the diameter of the fan
aperture, i. e., d = 600mm for the inlet and outlet of both test fans.
c) A parallelepiped similar to the one prescribed in DIN 45 635 Teil 38 (1986), only the
characteristic dimension was chosen to be d = 1000mm. This resulted in a larger distance
between reference box and measurement surface (1000mm instead of 600mm as before) and
a larger measurement surface. The number of measurement points was the same as under
b).
The environmental correction J{2 in equation (15.4) was determined by using a calibrated
reference sound source (Bruel & Kjaer type 4204) which was circulated with the test fans.
At the time of this writing, the final report of the round robin test was not completed,
only the test results of one of the participating laboratories (Neise et al. (1993)) was available
to the author. In these tests, the sound power measurements were made in a large laboratory
of 22.0 x 24.5m 2 floor space and an average height of 10m. There was very little other test
equipment present while the noise tests were done. To reduce sound refiections, 50mm thick
foam mats were spread over the refiecting surfaces closest to the measurement surfaces. For
the hemispherical measurement surface, a constant radius of R = 3000mm was used for all
configurations tested.
Despite the fact that the freefield tests were made under very favourable environmental
conditions, considering practical circumstances, it was found not possible to meet the requirement set in ISO 3744 (1981) for the environmental correction, J{2 ::; 2dB throughout
the frequency range of interest for engineering grade accuracy. In fact, values of as much as
4dB were measured, see Figure 15.8. On average, the lowest values for the environmental
correction J{2 were found for the smallest measurement surface which is a parallelepiped
according to DIN 45 635 Teil 38 (1986). Also, the smallest measurement surface proved to
be less sensitive to the infiuences of background noise than the larger ones.
The differences between the results of different measurement surfaces are largest at those
frequencies at which the environmental correction J{2 is also large. Level differences of up
to 4dB were found. Examples of the test results are given in Figure 15.9 for the inlet side of
the axialfiow fan and in Figure 15.10 for the outlet side of the centrifugal fan.
The differences measured between induct levels and freefield levels are consistent with
the findings of Holste & Neise (1992a): In the frequency range where only plane waves are
propagational in the test ducts, the induct levels are higher than the freefield levels, and in
the frequency range of higherorder mode sound propagation, the induct sound power levels
are lower than the freefield levels. Possible explanations for these differences were discussed
above.
287
OJ
ru
:L
Setup
for axial fan, inlet and
outlet side
1
1
dill' '6 3 1~5, '2~0 500' 'l:k' '2 k' '4:k 'B:k,A, lin
dill
OJ
31
G8Parallelepiped d= 600mm
0  oParallelepiped d=1000mm
eoHemisphere R=3000mm
Figure 15.8: Environmental correction ]{2 measured for the test configurations on the inlet
and outlet side of the test fans (after Neise et al. (1993)).
15.7
CONCLUSIONS
Three basic acoustical measurement methods of engineering grade accuracy were discussed
for the determination of the sound power emitted by fans: the reverberationroom method
ISO 37431 (1994) and ISO 37432 (1994), the freefield method over a reflecting plane ISO
3744 (1981), and the induct method ISO 5136 (1990). A test code specifically designed
for fan noise testing is presently prepared by ISO /TC 117/WG 2/N8 (1991). Both the
comparison method ISO 37431 (1994) and the direct method ISO 37432 (1994) are valid
for octave bands only and cover the frequency range from 125 to 8000Hz. In ISO 37431
(1994), the minimum number of measurement points in the reverberation room is three, but
six positions and more than one source location, possibly in two different test rooms, may
be required when tonal components are present in the spectrum. A qualification procedure
involving six microphone positions is necessary to determine the number of measurement
positions and source locations for the direct method ISO 37432 (1994). Both reverberationroom methods are applicable only to small to medium size fans because the maximum
dimension allowed in ISO 37431 (1994) is 1 to 2m, depending of the volume of the test
room, and in ISO 37432 (1994) the size of the test object is preferably less than 1% of the
room volume. Practical experience with a 200m3 room and an axial fan working at 3.3m3 /s
volume flow (Holste & Neise (1992a)) showed that the circulation of air in the room may
interfere with the sound measurements.
The freefield method ISO 3744 (1981) covers the onethird octave band frequency range
from 50 to 10 OOOHz and is applicable for all fan sizes and for all types of noise. The
288
3.~
I
J
o:J
D
"
<J
!)
'"
J \
<\(
\!>
_/'~:~
\
00<'1
<0
24
11031
53
lk
100
o:J
D
90
~~
3
"
80
Axial fan, inlet side
Q=5.1m 3/s (BEP, ori fice #1)
G8Parallelepiped d= 500mm
0  oParallelepiped d=1000mm
GBHemisphere R=3000mm
125 250 500 1k 2k 4k 8k A lin
f, Hz
A
70
60]1
53
Figure 15.9: Comparison of the sound power spectra obtained by using different measurement
surfaces. Axial fan; inlet side; Q = 5.1m3 js (best efficiency point); after Neise et al. (1993).
number of measurement points is fairly large, depending on the fan size tested. To give an
example, 15 to 17 microphone positions were required to measure the sound power radiated
from the inlet or outlet of a 500mm impeller diameter fan, see the experiments by Holste &
Neise (1992a). The same authors reported that, to meet the condition for the environmental
correction J{2 :::; 2dB over the entire frequency range of interest, which is one of the criteria
for engineering grade accuracy, it was necessary to increase the acoustic absorption in the
test room substantially. Another problem was the generation of background noise by the
fan drive system and by the throttle needed to control the fan operating condition, both of
which are difficult to separate from the fan noise to be measured. In an experimental study
as part of a European round robin test by Neise et al. (1993), it was found not possible to
meet the criterion for the environmental correction J{2 :::; 2dB specified in ISO 3744 (1981)
for engineering grade accuracy. Values of as much as 4 dB were measured despite the fact
that the environmental conditions were favourable, considering practical situations.
The induct method ISO 5136 (1990) requires that the sound pressure be measured
in an anechoically terminated test duct. Due to this, the method is largely insensitive
to environmental conditions and background noise problems. The measurement procedure
requires only three microphone positions in the duct or a circumferential microphone traverse.
The method has been tested specifically for fans in a European roundrobin test (Bolton
(1992b)) and was found very practical and to give reproducible results. On the other hand,
a few problems with the present standard remain and require revision in the near future:
(1) Swirl flow existing in the test duct may interfere with the microphone probe. Farzami
& Guedel (1992) showed that this problem can be overcome by placing a flow straightener
289
\ ,~:
co
u
30
'
<J
<S<
'00
&<1;
2
~
l1ciil 1
100
co
u
90
&
80
~
v
63
125
250
500 1k
f, Hz
2k
~~
~k
8k A. lin
Figure 15.10: Comparison of the sound power spectra obtained by using different measurement surfaces. Centrifugal fan; outlet side; Q = 3.5m3 /s (best efficiency point); after Neise
et al. (1993).
between fan and measurement plane.
(2) By comparison with other methods of determining the fan sound power, see Holste
& Neise (1992a), the data for the modal correction C 4 presented in the standard seem to
be in error and need to be revised. This statement is also supported by Davy (1994) who
presented theoretical considerations on the modal corection C4 .
Current experimental and theoretical investigations at DLR in Berlin are aimed at providing more realistic data for the modal correction to be implemented in the standard.
Moreover, a revised induct standard may also include measurement procedures for rectangular ducts, following the suggestions from a recent study by Neise & Holste (1991) on this
topic.
15.8
BIBLIOGRAPHY OF CHAPTER 15
AMCA 30085, 1985. Reverberation room method for sound testing of fans. American
standard, Air Movement and Control Association, Arlington Heights, Illinois, USA.
AMCA 330/ ASHRAE 68R SPC, 1994. in preparation, Air Movement and Control Association, Arlington Heights, Illinois, USA.
290
BOLLETER, U. & CROCKER, M. J., 1972. Theory and measurement of modal spectra in
hardwalled cylindrical ducts. Journal of the Acoustical Society of America 51, 1439~
1477.
BOLLETER, U., COHEN, R. & WANG, J., 1973. Design considerations for an induct sound
power measuring system. Journal of Sound and Vibration 28, 669685.
BOLTON, A. N., 1992a. Fan noise installation effects. In Proceedings fan Noise (Senlis,
France), Centre Technique des Industries Mecaniques (CETIM), pp. 7788.
BOLTON, A. N., 1992b. Intercomparison of induct fan noise measurements. Report EUR
13890 EN, Commision of the European Communities, Community Bureau of Reference,
Brussels, Belgium.
BS 848 PART 2,1985. Fans for general purposes. Methods of noise testing. British Standard,
British Standards Institution.
DAVY, J. L., in press, 1994. The modal and flow velocity corrections of microphone turbulence screens. Journal of Sound and Vibration.
DIN 45 635 TElL 38, 1986.
Gerauschmessung an Maschinen. Luftschallemission,
Hiillfliichen, Hallraum und Kanal Verfahren; Ventilatoren. Deutsche Norm, Deutsches
Institut fiir Normung e.V., Berlin, Germany.
DIN 45 635 TElL 9, 1989. Geriiuschmessung an Maschinen. Luftschallemission, KanalVerfahren; RahmenMeBverfahren fiir Genauigkeitsklasse 2. Deutsche Norm, Deutsches
Institut fiir Normung e.V., Berlin, Germany.
FARZAMI, R. & GUEDEL, A., 1992. Influence of a straightener on induct sound power measurements. In Proceedings Fan Noise (Senlis, France), Centre Technique des Industries
Mecaniques (CETIM), pp. 375~380.
HOLSTE, F. & NEISE, W., 1992. Experimental comparison of standardized sound power
measurement procedures for fans. Journal of Sound and Vibration 152, 1~26.
ISO 3741, 1975. Acoustics ~ Determination of sound power levels of noise sources ~ Precision methods for broadband sources in reverberation rooms. International Standard,
International Organisation for Standardization.
ISO 3742, 1975. Acoustics  Determination of sound power levels of noise sources ~ Precision methods for discretefrequency and narrowband sources in reverberation rooms.
International Standard, International Organisation for Standardization.
ISO 37431, 1994. Acoustics Determination of sound power levels of noise sources ~ Engineering methods for small movable sources in reverberant fields ~ Part 1: Comparison
method in hardwalled test rooms. International Standard, International Organisation
for Standardization.
ISO 37432, 1994. Acoustics Determination of sound power levels of noise sources ~ Engineering methods for small movable sources in reverberant fields Part 2: Special test
rooms. International Standard (in press), International Organisation for Standardization.
291
ISO 3744, 1981. Acoustics  Determination of sound power levels of noise sources  Engineering methods for freefield conditions over a reflecting plane. International Standard,
International Organisation for Standardization.
ISO 3745,1981. Acoustics  Determination of sound power levels of noise sources  Precision
methods for anechoic or semianechoic rooms. International Standard, International
Organisation for Standardization.
ISO 3746, 1992. Acoustics  Determination of sound power levels of noise sources Survey
method employing an enveloping measurement surface over a reflecting plane. Draft
International Standard, International Organisation for Standardization.
ISO 5136,1990. Acoustics Determination of sound power radiated into a duct by fans  Induct method. International Standard, International Organisation for Standardization.
ISO 5801, 1994. Performance testing of fans using standardized airways. International
Standard, in preparation, International Organisation for Standardization.
ISO /DIS 10302, 1993. Acoustics  Method for the measurement of airborne noise emitted
by small airmoving devices. Draft International Standard, International Organisation
for Standardization.
ISO/TC 117/WG 2/N8, 1991. The determination of fan sound power levels under standardized laboratory conditions. Draft Standard, International Organisation for Standardization.
MICHALKE, A., 1992. On the sensitivity of a slittube probe to the higherorder acoustic
modes and its consequences. In Proceedings DGLR/AIAA 14th Aeroacoustics Conference (Aachen, Germany), Deutsche Gesellschaft fiir Luft und Raumfahrt, Bonn,
Germany, pp. 163168.
NEISE, W. & HOLSTE, F., 1991. Determination of sound power in rectangular flow ducts.
In Proceedings Internoise'91 (Sydney, Australia), pp. 10091012.
NEISE, W., HOLSTE, F., MIRANDA, L. & HERRMANN, M., 1993. Intercomparison of openinlet/openoutlet sound power measurements on fans. Interner Forschungsbericht DLRIB 2221493/B5, Deutsche Forschungs und Versuchsanstalt fiir Luft und Raumfahrt,
Koln/Porz, Germany.
NEISE, W., 1975. Theoretical and experimental investigations of microphone probes for
sound measurements in turbulent flow. Journal of Sound and Vibration 39, 371400.
NEISE, W., 1991. Experimental determination of the acoustic sensitivity of a microphone
turbulence screen under mean flow conditions. In Proceedings Internoise '91 (Sydney,
Australia), pp. 10131016.
292
Chapter 16
293
. . . . . . . . . 294
298
16.4 SCALING FAN NOISE DATA FROM A MODEL TO A FULL SIZE FAN . 307
16.4.1 GENERAL REMARKS
. . . . . . . . . . . . . . . . . . . . . . . . . 307
. . . . . . . . . . . 308
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
16.1
INTRODUCTION
In recent years national and international organizations for standardization have proposed a
number of standards for the determination of the sound power of noise sources that are also
applicable to fans. Testing a large number of fan types and fan sizes is costly because of the
experimental facilities and running time involved. In this paper it is shown that the cost can
be reduced substantially by applying similarity laws to the aerodynamic fan noise. Thereby,
the sound power of a whole family of fans, which are different in size but dimensionally
similar, can be calculated based on experimental data on only one size of the fan series.
16.2
FORMULATION OF SIMILARITY
RELATIONSHIPS GOVERNING FAN NOISE
16.2.1
26pt
.
'ljJt = U2 pressure coefficIent
po
i.p
16.2.2
4Q
7r D2
fl
ffi .
U ow coe Clent
(16.1)
(16.2)
The earliest attempt to formulate a "fan sound law" for practical purposes was made by
Madison (1949): for a dimensionally similar fan series operating at the same point of the
pressurehead/volumeflow characteristic, the total sound power P was set proportional to
the volume flow Q and the square of the fan static pressure 6ps :
(16.3)
This relationship was derived under the assumption that the aerodynamic flow through the
rotor is similar for all members of the fan series and that the effect of viscosity can be
neglected. Although this relation is incorrect from a dimensional point of view, Madison's
fan sound law formed the basis for the most frequently used formula to estimate the sound
power level emitted by a fan:
Lw = LWsM
+ 10lg(Q/Qo) + 20lg(6ps/6pso),
(16.4)
Here LWsM is the specific sound power level which depends on the fan type and the operating
condition, and Qo and 6pso are reference quantities, e.g., Qo = Im 3 /s, 6pso = IPa. Using
the aerodynamic fan scaling laws (Q ex U D2 and 6ps ex PoU 2, where D and U are the
impeller diameter and tip speed, respectively, and po is the density of the working fluid),
equation (16.3) can be rearranged to give P ex p02D 2U 5 from which it becomes obvious that
a fixed tip speed exponent I = 5 is assumed.
294
The sound power P generated by an aerodynamic sound source can described by the
following relation, see e.g. Goldstein (1976):
(16.5)
where Land u mean a characteristic length and flow velocity of the aerodynamic source, and
ao is the speed of sound. The exponent m of the flow Mach number M a uj ao depends on
the source type. Under acoustic freefield conditions, m
1 for acoustic monopoles, m
3
for dipoles, and m
5 for quadrupoles. The U 5 dependence of the fan sound law agrees
with neither the u 6 dependence of the dipole sound radiation, nor with the u 4 dependence
of the monopole radiation. This lack of agreement stimulated Chanaud (1965) to perform
experiments with four uncased fan impellers with forward curved blades in the diameter range
D = 115 to 230mm. Measurements of the overall broadband sound power level confirmed
Madison's U 5 dependence but when frequencies below 100 Hz were excluded (because of the
insufficient measurement accuracy), the tip speed exponent became slightly greater than 6,
and Chanaud concluded that in general the dipole relationship applies to the rotor sound.
Note, however, that the speed exponent m is not only a function of the source type but
also of the type of sound propagation, as was pointed out by Stiiber & Heckl (in: Heckl &
Miiller (1975)) and Bartenwerfer (1977):
m = (ndim 
3)
+ (2e 
1)
(16.6)
(16.7)
where Poo is a suitable reference pressure. The spectral noise component pw is defined differently for the tonal noise and the random noise:
PWT
Clp
(6.jDjU)1/2
(16.8)
(16.9)
In case of the random noise, the filtered sound pressure 6.p has to be related to the filter
bandwidth Clj, which in turn is normalized by the impeller diameter D and tip speed U.
The nondimensional coordinates of the measuring position are denoted by xi/ D and the
isentropic exponent of the working fluid by !'C. The other nondimensional parameters are
defined by
Be
UD
  Reynolds number
v
295
(16.10)
Ma =
Sf
fD7r
Z
U
Mach number
ao
(16.11)
f
BP F Strouhal number
(16.12)
where v is the kinematic viscosity of the fluid, ao is the speed of sound and Z is the impeller
blade number. The foregoing definition of the Strouhal number is such that integer values
St = 1, 2, 3, ... correspond to the blade passage frequency (BPF) and its harmonics while
noninteger values indicate random noise components.
Weidemann (1971) studied the sound radiation from uncased centrifugal impellers under
freefield conditions with a rectangular wedge simulating the cutoff. From his results he
concluded that, for /)', !.p, x;/ D = const, the general relationship in equation (16.7) can be
replaced by a product of nondimensional terms:
Pw
/)',!.p,
x;/ D = canst
(16.13)
with the Helmholtz number appearing as a new parameter which can be derived from the
Mach number and the Strouhal number (sound wavelength A = ao/1) :
He =  = Ma . St . 
(16.14)
A
7r
The second and third term on the righthand side of equation (16.13) are well known from the
previous literature. They govern the influence of U, D, and von the sound pressure radiated.
The new and important feature of equation (16.13) is, however, that the sound spectrum
Pw is determined by two spectral functions F(St) and G(He) which depend on different
frequency parameters. The function F(St) is based on Strouhal number or, in other words,
is determined by the motion of the impeller, and it describes the spectral distribution of the
aerodynamic sound generated inside the fan. The second function, G( He), is a function of
a purely geometric parameter: the ratio of impeller diameter to sound wavelength radiated.
Therefore G( He) can be considered as an acoustic system response function, which accounts
for the acoustic resonance and radiation properties of the fan.
Weidemann's experimental results were consistent with the functional form of equation
(16.13). For the discrete frequency sound, values of ad = 2.6 and (3d = 0.2 were found for the
influence of Mach number and Reynolds number, and a r = 2.6 and (3r c:::'. 0 for the random
noise components.
It was shown by Neise (1975a) that there is a tremendous scattering of the tip speed
exponent, , = a + (3, measured by various experimenters. Not only does the tip speed exponent depend on the fan type considered, but it also varies with the fan operating condition.
It is believed that part of the scatter can be explained by the effect of the system response
function G( He): Consider some kind of sound source being inside the fan, the sound pressure of which increases as P ex U'Y (see the solid line in the top graph of Figure 16.1). The
frequency of the dominant spectral components of the noise, for example the blade passing
frequency, is simultaneously increased with the tip speed. If one imagines that the radiation
efficiency of, say, the fan outlet improves as the frequency is increased (dotted line in the
bottom graph), one would measure an apparently higher tipspeed exponent on the outlet
side than that of the sound source considered ({outlet> " see Figure 16.1). The frequency
dependence of the sound radiation efficiency of the fan inlet side may be just inverse, and one
would measure ,inlet < , (dashed line in Figure 16.1.). Since the radiation characteristics of
fan inlet and fan outlet depend on their acoustic loading, i.e., the acoustic impedances of the
296
Outlet
.~..~
~t
./
.:~~~Y
   ('
Inlet duct
Ig U
Outlet duct
20lgG [I)
O~~~~+
.__,,,
Ig 1
"
Inlet duct
Figure 16.1: Influence of the acoustic system frequency response G on the tip speed exponent
& Barsikow (1980, 1982)).
inlet duct system and the outlet duct system, the apparent tipspeed exponent may vary as
the duct system is altered. Obviously, the tip speed exponent is not a typical constant for
any particular fan.
Therefore, Bartenwerfer, Agnon & Gikadi (1976) and Neise & Barsikow (1982) used a
formulation of the acoustic similarity law in which the reference pressure Poo was chosen to
be POU 2 and the tipspeed exponent is no longer variable:
(16.15)
Any residual influence of the tip speed on the radiated sound pressure is included in the
functions F and G, and consequently the Mach number does not appear as an independent
variable anymore.
It should be emphasized that either one of the equations (16.13) or (16.15) may be used
when frequency spectra of dimensionally similar fans are to be scaled. The important point
is that the constants a and (3 and the functions F(St) and G(He), or F(St) and G(He),
respectively, are identical for the different fan sizes of the fan family.
16.3
16.3.1
Bartenwerfer & Agnon (1976) measured the noise of a centrifugal model fan with 140 mm
impeller diameter and six backward curved blades. The fan together with a short outlet duct
was placed inside a lowpressure chamber. With this arrangement the Reynolds number was
297
varied in the range 1.4.10 4 < Fe < 4.5.10 5 by changing the density of air, i.e., the pressure
in the chamber. Sound measurements were made at various positions in the chamber, in the
acoustic near field as well as the far field. In Figure 16.2 are shown the nondimensional sound
20,~
30
11.0 mm
Discrete noise
U =50 m/s
F or field
To = 298 K
~~~~~~:.
St 1
='
::J
1.0
a
Cl
(Weid~
"
ICl
<I
en
50
60
2.0
1.0
Re.1O 5
3.0
1..0
Figure 16.2: Nondimensional sound pressure level of the blade tone harmonics as functions
of the Reynolds number (after Bartenwerfer & Agnon (1976)).
pressure levels of the total noise (linear) and the blade tone harmonics (Si = 1, 2, 3, .. ) in
the acoustic far field (outlet duct) as a function of the Reynolds number. The experimental
data reveal that the various fan noise levels are a function of the Reynolds number only in
the range Fe < 10 5 while for Fe 2:: 10 5 the influence of the Reynolds number is negligibly
small.
16.3.2
Test Arrangement
Neise & Barsikow (1980) and Neise & Barsikow (1982) performed experiments with three
dimensionally similar centrifugal fans with impeller diameters of 140, 280, and 560 mm,
see the schematic given in Figure 16.3. Sound measurements were made in anechoically
terminated outlet ducts (see Figure 16.4), following the guidelines of the standardized induct method ISO 5136 (1990). For a more detailed description of the experimental facilities,
see Neise & Barsikow (1980).
Experimental Results
Overall sound pressure level In Figure 16.5, the overall sound pressure levels in the
outlet ducts of the three fans are shown for four aerodynamic loading conditions. Neise &
Barsikow (1980, 1982) used the following definition for the flow coefficient of a centrifugal
298
0.37 0
Cut  off
Outlet
0.440
_.
Inlet
0.410
0
6
= 0.008
b=0.150
51.5
70
61
i_g_~_:+2_1l_l_3_1~~g_:_,
ri;l
'0J
t!~!J
70
w,th j"tub,
:I
I
2960
280
560
50
20b:U
'@
7
I
261.,0~
f 2570
11.,0
Jt~ j~
r_ _M_icfro_p_h_on_e_ _
70r/J
"'"
520
An"hm, teeminati"
1.,520
I.,
I.,
570 11
960
Radial microphone position
with nosecone
d [mmJ
2?/d
2r/d
70
0.5
0.5
0.5
0.8
0.8
0.65
11.,0
d
with slittube
300
Figure 16.4: Schematic of the experimental setup (after Neise & Barsikow (1980, 1982))
fan: t.p = Q/,rrbDU, where b is the impeller width. The duct diameters were d 1 = 70rnm,
d2 = 140mm, and d3 = 300mm (see Figure 16.4), for the impeller diameters Dl = 140mm,
D2 = 280mm, and D3 = 560mm, respectively. In order to compare the experimental data
directly, the sound pressure levels measured in the duct with d3 = 300mm were converted
to levels in a 280mm diameter duct on the basis of constant sound power (Lp280 = Lp300 +
299
140~,~'
0) =140 mm
02= 280 mm
dB
130 f.
03 = 560 mm
120r~'~II
100
"_._
\.() = 0,09
\.() = 0,052
(design point)
90
140
dB
130 120
110
100
90
\.() = 0.166
\.() = 0.130
15
20
30
40
50 60
20
30
m/s
60
U 
Figure 16.5: Overall sound pressure levels in the fan outlet ducts of the three dimensionally
similar test fans, for four aerodynamic loading conditions (after Neise & Barsikow (1980,
1982)).
20 19(300/280) = Lp300 + 0.6 dB).
The curves for D3 = 560 mm in Figure 16.5 exhibit peaks at U ~ 26m/s (n = 880/min),
which corresponds to a blade passage frequency of 88 Hz. At this frequency, the nonreflecting
capability of the duct termination was insufficient (pressure reflection coefficient of 0.5), and
the axial standing wave in the test duct resulted in a peaked blade passage frequency level
and overall sound pressure level.
The curves measured for the different fans do not show a clear tendency that would
indicate an influence of the fan size or, in other words, the Reynolds number upon the
radiated sound pressure. This is in agreement with the experimental results of Bartenwerfer
& Agnon (1976) who varied the Reynolds number via the air density.
Note that the sound power emitted by the fan follows a D2law because the crosssectional areas of the three test ducts are related by Ad Ad A3 = Dl 2/ D2 2 / D3 2.
The mean slope of the curves in Figure 16.5 is not the same for the four different operating
conditions, and hence the tip speed exponent is a function of the flow coefficient. The sound
pressure is minimum at c.p = 0.09 which is near the point of optimum efficiency.
Tonal fan noise components In Figure 16.6( ad) are shown the levels of the blade
passage frequency (St = 1) and its harmonics (St
2, 3, 4 ... ) as functions of the Helmholtz
number He when the impeller tip speed is varied, for all three fan sizes and for four operating
300
conditions. The impeller tip speed is not explicitly shown on the abscissa but instead the
frequency that the various spectral components take as the fan speed is increased. This
frequency scale is different for the three fan sizes.
An important result in Figure 16.6 is that the curves of constant Strouhal number are
almost parallel to one another. This feature enables one to describe the filtered sound pressure in terms of a product of two nondimensional functions, one of which is a function of the
Strouhal number and the other is a function of the Helmholtz number, see equation (16.15).
The levels of the blade passage frequency and its harmonics do not increase monotonously
with the impeller tip speed but expose maxima and minima, which for all Strouhal numbers
occur at the same values of the Helmholtz number. For a constant fan size, this means
that the levels of all tonal components peak at the same frequency, however, at different fan
speeds. This behavior indicates the influence of the acoustic resonance and radiation characteristics of the impeller together with the casing and the duct system connected (system
frequency response function G in equation (16.15)); these effects will be discussed in more
detail in a later section.
The peaks of the blade passage frequency level of the largest fan (D3 = 560mm) at 88Hz
are caused by the anechoic duct termination, as was mentioned in the preceding section, and
must not be interpreted as a failure of the acoustic similarity laws.
At the point of optimum operation (<p = 0.09, see Figure 16.6b) the data for D2 = 280mm
do not agree very well with the data for the other two fan sizes and also, the curves of constant
Strouhal number for this fan are no longer parallel to one another. It was believed initially
that this effect was caused by a transition from laminar to turbulent boundary layer along
the blades as the impeller speed is raised. Attempts to destroy these phenomena by tripping
the flow at the impeller inlet were unsuccessful, and thus the reason for the different behavior
of the tone components of the fan size with D2 = 280mm remains unknown.
The deviation between the tone levels of the different fan sizes is larger than in case of
the overall sound pressure level, but again there is no clear trend that would indicate an
influence of the Reynolds number upon the fan noise. Hence, it seems justified to disregard
the Reynolds number in the acoustic similarity law also for the tonal components.
Random fan noise components The levels of a few random noise components as functions of the Helmholtz number are shown in Figures 16.7(ad). The presentation is similar
to that of the tonal components in Figure 16.6, the only difference being that the filtered
sound pressure is divided by the filter bandwidth, which, in turn is normalized by impeller
diameter D and  for convenience  by the sound speed ao instead of the tip speed U,
compare equation (16.9). Again the curves of constant Strouhal number are parallel to one
another which proves the validity of the product notation in equation (16.15). As in the case
of the overall sound pressure and the tonal components, an influence of the Reynolds number cannot be stated because the differences between the data for the three fan sizes do not
show any significant trend but are of random nature. Note that the acoustic resonance and
radiation characteristics of the system comprising impeller, casing and duct have a smaller
effect on the random noise than on the discrete components.
Acoustic system response function From equation (16.8) and (16.15) a relation for
the acoustic system response function for the discrete fan noise components can be derived,
which in logarithmic form reads
301
130
0) ljJ
= 0.052
1,\ .
120
110
   _ . _..
100
90
80
70
130
b)ljJ
= 0.09
120
110
100
90
80
1
0
l
..
10<J
01
120
0
N
100
90
80
70
130
120
110
90
0 1 = 140 mm
02 = 280 mm
03 = 560 mm
80
70
0.080.1
200 250
100 125
50 63
0.2
500
250
125
O.L.
1000
500
250
He
f D1
f D2
f D3

140 mm 
280 rnrn    
560 mm    
1.6
4000
2000
1000
3.2
8000
4000
2000
Figure 16.6: Levels of the blade passage frequency and its harmonics as functions of the
impeller tip speed, for four aerodynamic loading conditions (after Neise & Barsikow (1980,
1982) ).
302
130r,,,,~~
0) 'jJ = 0.052
120
1.5
110
100
r.
~
0
0
<J
l<ll
<J <l0
''
.E'
90
80
b)
'jJ
= 0.09
70
130
I'
120
I
I
,iSt = 0.5
0
N
110
1.5
100
5.5
90
80
c) tp = 0.130
70
130
0, = 140 mm
02 = 280 mm
03 = 560 mm
120
110
5.5
100
90
80
d)
70
'jJ
=0.166
0.080.1
200 250
100 125
50 63
0.2
500
250
125
0.4
1000
500
250
He
f D1
f D2
f D3
~
560 mm
.~
1.6
4000
2000
1000
3.2
8000
4000
2000
Figure 16.7: Levels of various random noise components as functions of the impeller tip
speed, for four aerodynamic loading conditions (after Neise & Barsikow (1980, 1982))
303
for if,
xd D
(16.16)
= const
Here Po = 20J.lPa is the usual acoustic reference pressure. For St = const, the last term
on the right hand side of equation (16.16) assumes a constant value, and one recognizes
that for each value of the Strouhal number, 20lg Gd(JIe) is given by the deviation of the
corresponding sound pressure level distribution 20lg(tlP!po) from a fictitious average (20
19 U 2 )dependence.
Figure 16.8 shows the acoustic system response function Gd(JIe) evaluated from the levels
St
1
2
<0
Ql
10
D3=560 mm
1\
'I
_2oLL~~~=~0~.0~9~0~~~~~~
0.08 0.1
0.2
0.4
0.8
He
1.6
3.2

Figure 16.8: Acoustic system response function of the tonal noise components at the fan
design point (if = 0.09, D3 = 560 mm) (after Neise & Barsikow (1980,1982)).
measured for D3 = 560mm (see Figure 16.6b). The vertical position of the curve St = 1 was
found by postulating
= 1,
(16.17)
i.e., the average of the frequency response function over its range of Helmholtz numbers
is to be zero, and by doing this, the spectral distribution function Fd ( St = 1) is assigned
a fixed value. The position of the curve St = 2 is found by shifting it vertically until
optimum agreement is obtained with the curve St = 1. The positions of the other curves
were determined accordingly.
A comparison of the acoustic system response functions of the tonal noise components
at four aerodynamic loading conditions is given in Figure 16.9. Each curve represents the
average over the Strouhal numbers St = 1 to 6 and over the three test fan sizes. There is a
distinct influence of the operating condition on the acoustic system response function, and
the reason for this behavior is as follows: First, like in the case of any system of cavities
with flow, the acoustic resonance characteristics of the impeller with the casing depend on
the flow velocity and the flow pattern, and second, the positions of the aerodynamic sound
sources in the fan are shifted as the operating condition is altered. In particular, the flow
around the cutoff region, which is the main source region for the tonal noise components, is
sensitive to changes in the aerodynamic loading of the fan.
The following relation for the acoustic frequency response function of the random noise
components was derived from equations (16.9) and (16.15)
304
20
dB
0
...uVi
10
<l')
Ol
'"
10
  \jl=0166 \jl=0.090
\jl= 0.130
\jl =0.052
20~~________~~______~________~________~~______~
0.08 0.1
0.2
04
0.8
He
16
3.2

Figure 16.9: Acoustic system response function of the tonal noise components, for four
aerodynamic loading conditions (after Neise & Barsikow (1980,1982)).
(16.18)
for tp, xd D = const
The procedure for determining GAHe) from the data presented in Figures 16.7 (ad) is
analogous to that of the discrete components just discussed, with a condition for the lowest
Strouhal number similar to equation (16.17):
< 20lgG'r(He)
>He=O.080.16=
0 for St = 0.5
(16.19)
Figure 16.10 shows that the acoustic system response function for the random components
20r,,,,~~
dB
10r+++~~~
o
Vi
'"
<l')
.Q1
1O~
________~________~_______~________~________~
0.08 0.1
0.2
0.4
0.8
1.6
3.2
20L~
He
Figure 16.10: Acoustic system response function of the random noise components, for four
aerodynamic loading conditions (after Neise & Barsikow (1980, 1982)).
also depends on the fan operating conditions. In contrast to the blade passage frequency
and its harmonics, the system response to the random noise components is much smoother;
this, however, is understandable because the tones are generated deterministically within
a concentrated region around the cutoff, while there is a large spatial distribution of the
random noise sources in which case maxima and minima in the frequency response of any
system are less pronounced.
305
Spectral distribution function Equations (16.8), (16.9), and (16.15) can also be used
to derive relationships for the determination of the spectral distribution functions Fd and Fr
for the discrete and random noise components:
(16.20)
for zp, x;j D = const
20 19 Fr(St) = 20 19( 6.pj Po)
+ 20 19(po/ POU2 )
 40
........0
dB
50
<LL"O
CJ'\
60
0
N
'"
70
3
St

Figure 16.11: Spectral distribution function for the discrete noise components, for four
aerodynamic loading conditions (after Neise & Barsikow (1980, 1982)).
50
1 dB
 60
0
'"
<LL~
.!21
70
0
N
'../
 80
0.5
1.5
2.5
St
3.5
4.5
5.5

Figure 16.12: Spectral distribution function of the random noise components, for four aerodynamic loading conditions (after Neise & Barsikow (1980, 1982)).
functions of both the discrete and the random noise components depend on the fan operating
306
condition (flow coefficient), as was to be expected. Each of the curves in Figures 16.11 and
16.12 was obtained by averaging the data for all three fan sizes. At optimum fan operation
(<p = 0.09), the spectral distribution function is minimum in both cases.
The acoustic similarity laws presented in equations (16.13) or (16.15) were used in recent work by Mongeau (1991), Mongeau, Thompson & McLaughlin (1993), and Bent &
McLaughlin (1993) to develop enhancement techniques for noise source measurements in
turbomachines. With these it is possible to remove the influence of purely acoustical sound
propagation or resonance effects, i.e., the effect of the acoustic system response function,
from measurement data to reveal the spectrum of the aeroacoustic source alone.
16.4
16.4.1
GENERAL REMARKS
One important precondition for the application of acoustic similarity laws to fan noise is
that the fan sizes in question are dimensionally similar. Strictly speaking, the duct system
connected to the model fan and the full size fan have to satisfy the condition of dimensional
similarity, too, because changes in the acoustic loading of a fan, i.e., the acoustic impedance
of the duct system, have in general an effect on the noise generated.
The experimental data presented in this paper and also earlier publications have shown
that the spectral distribution function F(St) and the acoustic system response function
G( He) are influenced by the fan operating condition, and the latter is also a function of
the measuring position xii D, see Bartenwerfer, Agnon & Gikadi (1976). Since these effects
cannot be accounted for by simple mathematical relationships, fan noise data can be scaled
only on the basis of constant aerodynamic loading condition (<p = const) and constant
measuring position (xii D = const).
According to equation (16.7), the isentropic exponent /'i, is one of the non dimensional
parameters involved in the acoustic similarity law. However, the actual influence of /'i, on the
fan noise was tested in neither the present nor any previous experiments. Thus, if a change
in the working fluid is involved, one cannot account for the influence of /'i, when scaling fan
noise data. Fortunately, the possible changes in the isentropic exponent are very small, and
up to now there is no indication whatsoever that /'i, has an influence on the noise, so that
disregarding the influence of /'i, seems uncritical.
16.4.2
Scaling fan noise data is' particularly easy when the working fluid, i.e., mean density and
sound speed of the medium, and the impeller tip speed for the two fan sizes considered are
the same (Po, a U
const). In that case, one has St = const and He = const in model and
full size fan for all frequencies, and according to equation (16.15) the spectral component
Pw is the same in the two cases. Hence, if <p, xii D, po, a U = const, the levels of the
tonal components are identical in the model fan and the full size version while the levels of
the random components have to be corrected for the influence of the impeller diameter, see
equation (16.9).
O)
O )
307
16.4.3
If sound pressure spectra are to be scaled for arbitrary conditions, knowledge of the acoustic
system response function G(He) and the spectral distribution function F(St) is required for
both discrete and random components. An experimental procedure for determining these
functions was described in previous sections.
From equation (16.15) one obtains for the filtered sound pressure levels of the discrete
and random components:
20 19(po/ POU
+ 20 19 Fd(St) + 20 19 Gd(He)
(16.22)
(16.23)
The explicit determination of F(St) and G(He) is not necessary, however, if the levels of
all spectral components of interest were measured as functions of the impeller tip speed and
are plotted as functions of the Helmholtz number, like the data presented in Figures 16.6
and 16.7. For the set of variables given, i.e., Po, a o, D and U, the Helmholtz number and
Strouhal number are determined for each frequency component of interest and their levels
are read from the model experiments.
Note that it is possible to extrapolate the measured data to higher impeller tip speeds.
For example, the blade passage frequency level distribution in FigurE; 16.6 (St = 1) can be
extended to up to twice the impeller tip speed measured by shifting the curve St = 2 in the
ordinate direction until optimum agreement with St = 1 is obtained.
It is strongly recommended, however, to cover the whole range of tip speeds in the model
experiments wherever possible because data extrapolation always bears the risk of higher
.
.
InaccuraCIes.
16.4.4
In some practical applications it is required to scale the total sound power of a fan without
knowing the frequency spectra. From equation (16.7) one obtains, upon omitting He and K,
and with Poo = POU2 ,
(16.24 )
00
St=1
St=1
(16.26)
308
(16.27)
Introducing equation (16.24) into equations (16.26) and (16.27) and performing the summation and integration, respectively, yields
(16.28)
(16.29)
where the factor Z/,rr is incorporated in the constant K;. The overall sound pressure is found
by adding the contributions of the tonal and random components,
(16.30)
When the fan operating condition and the measurement position are held constant (tp,
x;/D = const), the function K(Ma) is replaced by C1 Ma 2 t, and one obtains
(16.31)
Here E is the tip speed exponent in excess of 2 for the sound pressure (p ex: U +E). Equation
(16.25) can be rewritten by inserting equation (16.31), relating the measurement surface A
to the fan size, A = C2 D 2 , and by letting C = C1 C2 , to give the following expression for the
sound power:
2
(16.32)
Equation (16.32) can be expressed in logarithmic form for the sound power level Lw :
U
D
po
ao
Lw = Lws + 10(4 + 2E) 19 u, + 20lg D + 10lg   10(1 + 2E) 19  ,
0
P(Xl
a (Xl
o
(16.33)
where the various reference values are Uo = 1m/s, Do = 1m, and the data for air at 20C
temperature and 1bar barometric pressure: P(Xl = 1.19kg/m3 , a(Xl = 343m/s.
Equation (16.33) can be used to derive the required relation between the sound power
levels emitted by two dimensionally similar fans of different size, impeller tip speed, and
working fluid, for constant fan operating condition and measurement position (tp, x;/ D
const),
U2
D2
P02
a02
+ 20lg  + 10 19   10(1 + 2E) 19  .
(16.34)
U1
D1
POl
a01
According to the notation introduced in equation (16.31), the tip speed exponent for the
sound pressure is given by (2+E), i.e., P rv U2+E, while for the sound power P rv U 2 (2+c) = U',
b = 2(2 + E)). As was mentioned earlier, the tip speed exponent may vary with the
measurement position, e.g. may be different in the inlet and outlet duct.
Note that equation (16.34) is applicable only to the overall sound power level but not
the Aweighted level because the Aweighting curve is fixed in the frequency domain while
the fan sound spectra are shifted with the impeller diameter and tip speed. Hence, if the
Aweighted sound power level of a fan is to be scaled, one first has to scale the unweighted
sound spectra, then to apply the Aweighting and finally to compute the Aweighted sound
power level from the Aweighted spectrum.
LW2 
LWI
= 10(4 + 2E) 19 
309
16.5
CONCLUSIONS
Acoustic similarity laws are discussed to scale the sound power spectra of fans. The normalized filtered acoustic pressure is expressed as a product of two spectral functions of nondimensional parameters: The socalled spectral distribution function depends on the Strouhal
number and describes the aerodynamically generated sound spectrum. The second spectral
term (acoustic system response function) is the mathematical representation of the acoustic
resonance and radiation characteristics of the fan together with the duct system connected.
This acoustic system response is superimposed on the aerodynamically generated sound and
thereby influences the spectral content of the fan noise.
Experimental investigations are described to verify the acoustic similarity laws. The
results presented by Bartenwerfer & Agnon (1976), Neise & Barsikow (1980), and N eise
& Barsikow (1982) prove that the influence of the Reynolds number on the radiated noise
spectrum is negligible in the range 1.0 X 10 5 ::; Be ::; 2.2 x 10 6 ; this was found true for the
overall sound pressure as well as for the discrete and random noise components. From that
one concludes that the influence of the Reynolds number can be neglected for all practical
purposes.
The acoustic system response is a function of the Helmholtz number He, the measuring
position xii D and the flow coefficient !.p. The spectral distribution function depends on the
Strouhal number St and the flow coefficient !.p.
Based on the above experimental evidence, it is possible to scale fan noise data measured
on a model fan onto other fan sizes, fan speeds and working fluids, provided the fans in
question together with their duct systems are dimensionally similar and the fan operating
condition and the measurement position are the same.
16.6
NOMENCLATURE
f
6f
F,F
G,G
A
b
G, Gl , G2
d
D
He = D/A
f{, f{', f{
Lp
Lw
Ma = U/ao
n
p
Pw
Pwd
PWT
Re = U D j v
Reynolds number
Si
u
flow velocity
mj s)
Z
Greek symbols:
(3
,=
E+2
TJt
kinematic viscosity
po
'P
4Q j 'if D2 U
'ljJ = 26pt/ Po U
flow coefficient
2
pressure coefficient
Subscripts:
discrete component
random component
Miscellaneous:
q
rms val ue of q
< q >a
q averaged over a
19q
16.7
BIBLIOGRAPHY OF CHAPTER 16
BARTENWERFER, M. & AGNON, R., 1976. Der EinfluB der Zahigkeit des geforderten
Mediums auf die Schallerzeugung von Radialventilatoren. Forschungsbericht DLRFB
7630, Deutsche Forschungsanstalt fiir Luft und Raumfahrt, KolnjPorz, Germany.
BARTENWERFER, M., AGNON, R. & GIKADI, T., 1976. Eine experimentelle Untersuchung
zur Schallerzeugung und zur Schallausbreitung in Radialventilatoren. Forschungsbericht
DLItFB 7614, Deutsche Forschungsanstalt fiir Luft und Raumfahrt KolnjPorz, Germany.
BARTENWERFER, M., 1977. Letter to the editor. Noise Control Engineering 8,3.
BENT, P. H. & MCLAUGHLIN, D. K., 1993. Enhancements to noise source measurement
techniques for turbomachinery. AIAAPaper 934373, American Institute of Aeronautics
and Astronautics, Washington DC, USA. 15th AIAA Aeroacoustics Conference, Long
Beach, California, USA.
CHANAUD, R. C., 1965. Aerodynamic sound from centrifugal fan rotors. Journal of the
Acoustical Society of America 37, 969974.
GOLDSTEIN, M., 1976.
York.
ISO 5136,1990. Acoustics  Determination of sound power radiated into a duct by fans  Induct method. International Standard, International Organisation for Standardization.
MADISON, R., 1949. Fan Engineering (Handbook)) 5th Edition. Buffalo Forge Company,
Buffalo, New York.
MALING, G., 1963. Dimensional analysis of blower noise. Journal of the Acoustical Society
of America 35, 15561564.
MONGEAU, L., THOMPSON, D. & McLAUGHLIN, D., 1993. Sound generation by rotating
stall in centrifugal turbomachines. Journal of Sound and Vibration 163, 130.
MONGEAU, L., 1991. Experimental Study of the Sound Generation by Rotating Stall in
Centrifugal Turbomachines. PhD thesis, Penn State University, University Park, Pennsylvania, USA.
NEISE, W. & BARSIKOW, B., 1980. Akustische Ahnlichkeitsgesetze bei Radialventilatoren.
Forschungsbericht DFVLRFB 8036, Deutsche Forschungs und Versuchsanstalt fiir
Luft und Raumfahrt, KolnjPorz, Germany.
NEISE, W. & BARSIKOW, B., 1982. Acoustic similarity laws for fans. ASMETransactions)
Journal of Engineering for Industry 104, 162168.
NEISE, W., 1975. Application of similarity laws to the blade passage sound of centrifugal
fans. Journal of Sound and Vibration 43,6175.
312
J., 1971. Beitrag zm Analyse der Beziehungen zwischen den akustischen und
stromungstechnischen Parametern am Beispiel geometrisch ahnlicher RadialventilatorLaufrader. Forschungsbericht DLRFB 7112, Deutsche Forschungsanstalt fur Luft und
Raumfahrt, KolnjPorz, Germany.
WEIDEMANN,
B., 1964. Beitrag zm zweckmaBigen Bestimmung und Darstellung des Ventilatorgerausches als Grundlage fur die akustische Berechnung von Luftungsanlagen.
Doctoral dissertation, Technische Universitat Berlin, Germany.
WIKSTROM,
313
314
Chapter 17
17.1
INTRODUCTION
In the previous chapters the physical generation mechanisms of the aerodynamic noise of
turbomachines were discussed, and methods were outlined for calculating the sound field
315
radiated from these machines. This discussion has revealed that a very detailed knowledge
of the construction of the turbomachine and the surrounding flow field is necessary to be
able and predict the sound power and the noise spectrum.
In many practical applications, however, it is required for the design engineer to have
some initial knowledge about the sound power and the spectrum generated by a fan, based
on the aerodynamic duty, without specifically knowing what fan will be used. In this chapter,
various formulae will be presented for estimating the sound power level and the spectrum
of different fan types as used in heating, airconditioning, and ventilating duct systems or
for other industrial purposes. All of these equations are applicable only to the aerodynamic
noise of fans, other possible contributors are not accounted for.
The presentation below is by no means meant to be complete; only the most well known
prediction schemes are discussed.
It is useful for the purpose of this chapter to recall equation (16.5) for the sound power
P of an aerodynamic sound source as expressed by Goldstein (1976) (compare chapter 16
on acoustic similarity laws for fans),
(17.1)
where Land u mean a characteristic length and flow velocity of the aerodynamic source, and
the medium is characterized by the density po and the speed of sound ao. The exponent m
of the flow Mach number Ma = u/ ao depends on the source type. Under acoustic freefield
conditions, m = 1 for acoustic monopoles, m = 3 for dipoles, and m = 5 for quadrupoles. In
case of fans it is appropriate to choose impeller diameter D and tip speed U for the typical
length and flow velocity:
(17.2)
Upon considering the aerodynamic fan scaling laws (volume flow Q ex U D2 and fan total
pressure flpt ex POU2 , one finds the following proportionality for the fan aerodynamic power:
(17.3)
Comparing equations (17.2) and (17.3) shows that the fan sound power is proportional
to the fan aerodynamic power multiplied by a Mach number term which accounts for the
fact that the acoustic radiation efficiency of an aerodynamic source increases with flow Mach
number.
17.2
HISTORICAL BACKGROUND
17.2.1
Madison (1949) was the first to formulate a sound law for fans, based on the following
considerations: "As the fan blades intercept filaments of air there is a pressure built up on
one side of the blade and a rarefaction on the opposite side. These pressures are proportional
to the pressure developed by the fan since they are motivated by the same means and a change
in speed that affects one also affects the other. Since the intensity of sound varies as the
square of the effective sound pressure in the wave it is reasonable that it would also vary as
the square of the fan pressures developed. For a fan of given pressure the intensity varies as
316
the volume of air flowing through the fan inlet or outlet. This is because the intensity varies
with the area of the flow and area and capacity are proportional for a given velocity pressure.
On the above basis noise intensity for a given size fan will vary as capacity x (pressure)2 or
as the 5th power of speed.)) In the present notation Madison)s fan sound law reads
(17.4)
where P is the total fan sound power, Q the volume flow, and 6.ps the fan static pressure.
Madison (1949) emphasized that "the sound laws, like the fan laws, apply only at a particular
point of rating and cannot be used when moving along the curve from one rating to another".
Madison also introduced the concept of the specific sound power level by expressing equation
(17.4) in logarithmic form to obtain for the total fan sound power level Lw, again in the
present notation:
Lw = LWsM
6.ps
(17.5)
Here Lw sM is the specific sound power level which depends on the fan type and the operating
condition, and Qo and 6.pso are reference values, e.g., Qo = Im 3 /s, and 6.pso = IPa. The
specific sound power level can be seen as a normalized fan sound power level or, in other
words, as a sound power level for a unit fan aerodynamic duty. By using the aerodynamic
fan scaling laws, equation (17.4) can be rearranged to give
(17.6)
This relationship indicates what was quoted earlier from Madison, i.e., that a fixed tip speed
exponent I = 5 is assumed for the fan sound power. Comparing equations (17.2) and
(17.6) reveals that Madison's fan sound law is incorrect from a dimensional point of view,
nevertheless it formed the basis for the most frequently used formula to estimate the sound
power level emitted by a fan.
Madison (1949) did not attempt to predict the spectral distribution of the fan sound
power. This was done later by, among others, Beranek (1960), Bommes (1961), and Finkelstein (1972) by specifying frequency spectra relative to the total sound power level. Wikstrom
(1964) suggested to use a nondimensional frequency scale, the Strouhal number JU / D for
the relative spectra. All of the above researchers, among others not mentioned here, presented data for the specific sound power level based on measurements on different fan types.
17.2.2
Groff, Schreiner & Bullock (1967) presented a fairly sophisticated prediction formula for the
octaveband sound power level of centrifugal fans in the frequency range 125 to 8000Hz:
Lw
oct
= B
+ 10lg(Q 6.p~) + T
(17.7)
The term B has the character of a specific sound power level, but in this prediction scheme
it is a function of the octave band center frequency and the fan efficiency. Groff, Schreiner
& Bullock (1967) presented data for centrifugal fans with backward curved blades and centrifugal fans with forward curved blades. The exponent E is also frequency dependent, and
a nomogram was provided to facilitate calculation of the second term on the right hand side
of equation (17.7) as a function of frequency. T = 5dB is an increment to be added to the
octave band containing the blade passage frequency.
317
17.2.3
Graham (1972) employed Madison's equation (17.5) to estimate fan noise levels, but presented octave band data for the specific sound power level in the frequency range from 63
to 8000Hz, independent of the fan size. This was done for centrifugal fans with backward
curved and forward curved blades, for halfaxial fans, axial fans with outlet guide vanes,
vaneless axial fans, and propeller fans (without casing). In addition, different increments
were specified for all fan types for the octave band containing the blade passing frequency.
17.2.4
Regenscheit (see the book by Eck (1972)) related the fan sound power to the aerodynamic
power loss of the fan multiplied by a Mach number term similar to that in equation (17.1)
and obtained the following, dimensionally correct relationship:
1
U m
2 3( U m
P ex Q . 6.Pt (   1) (  ) ex PoD U )
'TIt
ao
ao
'TIt is the total fan efficiency. In logarithmic form equation (17.8) reads
(17.8)
Q
6.Pt
1
U m
Lw=LWsR+I0lgQ +10Ig~+10Ig[(1)() J
(17.9)
'TIt
ao
The reference pressure is the same as before, 6..pto = IPa. Regenscheit assumed fixed values
for the Mach number exponent: m = 2 for centrifugal fans and m = 2.5 for axialflow fans;
with the fan scaling laws one obtains tip speed exponents of I = 5 and I = 5.5, respectively,
for the two fan types. The spectral distribution of the noise of axial fans and centrifugal
fans was described by Regenscheit in terms of frequency spectra relative to the total sound
power.
o
LJ.PtO
17.3
17.3.1
In VDI 3731 Blatt 2 (1990) a prediction method for the fan sound power is presented which
is based on recent experimental data. When collecting the data, it was initially planned
to include only measurement data that were obtained under standardized measurement
conditions according to DIN 45635 Teil38 (1986), i.e., the induct method ISO 5136 (1990),
free field method ISO 3744 (1981), or reverberation room method ISO 3741 (1975), ISO
3742 (1975), ISO 37431 (1994), ISO 37432 (1994) (compare chapter 15). It turned out,
however, that only very few measurement data were made available under such premises, and
so it was decided to also accept data that were obtained by using measurement procedures
common to insitu conditions. Fan sound power data were supplied by various German
fan manufacturers, universities and research institutes for four fan categories: centrifugal
fans with backward curved blades, centrifugal fans with forward curved blades (scirocco
blowers), axial fans with outlet guide vanes, and vaneless axial fans. The measurements
were not necessarily made at the exact best efficiency, but close to it.
318
The data base for the four fan types was used to examine which of the fan sound power
prediction schemes available was best suited to "predict" the measurement data. Among the
prediction methods tested were the ones put forward by Madison (1949) (equation (17.5),
however based on the total fan pressure 6Pt instead of the static fan pressure 6ps), by
Groff, Schreiner & Bullock (1967), by Graham (1972), and by Regenscheit (equation (17.9)).
The result of this study was that Regenscheit's equation provided the best collapse of the
experimental data, when a variable Mach number exponent m was introduced. Thus the
following expression is the basis for the sound power prediction according to VDr 3731 Blatt
2 (1990):
Q 6Pt 1
Lw  10lg[
  1)] = LWspec
Qo 6ptO 'T/t
U
+ 10m 19( )
ao
(17.10)
This relationship is used for the linear as well as the Aweighted sound power level. The
expression on the left hand side of equation (17.10) can be seen as the level of the fan sound
power normalized by the aerodynamic flow losses (reduced sound power level). The specific
sound power level Lw spec and the Mach number exponent m were determined by a linear
regression analysis from the experimental data for the sound power levels Lw and Lw A
together with the flow rate Q, fan total pressure 6Pt and efficiency 'T/t, separately for the
four fan types mentioned above.
As an example, in Figure 17.1 is plotted the "reduced" linear sound power level of cen90~~
..,
;:::.
I
I
r:::: 80

0::10:
<l
<l
60
005
0,2
0,1
0,3
U/o o
Figure 17.1: Reduced linear sound power level in the outlet ducts of centrifugal fans with
backward curved blades as a function of the impeller tip speed Mach number; after VDr
3731 Blatt 2 (1990).
trifugal fans with backward curved blades as a function of the Mach number based on the
impeller tip speed. The data are for the fan outlet duct. The solid line represents the mean
which is described by the relation
Lw
10lg[~ 6Pt (~
Qo 6pto
1)]
17t
85.2dB
+ 15.3lg( U),
ao
(17.11)
i.e., LWspec = 85.2dB and m = 1.53. The dashed lines indicate the range described by the
standard deviation of the individual data from the mean (for this particular fan type, data
for 64 members were available; the standard deviation is s = 4.1dB). The corresponding
319
data for the Aweighted sound power level of this fan category are LWspecA = 85.5dB(A),
= 2.77, and SA = 3.4dB. Corresponding information is given in VDI 3731 Blatt 2 (1990)
for the other three fan types.
The usefulness of including the fan efficiency in the fan sound power prediction formula, or
in other word the usefulness of relating the sound power to the fan losses, is often questioned
with the argument that only a tiny fraction of the losses is converted to sound energy. There
is certainly no conclusive answer to this question but the analysis of the experimental has
undoubtedly shown that the scatter of the individual data around the mean is substantially
larger than shown in Figure 17.1 when the efficiency term in equation (17.10) is omitted.
The standard deviation of the data for all four fan types covered by VDr 3731 Blatt 2
(1990) is in the range s = 2.7 to 4.1dB for the linear sound power level and s
3 to 4.7dB
for the Aweighted sound power level. This indicates the accuracy one has to expect when
using the prediction method of VDI 3731 Blatt 2 (1990).
mA
17.3.2
The spectral distribution of the fan sound power is expressed in VDI 3731 Blatt 2 (1990) by
means of an octave spectrum relative to the linear sound power level. The general form of
this relative spectrum is given by the following equation:
(17.12)
Here too, the constants G1 , G2 , G3 were determined from experimental data. In addition to
that, the level of the octave band containing the blade passing frequency is subjected to an
increment between 0 and 8dB. As an example, the relative spectrum of centrifugal fans with
backward curved blades is shown in Figure 17.2. The data are for fans within the range of the
Or~==~~~~
.';'I.,~=t:i~:!~:e~::le:l
ri  1~fII::J.=:':J:'
dB
.:'S ::I~I~
IJ ::::Il..~:...... I.
.~;"""L~I
.......... 
...
~,.
60 0 ,1
0,2
0,4
0,8
1,6
3,15
6,3
12,5 25
50
100 200
fO/U
Figure 17.2: Relative octave band spectrum of centrifugal fans with backward curved blades;
range of fan specific speed (J' = 0.21 to 0.56; after VDI 3731 Blatt 2 (1990).
fan specific speed (J' = cpO.5 NJtO.75 = 0.21 to 0.56 (flow coefficient cp = 4Q/7fD 2 U; pressure
coefficient 'ljJt = 2 . 6pt/ POU 2 ). The three curves in Figure 17.2 indicate the mean of the
experimental data and the standard deviation, similarly to the presentation in Figure 17.l.
The mean is described by equation (17.12) with the following constants: G1 = G2 = 5dB
and C3 = 0.39, and the standard deviation is s = 2.9dB. For this class of fans, the blade
320
passage frequency increment is specified as 6.L BPF = 0 ... 6dB. The choice of the actual value
of the increment for a given application is left to the experience of the fan engineer.
17.3.3
The data in Figures 17.1 and 17.2 apply to the induct fan sound power level on the outlet
side. The corresponding levels on the inlet side are assumed to be approximately equal to
the outlet side.
17.3.4
The sound power radiated by a fan into the free field is generally lower than that radiated
into the duct; this is due to the refiection of the sound waves at t1:..e inlet and outlet apertures
of the fan casing in the plane wave frequency range, compare chapters 14 and 15. This effect
is accounted for in VDI 3731 Blatt 2 (1990) by the data shown in Figure 17.3, which are
20r~
dB
Inlet side
10
..
..
... .......
I_!
~ .~
o ~~~~.~.~~_.~~.~~.~
"
Cl
I
~'10
..,
~20LLL~~liL~~~~L~L~~~~
"i' 2 0 ,,
OJ
Outlet side
.f dB
J'10
.I
dI Ji' .
In
~~~~~1~_1
10 _20L~LJJ~liL~LJ~~L~LJJ~~
0,1
0.2
0.4
kRh = 2rrfRh
10
20
40
100
/0 0
Figure 17.3: Difference between freefield fan sound power levels and induct levels; after
VDI 3731 Blatt 2 (1990).
based on experimental results by Holste & Neise (1992a). There, the difference between
free field levels and induct level Lw, freefield  L w, induct is plotted as a function of the
dimensionless wave number kR h, where k = 21f f / ao, and Rh = 2A/ C is the hydraulic radius
of the fan aperture (A = cross sectional area, C = circumference). The average curve in the
low frequency range is approximated by
L w,
freefield 
2.3(kRh)2
L w, induct = 10 1g 1 + 2.3(kR )2
h
(17.13)
Similar results are presented in VDI 3731 Blatt 2 (1990) for centrifugal fans with backward
curved blades in the specific speed range (J' = 0.07 to 0.20, for scirocco blowers, for axial fans
321
17.4
The present description of this method very closely follows that by Graham (1992). The
method published by ASHRAE (1991) is basically the same as the one proposed by Graham
(1972). Madison's fan sound law (equation (17.5)) is modified in that the fan static pressure
is replaced by the total fan pressure. In the present notation one has
Lw = LWspec
Q
6Pt
+ 10 19( Q
) + 20 19( ~),
o
UPtk
(17.14)
Note that the reference pressure chosen in the ASHRAEmethod is 6Ptk = lkPa. Graham
(1992) characterized the specific sound power level as "the noise a fan would make if it were
operating at a flow rate of one m 3 /s and a total pressure of one kPa." In the ASHRAEhandbook specific sound power levels in octave bands in the frequency range from 63 to
8000Hz are listed for various fan types and fan sizes, see the data reproduced in Table
17.l. In addition, an increment (BFI) is specified for the octave band containing the blade
passage frequency. All data are applicable to fans operating at or near their best efficiency
point. According to Graham (1992), the specific sound power levels given in the ASHRAEhandbook do not represent the average over a certain number of fan noise data but were
selected on the basis of the following consideration: "In most cases, fans of good design were
used so the resulting levels found in the table represent fans that are better than what the
average would be if a simple numerical average were obtained". The values given in the
table are total sound power levels, i.e., the level of the sum of inlet and outlet sound power.
If the power levels of either inlet or outlet are of interest, the levels listed in Table 1 have to
be reduced by 3dB, except for the blade tone increment.
17.5
Wright (1982) argued that it is the maximum flow velocity in a blade row that determines
the sound power output of a fan and, therefore, suggested to relate the sound power level of
subsonic axial fans to the maximum flow velocity Vp ("peak velocity") on the blade surface
in the rotating frame of reference:
Lw =
J{l
+ J{2lg Vp.
(17.15)
[VI +
tp2
+ ~ 7r ZD]
41]t
(17.16)
Here c is the blade chord length at the outer impeller circumference, and Z the impeller
blade number. The factor J{ ::::: 1 is a socalled peak velocity or spiking parameter. For
1 is assumed.
welldesigned fans which are operated at or near the best efficiency point, J{
322
63
(Hz)
125
(Hz)
250
(Hz)
85
90
85
90
84
88
79
84
75
79
68
73
64
69
62
64
3
3
>1.0
<1.0
101
112
92
104
88
98
84
88
82
87
77
84
74
79
71
83
7
7
Medium press.
(2.5  5kPa)
>1.0
<1.0
103
113
99
108
90
96
87
93
83
91
78
86
74
82
71
79
8
8
High pressure
(5  15kPa)
>1.0
<1.0
106
116
103
112
98
104
93
99
91
99
89
97
86
94
83
91
Forward curved
All
98
98
88
81
81
76
71
66
Axial Fans
Vaneaxial
Dd D =0.30.4
Di/ D =0.40.6
Dd D =0.60.8
All
All
All
94
94
98
88
88
97
88
91
96
93
88
96
92
86
94
90
81
92
83
75
88
79
73
85
6
6
6
Tubeaxial
>1.0
<1.0
96
93
91
92
92
94
94
98
92
97
91
96
84
88
82
85
7
7
Propeller
All
93
96
103
101
100
97
91
87
Fan Type
(m)
Centrifugal Fans
Airfoil or
>0.75
backward curved <0.75
Radial fans
Low pressure
(1  2.5kPa)
Table 17.1: The Relative Sound Power Produced by Various Types of Fans. Specific Sound
Power Levels in dB for Octave Bands (after ASHRAE (1991)).
For comparison, the onset velocity of the relative flow entering the blade row is governed
by the following expression:
(17.17)
The constants J{l and J{2 in equation (17.15) were determined by experimental data for
36 axial fans of different designs, including house hold fans, model compressor rotors with
outlet guide vanes, and aircraft propulsion fan models. Figure 17.4 shows the sound power
levels of these fans as functions of the calculated peak velocity Vp. In all cases J{ = 1 was
assumed in equation (17.16). The constants J{l and J{2 are different for the three different
fan types included in the test, as is obvious from the three groups of data. The symbol
"1 RO" stands for fans with inlet guide vanes, rotor and outlet guide vanes, "RO" for fans
with outlet guide vanes, and "R" for vaneless fans.
To compare his own method with other ways of correlating axial fan noise data, Wright
(1982) used the same set of fan noise data to determine the constants J{i for the following
323
140
IRO
v
v
v:;~
120
RO
R
6.
OJ
""0,
3::
100
.J
6.
80
0
0
60 20
30 40 50
100
Vp. m/s
200
300 400
Figure 17.4: Axial fan sound power level as function of the peak velocity (after Wright
(1982)).)
other prediction formulas: (1) Lw = K3 + K4lg Vi; (2) Lw = Ks + K6lg U; (3) Lw =
K7 + 10 19( Q~pn; (4) Lw = Kg + 10 19 ~Pt; (5) Lw = Kg + 10 19 Paero . The result of this
exercise is shown in Figure 17.5. Clearly the peak velocity parameter provides the best
collapse of the data, followed by the correlation based on the onset velocity Vi instead of the
peak velocity v;,.
While Wright's fan noise correlation scheme seems to work very well for well designed fans
operating near the best efficiency point, yet more evidence is needed to prove the suitability
of his method for fans of lower aerodynamic quality or offdesign operation. The crucial
point here is how to determine or estimate the spiking parameter K in equation (17.16).
Wright presented an empirical expression for K as function of the relative flow incidence
angle with respect to the blade row. For blades operating well away from their design angle,
K may approach values near 2.5 which results in drastic changes of the sound power level,
depending on the constant Kl in equation (17.15).
The peak velocity method put forward by Wright (1982) seems to provide a better collapse of experimental fan noise data than the other prediction schemes discussed before, it
does however require more extensive information about the geometric details and operating
characteristics of the fan in question. No adaption of Wright's concept to other fan types
are known.
17.6
CONCLUSIONS
Various methods for predicting or estimating the sound power level and spectra of ventilating
and industrial fans are discussed. These methods are necessary for practical purposes because
the theoretical determination of fan noise data based on the aeroacoustic noise generation
324
~
!:::!
....
ro
12dB
6dB"J
'0
140
140
/"{
120
120
100
100
BO
80
~
::E
60~__~~__~__._
"0
..J
a..
"0
..
::l
60
80
Calculated
100
120
60~~__~__ .__~_
60
140
PWL, dB re 10_12 W
100 120
140
60
Velocity, Vp.
~
!:::!
80
on Onset
80
100
120
140
Velaclty, VI'
IIdB
140
16 dB
140
140
120
120
100
100
100
80
80
80
o
~ 120
00
ro
"0
.1
~
a..
.......
;: 6 0 L..!..._.._,.,,_,.60 80
100 120 140
:::E
6 0 '''.,,....,,....._
60
80
100
120
140
Squared, Cc.pt
60
80
100
120
140
Ca leu Ia Ie d P WL , d B rei 0  12 W
T
(0 Based on Power, HP
Figure 17.5: Comparison of calculated and measured fan sound power levels for six correlation schemes (after Wright (1982)).
mechanisms requires very detailed input data regarding the fan geometry and surrounding
flow field which is generally not available. The prediction schemes published in VDr 3731
Blatt 2 (1990) and ASHRAE (1991) are based on simple expressions for the fan sound power
level as a function of volume flow, total pressure head, efficiency, and impeller tip speed. The
free constants in these expressions are determined by comparison with measured fan noise
data. The spectral distribution of the sound energy is described by means of octave band
spectra relative to the total sound power level. The accuracy of these methods is largely
dependent on the quality of the data base used to determine the free constants.
17.7
BIBLIOGRAPHY OF CHAPTER 17
BOMMES, L., 1961. Geriiuschentwicklung bei Ventilatoren kleiner und mittlerer Umfangsgeschwindigkeit. Liirmbekiimpfung 5, 6975.
DIN 45 635 TElL 38, 1986.
Geriiuschmessung an Maschinen. Luftschallemission,
Hiillfliichen, Hallraum und Kanal Verfahren; Ventilatoren. Deutsche Norm, Deutsches
Institut fiir Normung e.V., Berlin, Germany.
ECK, B., 1972. Ventilatoren, 5th ed. Springer Verlag, Berlin, Germany.
FINKELSTEIN, W., 1972. Vorausberechnung des Ventilatorgeriiusches.
Haustechnik 23, 211215.
Heizung Liiftung
1964. Beitrag zur zweckmiifiigen Bestimmung und Darstellung des Ventilatorgeriiusches als Grundlage fur die akustische Berechnung von Luftungsanlagen.
Doctoral dissertation, Technische Universitiit Berlin, Germany.
WIKSTROM, B.,
T., 1982. A velocity parameter for the correlation of axial fan noise. Noise Control
Engineering 19, 1725.
WRIGHT,
327
328
Chapter 18
18.1
341
INTRODUCTION
This chapter is concerned with fan selection from a noise point of view. The main criterion
of fan selection is the aerodynamic duty described by volume flow Q and total pressure rise
6pt. In many practical situations a given amount of air has to be pumped through a given
duct system, and then Q and 6pt are fixed. In this case there is generally only a limited
choice of fan types that are suitable for the given task. However, there are also technical
problems which can be solved with different combinations of Q and 6pt, and in these cases
fans of different designs can be employed. For example, a heat exchanger for a given heat
transfer rate can be designed to have a large flow area and a small depth or, vice versa, a
329
small flow area and a large depth. In the first case, the cooling fan would have to deliver a
large flow rate at a low flow resistance, and in the second case a low flow rate against a high
back pressure.
A direct fan noise comparison was attempted by Sigel, Schilling & Zierep (1980) and Sigel
& Felsch (1981), who compared the sound power of six fans for a constant volume flow and
for almost the same fan pressure. The drawback of this procedure is that the aerodynamic
duties of the test fans are not exactly equal, that not all fans can be operated at their best
efficiency point, and that the comparison can be made for only one operating condition.
In this chapter the noise characteristics of various fan types are compared using the
concept of the specific sound power, i.e., the fan sound power normalized by its aerodynamic
duty, see Neise (1988b) and Neise (1989). In this way it is possible to compare different fan
types over the whole range of their aerodynamic performance curve.
18.2
The sound power P generated by an aerodynamic sound source can be described using the
following expression, see e.g. Goldstein (1976) (compare chapters 16 and 17):
(18.1)
where Land u mean a characteristic length and flow velocity of the source, and po and
ao are density and speed of sound of the fluid. The exponent m of the flow Mach number
M a = u/ ao depends on the source type and on the type of sound propagation, compare
chapter 17. For acoustic freefield conditions, m = I, 2, and 3 for monopoles, dipoles, and
quadru poles.
In case of fans it is useful to choose impeller diameter D and tip speed U for the typical
length and flow velocity. According to equation (18.1), the fan sound power is proportional
to the aerodynamic power multiplied by a Mach number term which accounts for the fact
that the acoustic radiation efficiency of an aerodynamic source increases with flow Mach
number.
The first attempt to formulate a fan sound law for practical purposes was made by
Madison (1949) who related the sound power emitted by the fan to the volume flow and
the square of the fan pressure rise: P ex Q6.p; ex P6D2 US (compare chapters 16 and 17).
Although this expression is incorrect from a dimensional point of view, compare equation
(18.1), it is the basis for the most frequently used formula to predict the sound power level
Lw of a fan:
Lw = LWsM
6.Pt
+ 101g QQo + 201g ~
u..pw
(18.2)
Here LWsM is the specific sound power level which depends on the fan type and operating
condition, and Qo
Im 3 /s und 6.pw
IPa. A fixed tip speed exponent I
5 is assumed
in equation (18.2) for all fan types.
Regenscheit (see the book by Eck (1972)) related the fan sound power to the aerodynamic
power loss of the fan multiplied by a Mach number term similar to that in equation (18.1)
and obtained the following relation for predicting the sound power level of a fan:
330
Lw
LWsR
+ 10lg!:L + 10lg
Qo
llPt
llptO
+ 101g[( 1
'TJt
1)( U )m]
ao
(18.3)
The specific sound power level in the definition of equation (18.2) or (18.3) can be seen as
a normalized fan sound power level or, in other words, as a sound power level for a unit fan
aerodynamic duty. Hence, the specific sound power level is used here to compare the noise
of different fan types. The two different formulations according to Madison and Regenscheit
are used to ensure that the ranking of the various fans obtained is not a mere result of the
way the specific sound power is defined. In the following, equation (18.2) and equation (18.3)
with the efficiency term omitted are applied to the total, Aweighted and onethird octave
band sound power levels.
18.3
EXPERIMENTAL DATA
18.3.1
The experimental data for the noise comparison are from Neise, Hoppe & Herrmann (1986),
Neise, Hoppe & Herrmann (1987), and Hoppe & Neise (1987). Ten fans were measured
using the standardized inductmethod ISO 5136 (1990). Anecoically terminated test ducts
were connected to both inlet and outlet of each fan. The impeller diameters of the test
fans ranged from 450 to 600mm, and their tip speeds from 18 to 92m/s. Figure 18.1 shows
the performance characteristics of the test fans in nondimensional form,i.e., in terms of
the pressure coefficient 'IjJ = 2llPt/ POU 2 as function of the flow coefficient i.p = 4Q /,1[' D 2 U.
The points of optimum operation are marked by filled or circled symbols. The performance
curve of the scirocco blower is only shown in part because it is almost flat over the range
i.p = 0.650.95. More details about the test arrangements and test fans are given in the
original references and by Neise (1988, 1989).
18.3.2
Figure 18.2 shows the specific sound power levels of the various fans as functions of the
flow coefficient. The upper curves are based on equation (18.2) after Madison and the lower
ones on equation (18.3) after Regenscheit. The total sound power is the sum of inlet and
outlet duct power over the frequency range 5010000Hz. The points of optimum operation
are marked as in Figure 18.1. With the exception of the axial fan without guide vanes, all
fans show the well known Ushape distribution of the specific sound power level as function
of the flow coefficient with their minima close to optimum efficiency operation.
If only the maximum efficiency points are considered, the centrifugal fans with backward
curved blades are quietest, and the centrifugal fan with radial blades is noisiest with a
specific sound power level of 17dB above the quietest. While the ranking of these two fan
types is independent of the specific sound power level definition, this is not quite so in case
of the other fans. In Madison's definition (lower group of curves), the scirocco blower is a
331
3
__ 0
+ ,;y
0_<27W
Q)
or'
U
_M
44
ill
0
U
ill
'
::J
(J)
(J)
ill
'
0..
.J(*'";+'"
+ "!&,
..."",
~"
"
.3
.~
Flow coefficient
n
1/min
++Axial fan,
OGV, airfoil blades 3000
++Axial fan,
OGV, airfoil blades 3000
*"""""*Axial fan, no GV, airfoil blades 2925
OGV, airfoil blades 2943
*"   *Ax ial fan,
G<>Centr. fan, b.c. airfoil blades 1820
'i''i'Centr. fan, b.c. airfoil blades 1600
6t.Centr. fan, b.c. blades (f lat)
1600
~Centr. fan,
flat radial blades 1800
0<:>Scirrocco blower (f. c. blades)
700
G8Mixedflow fan (halfradial)
1000
DO
.1
.2
.5
0
m
0.45
0.45
0.60
0.50
0.60
0.51
0.51
0.51
0.50
0.45
.6
U
Z m
m/s
70.7 24 2.5
70.7 12 2.5
91.9 8 2.5
77 .0 12 2.5
57.2 12 2.0
42.7 12 2.0
42.7 12 2.0
48.0 15 2.0
18.3 38 2.0
23.5 8 2.0
impeller
little quieter than the halfradial fan and all the axialflow fans, but compared on the basis
of Regenscheits definition it ranks behind the halfradial and one of the axial fans. The
reason for this behaviour is that in Regenscheit's definition volume flow and pressure rise are
assigned equal influence on the specific sound power level, while in Madison's definition the
pressure rise is given a much larger weight. Therefore the large pressure rise of the scirocco
blower helps make its specific sound power level according to Madison particularly small.
Over a fairly broad range of their performance curves, the axialflow fans have, on average,
higher specific sound power levels than the centrifugal fans with backward curved blades.
This is particularly true at the respective points of optimum operation, and if the quietest
ones in each of the two classes of fans are compared, the difference is between 6 to 8dB,
depending on the specific sound power level definition. Note that the maximum efficiency
points of all axial and centrifugal fans compared in Figure 18.2 lie in a small range of the
flow coefficient.
The two centrifugal fans with backward curved blades that are labelled with triangle
symbols have almost identical geometries, except for the blade profile, i.e., airfoil blades
versus flat blades. The first proves only marginally quieter in Figure 18.2 than the latter so
that from an acoustic point of view flat blades seem to be just as good for centrifugal fans
as airfoil blades
332
90
+
<$
"0
~
CD
80
>
CD
70
a
0.
"0
3o
60
OJ
U
M
4M
50
U
CD
0.
OJ
ro
40
"'o""'
f
30
.1
.4
.3
Flow coefficient
n
1/min
ItAxial fan,
OGV, airfoil blades 3000
++Axial fan,
OGV, airfoil blades 3000
l<*Axial fan, no GV, airfoil blades 2925
l<*Axial fan,
OGV, airfoil blades 2943
<><>Centr. fan, b.c. airfoil blades 1820
Centr. fan, b.c. airfoil blades 1600
t>t.Centr. fan, b.c. blades (f lat)
1600
~Centr. fan,
flat radial blades 1800
""  <>Sc i rrocco blower (f.c. blades)
700
88Mixedflow fan (halfradial)
1000
.2
.6
.5
0
m
0.45
0.45
0.60
0.50
0.60
0.51
0.51
0.51
0.50
0.45
m/s
70.7
70.7
91.9
77.0
57.2
42.7
42.7
48.0
18.3
23.5
24
12
8
12
12
2.5
2.5
2.5
2.5
2.0
2.0
2.0
2.0
2.0
2.0
12
12
15
38
8
Figure 18.2: Total specific sound power level of various fan types (after Neise (1988, 1989)).
18.3.3
Aweighting reduces the level of the low frequency components, and the resulting effect on
the overall noise level is the stronger the lower the fan speed. Therefore one has to be careful
in discussing the acoustic performance of fans based on the Aweighted levels as shown in
Figure 18.3, in particular when the fans to be compared were measured at. different impeller
speeds.
Because of its low impeller speed, t.he halfradial fan generates t.he lowest Aweighted
specific sound power level of all fans, followed by the centrifugal fans with backward curved
blades. As before, the ranking of the scirocco blower depends on the specific sound power
level definition.
With Aweight.ing, t.he specific sound power levels of axial and cent.rifugal fans differ even
more t.han before, and this is not only due t.o the different impeller speeds but mainly to t.he
fact t.hat. t.he spectral characteristics of t.he two fan t.ypes are entirely different: centrifugal
fans radiate predominant.ly low frequency sound, while the noise spectra of axialflow fans
show a maximum in a medium frequency range where the effect of Aweight.ing is small. The
result is t.hat. centrifugal fans appear quieter over almost t.he entire range of flow coefficient.s,
and if only t.he quietest of two classes of fans are compared at their respective design point.s,
the difference in t.he Aweighted specific sound power level is 1314dB.
333
100
m
LJ
90
OJ
>
OJ
80
L
OJ
'"
Cl
70
LJ
c:
::J
0
Ul
60
u
.~
~
.~
u
(l)
50
Cl
Ul
LJ
OJ
oW
.c:
I
40
en
.~
OJ
'"
I
30
.1
.2
.3
.4
.5
.6
Flow coefficient
OGV, airfoil blades
++Axial fan,
++Axial fan,
OGV, airfoil blades
lI*Axial fan, no GV, airfoil blades
It*Axial fan,
OGV, airfoil blades
<><>Centr. fan, b.c. airfoil blades
Centro fan, b.c. airfoil blades
66Centr. fan, b.c. blades (flat)
~Centr. fan,
flat radial blades
<)~Scirrocco blower (f.c. blades)
GlMixedflow fan (halfr'adiaJ)
n
l/min
m/s
3000
3000
2925
2943
1820
1600
1600
1800
700
1000
0.45
0.45
0.60
0.50
0.60
0.51
0.51
0.51
0.50
0.45
70.7
70.7
91.9
77.0
57.2
42.7
42.7
48.0
18.3
23.5
24
12
8
12
12
12
12
15
38
8
2.5
2.5
2.5
2.5
2.0
2.0
2.0
2.0
2.0
2.0
Figure 18.3: Aweighted specific sound power level of various fan types (after Neise (1988,
1989)).
18.3.4
Figure 18.4 shows the normalized onethird octave spectra of the specific sound power of
the various fans at their respective points of optimum operation using again the definitions
according to Madison and Regenscheit. The sound power in each frequency band is the
sum of inlet and outlet noise. The spectra are plotted versus the Strouhal number Si =
U D jU)( njZ). In this presentation, the blade passing frequencies of all fans lie in the onethird octave band at Si = 1.
The spectra of the centrifugal fans are highest at low frequencies and fall off towards
higher frequencies, while the axial fans reach their maximum levels at mid frequencies. As a
consequence, only in the very low frequency range are the specific sound power levels of the
axial fan type lower than those of the centrifugal fans, but they are substantially higher at
mid and high frequencies and in particular at the blade passing frequency and its harmonics.
Only the centrifugal fan with radial blades which is known for its poor aerodynamic design
generates specific blade tone levels as high as those of axial fans.
The spectrum of the vaneless axial fan peaks also at half the blade passing frequency
because its eight blades are arranged unevenly in four pairs.
334
90
80 70
*lh
60
Q)
>
Q)
<
50
ill
'"
0
Cl
40
::J
0
UJ
u
.~
4
30
.~
U
Q)
0.
(J)
20
Figure 18.4: Nondimensional specific sound power spectra of various fan types (after Neise
(1988, 1989)).
The spectral characteristic of the halfradial fan resembles that of centrifugal fans. The
effect of the blade profile of centrifugal fans on the noise spectrum is as small as on the
total noise. Comparing the curves with the triangle symbol, with airfoil blades only the high
frequency random noise components are about 3dB lower than with flat sheet metal blades.
The presentation in Figure 18.4 is useful when comparing the levels of blade tones because
due to the definition of the Strouhal number used they appear at integer values Si = 1,
2, ... for all fans. For the comparison of random noise components, however, it is more
advantageous to plot the spectra as functions of f D / U, i.e., without involving the number of
impeller blades, see Figure 18.5. Here the large level differences between axial and centrifugal
fans at medium and high frequencies are particularly evident.
335
90
80
70
(1]
u
.<
60
Qj
>
Qj
"Qj
LWA 
50
m
10 19 10/0 0 I  10 19 IlIPI / lIPlo I  10 19 IU/o o1
J:
0
co.
LJ
C
~O
:::J
0
U1
u
.~,
4
30
,,..1
U
Qj
Cl.
en
20
10
LWA
0
10
0.1
100
10
fO
n
l/min
3000
3000
2925
2943
1820
1600
1600
1800
700
1000
m
0.45
0.45
0.60
0.50
0.60
0.51
0.51
0.51
0.50
0.45
U
Z
m/s
70.7 24
70.7 12
91.9 8
77 .0 12
57.2 12
42.7 12
42.7 12
48.0 15
18.3 38
23.5 8
2.5
2.5
2.5
2.5
2.0
2.0
2.0
2.0
2.0
2.0
Figure 18.5: Nondimensional specific sound power spectra of various fan types (after Neise
(1988, 1989)),
18.3.5
In Figure 18,6 the total specific sound power levels of the various fans at optimum operation
are plotted versus the specific speed (J' = 'P 0 .5 /1jJO.75 (open symbols). Also included are results
from Sigel, Schilling & Zierep (1980), Sigel & Felsch (1981), and Barsikow & Neise (1978)
(filled symbols). The fan noise measurements in these two studies were made only in the fan
outlet ducts, In order to compare the data with the total sound power levels discussed here,
2dB were added (this means that the inlet sound power levels are assumed 2.3dB lower than
the outlet levels).
A general observation is that the scatter of the data for each class of fans is quite substantial, i.e., 57dB for the axial fan type and even 1012dB for the centrifugal fans with
backward curved blades, although the data do not cover an extremely wide range of specific
speeds.
336
100
90
.. 0
co
TI
~,
OJ
80
>
OJ
'OJ
30
70
0
Cl.
TI
C
::0
60
0
trJ
U
.r<
4_
.~,
50
*Axial fan, no GV
oAxial fan, OGV
Axial fan, OGV (Sigel et al.)
vCentr. fan, b.c. blades
TCentr. fan, b.c. blades (Neise&8arsikow)
"'Centr. fan, b.c. blades (Sigel et al.)
+Centr. fan, flat radial blades
oScirrocco runner
+Scirrocco runner (Sigel et al.)
DMixedf low fan (hal fradial)
OJ
Cl.
trJ
co
... 0
40
W
f
30
200
.5
1.5
Specific speed
Figure 18.6: Total specific sound power level as function ofthe fan specific speed (after Neise
(1988, 1989).
Among the fans tested by Sigel, Schilling & Zierep (1980) and Sigel & Felsch (1981), the
axial fan is one of the quietest ones, which is in contradiction to the findings of the present
study where the ranking is quite different.
Taking all data into consideration, the general trend is as stated before, namely that axial
fans tend to produce higher specific sound power levels than centrifugal fans with backward
curved blades. However the differences between the two classes of fans are smaller than
found before.
18.4
CONCLUSIONS
The aerodynamic noise of fans of different designs is compared on the basis of the specific
sound power level, which is equal to the sum of inlet and outlet sound power reduced in a
certain way by the aerodynamic power of the fan. With this concept it is possible to compare
different fan types over the whole range of their performance characteristics. To ensure that
the result of the comparison is not influenced too much by the choice of the specific sound
power level definition, two different formulations are used.
Among the fan types tested, the centrifugal fans with backward curved fans generate
the lowest specific sound power levels, followed by the mixedflow fan (halfradial fan). The
centrifugal fan with straight radial blades is the noisiest, as was to be expected. Airfoil
shaped blades in centrifugal fans with backward curved blades have only a small advantage
over flat sheet metal blades.
Over a fairly wide range of flow coefficients, axial fans have, on average, a higher specific
sound power level than centrifugal fans with backward curved blades. At the respective
points of optimum operation, the difference between the quietest in each class is 68dB for
the total and 1314dB for the Aweighted level. If the results of an earlier study by Sigel,
337
Schilling & Zierep (1980) and Sigel & Felsch (1981) are taken into account, the general
trend is as stated before, however, the differences between axial and centrifugal fans become
smaller.
The scirocco blower ranks between halfradial and axial fans, depending somewhat on
the definition of the specific sound power level.
The spectral shape of the noise of axialflow fans is characterized by a maximum at
medium to high frequencies while centrifugal fans and halfradial fans radiate noise predominantly at low frequencies with a continous fall off towards high frequencies. When
normalized specific sound power spectra of different fan types are compared at their individual optimum operation points, the centrifugal fans with backward curved blades show the
lowest levels from mid to high Strouhal numbers.
Only a limited number of fans has been used for the present comparison, and therefore
the results of the comparison described above cannot be generalized. However it is felt that
it shows a first significant trend. Further tests under comparable measurement conditions
are necessary to validate the present findings.
18.5
BIBLIOGRAPHY OF CHAPTER 18
BARSIKOW, B. & NEISE, W., 1978. Der EinfluB ungleichmaBiger Zustromung auf das
Gerausch von Radialventilatoren. In Fortschritte der Akustik) DAGA )78 (Bochum,
Germany), VDEVerlag Berlin, Germany, pp. 411414.
ECK,
HOPPE, G. & NEISE, W., 1987. Vergleich verschiedener GerauschmeBverfahren fur Ventilatoren. Forschungsbericht FLT 3/1/31/87, Forschungsvereinigung fur Luft und Trocknungstechnik e.V., Frankfurt/Main, Germany.
ISO 5136,1990. Acoustics  Determination of sound power radiated into a duct by fans  Induct method. International Standard, International Organisation for Standardization.
MADISON, R., 1949. Fan Engineering (Handbook)) 5th Edition. Buffalo Forge Company,
Buffalo, New York.
NEISE, W., HOPPE, G. & HERRMANN,!., 1986. Intercomparison of fan noise measurements using the induct method ISO/DIS 5136. Interner Forschungsbericht DFVLRIB 2221486 /B 1, Deutsche Forschungs und Versuchsanstalt fur Luft und Raumfahrt,
Koln/Porz, Germany.
NEISE, W., HOPPE, G. & HERRMANN,!., 1987. Gerauschmessungen an Ventilatoren
 Vergleichende Untersuchungen nach dem KanalVerfahren DIN 45 635 Teil 9 bzw.
ISO/DIS 5136. Heizung L7"iftung Haustechnik 38, 343351.
NEISE, W., 1988.
Gerauschvergleich von Ventilatoren. Die spezifische Schalleistung
zur Beurteilung des Gerausches unterschiedlicher Ventilatortypen. Heizung Luftung
Haustechnik 39, 392399.
338
NEISE, W., 1989. Noise rating of fans on the basis of the specific sound power level. In
Proceedings Tenth Australasian Fluid Mechanics Conference (Melbourne, Australia),
pp. 1.411.44.
SIGEL, T. & FELSCH, K.O., 1981. GerauschminderungdurchgeeigneteVentilatorauswahl.
Heizung Liiftung Haustechnik 32, 441446.
SIGEL, T., SCHILLING, R. & ZIEREP, J., 1980. Gerauschvergleich verschiedener Ventilatoren bei gleicher stromungstechnischer Aufgabe. Forschungsbericht FLT 3/1/31/80,
Forschungsvereinigung fur Luft und Trocknungstechnik e.V., Frankfurt/Main, Germany.
339
340
Chapter 19
BIBLIOGRAPHY
ABOM, M., BODEN, H. & LAVRENTJEV, .J., 1992. Source characterization of fans using acoustic 2port models. In Proceedings Fan Noise (Senlis, France, 1992), Centre
Technique des Industries Mecaniques (CETIM), pp. 359364.
ABRAMOWITZ, M. & STEGUN, 1. A., 1970. Handbook of Mathematical Functions. Dover
Publications, Inc., New York.
AMCA 30085, 1985. Reverberation room method for sound testing of fans. American
standard, Air Movement and Control Association, Arlington Heights, Illinois, USA.
AMCA 330/ ASHRAE 68R SPC, 1994. in preparation, Air Movement and Control Association, Arlington Heights, Illinois, USA.
ARCHIBALD, F. S., 1975. The laminar boundary layer instability excitation of an acoustic
resonance. Journal of Sound and Vibration 38, 387402.
ASHRAE, 1991. ASHRAEHandbook. American Society of Heating Refrigerating and Air
Conditioning Engineers.
BAADE, P. K., 1971. Die Behandlung des Axialventilators als akustisches Zweitor. Doctoral
dissertation, Technische Universitat Berlin, Germany.
BAADE, P. K., 1977. Effects of acoustic loading on axial flow fan noise generation. Noise
Control Engineering 8, 515.
BAEHR, H. D., 1962. Thermodynamik. Springer Verlag, Berlin, Germany.
BARD, H. & KLEE, D., 1986. Axialventilatoren mit kontinuierlicher Kennlinie ohne Pumpgrenze fur VLVSysteme in der Elektronikindustrie. In Ventilatoren im industriellen
Einsatz (Dusseldorf, Germany, 1986), vol. 594 of VDIBerichte, VDI Verlag, Dusseldorf,
pp. 247255.
BARSIKOW, B. & NEISE, W., 1978. Der EinfluB ungleichmaBiger Zustromung auf das
Gerausch von Radialventilatoren. In Fortschritte der Akustik) DA GA )78 (Bochum,
Germany, 1978), VDE Verlag Berlin, Germany, pp. 411414.
BARSIKOW, B. & NEISE, W., 1979. The influence of nonuniform inflow conditions on
centrifugal fan noise. In Proceedings Internoise )79 (Warsaw, Poland, 1979), pp. 8993.
341
BARTENWERFER, M. & AGNON, R., 1976. Der Einflui3 der Zahigkeit des geforderten
Mediums auf die Schallerzeugung von Radialventilatoren. Forschungsbericht DLRFB
7630, Deutsche Forschungsanstalt fur Luft und Raumfahrt, Koln/Porz, Germany.
BARTENWERFER, M., AGNON, R. & GIKADI, T., 1976. Eine experimentelle Untersuchung
zur Schallerzeugung und zur Schallausbreitung in Radialventilatoren. Forschungsbericht
DLRFB 7614, Deutsche Forschungsanstalt fur Luft und Raumfahrt Koln/Porz, Germany.
BARTENWERFER, M., GIKADI, T., NEISE, W. & AGNON, R., 1977. Noise reduction in
centrifugal fans by means of an acoustically lined casing. Noise Control Engineering 8,
100107.
BARTENWERFER, M., 1977. Letter to the editor. Noise Control Engineering 8, 3.
BENT, p, H. & McLAUGHLIN, D. K., 1993. Enhancements to noise source measurement
techniques for turbomachinery. AIAAPaper 934373, American Institute of Aeronautics
and Astronautics, Washington DC, USA. 15th AIAA Aeroacoustics Conference, Long
Beach, California, USA.
BENZAKEIN, M. J" 1972. Research on fan noise generation.
Society of America 51, 14271438.
BERANEK, L. L" 1960, Noise Reduction, 4th ed. McGrawHill Book Company, New York.
BLAKE, W. K., 1986. Mechanics of FluidInduced Sound and Vibration. Academic Press
Inc. New York, USA.
BOLLETER, U. & CROCKER, M. J., 1972, Theory and measurement of modal spectra in
hardwalled cylindrical ducts. Journal of the Acoustical Society of America 51, 14391477.
BOLLETER, U" COHEN, R. & WANG, J., 1973. Design considerations for an induct sound
power measuring system. Journal of Sound and Vibration 28, 669685.
BOLTON, A. N" 1992a. Fan noise installation effects. In Proceedings Fan Noise (Senlis,
France, 1992a), Centre Technique des Industries Mecaniques (CETIM), pp. 7788.
BOLTON, A. N" 1992b. Intercomparison of induct fan noise measurements. Report EUR
13890 EN, Commision of the European Communities, Community Bureau of Reference,
Brussels, Belgium.
BOMMES, L., 1961. Gerauschentwicklung bei Ventilatoren kleiner und mittlerer Umfangsgeschwindigkeit. Larmbekampfung 5, 6975.
BOMMES, L., 1980. Larmminderung bei einem Radialventilator kleiner Schnellaufigkeit
unter Berucksichtigung von Zungenform, Zungenabstand und Schaufelzahl. Heizung
Liiftung Haustechnik 31, 173179 and 210218.
BOTTCHER, J. & GEHLHAR, B., 1993. Experimental investigation of propeller noise under
highly turbulent inflow condition. DLRIB 12993/21, Deutsche Forschungsanstalt fur
Luft und Raumfahrt e.V.
342
BRENTNER, K. S. & FARASSAT, F., 1992. Helicopter noise prediction: The current status
and future direction. In DGLR/AIAA) 14th Aeroacoustics Conference (1992), pp. 724735. DGLR/ AIAA Paper 9202122, DGLRBericht 9203.
BRIDELANCE, J. P., 1986. Aeroacoustic study of axial fans with small diameter. Analysis
and suppression of instability noise. In Proceedings Internoise )86 (Cambridge, USA,
1986), pp. 141146.
BROOKS, T. F. & HODGSON, T. H., 1981. Trailing edge noise prediction using measured
surface pressures. Journal of Sound and Vibration 78, 69117.
BROWN, N. A., 1977. The use of skewed blades for ship propellers and truck fans. In Proceedings 98th ASME Winter Annual Meeting {Noise and Fluids Engineering} (Atlanta,
Georgia, USA, 1977), pp. 201207.
BS 848 PART 2,1985. Fans for general purposes. Methods of noise testing. British Standard,
British Standards Institution.
CAROL us, T., 1992. Acoustic performance of low pressure axial fan rotors with different
blade chord lengths and radial load distributions. In Proceedings DGLR/AIAA 14th
Aeroacoustics Conference (Aachen, Germany, 1992), Deutsche Gesellschaft fiir Luftund Raumfahrt, Bonn, Germany, pp. 809815.
CHANAUD, R. C., KONG, N. & SITTERDING, R. B., 1976. Experiments on porous blades
as a means of reducing fan noise. Journal of the Acoustical Society of America 59,
564575.
CHANAUD, R. C., 1965. Aerodynamic sound from centrifugal fan rotors. Journal of the
Acoustical Society of America 37, 969974.
CHOU, S. R., 1990. A study of rotor broadband noise mechanisms and helicopter tail rotor
noise. NASACR 177565, National Aeronautics and Space Administration, USA.
CHUNG, J. Y. & BLASER, D. A., 1980. Transfer function method of measuring induct
acoustic properties (part 1 and part 2). Journal of the Acoustical Society of America
68, 907921.
CORDIER, 0., 1953. Ahnlichkeitsbedingungen bei Ventilatoren. BrennstojJ Warme Kraft 5,
337340.
CREMER, L., 1971. The second annual fairey lecture: The treatment of fans as black boxes.
Journal of Sound and Vibration 16, 115.
CRIGHTON, D. G. & PARRY, A. B., 1991. Asymptotic theory of propeller nose part ii:
Supersonic singlerotation propeller. AIAA Journal 29, 12, 20312037.
CRIGHTON, D. G. & PARRY, A. B., 1992. Higher approximations in the asymptotic theory
of propeller nose. AIAA Journal 30, I, 2328.
CUMPSTY, N. A., 1977. Review  A critical review of turbomachinery noise.
Transactions) Journal of Fluids Engineering 99, 278293.
ASME
CURLE, N., 1955. The influence of solid boundaries upon aerodynamic sound. Proceedings
of the Royal Society {London} A 231, 505514.
343
CURVERS, J. M. & DUPRE, M., 1992. Reduction of centrifugal fan noise by choice of the
rotor geometry. In Proceedings Fan Noise (Senlis, France, 1992), Centre Technique des
Industries Mecaniques (CETIM), pp. 261266.
DAHLEN, H., DOBRZYNSKI, W. & HELLER, H., 1988. Aeroakustische Untersuchungen
zum Ui,rm von Ultraleichtflugzeugen. Forschungsbericht DFVLRFB 8803, Deutsche
Forschungsanstalt fiir Luft und Raumfahrt e.V., KolnjPorz, Germany.
DAVY, J. L., in press, 1994. The modal and flow velocity corrections of microphone turbulence screens. Journal of Sound and Vibmtion.
DEMING, A. F., 1937. Noise from propellers with symmetrical sections at zero blade angle.
NACA Technical Memorandum 605, NACA.
DEMING, A. F., 1938. Noise from propellers with symmetrical sections at zero blade angle,
II. NACA Technical Memorandum 679, NACA.
DIN 45 635 TElL 38, 1986. Geriiuschmessung an Maschinen. Luftschallemission, Hiillfliichen, Hallraum und Kanal Verfahren; Ventilatoren. Deutsche Norm, Deutsches
Institut fiir Normung e.V., Berlin, Germany.
DIN 45 635 TElL 9, 1989. Geriiuschmessung an Maschinen. Luftschallemission, KanalVerfahren; RahmenMeBverfahren fiir Genauigkeitsklasse 2. Deutsche Norm, Deutsches
Institut fiir Normung e.V., Berlin, Germany.
DOAK, P. E., 1960. Acoustic radiation from a turbulent fluid containing foreign bodies.
Proceedings of the Royal Society (London) A254, 129145.
DOAK, P. E., 1973. Excitation, transmission and radiation of sound from source distributions in hardwalled ducts of finite length. Part I: The effects of duct crosssection
geometry and source distribution spacetime pattern. Part II: The effects of duct length.
Journal of Sound and Vibmtion 31, 172 and 137174.
DOBRZYNSKI, W. & GEHLHAR, B., 1993. Propeller blade number: A parameter for flyover
noise reduction. In International Noise and Vibration Control Conference (1993), M. J.
Crocker & N. I. Ivanov, Eds., vol. 1, pp. 6570.
DOBRZYNSKI, W., 1986. The effect on radiated noise of nonzero propeller rotational plane
attitude. AIAA Paper 861926. AIAA Aeroacoustics Conference, Seattle, Washington.
DOBRZYNSKI, W., 1993. Propeller noise reduction by means of unsymmetrical bladespacing. Journal of Sound and Vibmtion 163, 123136.
DREVET, P., DUPONCHEL, J. P. & JACQUES, J. R., 1977. The effect of flight onjet noise
as observed on the Bertin Aerotrain. Journal of Sound and Vibration 54, 173201.
DUNCAN, P. E. & DAWSON, B., 1974. Reduction of interaction tones from axial flow fans
by suitable design of rotor configuration. Journal of Sound and Vibration 33,143154.
DUNCAN, P. E. & DAWSON, B., 1975. Reduction of interaction tones from axial flow fans
by nonuniform distribution of the stator vanes. Journal of Sound and Vibration 38,
357371.
344
DUNCAN, P. E., DAWSON, B. & HAWES, S. P., 1975. Design techniques for the reduction
of interaction tonal noise from axial flow fans. In Proceedings Conference on Vibrations
and Noise in Pump) Fan) and Compressor Installations (University of Southampton,
U.K., 1975), pp. 143161.
ECK, B., 1962. Ventilatoren, 4th ed. Springer Verlag, Berlin, Germany.
ECK, B., 1972. Ventilatoren, 5th ed. Springer Verlag, Berlin, Germany.
ECK,
B., 1973. Fans, 1st English ed. Pergamon Press Ltd., Oxford.
EMBLETON, T., 1963. Experimental study of noise reduction in centrifugal blowers. Journal
of the Acoustical Society of America 35, 700705.
ERNSTHAUSEN, W., 1936. Die Quelle des Propellergerausches. Luftfahrtforschung 8, 433440. Translated as "On the Origin of Propeller Noise" NACA TM 825, 1937.
ERNSTHAUSEN, W., 1938. Untersuchungen tiber das Luftschraubengerausch. Akustische
Zeitschrift 3,141146.
EVTEEV, 1. & PADALKIN, A., 1992. The new active method of decreasing the siren noise
of radial fans. In Proceedings Fan Noise (Senlis, France, 1992), Centre Technique des
Industries Mecaniques (CETIM), pp. 267270.
FARASSAT, F. & SUCCI, G. P., 1980. A review of propeller discrete frequency noise prediction technology with emphasis on two current methods for time domain calculations.
Journal of Sound and Viobration 71, 399419.
FARASSAT, F., 1986. Production of advanced propeller noise in the time domain. AIAA
Journal 24, 4, 578584.
FARZAMI, R. & GUEDEL, A., 1992. Influence of a straightener on induct sound power
measurements. In Proceedings Fan Noise (Senlis, France, 1992), Centre Technique des
Industries Mecaniques (CETIM), pp. 375380.
FFOWCS WILLIAMS, J. E. & HALL, L. H., 1970. Aerodynamic sound generation by
turbulent flow in the vicinity of a scattering halfplane. Journal of Fluid Mechanics 40,
657670.
FFOWCS WILLIAMS, J. E. & HAWKINGS, D. L., 1969a. Sound generated by turbulence and
surfaces in arbitrary motion. Philosophical Transactions of the Royal Society (London)
A 264, 321342.
FFOWCS WILLIAMS, J. E. & HAWKINGS, D. L., 1969b. Theory relating to the noise of
rotating machinery. Journal of Sound and Vibration 10, 1021.
FFOWCS WILLIAMS, J. E. & KEMPTON, A. J., 1978. The noise from the largescale
structure of a jet. Journal of Fluid Mechanics 84, 673694.
FFOWCS WILLIAMS, J. E., 1963. The noise from turbulence convected at high speed.
Philosophical Transactions of the Royal Society A 225, 469503.
FFOWCS WILLIAMS, J. E., 1969. Hydrodynamic noise. Annual Review of Fluid Mechanics
1, 197222.
345
Heizung Liijtung
FUCHS, H. V. & MICHALKE, A., 1973. Introduction to aerodynamic noise theory. Progress
in Aerospace Science 14, 227297.
FUCHS, H. V., MANTE, J., NEISE, W. & STAHL, B., 1978. Untersuchungen des Larms
von Axialgeblasen in Kraftfahrzeugen. In FOTtschritte deT Akustik} DAGA }'l8 (Bochum,
Germany, 1978), VDEVerlag Berlin, Germany, pp. 403406.
FUCIIS, H. V., 1976. Basic aerodynamic noise theory. AGARDVKI Lecture Series 80 on
"Aerodynamic Noise" .
FUKANO, T., KODAMA, Y. & SENOO, Y., 1977. Noise generated by low pressure axial
flow fans. Journal of Sound and Vibration 50, 6374.
FUKANO, T., TAKAMATSU, Y. & KODAMA, Y., 1986. The effects of tip clearance on the
noise of low pressure axial and mixed flow fans. Journal of Sound and Vibration 105,
291308.
GHILADI, A., 1981. Drehklangentstehung in axialen Turbomaschinen und ausbreitung in
angeschlossenen Rohrleitungen. Doctoral dissertation, RWTH Aachen, Germany.
GOLDSTEIN, M. E., ROSENBAUM, B. M. & ALBERS, L. U., 1974. Sound radiation from
a high speed axialflow due to the inlet turbulence quadrupole interaction. NASATN
D7667, National Aeronautics and Space Administration, USA.
GOLDSTEIN, M. E., 1974. Unified approach to aerodynamic sound generation in the presence of solid boundaries. Journal of the Acoustical Society of America 56, 497509.
GOLDSTEIN, M., 1976. Aeroacoustics. McGrawHill International Book Company, New
York.
GRAHAM, J. B., 1972. How to estimate fan noise. Sound and Vibration 6, 2427.
GRAHAM, J. B., 1992. The status of fan noise measurement in North America. In Proceedings Fan Noise (Senlis, France, 1992), Centre Technique des Industries Mecaniques
(CETIM), pp. 293300.
GROFF, G. C., SCHREINER, J. R. & BULLOCK, M. J., 1967. Centrifugal sound power
level prediction. ASHRAEJournal 9, 7177.
GUTIN, L., 1936. On the sound field of a rotating airscrew (English translation of Russian
title). Zhurnal Technicheskoi Fiziki 6, 889909. English translation of Russian original
published as NACATM 1195,1948.
GUTIN, L., 1942. On the rotational sound of an airscrew (English translation of Russian
title). Zhurnal Technicheskoi Fiziki 12, 7683. Translated as British National Lending
Library for Science and Technology RTS 7543.
HANSON, D. B., 1979. The influence of propeller design parameters on farfield harmonic
noise in forward flight. AIAA Paper 790609.
HANSON, D. B., 1980. Helicoidal surface theory for harmonic noise of propellers in the far
field. AIAA Journal 18, 10, 12131220.
346
Siemens
ISO 3741, 1975. Acoustics Determination of sound power levels of noise sources  Precision methods for broadband sources in reverberation rooms. International Standard,
International Organisation for Standardization.
ISO 3742, 1975. Acoustics  Determination of sound power levels of noise sources  Precision methods for discretefrequency and narrowband sources in reverberation rooms.
International Standard, International Organisation for Standardization.
ISO 37431, 1994. Acoustics  Determination of sound power levels of noise sources  Engineering methods for small movable sources in reverberant fields  Part 1: Comparison
method in hardwalled test rooms. International Standard, International Organisation
for Standardization.
IS 0 3743 2, 1994. Acoustics  Determination of sound power levels of noise sources Engineering methods for small movable sources in reverberant fields  Part 2: Direct
method for special reverberation test rooms. International Standard (in press), International Organisation for Standardization.
ISO 3744, 1981. Acoustics  Determination of sound power levels of noise sources  Engineering methods for freefield conditions over a reflecting plane. International Standard,
International Organisation for Standardization.
ISO 3745, 1977. Acoustics  Determination of sound power levels of noise sources  Precision
methods for anechoic or semianechoic rooms. International Standard, International
Organisation for Standardization.
ISO 3746, 1992. Acoustics  Determination of sound power levels of noise sources  Survey
method employing an enveloping measurement surface over a reflecting plane. Draft
International Standard, International Organisation for Standardization.
ISO 5136,1990. Acoustics Determination of sound power radiated into a duct by fans  Induct method. International Standard, International Organisation for Standardization.
ISO 5801, 1994. Performance testing of fans using standardized airways. International
Standard, in preparation, International Organisation for Standardization.
ISO /DIS 10302, 1993. Acoustics  Method for the measurement of airborne noise emitted
by small airmoving devices. Draft International Standard, International Organisation
for Standardization.
348
ISO /TC 117/WG 2/N8, 1991. The determination of fan sound power levels under standardized laboratory conditions. Draft Standard, International Organisation for Standardization.
KAISER, L. & WOHRLE, K., 1989. Variable speed blowers  a solution to control the
noise of computers. In Proceedings Internoise)89 (Newport Beach, Ca., USA, 1989),
pp. 157162.
KAMEIER, F. & NEISE, W., 1993. Verfahren zur Reduzierung der Schallemission
sowie zur Verbesserung der Luftleistung und des Wirkungsgrades bei einer axialen
Stromungsmaschine und Stromungsmaschine. Deutsche Patentanmeldung P 43 10 104.6,
Deutsches Patenamt Berlin, Germany.
KAMEIER, F., NAWROT, T. & NEISE, W., 1992. Experimental investigation of tip clearance noise in axial flow machines. In Proceedings DGLR/AIAA 14th Aeroacoustics
Conference (Aachen, Germany, 1992), Deutsche Gesellschaft fur Luft und Raumfahrt,
Bonn, Germany, pp. 250259.
KAMEIER, F., 1994. Experimentelle Untersuchung zur Entstehung und Minderung des
Blattspitzenwirbellarms axialer Stromungsmaschinen. FortschrittBericht VDI Reihe 7,
Nr. 243, Verein Deutscher Ingenieure, VDIVerlag GmbH, Dusseldorf, Germany.
KAPUR, A. & MUNGUR, P., 1972. On the propagation of sound in a rectangular duct with
gradients of mean flow and temperature in both transverse directions. Journal of Sound
and Vibration 23, 401404.
KHOROSHEV, G. & PETROV, Y., 1965. Decreasing the noise level of centrifugal ventilators.
Sudostroenie 3,1517.
KHOROSHEV, G. & PETROV, Y., 1971. Some new methods of fan noise reduction. In
Proceedings of the 7th International Congress on Acoustics (Budapest, 1971), no. Paper
19N16.
KING III, W. F., 1989. Aeroacoustics of highspeed tracked vehicles. DLRIB 2221489/B4,
Deutsche Forschungsanstalt fur Luft und Raumfahrt e.V.
Ko, S., 1972. Sound attenuation in acoustically lined circular ducts in the presence of
uniform flow and shear flow. Journal of Sound and Vibration 22, 193210.
KONIECZNY, P. & BOLTON, J. S., 1990. The influence of design parameters on broadband
noise generation by small centrifugal blowers. In Proceedings Internoise )90 (Gothenburg, Sweden, 1990), pp. 717720.
KOOPMANN, G. H., Fox, D. J. & NEISE, W., 1988. Active source cancellation ofthe blade
tone fundamental and harmonics in centrifugal fans. Journal of Sound and Vibration
126, 209220.
KOOPMANN, G. H., NEISE, W. & HOOVER, R. M., 1985. Active cancellation of blade
tones in power plant fans. Final Report Project 16497 EPRI 3260, Electric Power
Research Institute, Palo Alto, California.
KRISHNAPPA, G., 1979. Some experimental studies on centrifugal blower noise.
Control Engineering 12, 8290.
349
Noise
KRISHNAPPA, G., 1980. Effect of modulated blade spacing on centrifugal fan noise. In
Proceedings Internoise 1980) December 810 (Miami, Fla., 1980), pp. 215218.
KUROSAKA, M., 1971. A note on multiple pure tone noise. Journal of Sound and Vibration
19, 453462.
LANDSKRON, R. & NEISE, W., 1989. Gerauschmessungen an Propellern mit ungleichformiger Blatteilung und variabler Blattlange. In Fortschritte der Akustik) DAGA )89
(1989), DPGGmbH, Bad Honnef, Germany, pp. 587590.
LANSING, D. L. & ZORUMSKI, W. E., 1972. Transmission and radiation of sound from
ducts with axial variations in wall impedance. In Symposium on the Acoustics of Flow
Ducts) 1014 January (Southampton, England, 1972), Southampton University, Institute of Sound and Vibration Research.
LANSING, D. L., DRISCHLER, J. A. & PUSEY, C. G., 1970. Radiation of sound from an
unflanged circular duct with flow. In 19th Meeting of the Acoustical Society of America)
2124 April (Atlantic City, USA, 1970), The Acoustical Society of America.
LEIDEL, W., 1969. Einfluf3 von Zungenabstand und Zungenradius auf Kennlinie und Gerausch eines Radialventilators. Forschungsbericht DLRFB 6916, Deutsche Forschungsanstalt fiir Luft und Raumfahrt, Koln/Porz, Germany.
LEVERTON, J. W., 1989. Twentyfive years of rotorcraft aeroacoustics: Historical prospective and important issues. Journal of Sound and Vibration 133, 261287.
LEVINE, H. & SCHWINGER, J., 1948. On the radiation of sound from an unflanged circular
pipe. Physical Review 73, 383406.
LIGHTHILL, M., 1952. On sound generated aerodynamically. I. general theory. Proceedings
of the Royal Society (London) A 211,564587.
LIGHTHILL, M. J., 1954. On sound generated aerodynamically. II. Turbulence as a source
of sound. Proc. Royal Soc. London A 222, 132.
LIN, H. & MARTENSEN, A., 1969. Optimum lining configurations. Tech. Rep. SP207,
National Aeronautics and Space Administration, USA.
LIPKENS, B. & TOURNOY, D., 1990. The use of a quarter wavelength resonator in large
centrifugal fans. In Proceedings Internoise )90 (Gothenburg, Sweden, 1990), pp. 497500.
LOHMANN, D., 1993. Numerical optimization of propeller aeroacoustics  using evolution
strategy. In Proceedings Noise )93 (St. Petersburg, Russia, 1993), Interpublish Ltd, St.
Petersburg, Russia, pp. 103114.
LONGHOUSE, R., 1977. Vortex shedding noise of low tip speed, axial flow fans. Journal of
Sound and Vibration 53, 2546.
LONGHOUSE, R., 1978. Control of tip clearance noise of axial flow fans by rotating shrouds.
Journal of Sound and Vibration 58, 201214.
LOWSON, M. V., 1968. Reduction of compressor noise radiation. Journal of the Acoustical
Society of America 43, 3750.
350
LUSH, P. A. & BURRIN, R. H., 1972. The generation and radiation of supersonic jet
noise, Volume V, An experimental investigation of jet noise variation with velocity and
temperature. AFAPL TR7253Volume V, Air Force Aero Propulsion Laboratory.
LYONS, L. & PLATTER, S., 1963. Effect of cutoff configuration on pure tones generated by
small centrifugal blowers. Journal of the Acoustical Society of America 35, 14551456.
MADISON, R., 1949. Fan Engineering (Handbook); 5th Edition. Buffalo Forge Company,
Buffalo, New York.
MALING, G., 1963. Dimensional analysis of blower noise. Journal of the Acoustical Society
of America 35, 15561564.
MARCINOWSKI, H., 1953. Einfluss des Laufradspaltes und der Luftfuhrung bei einem
Kuhlgebliise axialer Bauart. Motortechnische Zeitschrijt (MTZ) 14, 259262.
MARGETTS, E. J., 1987. A demonstration that an axial fan in a ducted inlet ducted outlet
configuration generates predominantly dipole noise. Journal of Sound and Vibration
117, 399406.
MARIANO, S., 1971. Effect of wall shear layers on the sound attenuation in acoustically
lined rectangular ducts. Journal of Sound and Vibration 19, 261275.
MARIANO, S., 1975. Sound source location effects on the attenuation in acoustically lined
rectangular ducts. Journal of Sound and Vibration 41, 473491.
MCCROSKEY, W. J., 1982.
285311.
Unsteady airfoils.
MELLIN, R. & SOVRAN, G., 1970. Controlling the tonal characteristics of the aerodynamic
noise generated by fan rotors. ASMETransactions; Journal of Basic Engineering 92,
143154.
MICHALKE, A. & HERMANN, C., 1982. On the inviscid instability of a circular jet with
external flow. Journal of Fluid Mechanics 114, 343359.
MICHALKE, A. & MICHEL, U., 1979. Prediction of jetnoise in flight from static tests.
Journal of Sound and Vibration 67, 341367.
MICHALKE, A., 1970. A wave model for sound generation in circular jets. Deutsche Luftund Raumfahrt, DLR FB 7057.
MICHALKE, A., 1972. An expansion scheme for the noise from circular jets. Zeitschrift fur
Flugwissenschaften 20, 229237.
MICHALKE, A., 1977. On the effect of spatial source coherence on the radiation of jet noise.
Journal of Sound and Vibration 55, 377394.
MICHALKE, A., 1992. On the sensitivity of a slittube probe to the higherorder acoustic
modes and its consequences. In Proceedings DGLR/AIAA 14th Aeroacoustics Conference (Aachen, Germany, 1992), Deutsche Gesellschaft fur Luft und Raumfahrt, Bonn,
Germany, pp. 163168.
MICHEL, U. & BOTTCHER, J., 1992. Prediction of jet mixing noise for high subsonic
flight speeds. In DGLR/AJAA; 14th Aeroacoustics Conference (1992), pp. 846853.
DGLRj AIAA Paper 9202145, DGLRBericht 9203.
351
MICHEL, U. & MICHALKE, A., 1981a. Prediction of flyover jet noise spectra. AIAA Paper
812025. AIAA 7th Aeroacoustics Conference.
MICHEL, U. & MICHALKE, A., December 1981b. Prediction of flyover jet noise spectra
from static tests. NASA Technical Memorandum 83219.
MICHEL, U., 1984. Berechnung des aerodynamischen Liirms von Windkraftanlagen. In
Fortschritte der Akustik) DAGA )84 (Bad Honnef, Germany, 1984), DPGGmbH, Bad
Honnef, Germany, pp. 521524.
MONGEAU, L., THOMPSON, D. & McLAUGHLIN, D., 1993. Sound generation by rotating
stall in centrifugal turbomachines. Journal of Sound and Vibration 163, 130.
MONGEAU, L., 1991. Experimental Study of the Sound Generation by Rotating Stall in
Centrifugal Turbomachines. PhD thesis, Penn State University, University Park, Pennsylvania, USA.
MOORE, D., 1975. Reduction of fan noise by annulus boundary layer removal. Journal of
Sound and Vibration 43, 671681.
MORFEY, C., 1964. Rotating pressure patterns in ducts; their generation and transmission.
Journal of Sound and Vibration I, 6087.
MORFEY, C., 1969. A note on the radiation efficiency of acoustic duct modes. Journal of
Sound and Vibration 9, 367372.
MORFEY, C., 1971. Tone radiation from an isolated subsonic rotor. Journal of the Acoustical
Society of America 49, 16901692.
MORFEY, C. L., 1973a. Amplification of aerodynamic noise by convected flow inhomogeneities. Journal of Sound and Vibration 31, 391397.
MORFEY, C. L., 1973b. Rotating blades and aerodynamic sound. Journal of Sound and
Vibration 28, 578617.
MORFEY, C. L., 1979. Propagation from moving sources in flows. In Special course on
Acoustic Wave Propagation (1979), vol. AGARDR686, pp. 111  1113.
MORSE, P. & INGARD, K., 1968. Theoretical Acoustics. McGrawHill Book Company, New
York.
MUGRIDGE, B., 1971. Acoustic radiation from aerofoils with turbulent boundary layers.
Journal of Sound and Vibration 16, 593614.
MUGRIDGE, B. L., 1975. Axial flow fan noise caused by inlet flow distortion. Journal of
Sound and Vibration 40, 497512.
MULLER, K.D., 1986. Optimieren mit der Evolutionsstrategie in der Industrie anhand von
Beispielen. Doctoral dissertation, Technische Universitiit Berlin, Germany.
MUNGUR, P. & GLADWELL, G., 1969. Acoustic wave propagation in a sheared fluid
contained in a duct. Journal of Sound and Vibration 9, 2848.
MUNGUR, P. & PLUMBLEE JR., H., 1969. Propagation and attenuation of sound in a softwalled annular duct containing a sheared flow. Tech. Rep. SP207, National Aeronautics
and Space Administration, USA.
352
MUNGUR, P., PLUMBLEE, H. & DOAK, P. E., 1974. Analysis of acoustic radiation in a
jet flow environment. Journal of Sound and Vibration 36, 2152.
NEISE, W. & BARSIKOW, B., 1980. Akustische Ahnlichkeitsgesetze bei Radialventilatoren.
Forschungsbericht DFVLRFB 8036, Deutsche Forschungs und Versuchsanstalt fiir
Luft und Raumfahrt, Koln/Porz, Germany.
NEISE, W. & BARSIKOW, B., 1982. Acoustic similarity laws for fans. ASMETransactions)
Journal of Engineering for Industry 104, 162168.
NEISE, W. & HOLSTE, F., 1991. Determination of sound power in rectangular flow ducts.
In Proceedings Internoise)91 (Sydney, Australia, 1991), pp. 10091012.
NEISE, W. & HOPPE, G., 1986. Effect of inflow conditions on centrifugal fan noise. In Proceedings First European Symposium on Air Conditioning and Refrigeration (Brussels,
Belgium, 1986), pp. 165172.
NEISE, W. & KOOPMANN, G., 1982. Anwendung von A/4Resonatoren zur Larmminderung
bei Radialventilatoren. In Fortschritte der Akustik  FASE/DAGA '82 (Gottingen,
Germany, 1982), pp. 801804.
NEISE, W. & KOOPMANN, G., 1984. Noise reduction on the centrifugal suction fan of a
Berlin street sweeper truck. Noise Contml Engineering 23, 7888.
NEISE, W. & KOOPMANN, G., 1991. Active sources in the cutoff of centrifugal fans to
reduce the blade tones at higherorder duct mode frequencies. ASME Transactions)
Journal of Vibration and Acoustics 113, 123131.
NEISE, W., HOLSTE, F., MIRANDA, L. & HERRMANN, M., 1993. Intercomparison of openinlet/openoutlet sound power measurements on fans. Interner Forschungsbericht DLRIB 2221493/B5, Deutsche Forschungs und Versuchsanstalt fiir Luft und Raumfahrt,
Koln/Porz, Germany.
NEISE, W., HOLSTE, F. & HOPPE, G., 1988. Experimental comparison of standardized
sound power measurement procedures for fans. In Proceedings Internoise '88 (A vignon,
France, 1988), pp. 10971100.
NEISE, W., HOPPE, G. & HERRMANN,!., 1986. Intercomparison of fan noise measurements using the induct method ISO/DIS 5136. InterneI' Forschungsbericht DFVLRIB 2221486/B1, Deutsche Forschungs und Versuchsanstalt fiir Luft und Raumfahrt,
Koln/Porz, Germany.
NEISE, W., HOPPE, G. & HERRMANN,!., 1987. Gerauschmessungen an Ventilatoren
 Vergleichende Untersuchungen nach dem KanalVerfahren DIN 45 635 Teil 9 bzw.
ISO/DIS 5136. Heizung Liiftung Haustechnik 38, 343351.
NEISE, W., 1975a. Application of similarity laws to the blade passage sound of centrifugal
fans. Journal of Sound and Vibration 43,6175.
NEISE, W., 1975b. Theoretical and experimental investigations of microphone probes for
sound measurements in turbulent flow. Journal of Sound and Vibration 39, 371400.
NEISE, W., 1976. Noise reduction methods in centrifugal fans  a literature survey. Journal
of Sound and Vibration 45, 375403.
353
NEISE, W., 1983. Turbulenz und Geriiuschminderung bei Ventilatoren. VGB Kraftwerkstechnik 63, 957969.
NEISE, W., 1988a. Fan noise Generation mechanisms and control methods. In Proceedings
Internoise '88 (Avignon, France, 1988a), pp. 767776.
NEISE, W., 1988b. Geriiuschvergleich von Ventilatoren. Die spezifische Schalleistung
zur Beurteilung des Geriiusches unterschiedlicher Ventilatortypen. Heizung Liiftung
Haustechnik 39, 392399.
NEISE, W., 1989. Noise rating of fans on the basis of the specific sound power level.
In Proceedings Tenth Australasian Fluid Mechanics Conference (Melbourne, Australia,
1989), pp. 1.411.44.
NEISE, W., 1991. Experimental determination of the acoustic sensitivity of a microphone
turbulence screen under mean flow conditions. In Proceedings Internoise '91 (Sydney,
Australia, 1991), pp. 10131016.
NEISE, W., 1992. Review of fan noise generation mechanisms and control methods. In Proceedings Fan Noise (Senlis, France, 1992), Centre Technique des Industries Mecaniques
(CETIM), pp. 4556.
NEMEC, J., 1967. Noise of axial fans and compressors: Study of its radiation and reduction.
Joumal of Sound and Vibration 6, 230236.
NORUM, T. D. & SEINER, J. M., 1982. Measurement of mean static pressure and far field
acoustics of shockcontaining supersonic jets. NASA Technical Memorandum 84521.
OHTSUTA, K. & AKISHITA, S., 1990. Noise reduction of shortly ducted fans by using
forward swept and inclined blades. AIAAPaper 903986.
PARRY, A. B. & CRIGHTON, D. G., 1989. Asymptotic theory of propeller nose part i:
Singlerotation propeller. AIAA Joumal 27, 9, 11841190.
PETROV, Y. 1. & KHOROSHEV, G. A., 1971. Improving the noise level of centrifugal fans.
Russian Engineering Joumal 51, 4244.
PETROV, Y. 1. & KHOROSHEV, G. A., 1992. Fan noise reduction by affecting the blade
air flow structure. In Proceedings Fan Noise (Senlis, France, 1992), Centre Technique
des Industries Mecaniques (CETIM), pp. 247252.
PETROV, Y. 1., KHOROSHEV, G. A. & NOVOSHILOV, S. Y., 1970. Reduction of the noise
level in centrifugal fans by means of transition meshes. Sudostroenie 5, 2225.
PHILLIPS, O. M., 1956a. The intensity of aeolian tones. J. Fluid Mech. 1,607624.
PHILLIPS, O. M., 1956b. On the aerodynamic surface sound from a plane turbulent baudary
layer. PT'Oceedings of the Royal Society (London) A234, 327335.
PILTZ, E., 1974. Energiebedarf und Schallerzeugung bei verschiedenen Methoden der Volumenstromvariation in Ventilatoranlagen. Heizung Liiftung Haustechnik 25, 207214.
PLONER, B. & HERZ, F., 1969. New design measures to reduce siren tones caused by
centrifugal fans in rotating machines. BT'Own Boveri Revue 56, 280287.
354
POWELL, A., 1959. On the aerodynamic noise of a rigid flat plate moving at zero incidence.
Journal of the Acoustical Society of America 31, 16491653.
POWELL, A., 1960. Aerodynamic noise and the plane boundary. Journal of the Acosutical
Society of America 32, 982990.
PRANDTL, L., 1904. Uber stationare Wellen in einem Gasstrahl. Phys. Zeitschrijt 5, 599601.
PRIDMOREBROWN, D. C., 1958. Sound propagation in a fluid flowing through an attenuating duct. Journal of Fluid Mechanics 4, 393.
RAYLEIGH, .T. W. S., 1945. Theory of Sound, vol. 2. Dover Publications New York.
RECHENBERG, 1., 1973. Evolutionsstrategie. Frommann Verlag Stuttgart.
RIBNER, H. S., 1959. New theory of jetnoise generation, directionality, and spectra. The
Journal of the Acoustical Society of America 31, 245246.
RICHARDS, E. .T. & MEAD, D . .T., 1968. Noise and Acoustic Fatigue in Aeronautics . .Tohn
Wiley and Sons Ltd., London.
SAE ARP 867C, 1985. Gas turbine jet exhaust noise prediction. Aerospace Recommended
Practice, Society of Automotive Engineers.
SAVKAR, S. D., 1971. Propagation of sound in ducts with shear flow. Journal of Sound and
Vibration 19, 355372.
SCHAUB, U. W. & KRISHNAPPA, G., 1977. The stepped stator concept: Aerodynamic
and acoustic evaluation of a thrust fan. Paper presented at 4th AIAAAeroacoustics
Conference, Atlanta, Georgia, USA 771362, American Institution of Aeronautics and
Astronautics.
SCHLINKER, R. H. & BROOKS, T. F., 1982. Progress in rotor broadband noise research.
Preprint A823851D, 38th Forum of the American Helicopter Society, Anaheim, California, USA.
SCHMIDT, L., 1976. Der Einfluf3 eines 90Grad Rohrkriimmers auf die Schallemission eines
Axialventilators. Lujt und Kiiltetechnik, 287290.
SCHRECK, C., 1961. Grundlagen der hydrodynamischen Maschinen. Habilitation, Technische Universitat Berlin, Germany.
SEINER, .T. M., DASH, S. M. & WOLF, D. E., 1983. Shock noise features using the
SCIPVIS code. AIAA Paper 830705.
SELLMANN, M. & KOCH, W., 1984. Leitradloser Axialventilator, insbesondere zur
Beliiftung von Warmetauschern. European Patent Application EP 0 143 235 AI, European Patent Office, Munich, Germany.
SHANKAR, P. N., 1971. On acoustic refraction by duct shear layers.
Mechanics 47, 8191.
Journal of Fluid
SHARLAND, 1. .T., 1964. Sources of noise in axial flow fans. Journal of Sound and Vibration
1, 302322.
355
TAM, C. K. M., JACKSON, .1. A. & SEINER, .1. M., 1985. A multiplescales model of the
shockcell structure of imperfectly expanded supersonic jets. Journal of Fluid Mechanics
153, 123149.
TANAKA, S. & MURATA, S., 1975. On the partial flow rate performance of axialflow
compressor and rotating stall. Bulletin of the Japanese Society of Mechanical Engineers
18, 256271.
TERAO, M. & SEKINE, S., 1989. Induct pressure measurements to determine sound
generation, characteristic reflection and transmission factors of an air moving device in
airflow. In Proceedings Internoise}89 (Newport Beach, Ca., USA, 1989), pp. 143146.
TOURNOY, D. & BRUYERE, E., 1992. Reduction of the noise generated by industrial
centrifugal fans: Use of a quarter wavelength resonator. In Proceedings Fan Noise
(Senlis, France, 1992), Centre Technique des Industries Mecaniques (CETIM), pp. 279285.
TYLER, J. M. & SOFRIN, T. G., 1962. Axial flow compressor noise studies. Transactions
of the Society of Automotive Engineers 70, 309332.
UNRUH, .1. F. & EVERSMAN, W., 1972. The utility of the galerkin method for the acoustic
transmission in an attenuating duct. Journal of Sound and Vibration 23, 187197.
VAVRA, M. H., 1969. AeroThermodynamics and Flow in Turbomachines. John Wiley &
Sons, New York, USA.
VDr 3731 BLATT 2, 1990. Emissionskennwerte technischer Schallquellen, Ventilatoren.
VDIRichtlinie, Verein Deutscher Ingenieure, Dusseldorf, Germany.
VON HEESEN, W. & REISER, P., 1991. Einsatz eines Kanalprufstandes nach DIN 45635,
Teil 9, bei der Parameteroptimierung von Axialventilatoren. In Proceedings Ventilatoren
im industriellen Einsatz (Dusseldorf, Germany, 1991), vol. 872 of VDIBerichte, VDIVerlag, Dusseldorf, pp. 275290.
WEIDEMANN, .1., 1971. Beitrag zur Analyse der Beziehungen zwischen den akustischen und
stromungstechnischen Parametern am Beispiel geometrisch ahnlicher RadialventilatorLaufrader. Forschungsbericht DLRFB 7112, Deutsche Forschungsanstalt fur Luft und
Raumfahrt, Koln/Porz, Germany.
WIELAND, H., FUCHS, H., ACKERMANN, U., JACOBS, A. & MOHR, .1., 1989. Axialventilator. Deutsche Offenlegungsschrift DE 39 27 791 AI, Deutsches Patenamt Berlin,
Germany.
WIELAND, H., 1986. Vergleich verschiedener Systeme zum Verandern der Forderleistung
bei Radialventilatoren. In Proceedings Ventilatoren im industriellen Einsatz (Dusseldorf,
Germany, 1986), vol. 594 of VDIBerichte, VDI Verlag, Dusseldorf, pp. 267281.
WIKSTROM, B., 1964. Beitrag zur zweckmaBigen Bestimmung und Darstellung des
Ventilatorgerausches als Grundlage fur die akustische Berechnung von Luftungsanlagen.
Doctoral dissertation, Technische Universitat Berlin, Germany.
WILLE, R., 1972. Gehause fur eine radiale TurboArbeitsmaschine. Deutsche Patentanmeldung 1503624, Deutsches Patenamt Berlin, Germany.
357
WOLLHERR,
S. E., 1969. Sound radiation from a lifting rotor generated by asymmetric disk
loading. Journal of Sound and Vibration 9, 223240.
WRIGHT,
S. E., 1972. Waveguides and rotating sources. Journal of Sound and Vibration
25, 163178.
WRIGHT,
S. E., 1976. The acoustic spectrum of axial flow machines. Journal of Sound and
Vibration 45, 165223.
WRIGHT,
T., 1982. A velocity parameter for the correlation of axial fan noise. Noise Control
Engineering 19, 1725.
WRIGHT,
T., LAAN, J. N. & ZEEGMANS, H. J., 1983. Inflight acoustic measurements in the engine intake of a Fokker F28 aircraft. AIAAPaper 830677, American
Institute of Aeronautics and Astronautics, Washington DC, USA.
ZANDBERGEN,
358
Bien plus que des documents.
Découvrez tout ce que Scribd a à offrir, dont les livres et les livres audio des principaux éditeurs.
Annulez à tout moment.