Journal of Sound and Vibration (1976) 45(3), 375-403

NOISE REDUCTION IN CENTRIFUGAL FANS: A LITERATURE

SURVEY
W. NEISE
Deutsche Forschungs- und Versuchsanstalt fiir LuJi- und Raumfahrt e. V.,
Institut fiir Turbulenzforschung, Berlin, Germany

(Received 28 June 1975, and in revisedform 25 September 1975)

This paper summarizes the research work done by various experimenters during the
last fifteen years. Only those investigations are included in which an attempt was made to
reduce the generated sound by modifying the fan itself. Noise attenuation by applying any
kind of silencer is not the subject of this paper. Most of the work described was aimed at
reducing the blade passage sound, but in some cases a decrease of the broad band noise was
also obtained. Three different centrifugal impeller designs are dealt with: multi-vane
impellers with forward curved blades, impellers with backward curved blades, and impellers
with radial blades.

1. INTRODUCTION

Only very few studies on noise reduction methods in centrifugal fans are known from the
literature, despite the frequent use of this fan type in air-conditioning and ventilating equip-
ment. However, it is even more surprising that most of the successful methods have still not
been commercially exploited even though some of the fan modifications suggested do not
affect its aerodynamic performance, nor are they very costly in fan production. It is the
purpose of this paper, therefore, to summarize the research work known from the literature
during the last fifteen years, to help the industrial engineer in constructing quieter fans.
The paper is subdivided into three main parts with respect to the three fan types considered :
multi-vane fans with impellers having forward curved blades, impellers with backward
curved blades, and impellers with radial blades. In each section the relevant investigations
are described in historical sequence.
The aerodynamic sound from centrifugal fans can be formally divided into two com-
ponents.

(1) The harmonic part is caused by the interaction between the mean air flow leaving the
impeller and the fan casing. The main source region can be located at the volute cut-off which
is nearest to the rotating impeller blades. Very close to the rotor, the mean velocity profile
with respect to the circumferential direction exposes sharp minima and maxima due to the
blade wakes; see the experimental data reported by Giese and Schreck [ 11. This non-uniformity
of the velocity profile produces strong pressure fluctuations at the cut-off which result in an
effective sound radiation at the blade passage frequency and its harmonics. In a very recent
paper Piltz [2] discussed the effects of blade number and blade exit angle upon the sound
generated by the interaction between blade wakes and cut-off. When the distance from the
rotor tip is enlarged, the circumferential velocity profile becomes smoother, which clearly
shows a possible method for reducing the harmonic part of centrifugal fan noise. Increasing
the cut-off clearance was a very early recommendation for noise control; see the handbooks
by Madison [3] and Harris [4]. The first experimental results on noise reduction obtained by
this means were reported in the book on fans by Eck [5].
375
376 W. NEISE

In conventional designs of centrifugal fans the volute cut-off is parallel to the impeller
blades. Thus, the pressures along the cut-off edge are in phase. Inclining either the cut-off
or the rotor blades would result in a phase shift and local cancellation of the pressures at the
cut-off. This represents a second possible method for noise reduction.
(2) The broad band noise appears to be generated by the turbulent flow acting on the solid
surfaces of the impeller as well as of the fan casing. Another contribution comes from the
vortex shedding at the impeller blade trailing edges. The sound radiation from the turbulent
fluctuating quantities in the flow itself is presumed to be of minor importance since the mean
flow Mach numbers involved are small.
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
concentrated on the harmonic part of the spectrum only. It appears much more difficult to
influence the turbulence in the rotor and the casing itself in order to gain a lower broad band
noise level. The only successful attempt in this direction was made by Russian investigators
by mounting meshes along both the inner and the outer circumference of the impeller. The
mesh at the leading edges of the rotor blades was meant to shift the point of flow separation
towards the blade trailing edges by generating a more stabilized turbulent boundary layer.
The mesh around the impeller was to smooth the rotor outlet velocity field and create a
smaller turbulence scale.

2. FAN NOISE REDUCTION FOR IMPELLERS WITH FORWARD CURVED BLADES

2.1. INCLINED ROTOR BLADES; EMBLETON [6]

Embleton measured properties of two double inlet impellers of a common household

furnace, a schematic of which is shown in Figure 1. The standard form had blades parallel
to the axis, whereas the blades of the modified rotor were inclined by about 20” such that
they opposed the inward flow in the axial direction. The fan casing was of the volute type. More
detailed information about the dimensions of the impeller and the casing, about the clearance
between impeller and cut-off, and about the measuring method applied, are missing in
reference [6].
DIrectIon of rotation

(a) 42 blades (bl

Figure 1. Schematic of Embleton’s [6] double inlet blowers with forward curved blades. (a) Standard
impeller; (b) modified impeller.

Figure 2 shows the change in aerodynamic performance due to the blade sloping. At all
rotor speeds, both the electrical power input and the volume flow were decreased. Above
600 rev/min, however, any given volume of air was delivered with less power by the impeller
with inclined blades, as indicated by the dashed lines in Figure 2.
At constant rotor speeds in the range from 400 to 800 rev/min, Embleton reported a sound
 NOISE REDUCTION IN CENTRIFUGAL FANS 377

400 600 000

n (rev /mid

Figure 2. Aerodynamic performance of the impellers shown in Figure 1. At constant speed, the level of
the blade passage frequency is reduced by 12 dB when the blades are sloped. Embleton [6]. -O-O-, Standard
impeller; -O-O-, modified impeller.

pressure reduction of 9 to 12 dB at both the blade passage frequency and at half this frequency.
However, no comparison of the sound levels was made for a constant volume flow, for which
the speed of the modified impeller had to be higher than that of the standard impeller. Thus,
one has to assume that the broad band noise-which mostly determines the overall noise from
this type of fan-was increased by the blade sloping, when a given aerodynamic load is
considered. The main disadvantage of Embleton’s impeller modification is its being difficult
to manufacture economically.

2.2. SINGLE INLET ROTOR: INCLINED CUT-OFF EDGE; DOUBLE INLET ROTOR: ANGULAR DIS-
PLACEMENT BETWEEN THE BLADES OF THE TWO ROTOR HALVES; LYONS AND PLATTER [7]

Like Embleton, and at about the same time, Lyons and Platter tried to reduce the blade
passage tone by local cancellation of the pressure fluctuations excited at the volute cut-off.
They investigated a single inlet impeller as well as a double inlet impeller. Measurements
were made in a discharge duct having a sound absorbent end section to reduce axial standing
waves.
The outer diameter of the single inlet rotor was 101.6 mm, and its axial length was 63.5 mm.
Unfortunately, Lyons and Platter did not report more details of the geometric configuration.
The main experimental results concerning the pure tone level was described as follows:
“Of the geometries studied, the maximum reduction, 10 dB, was measured by skewing a
sharp cut-off close to the wheel across two blade-to-blade spacings. There was no loss in
blower efficiency”. The inclined cut-off edge was found to be most effective in reducing
noise when the fan was operating near the point of maximum efficiency. In the case of double
inlet fans, local cancellation of the pressures excited along the cut-off edge can be achieved
by an angular displacement between the two rotor halves, so that the blades of one rotor half
lie half-way between those of the other rotor half. This was first mentioned by Eck [5]. A
maximum reduction of the blade passage tone of 10 dB, with no loss in the fan efficiency, was
reported by Lyons and Platter. Again, no detailed description of the geometrical configuration
studied was given.
Both the noise reduction methods suggested by Lyons and Platter look rather promising
and acceptable for industrial mass production. The second, slightly modified, can also be
used for single inlet fans, as Yeow [8] pointed out, by simply splitting the cut-off into two
halves. The two halves are then separated in the circumferential direction by half a blade-to-
378 W. NEISE

blade spacing. The effect of local pressure cancellation would be identical to that observed by
Lyons and Platter. However, Yeow did not measure the actual noise reduction effect of the
cut-off splitting.
The effectiveness of the noise control procedures described above should be investigated
in more detail to get further information about the noise reduction at different rotor speeds and
different operating points. Another open question is whether or not inclining either the
blades or the cut-off yields a lower pure tone level only at very small clearances between
impeller tip and cut-off.

2.3. MESHES AT THE LEADING EDGES AND AT THE TRAILING EDGES OF THE ROTOR BLADES;
PETROV, KHOROSHEV AND NOVOSHILOV [9-111

The noise reduction studies described in the following were concerned with impellers which
were not exactly of the usual multi-vane type having forward curved blades : only the intake
parts of the blades were forward curved, the trailing edges were terminated radially; see
Figure 3. The number of blades was Z = 24. (A list of notation is given in the Appendix.)

_ITrans~l~on meshes

Figure 3. Schematic of a rotor with transition meshes at both leading and trailing edges of the blades;
Petrov el al. [9-l 11.

The basic idea of reducing noise by means of transition meshes was to improve the mean
flow and the turbulence in the impeller ducts as well as outside the rotor. The effect of a
mesh, placed near the blade leading edges, was explained by the formation of a small-scale
turbulent flow behind the mesh. The blade boundary layer became turbulent, and thereby
the point of flow separation was shifted towards the trailing edge. The second transition
mesh, around the outer circumference of the rotor, was to smooth the outlet velocity field
and to further reduce the turbulence scale. Figure 3 shows a rotor with transition meshes
mounted at both the leading and the trailing edges.
The total sound power emitted by a fan with transition meshes, it was argued, may remain
the same or even increase slightly, but the noise spectrum will be shifted to a higher frequency
range. The use of the meshes increased the drag of the rotor as a whole, which resulted in a
reduced pressure head. In order to compensate for this factor, the rotor speed had to be
increased which also led to a slightly increased noise level. However, there was still a gain
in noise reduction, but the fan efficiency decreased considerably.
In Figure 4 the measured noise reduction is shown for two different types of meshes. The
finer mesh (I.3 x I.3 x 0.25 mm) reduced the sound pressure in the frequency range from
100 Hz to 10 kHz by 4 to 9 dB. The penalty to pay was that the fan efficiency dropped from
r] = O-626 to q = 0.545. The use of the coarse mesh (5.4 x 5.4 x 1.4 mm) yielded significantly
lower noise levels between 100 and 2000 Hz, whereas the high frequency components above
2000 Hz were intensified. The decrease of the fan efficiency was slightly greater than before,
r] = 0.536.
The curves in Figure 4 clearly show that the mesh dimensions, free cross-section and wire
diameter, are important parameters with respect to the noise reduction. Although no exact
 NOISE REDUCTION IN CENTRIFUGAL FANS 379

f (Hz)
Figure 4. Effect of transition meshes on the noise from a centrifugal fan; Petrov et al. [9]. D = 325 mm,
n = 2870 rev/min, Z = 24.

Mesh tvne p fm3/h) dP (kulm’) II

-A-A- 4100 146 0626
-o-o- 1.3 x 1.3 x 0.25 mm 4100 138 0.545
---- 5.4 x 5.4 x 1.4 mm 4100 135 0.536

data were available, which would allow a selection of the meshes at the stage of the fan
design, general recommendations were given : the free mesh area should be 50 to 60 % of the
total area. The wire diameter, d, should be chosen such that the frequency of vortex shedding
behind the mesh,f= O-2v/d, lies outside the interesting frequency range; v is the flow velocity
before the mesh. However, the optimum mesh type has to be found by experiments. In order
to increase the effectiveness of the meshes, it was recommended to cut windows of half the
blade-to-blade spacing into the mesh at the blades’ leading edges, so that only the flow along
the blade suction side has to pass through the mesh. The actual effect of this modification was
not measured.
Generally, Petrov et al. stated, the transition meshes reduced noise most effectively when
the mean flow conditions in the impeller ducts were bad. Thus, the use of meshes was primarily
recommended for impellers with large regions of flow separation.
In Figure 5 the noise reduction to be expected, when transition meshes are used, is plotted
against the velocity ratio vM/vz; vMis the maximum local velocity at the blade suction side,
and vz is “the flow rate component of the intake velocity”. Large values of vMindicate strong
flow separation, because the flow has to decelerate after passing the point of maximum
velocity. Figure 5 shows that the possible noise reduction is large when flow separation occurs,
which was already mentioned before.

0
3 4 6 6

Figure 5. Estimate of noise reduction when transition meshes are used. Us is the maximum local velocity
along the blade suction side, vz is the flow rate component of the intake velocity; Petrov et al. [l 11.

It is the merit of Petrov et al. to have found a simple method for reducing the broad band
noise from centrifugal fans over a wide frequency range. The disadvantage of the transition
meshes lies in the decreased fan efficiency, at least as far as impellers with forward curved
380 W. NEISE

blades are concerned. Petrov et ul. also applied turbulence creating meshes to centrifugal
fans having rotors with radial blades, as will be described in section 4.4; for two examples of
this impeller type, the fan efficiency remained almost unaffected by the presence of the meshes.

3. FAN NOISE REDUCTION FOR IMPELLERS WITH BACKWARD CURVED BLADES

3.1. INCREASING THE CUT-OFF CLEARANCE; ECK [5]

Experimental results concerning fan noise reduction by increasing the cut-off clearance
were first described by Regenscheit in the book by Eck. When the distance between impeller
tip and cut-off was varied from 3 to 70 mm, the following relation between the “loudness”, L,
and the cut-off distance, Ar, was found :
L = 100(Ar/mm)-0’063 phon. (1)
The value L = 100 phon was the loudness of the fan at Ar = 1 mm. Further details concerning
fan dimensions, operating point, rotor speed, or change in aerodynamic fan performance
were not reported.

3.2. INCLINED IMPELLER BLADES, SLOTS IN THE IMPELLER BLADES; EMBLETON [6]

The noise reduction effect of inclining either the impeller blades or the cut-off edge of the

/
volute has already been discussed in previous sections. As in the case of fans with forward
curved blades, Embleton applied blade slope to the centrifugal fan of a domestic vacuum
cleaner. A schematic of the blade sloping is given in Figure 6.

25.4
-__

?6!!FEt
,_Jjde slope y
\G

‘I I= 12

Figure 6. Principle of blade sloping applied to an unshrouded impeller of a domestic vacuum cleaner;
Embleton [6].
As the slope of the blades was increased, the capacity decreased, when a constant pressure
was considered; see Figure 7. The loss in volume flow was small only for the blade slope of 12”,
which yielded the best fan efficiency of all impellers investigated, as is shown in Figure 8.
I I I I

I 1 1 i
0 0.1 O-2 0.3 0.4 0.5

tihl3/s)

Figure 7. Effect of blade sloping upon the aerodynamic fan performance; Embleton [6]. Arctan t/b = 52.6”.
 NOISE REDUCTION IN CENTRIFUGAL FANS 381
The noise reduction effect was reported as “This optimum slope is associated, under all
operating conditions, with a slight reduction in the broad band aerodynamic noise, and
2 to 3 dB in the blade-frequency spectrum”. The changes in noise, capacity and efficiency
described above were reported to be typical of the results for rotor speeds between 4000 and
9000 rev/min.
0.6_ I I I I

0.3 I 1 1 ,
0 0.2 0.4 0.3 0.8

w,.,
Figure 8. Effect of blade sloping upon the fan efficiency. Broad band noise was “slightly” reduced, the level
of the blade passage frequency was reduced by 12 dB when the blade slope was 12”; Embleton [6].

Embleton also made an attempt to improve the flow in the impeller ducts itself by cutting
slots into the blades, which is schematically shown in Figure 9. The basic idea was to lead high-
energy flow from the pressure side of the blade to the suction side in order to displace the

Figure 9. Principle of slots in impeller blades suggested by Embleton [6].

li (m3/s)
Figure 10. Effect ofslotsupon the blower efficiency. The curves are for an industrial blower having standard
blades, and with 1 or with 3 slots cut into each blade, parallel to the axis of rotation. The broad band noise was
reduced by 1 to 2 dB; Embleton [6]. Slots per blade: -, 0; ---, 1; ....., 3; n = 1500 rev/min.
382 W. NEISE

point of boundary layer separation further downstream. However, the measured effect shown
in Figure 10 is only small. At low flow rates, the fan efficiency was marginally increased.
The broad band noise was reduced by 1 to 2 dB.
Both noise reduction methods tried by Embleton did not change the noise output of the
fans significantly, and the aerodynamic fan performance became worse in both cases. From
this point of view, a practical application of both methods seems rather unlikely. On the other
hand, inclined rotor blades are not necessarily accompanied by an unsatisfactory fan
efficiency. The development of high performance fans led to impellers with inclined blades,
and the fan efficiency was raised near 90%; see the paper by Fluegel [12] and the book by
Eck [5]. A lower noise level associated with extremely high efficiency may make it worthwhile
to manufacture impellers of rather complex geometry. To the author’s knowledge, however,
no sound measurement data concerning the high performance impellers have been published
yet.

3.3. VARIATION OF CUT-OFF RADIUS AND CUT-OFF CLEARANCE; LEIDEL [13]

Leidel studied the effect of the cut-off configuration on both the aerodynamic performance
and the noise of a centrifugal fan; a schematic of it is shown in Figure 11. The volute of the
fan casing was adjustable so that the cut-off radius as well as the clearance between cut-off
and impeller tip could be varied independently. The cut-off clearance, Ar, was changed within
the range from Ar/R = 0.016 to 0.4, the cut-off radius, r, from r/R = 0.01 to O-2; R was the
impeller radius. The width of the fan casing, 125 mm, as well as the pitch angle, 7*2”, and the
looping angle, 360”, of the logarithmic spiral were kept constant. That means that the whole
fan casing was enlarged when the cut-off clearance was increased. The impeller blades were
shaped as an arc of a circle.

Figure 11. Schematic of the fan investigated by Leidel [13].

Sound measurements were made in an anechoically terminated discharge duct for only
one impeller tip speed of 3150 rev/min. The pressure reflection coefficient measured on the
duct axis was smaller than O-2 in the interesting frequency range above 200 Hz. The inner duct
diameter was 152 mm which allowed plane wave propagation up to 13 10 Hz. The operating
point of the fan was controlled by a throttle mounted in the inlet duct.

3.3. I. Aerodynamic performance of the fan

Figure 12 shows the influence of the cut-off clearance upon the aerodynamic performance
of the fan: the cut-off radius was kept constant, r/R = O-125. Fan efficiency, v, and pressure
 NOISE REDUCTION IN CENTRIFUGAL FANS 383

Figure 12. Influence of cut-off clearance upon the aerodynamic fan performance; Leidel[13]. r/R = 0.125;
Ar/R = 0, 0.016; A, 0.025; ?,0.060;
? 0, 0.100; V, 0.160; D, 0.250; @, 0400.

$, are plotted
coefficient, as functions of the flow coefficient cp. The non-dimensional par-
ameters were defined by
cp = 41+Dz U, (2)
II/= 2AP,lpU2, (3)
v = AP, PiPmcc,,. (4)
Here ri, AP, and Pmech meant volume flow, total pressure head and the mechanical input
power of the fan, respectively. D and CJ were impeller diameter and impeller tip speed, the
density of air was denoted by p. Both the flow coefficient and the fan efficiency were not
significantly affected by the variation of the cut-off clearance. The optimum efficiency was
changed by about 6%; its maximum value was measured at At-/R = 0.25.
When the cut-off radius was varied, no influence upon the fan aerodynamic performance
could be measured, which is shown in Figure 13 for three different cut-off clearances.

3.3.2. Blade passage tone

The cut-off clearance strongly influenced the sound power level at the blade passage
frequency, which is plotted versus the flow coefficient in Figure 14. The cut-off clearance was
varied in the range from Ar/R = 0.016 to 0.4, whereas the cut-off radius was kept constant,
r/R = 0.125. The sound level was highest for high flow rates and for small cut-off clearances;
it monotonically decreased when the cut-off was displaced from the impeller tip. The maximum
reduction measured was 20 dB; the reduction at the optimum operating point was 18 dB.
The influence of the cut-off radius upon the level of the blade passage frequency was much
less significant than that of the cut-off clearance; see Figure 15. There the maximum noise
reduction was about 6 dB for each of the three different cut-off clearances.
The lowest sound power level at the blade passage frequency was achieved by combining
the cut-off clearance of Ar/R = 0.25 with the cut-off radius of r/R = 0.2; compare the bottom
graph in Figure 15 with Figure 14. This minimum noise radiation is associated with the
maximum fan efficiency as was shown in the previous section. Thus, no penalty had to be
paid for gaining a lower sound level. The only drawback of Leidel’s results can be seen in the
somewhat enlarged fan casing.
Figure 16 shows that the second and third harmonic of the blade passage frequency were
reduced even more than the fundamental. This is clear, because the velocity profile close to
the impeller tip exhibits distinct minima and maxima which cause impulse-like time variation
of the pressures at the cut-off; the resultant far-field sound spectrum has many harmonics of
384 W. NEISE
0.8

0.8

O-6

1 0.4

0.2

(c)/ j
I
0 0.02 0.04 O-06 O-08 0 40 0.12 0.14 0.16 0-m 0.20

P
Figure 13. Influence of cut-off radius upon the aerodynamic fan performance, for different cut-off Clear-
ances; Leidel [13]. (a) Ar/R = 0.06; r/R = ;7,0.08; 0,0.125; 0,0.20; (b) At-/R = 0.10; r/R = A, 0.05; ?,0.08;
?
0,0.125; 0, 0.20; (c) Ar/R = 0.25; r/R = 0, 0.01; A, 0.05; 0,0.125; 0,0.20.

equal importance. With increasing radial distance from the impeller tip, the velocity profile
and, therefore, the time variation of the pressures at the cut-off become smoother. Thus, the
higher harmonics are attenuated more than the blade passage frequency itself.
 NOISE REDUCTION IN CENTRIFUGAL FANS 385

Figure 14. Influence of cut-off clearance upon the level of the blade passage frequency; Leidel [13]. r/R =
0.125, n = 3150 rev/min. Ar/R = 0, 0.016; A, 0.025; ?,0.060;
? 0, 0.100; V, 0.160; D, @2X);*, 0.400.

75

70
0 0.02 0.04 0.06 O-06 0.10 0.12 0.14

Figure 15. Influence ofcut-off radius upon the level of the blade passage frequency; Leidel[l3]. (a) Au/R =
0.06; (b) At-/R = 0.10;-(c) Ar/R = 0.25. n = 3150 rev/min. r/R = v, 0.01; A, 0.05; 0, 0.08; 0, 0.125; o, 020.
386 W. NEISE

/‘- ,,Ar

Level of blode passagefrequency

002 0.04 0.06 0.06 0.10 0.12 0.14 0.16

Figure 16. Difference in the levels of the blade passage frequency and its 2nd and 3rd harmonic; Leidel[13].
rJR = 0.125, n = 3150 rev/min. Ar/R = O, 0.016; A, 0.025; fl, 0.06: 0, 0.10.

3.3.3. Broad band noise

Enlarging the cut-off clearance and the cut-off radius was much less effective in reducing
the broad band noise than in reducing the harmonic component. Figures 17 and 18 show the
sound power level within three different frequency bands as a function of cut-off clearance
or cut-off radius, respectively. Measurements were made at four operating points. The point
of maximum efficiency was indicated by cpoP,. There was a considerable scattering of the
experimental data; however, the general tendency came out quite clearly: minimum broad
band noise occurred at ArlR = 0.25 and r/R = 0.2. That was the cut-off configuration of
maximum fan efficiency and minimum harmonic noise, as was shown before.
The data shown in Figures 17 and 18 were measured at constant fan operating points.
This means that the flow through the impeller ducts remained unchanged when the cut-off
configuration was modified. Nevertheless, an effect upon the broad band noise was measured.
From that one can conclude that the random noise originated at the volute casing, including
the cut-off, at least at small cut-off clearances. Obviously, not only the mean velocity profile
became smoother, at larger distances from the impeller tip, but also the turbulence intensity
decreased. Additionally, the mean velocities in the larger fan casing are lower.
Finally, Leidel’s most important results are summarized in Figure 19. The blade passage
frequency remained dominant over the whole range of cut-off clearances. The second har-
monic of the blade passage frequency was reduced most significantly; above Ar/R = 0.06 its
level fell below the random noise. Since these noise reductions are associated with a maximum
fan efficiency, and since the somewhat enlarged fan casing is not supposed to cause signifi-
cantly higher costs of production, no reasons opposing an industrial application can be seen
here.
 NOISE REDUCTION IN CENTRIFUGAL FANS 387

65

60

66

60

65. I I 1 /
o-01 0.02 0.05 0.1 0.2 0.4

Figure 17. Influence of cut-off clearance upon the broad band noise; Leidel [I 31.a, 0.1 pop,; QO.35 q~,,,
;
0,065 post;o,qop,.r/R= O.l25,n= 3150 rev/min.

3.4. VARIATION OF THE INNER IMPEDANCE OF THE FAN; WOLLHERR [14]

Wollherr’s study does not fit exactly into the scope of the present paper, because the noise
reduction was not obtained by influencing the aerodynamic sound source itself, but by
mismatching the impedances of fan casing and duct system to that of the aerodynamic
sources. The fan investigated had an impeller of 400 mm outer diameter with 6 backward
curved blades. Figure 20 shows a schematic of Wollherr’s experimental rig. Sound measure-
ments were made in an anechoic inlet duct. The pressure reflection coefficient was smaller
than 0.15 for frequencies above 80 Hz.
The reduction of the noise radiated into the inlet duct was obtained by mounting two ducts
of different lengths at the fan outlet. By this means, the inner impedance of the fan as a whole.
including the outlet duct, was changed. In the case of the 1.5 m long outlet duct, the level of
the blade passage frequency was indistinguishable from the random noise, whereas it was
about 17 dB higher when the O-9 m long duct was mounted; see Figure 21. This effect will only
occur at a fixed frequency; the practical application of this method is therefore restricted to
cases where the fan operates at only one rotational speed. It should be mentioned that the
above results were only a by-product of Wollherr’s investigations on the acoustic source
characteristics of centrifugal fans.
388 W. NEISE

66
0 0.m 0.H 0.20

r/R

Figure 18. Influence of cut-off radius upon the broad band noise; Leidel [13]. :., 0.10 q,,,; 0, 0.35 qopI;
c, 0.65 q~ ; o, tp,,,. Ar/R = 0.25, n = 3150 rev/min.

3.5. VARIATION OF THE CUT-OFF CLEARANCE; SMITH, O’MALLEY AND PHELPS [151
Similarly to Leidel’s [13] study, Smith ef al. tried to reduce the blade passage tone by
increasing the cut-off clearance. Differently to Leidel’s work, however, the size of the volute
casing remained unaltered when the cut-off clearance was varied. This was accomplished by

TABLE 1
Influence of cut-off clearance upon the aero-
dynamic fan performance at near-peak ef-
jiciency operation and at 1250 revjmin; Smith
et al. [I51
P
(l& W/s) tlr
0.16 42.03 45.67 14.02 0.760
0.14 41.91 45.67 14.25 0.770
0.12 41.91 45.54 14.02 0.750
0.10 41.91 45.54 14.02 0.745
0.08 41.86 45.34 13.78 0.730
0.06 41.78 45.09 13.40 0.705
 NOISE REDUCTION IN CENTRIFUGAL FANS * 389

90

a5

a0

T 65

a0

65

60
0.01 0.02 0.05 0.1 0.2 0.4

Ar/R
Figure 19. Influence of cut-off clearance upon the most important components of fan noise; Leidel [13].
(a) papt; ON065 %pt; (4 0.35 v)~,,(.o, Blade passage frequency; ‘>, 2 x blade passage frequency; V, broad band,
180_1120Hz;r/R=0~125;n=3150rev/min.

probe

: termination

Figure 20. Schematic of Wollherr’s [14] test set-up. D = 400 mm; Z = 6.

390 W. NEISE

too 200 300 400 500

f(Hz)

Figure 21. Sound pressure spectra measured in the inlet duct; Wollherr [14]. n = 1400 rev/min; cut-off
clearance Ar/R = 0.35; ---, outlet duct 0.9 m; -, outlet duct 1.5 m.

adjusting a separate cut-off plate to the desired distance from the periphery of the impeller.
The impeller of the fan used in the experiments was 464 mm in diameter and had 12 backward
curved blades. Sound measurements were made in a reverberation chamber.

TABLE 2
Injuence of cut-08 clearance upon the aero-
dynamic fan performance at near-peak ef-
ficiency operation and at 1500 revjmin; Smith
et al. [15]
ri
Ar/R (k$‘) (k$h) (m3/s) t1r
0.16 65.41 71.32 17.93 0.825
0.14 65.28 71.12 17.84 0.820
0.12 65.10 70.86 17.75 0.810
0.10 64.90 70.82 17.93 0.815
0.08 64.90 70.43 17.13 0.780
0.06 64.97 70.10 16.71 0.750

Tables 1 and 2 describe the change in aerodynamic fan performance when the cut-off
clearance was varied from Ar/R = 0.06 to 0.16. At both rotor speeds, 1250 and 1500 rev/min,
the fan was operating near the maximum efficiency. Both the volume flow, ri, and the fan
efficiency based on the total pressure head AP,, were increased as the cut-off clearance was
enlarged, which qualitatively agrees very well with the results obtained by Leidel. Smith et al.
did not give reasons why the maximum efficiencies measured for different rotor speeds were
not the same. One can assume only that for some reasons the operating point of the fan was
shifted as the rotor speed varied.
The influence of the cut-off clearance on fan noise is presented in Figures 22 and 23, for
rotor speeds of 1250 rpm and 1500 rpm, respectively. The solid curves are for a cut-off
clearance of Ar/R = 0.06, and the dashed curves are for Ar/R = 0.16. All measurements were
made at both near-peak and off-peak efficiency operation. At both rotor speeds, the levels
of the blade passage frequency and the second harmonic were reduced by 8 to 11 dB, when the
fan was operating near its maximum efficiency. The harmonic components could not be
 NOISE REDUCTION IN CENTRIFUGAL FANS 391

(al
ao-

(b)

60-

I
?O-

60. 250 looo

Figure 22. Influence of cut-off clearance upon third octave sound pressure spectra at 1250 rev/min; Smith
et al. [15]. (a) Near-peak efficiency; (b) off-peak efficiency.

(b)
60
,~Ar/R=Oo6
-_

60 ’ ’ j ’
250 500 1000 2000

f (Hz)

Figure 23. Influence of cut-off clearance upon third octave sound pressure spectra at 1500 rev/min; Smith
et al. [15]. (a) Near-peak efficiency; (b) off-peak efficiency.
392 W. NEISE

distinguished from the random noise after increasing the cut-off clearance, at least not by
means of a third octave frequency analysis. The upper graphs in Figures 22 and 23 show that
the broad band noise was also affected by the cut-off clearance, but the reduction achieved is
not very significant.
At off-peak efficiency operation, see the lower graphs in Figures 22 and 23, the effect of
enlarging the cut-off clearance is much less pronounced. The maximum reduction at the blade
passage frequency was smaller than 4 dB.
Another interesting result from Figures 22 and 23 is that, for the large cut-off clearance,
the levels of the harmonic component as well as of the random component are lowest at the
operating point of maximum efficiency, which clearly shows the importance of proper fan
selection.
The results of Smith et al. confirm those reported by Leidel, and they furthermore show that
noise reduction can be achieved without enlarging the whole fan casing. Later on the results
of both studies will be directly compared in order to find out whether or not one of the two
methods is more effective than the other.

4. FAN NOISE REDUCTION FOR IMPELLERS WITH RADIAL BLADES

One way to reduce the level of the blade passage tone which has not yet been mentioned here
is an irregular spacing of the impeller blades. By this means, however, the total sound energy
generated is not reduced but distributed over a broader frequency band. Huebner [16]
investigated irregular blade spacing on impellers with radial blades used for cooling electrical
machines. The harmonic noise component was found masked by the random component.
The aerodynamic efficiency was reported to have remained the same for the type of centrifugal
fans used in electric motors. Further details were not provided in reference [16]. Other authors
[17], however, mentioned that the efficiency of an impeller with irregular blade spacing was
worse than with regular spacing. Furthermore the rotor imbalance caused by the irregularly
spaced blades is often a great problem.

4.1. OPTIMUM CUT-OFF CLEARANCE AND INCLINED CUT-OFF; EMBLETON [6]

The fan used in Embleton’s investigation had a normal volute casing and an impeller of
254 mm diameter and 76 mm axial width. Eight radial blades extended from a 25.4 mm
diameter hub to the periphery of the rotor. Measurements were made under free discharge
conditions at an impeller speed of 2400 rev/min. The cut-off clearance was varied only in a
small range from Ar/R = 0 to 0.08. It was not explicitly said that the size of the volute remained
the same for all cut-off positions but it seems likely that this was true.
Embleton’s experimental results are shown in Figure 24. The volume flow was slightly
decreased as the cut-off clearance became larger, but the efficiency was found constant. In
contrast to the results reported in references [5], [13], [15] and [I 71, no monotonic decrease
of the overall sound pressure was measured but a minimum was found at At-/R = 0.034.
The sound pressure at the blade passage frequency and at its harmonics showed the same
characteristic behaviour as the overall sound. The minima of odd harmonics occurred at
ArlR = 0.02, while those of even harmonics occurred at ArlR = 0.034. Sound pressure
measurements were made at a distance of 152 mm from the cut-off.
When the cut-off edge was inclined so as to span two blades simultaneously, there was no
minimum of sound pressure any more but a monotonic decrease as the cut-off clearance was
enlarged. The noise reduction effect of the sloped cut-off was 10 to 12 dB “at small clearances,
when the blade passage frequency is strongly emitted”.
The minima in sound pressure level measured with the cut-off edge parallel to the impeller
blades are considered of no universal meaning, here. It is believed that the minima were caused
 NOISE REDUCTION IN CENTRIFUGAL FANS 393
0.70

Figure 24. Changes of mechanical and acoustical properties at free discharge when the cut-off clearance
is varied. D = 254 mm; b = 76 mm; Z= 8; n = 2400 rev/min; Embleton [6].

by secondary effects : e.g., by the directionality of the sound source depending on the position
of the cut-off plate. Generally, one has to expect a monotonic decrease of the radiated
sound as the cut-off clearance increases, as was measured with the sloped cut-off and as was
described in references [5], [13], [15] and [I 71.

4.2. INCLINED CUT-OFF; KHOROSHEV AND PETROV [lo, 181

Khoroshev and Petrov gave a formula to determine the angle of inclination of the cut-
off, u:
tan tl = (r - 2r)/b, (5)
where t is the blade spacing, r is the cut-off radius, and b is the impeller width. This expression
was obtained by estimating the unsteady aerodynamic force caused by the radial velocity, V,
leaving the impeller; see Figure 25. The condition of minimum noise radiation led to the

Figure 25. Schematic of the fan studied by Khoroshev and Petrov 1181.
394 W. NEISE

above equation. The cut-off radius, r, is normally small compared to the blade spacing, t.
Therefore equation (5) proves what was earlier found by Embleton [6]: namely, that the
cut-off should encompass one blade spacing.
Figure 26 shows the experimental results for a fan, the dimensions and the rotor speed of
which were not reported. The level of the blade passage frequency was dramatically reduced
over the whole range of the pressure head-volume flow characteristic by inclining the cut-off
by 70”. No information was given about the cut-off clearance. Maximum noise reduction
occurred at off-design operation which contrasts with the results obtained by increasing the
cut-off clearance; see Figures 14, 15, 19, 22 and 23. inclining the cut-off also gained a sub-
stantially lower broad band noise level which is apparent from the spectra in Figure 26(b).

70 -

40’ ’ ““’
100 1000 IC x)0

f (Hz)

300 400 500 600

p (d/h1

(0)

Figure 26. Influence of angle of inclination of the cut-off upon (a) the blade passage frequency level and the
aerodynamic fan performance and (b) the noise spectrum. -, Straight cut-off; ---, cut-off inclined by 70”;
Khoroshev and Petrov [18].

On the other hand, the optimum fan efficiency decreased from O-77 to 0.72. This was referred
to flow losses in the outlet duct owing to the steep increase in cross-section beyond the
inclined cut-off.

4.3. VARIATION OF THE DISTANCE BETWEEN IMPELLER TIP AND ADJACENT RIBS; INCLINED RIBS;
PLONER AND HERZ [17]
The paper by Ploner and Herz was addressed to the reduction of harmonic sound radiated
from a fan used for cooling electrical machines. In this sort of application, the impeller
discharges the heated air to the outside through openings in an axisymmetric housing. Planer
and Herz simulated the housing by wooden ribs of different shape. Figure 27 shows details
of the experimental rig and of the single intake impeller with two rows of blades. The 21
blades of each row were in line with each other and so performed acoustically as those of a
single-row impeller. Sound measurements were made in an anechoic room. The microphone
was traversed horizontally along 90 measuring positions, the sound pressure readings at
which were averaged.
The operating point of the impeller remained nearly unchanged as different single rib
models were studied at various distances from the impeller tip, 6/R= 0.03 to 0.09. In all
cases the fan was working above the point of optimum efficiency at about cp = 0.16. The flow
 NOISE REDUCTION IN CENTRIFUGAL FANS 395

Measuring path MIcrophone

y.- - -.- T -. . ___.-_cc1

% ,/

Figure 27. Experimental setup used by Planer and Herz [17]. 2 = 21.

coefficient was defined as in equation (2). The mechanical power consumption and the
efficiency of the impeller were not measured.
Increasing the rib clearance resulted in a monotonic decrease of the level of the blade
passage frequency, as is apparent from Figure 28, which also shows a strong influence of the
rib shape. When the rib of trapezoidal cross-section was inclined by 50” against the axis of
rotation, a significant reduction in sound level was measured at small rib clearances. The
angle of 50” corresponded to a rib encompassing two blades simultaneously. In this case, the
sound level was only slightly reduced by enlarging the rib clearance. This implies that the
effects of rib inclination and rib clearance are not additive. Once the noise is lowered by
either one of the two methods, almost no further reduction seems possible. It is believed here
that this statement also holds for the cut-off configuration of volute casings.

___--_---_----_-------
Without rib

Figure 28. Sound pressure level of the blade passage frequency as a function of rib clearance. n = 2500
rev/min; 9 g 0.16; Ploner and Herz [17].

The effect of rib inclination is shown more thoroughly in Figure 29. The rib used for this
study was made of sheet metal which could be bent in order to maintain a constant rib
clearance along the impeller width for all slope angles. Figure 29 clearly indicates that
inclining the rib, or the cut-off of a volute casing, is effective only if one blade spacing is
396 W. NEISE

-10
I
0 IO 20 30 40 50 60
Q (degrees)
Figure 29. Influence of the angle of inclination of a sheet metal rib (300 x 40 x 4 mm) upon the sound level
at the blade passage frequency. n = 2500 rev/min ;t-p ;6/R= 0.0465 ; Planer and Herz [ 171.
r 0.16

encompassed. Unfortunately, one cannot directly compare the results presented in Figures
28 and 29 with those of Figure 26, because Khoroshev and Petrov [18] did not report the
magnitude of the cut-off clearance used in their experiments. The distance between impeller
tip and cut-off, however, strongly influences the effectiveness of the cut-off inclination which
is obvious from Figure 28.

110 , I ,

1t
loo

90

60

70-

60- 4

m
0
-50 ’ I
b..T
110 , I I I

/ (bl I

601 ’ I
DO0 1500 2000 2500 3ooo 3500
n (rev/n-h)
Figure 30. Influence of (a) rotor speed, and (b) rib size upon the level of the blade passage frequency.
6/R= 0.0465; Planer and Herz [17]. Dimensions in parentheses are “length x length x height” in mm.
 NOISE REDUCTION IN CENTRIFUGAL FANS 397
The noise reduction effects of rib inclination and rib shape were found to be independent of
rotor speed, as is shown in Figure 30(a). Tests were also made with ribs of different size, the
same geometrical shape being maintained. The results of these measurements are presented
in Figure 30(b). There was only a small effect but the highest noise level was measured for the
smallest size, presumably owing to the lesser blockage of air and, therefore, higher velocity
at the outlet of the impeller.

1000 2000 dB(A)

f (Hz1
Figure 31. Third octave spectrum of sound pressure for various rib shapes when four equally spaced ribs
are mounted at the impeller periphery. 6/R = 0.0465; n = 2500 rev/min; Ploner and Herz [17].

Figure 31 shows spectra of sound pressure when four ribs were used instead of only one
as in the previous experiments. The blade passage frequency, fi = 875 Hz, mainly appeared
in the third octave band centred at 800 Hz, but obviously influenced the 1000 Hz band to
some extent, too. The absolute noise levels were naturally higher than with one rib only;
however, the differences between the different rib models were the same as in Figure 28.

80-

Optimum efficiency
i D
70 1 I I 1
0.2 0.4 0.6 0.8 I.0 l-2 1.4
ti (mVs 1
Figure 32. Influence of the operating point of the fan upon the level of the blade passage frequency when
four equally spaced ribs are mounted at the impeller periphery. 6/R = 00465; n = 2500 rev/min; Ploner and
Herz [17].

Finally, the influence of the fan operating point was examined by throttling the inlet duct.
The ends of the curves at the right-hand side of Figure 32 corresponded to the unthrottled
duct; these measurement data were already shown in Figure 3 1. The main result of Figure 32
is that inclining the ribs is most effective at off-design operation of the fan which agrees very
well with the effect of sloping the cut-off of a volute casing; see Figure 26(a).
398 W. NEISE

4.4. MESHES AT THE LEADING EDGES AND THE TRAILING EDGES OF THE ROTOR BLADES; PETROV,
KHOROSHEV AND NOVOSHILOV [9-111

f (Hz)
90

90
I !
(d) /
60

50 63 I
125 250 500 1000 2000 4000
6000
f (Hz)

Figure 33. Several examples for the effect of transition meshes, mounted at both inlet and outlet of the
impeller, upon the sound radiation from centrifugal fans with radial bladed impellers; Petrov et al. [9-l 11.
$tte; 3000 m’/h; AP= 165 kp/m’; D = 300 mm; n = 2870 rev/min; -o-, without meshes; --0--, with

(b) D = 325 mm; n = 3870 rev/min; Z= 12.

3(m3/h) AP (kplm’) I
-a--, without meshes 4100 131 0.59
--o--, with meshes 4100 134 0.585
Cc)
Mesh type AP (kp/m2) tf
-a- no mesh 264 0.67
--o-- fs=0~5,d=1~4mm.inlet \ 233 0.56
1 s = 0.575, d = 0.6 mm, outlet (
Cd) v= 2870 m3/h
Mesh type AP (kplm’) v
-a- no mesh 342 0.69
--o-- s = 0.575, d = @6 mm 344 0.662
 NOISE REDUCTION IN CENTRIFUGAL FANS 399
The general aspects of transition meshes were already mentioned in section 2.3, and they
also apply to impellers with radial blades. Several examples for the effect of transition meshes
on noise are presented in Figure 33. A comparison of Figures 33(a) and 33(b) shows that in
some cases meshes, mounted at both inlet and outlet of the rotor, did not reduce the sound
radiation significantly, whereas they were very effective in other cases. As mentioned before,
the application of transition meshes were mainly recommended for impellers with un-
satisfactory flow conditions.
The additional drag caused by the turbulence creating grids normally resulted in a reduced
fan efficiency; see Figure 33(c). For some fan types, however, the impairment of aerodynamic
fan performance is only marginal, two examples of which are shown in Figures 33(b) and
33(d).
Altogether, the effect of transition meshes seems almost unpredictable, and one has to
select them by means of experiments, separately for any particular fan type. Yet, experiments
were only done with fans which had a low efficiency already before the meshes were mounted.
It is possible that high performance impellers would suffer a considerable reduction in
efficiency when grids are used. On the other hand, the great advantage of transition meshes
is that they reduce the noise level over a broad frequency range, and not only that of single
frequency components.
In order to compensate for the loss in pressure head and efficiency, Petrov et al. suggested
reducing the cut-off clearance. The resultant increase in the harmonic noise, it was argued,
could be effectively reduced by inclining the cut-off. However, the success of this procedure
is doubtful, because the fan efficiency reaches its maximum at cut-off clearances of Ar/R = 0.16
to 0.25, as was shown in references [ 131and [ 151and an inclined cut-off again may reduce the
fan efficiency owing to the complex geometry of the outlet duct; see reference [18].

5. CONCLUSIONS
In this paper the following methods for reducing the noise from centrifugal fans have been
described.
(1) Increasing the distance between impeller tip and cut-off
A comparison of the noise reduction achieved by enlarging the cut-off clearance is made in
Figure 34. In the upper graph, the difference in blade passage frequency level is plotted versus
the normalised cut-off clearance or rib clearance, respectively. The reference point was chosen
at At-/R = 0.06, because this distance lies within the range covered by all studies involved. The
lower graph presents the difference in blade passage frequency level as a function of the
absolute value of the cut-off clearance. This is to include the results from reference [5] where
the impeller diameter was not reported. Embleton’s [6] measurements do not agree with the
other results, and it was stated before that the minimum of the curve is supposed to be
caused by some unknown secondary effects. The decrease in blade passage frequency level
obtained by enlarging the cut-off clearance depends on the shape of the cut-off or the rib,
respectively, which is apparent from the two curves quoted from Ploner and Herz [17]. The
results given by Leidel [13] and Smith et al. [ 151agree very well with each other. From these
it seems not necessary to enlarge the whole fan casing together with the cut-off clearance, as
was the case in Leidel’s work, since the same effect was measured by Smith et al. with a casing
of constant size. On the other hand, there was rarely an effect on the broad band noise in
reference [15] whereas Leidel found a reduction of 3 to 5 dB (see Figures 17 and 18), pre-
sumably owing to the lower flow velocity in the larger casing. This appears to make the en-
larged casing worthwhile.
In both references [13] and [15] the fan efficiency was not reduced by the increase in cut-off
400

_g -20 I I 1 1 I I 1 I I ,1 I
tk 0.01 0.02 0.03 0.04 0.06 0.060.1 0.2 0.3 0.4 0.6 0.6 /
Ar/R, S/R
z
0”
0
d IO- I I I /Ill,, 1 I /II 1
=
a
E (b)

Plow and Herz w

-20 I I I,,,,, I I 11, I,

2 3 4 5 6 78910 20 30 40 50 60 60 100

Ar_8 (mm)
Figure 34. Effect of enlarging cut-off clearance or rib clearance, Ar and 6, respectively: (a) Ar/R, 6/R
and (b) Ar, 6 as abscissa. Ploner and Herz, D = 430 mm, Z= 21 (radial), n = 2500 rev/min, a, > v.,~; Smith
et al., D = 464 mm, Z= 12 (backwards), n = 1500 rev/min, v),,~~;Embleton, D = 254 mm, 2 = 8 (radial),
n = 2400 rev/min, q > pop, ; Leidel, D = 199 mm, Z= 6 (backwards), n = 3150 rev/min, popt.

clearance. In reference [13] the maximum efficiency was associated with minimum fan noise
at a cut-off clearance of ArlR = 0.25 and at a cut-off radius of r/R = 0.2. Thus, shape and
position of the cut-off are two main design features with respect to centrifugal fan noise.

(2) Angle of inclination between impeller blades and cut-ofSedge

This can be achieved by sloping either the blades or the cut-off. The first method, however,
has the drawback of being difficult to manufacture. Embleton [6] studied inclined blades for
multi-vane impellers as well as for impellers with backward curved blades. In the first case
a reduction of 12 dB in blade passage frequency level was found for constant rotor speed,
but the modified impeller had to run faster for a given volume flow which increased the broad
band noise and, thereby, possibly the overall noise, too. In the case of the backward curved
blades, there was only a small reduction in the harmonic sound, 2 to 3 dB, for a slope angle
of 12” which also resulted in a slightly reduced pressure head.
An inclined cut-off is easier to produce, and it was also found more effective in reducing
noise. The cut-off has to be inclined so as to span at least two blades simultaneously. The
noise reduction obtainable by this means strongly depends on the distance between rotor
and cut-off which was shown by Ploner and Herz [17] for the case of an inclined rib; see
 NOISE REDUCTION IN CENTRIFUGAL FANS 401

Figure 28. Lyons and Platter [7] measured a 10 dB reduction at the blade passage frequency
for a sharp cut-off placed close to a multi-vane impeller. This was measured for the fan
operating at optimum efficiency; at other operating points, inclining the cut-off was found
to be less effective. The fan efficiency remained unchanged. The slope angle of the cut-off
required to encompass one blade spacing is small in case of multi-vane impellers, but it
becomes large in the case of impellers with 6 to 12 blades. This results in a rather complex
geometry of the fan outlet which causes additional losses. Khoroshev and Petrov [18]
measured a decrease in efficiency from 0.77 to 0.72 for a cut-off slope angle of 70”. The
reduction in blade passage noise, however, was considerable (see Figure 26) and also the
broad band noise was lowered by 8 to 10 dB. In contrast with Lyons and Platter’s result, but
in agreement with the data reported by Ploner and Herz [17], the inclined cut-off reduced
the harmonic sound most strongly at off-design operation.
The angle of inclination of the cut-off can be reduced if the impeller blades are inclined,
too, but in the opposite direction. This could be applied to the high performance fans described
in references [5] and [12] which already have inclined blades.
Both increasing the cut-off clearance and inclining the cut-off edge affect the blade passage
noise and to some extent the broad band noise, too; however, their effects are not additive.
Once a substantial noise reduction is achieved by either one of the two methods, the other
method will be much less effective.

(3) Staggering of blades of double inlet blowers and double row impellers
Multi-vane double inlet rotors normally consist of two separately produced halves mounted
together. If the blades of one rotor half lie between the blades of the other half, the blade
passage noise can be reduced by 8 to 10 dB, as was shown by Lyons and Platter [7]. It is
believed here that a similar effect can be achieved by staggering the blades of double row
impellers, but no studies on this subject have been published yet.

(4) Transition meshes at the leading and the trailing edges of the rotor blades
This noise control procedure was investigated for rotors with forward curved blades and
for radial bladed impellers. The use of meshes was most effective for fans with bad mean flow
conditions in the impeller ducts; the noise was reduced over a wide frequency range. For nearly
all cases, however, the fan performance was impaired, although the efficiencies of the fans
without meshes were not very high. It is expected that transition meshes would cause a much
larger drop in efficiency in the case of impellers of higher aerodynamic quality.

(5) Mismatch between the acoustic impedances offan and ducted system
Wollherr’s [14] experimental results showed that manipulation of the fan outlet duct
resulted in a 17 dB variation of the sound radiated into the inlet duct. Likewise, one can
influence the sound pressure in the outlet duct by changing the geometry of the fan inlet.
However, this procedure will not be a standard one for a given fan because the effect of
any modification of, say, the fan inlet also depends on the acoustic impedance of the outlet
duct system. Therefore one has to find the optimum configuration separately for each case.
The noise reduction effect only occurs at fixed frequencies which restricts the application of
this method to fans running at constant speed.

The present paper has shown that several noise reduction methods have been established
for centrifugal fans. There is still much work to do in this field because most of the studies
described were made with only one fan type and for a few rotor speeds and operating points.
The effect of combining two or three different noise reduction methods in one fan has not
been examined either.
The most important task of future research work is to reduce the random component of
centrifugal fan noise without losing fan efficiency. The first step in this direction would be to
402 W. NElSE

locate the main source regions of broad band noise. There are some hints that the rotating
impeller is not the dominant source, but that a significant contribution comes from the fan
casing: some reduction in broad band noise was gained by modifying the casing while the
flow through the impeller remained unchanged; see references [13] and [ 181. In a discussion
of fan sound generation as a function of fan operating point, Bommes [19] showed that for
small values of the flow coefficient the sound radiated from stationary parts of the fan tends
to dominate, whereas the sound due to the flow past the rotating impeller surfaces is more
important at high flow coefficients. The action of both these source regions was assumed to
cause the typical minimum of sound pressure as a function of flow coefficient; an example
for this behaviour is depicted here in Figure 26(a). Finally, Moreland [20] recently reported
that the noise from an unthrottled multi-vane blower with casing was some 10 dB higher than
from the impeller alone.

ACKNOWLEDGMENTS

This paper was written during a Visiting Research Fellowship at the Institute of Sound and
Vibration Research, Southampton, under the auspices of Professor P. E. Doak. The author
wishes to thank Mr L. R. A. Melton, librarian at the University of Southampton, for running
a computer search for relevant literature. Thanks are also due to Dr B. D. Mugridge for his
helpful discussions and comments on the paper.

REFERENCES
1. K. GIESE and C. SCHRECK1968 Technica 17, 2339-2344; 1969 Technica 18, 103-108; 495-508.
Hitzdrahtmessungen an rotierenden Laufrldern von Stromungsmaschinen.
2. E. PILTZ 1975 Zeitschrif fur Heizung Luftung Klimatechnik Haustechnik 26, 165-170. Ge-
rauscherzeugung von Radialventilatoren bei unterschiedlicher Laufradbeschaufelung.
3. R. D. MADISON1949 Fan Engineering. New York: Buffalo Forge Company, Buffalo, 5th edition.
4. C. M. HARRIS1957 Handbook of Noise Control. New York: McGraw-Hill Book Company, Inc.
5. B. ECK 1962 Ventiiatoren. Berlin: Springer Verlag, 4. Auflage; 1973 Fans. Oxford: Pergamon
Press Limited, 1st English edition.
6. T. F. W. EMBLETON1963 Journal of the Acoustical Society of America 35,700-705. Experimental
study of noise reduction in centrifugal blowers.
7. L. A. LYONS and S. PLATTER 1963 Journal of the Acoustical Society of America 35, 1455-1456.
Effect of cutoff configuration on pure tones generated by small centrifugal blowers.
8. K. W. YEOW 1966 Institute of Sound and Vibration Research, Southampton, Memorandum No.
143. Centrifugal fan noise research: A brief survey of previous literature.
9. Yu. I. PETROV,G. A. KHOROSHEV and S. YA. NOVOSHILOV 1970 Sudostroenie 5,22-25. Reduction
of the noise level in centrifugal fans by means of transition meshes. (In Russian.) (Translated
into English: NAVSTIC-Translation 3322, 1972.)
10. G. A. KHOROSHEVand Yu. I. PETROV1971 Proceedings of the 7th International Congress on
Acoustics, Budapest, Paper 19N16. Some new methods of fan noise reduction.
11. Yu. I. PETROVand G. A. KHOROSHEV 1971 Russian Engineering Journal51,42244. Improving the
noise level of centrifugal fans.
12. G. FLUEGEL1954 Gliickauf 90,919-922. Kreiselgeblhe hohen Wirkungsgrades.
13. W. LEIDEL1969 DLR-FB 69-16. Einfluss von Zungenabstand und Zungenradius auf Kennlinie
und Gerlusch eines Radialventilators.
14. H. WOLLHERR1973 Ph.D. Dissertation, Technische UniversittitBerlin. Akustische Untersuchungen
an Radialventilatoren unter Verwendung der Vierpoltheorie.
15. W. A. SMITH,J. K. O’MALLEYand A. H. PHELPS1974 American SocietyofHeating, Refrigeration
and Air Conditioning Engineers Transactions Part II 80,45-51. Reducing blade passage noise in
centrifugal fans.
16. G. HUEBNER1961 Siemenszeitschrift 35, 11 l-121. Gerguschminderung bei grossen elektrischen
Maschinen.
17. B. PLONERand F. HERZ 1969 Brown Boveri Revue 56,280-287. New design measures to reduce
siren tones caused by centrifugal fans in rotating machines.
 NOISE REDUCTION IN CENTRIFUGAL FANS 403

18. G. A. KHOROSHEV and Yu. I. PETROV 1965 Sudostroenie 3, 15-17. Decreasing the noise level of
centrifugal ventilators (in Russian). (Translated into English : Joint Publication Research Service
31116, Z-T-65-31614, 1965.)
19. L. BOMMES 1967 74th Annual Meeting of the American Society of Heating, Refrigeratingand Air-
Conditioning Engineers, Minneapolis, Minnesota, June 26-28, 1967, Reprint No. 2057. Effect of
fan configuration on the variation of fan sound generation with operating condition.
20. J. B. MORELAND 1974 Journalof Soundand Vibration 36,191-205. Housing effects on centrifugal
blower noise.

APPENDIX: LIST OF NOTATION

b impeller width
d diameter of mesh wire
D impeller diameter
f frequency
L sound pressure level
L o) sound pressure level at the blade passage frequency, averaged from readings at 90 measuring
positions (reference [17])
n impeller revolutions per minute
PE electrical input power of the fan
AP, static pressure head
AP, total pressure head
r cut-off radius
Ar clearance between cut-off and impeller periphery
R = D/2, impeller radius
s ratio of free to total area of transition mesh
t spacing of impeller blades
U impeller tip speed
v velocity
P volume flow
Z blade number
a angle of inclination of cut-off
y angle of blade sloping
6 clearance between rib and impeller periphery
q fan efficiency
p air density
v, flow coefficient
v pressure coefficient