CEAS Aeronautical Journal

https://doi.org/10.1007/s13272-019-00372-7

ORIGINAL PAPER

Aircraft noise generation and assessment

Aeroacoustic wind tunnel design

J. Pereira Gomes1   · A. Bergmann1 · H. Holthusen2

Received: 1 December 2017 / Revised: 2 November 2018 / Accepted: 11 February 2019

© Deutsches Zentrum für Luft- und Raumfahrt e.V. 2019

Abstract
Further progress in airframe noise research including noise prediction and noise reduction solutions depends on the avail-
ability of aeroacoustic wind tunnel test facilities with superior aerodynamic and acoustic quality. The demand for aeroacoustic
wind tunnels with extremely low background noise and pressure fluctuations, yet with a relevant test section cross-section
area and flow velocity, increased significantly over the last decade. In the future, this demand will continue to grow to cope
with the challenging noise reduction objectives for aviation noise defined for the years to come. The present text is focused
on the state-of-the-art aeroacoustic wind tunnels available today and their design. The design guidelines discussed here
assume a classic aerodynamic wind tunnel as a baseline. Therefore, the present text is addressed to both those who are
interested in the design of a completely new aeroacoustic wind tunnel as well as those interested in the acoustic upgrade of
an existing aerodynamic wind tunnel. As a direct consequence of the multi-disciplinary nature of this complex task, and the
multitude of solutions and design tools that are required to complete it, the approach followed here subdivides the design of
the aeroacoustic wind tunnel into four main sections: wind tunnel airline circuit (includes the first and second airline cross
legs), drive unit and anechoic plenum. While the design approach for the airline circuit and the drive unit is strongly based
on coupled numerical solutions of CFD and acoustic solvers, the design of the acoustic plenum gives more emphasis to
in situ observations and to experimental results. The main sections of the aeroacoustic wind tunnel and their best design are
discussed separately in this contribution.

Keywords  Aeroacoustic · Aerodynamic · Wind tunnel · Acoustics

List of symbols Cp,RMS Pulsating coefficient
A Area, m 2 d Duct diameter, m
A0 Reference area ( = 1 m2 ), m 2 f Frequency, Hz
Ap Wall partition area, m 2 h Brightness per histogram level
c Speed of sound, m/s 𝜅 Ratio of specific heats
L Length, m
l Duct length, m
This paper is part of a Special Issue on Aircraft Noise Generation lef Effective duct length, m
and Assessment. Lm Overall brightness level
Lp Sound pressure level
* J. Pereira Gomes
jorge.gomes@dnw.aero Lp,OSPL Overall sound pressure level
𝛥f Frequency bandwidth, Hz
A. Bergmann
andreas.bergmann@dnw.aero 𝛥Lp Acoustic damping
𝛥ploss Pressure loss
H. Holthusen
hermann.holthusen@dnw.aero 𝜆 Wavelength, m
ṁ Mass flow rate, kg/s
1
German-Dutch Wind Tunnels (DNW), Lilienthalplatz 7, Ma Mach number
38108 Braunschweig, Germany p Pressure, N/m2
2
German-Dutch Wind Tunnels (DNW), Voorsterweg 31, p0 Reference pressure ( = 20 𝜇Pa), N/m2
8316 PR Marknesse, The Netherlands

13
Vol.:(0123456789)
 J. P. Gomes et al.

ptot Total pressure, N/m2 range [12]. More silent aerofoils, such as those designed for
r Distance, m wind turbines, require extremely low background noise lev-
r0 Offset of the acoustic centre along the measure- els in the test section and the surrounding chamber to maxi-
ment path, m mize the signal-to-noise ratio of the experimental results. In
RI Apparent intensity sound reduction index
′ contrast, aircraft models in high-lift configuration are typi-
𝜌 Flow density, kg/m3 cally much noisier and consequently more forgiving of the
U Flow velocity, m/s background noise around them.
V Volume, m  3 The current contribution approaches the aeroacoustic
wind tunnel design under the assumption that the reader is
Abbreviations
familiar with the design of classic aerodynamic wind tun-
ARC​ Active resonance control
nels. This way, instead of covering the design of an aeroa-
atm Atmospheric
coustic wind tunnel from scratch, the present contribution
BBN Broadband noise
discusses the major guidelines for providing an aerodynamic
BPF Blade passing frequency
wind tunnel with excellent acoustic capabilities. It discusses
CFD Computational fluid dynamics
the upgrade of a classic design of an aerodynamic wind tun-
DLR German Aerospace Centre
nel into an optimized aeroacoustic wind tunnel design, using
DNW German–Dutch Wind Tunnels foundation of DLR
the low-speed wind tunnel DNW-NWB as example.
and NLR
In the approach suggested here, the wind tunnel design
in Inlet
is subdivided into four major sections that can be designed
inc Incident
independently (see Fig. 1). For the design of both cross legs
LLF Large low-speed facility
of the airline circuit, an interactive optimization approach
NLR Netherlands Aerospace Centre
that combines CFD tools and a modern calculation method
NWB Low speed wind tunnel Braunschweig
for the propagation of acoustic waves inside the airline is
OSPL Overall sound pressure level
suggested. The solution for the flow field is provided by CFD
out Outlet
solvers that produce solutions of the Navier–Stokes equa-
SPL Sound pressure level
tions. A photon-based ray tracing technique is suggested
trs Transmitted
as an acoustic analogy to simulate the propagation of the
acoustic waves inside the same circuit and their interaction
with the solid boundaries. The focus of this two-point design
1 Introduction optimization process is to minimize the pressure loss inside
the airline circuit at the same time it maximises the acoustic
The design process of an aeroacoustic wind tunnel, with attenuation of the noise generated inside it, like that gener-
rare exceptions, is very complex, as are the multi-physical ated by the wind tunnel fan and heat exchanger.
disciplines required to guarantee a successful design. The The acoustics of the wind tunnel fan are optimized using
complexity of such design is further increased, because it is the inverse cutoff design [3, 4]. The main purpose of this
absolutely essential that a good acoustically oriented wind design is to minimize the broadband noise generated by the
tunnel should provide superior aerodynamic characteristics fan. To mitigate the higher harmonics of the blade passing
as well. In this respect, an ideal aeroacoustic wind tunnel frequency generated by the fan in the inverse cutoff design, a
must offer the possibility of testing in a closed test section to systematic evaluation of the impact of the most relevant fan
guarantee the best representation of the flow field, and also parameters on the different noise contributions is performed.
in an open or partially open test section to measure the com- The impact is predicted using computational tools based on
plete acoustic far field generated by the test model in free- semi-analytical acoustic models. The prediction of noise is
field conditions. From the mere point of view of the acoustic determined by semi-empirical scaling laws that describe the
design, the requirements for the background noise level in evolution of the sound pressure level of a noise source as a
the wind tunnel test section and its spectral distribution are function of flow and geometry parameters. The changes in
primarily defined by the size of the wind tunnel and by the the aerodynamic performance introduced by those design
type of acoustic tests to be performed. In large wind tunnels modifications are verified with turbomachinery-oriented
dedicated to full-scale model testing, the sound frequency CFD solvers.
range below 4 kHz is of special interest. For smaller models, As far as the test section is concerned, several investiga-
such as 1 / 10 scaled models, the sound frequency range tions have shown that it is not yet possible to reliably meas-
between 10 and 40 kHz is of more importance. Because ure absolute noise pressure levels in closed test sections
this is a very common model scale tested in mid-sized wind with hard walls. This is true for measurements of discrete
tunnels, their design should be optimized to this frequency tonal noise and broadband noise contributions alike. The

13
Aircraft noise generation and assessment

Fig. 1  Sketch of a world-class
aeroacoustic wind tun-
nel: Low speed wind tunnel
Braunschweig (NWB) of the
German–Dutch Wind Tunnels
Foundation (DNW) (dark blue
areas correspond to acoustically
treated surfaces)

best alternatives for producing quantitative acoustic results field and acoustic assumptions. Then, in each iteration, the
are still open-jet test sections and 3/4 open test sections sur- individual design of the sections is modified towards the
rounded by large anechoic plenums. The design and later optimum point and a new set of boundary conditions is cal-
assessment of the acoustic characteristics of the anechoic culated. These boundary conditions will be used as input
plenum is performed in both wind-off and wind-on condi- in the subsequent iteration. The optimization process ends
tions and makes use of room acoustics principles. Room when both flow field and acoustic boundary conditions of
acoustics is a subdiscipline of the acoustics branch that stud- all adjacent sections match and when all the global require-
ies the sound behaviour in enclosed spaces, such as concert ments of the aeroacoustic wind tunnel design are fulfilled.
halls and recording studios, theatres, railway stations, and
office buildings, to name a few. It is also known as archi-
tectural acoustics or building acoustics. The design aims 2 Airline
to achieve two main objectives: to isolate the test section
against external noise so that the resulting internal noise The design of a wind tunnel airline with superior acoustic
does not invalidate the measurements and to reproduce performance must be focused on the reduction of the flow-
with high fidelity the acoustic free-field conditions for the induced noise caused by high local flow velocities. Thus,
acoustic measurements. Finally, the background noise and as a first measure, the diffuser after the test section must be
pressure fluctuation requirements in wind-on conditions are designed as large as possible to maximize the cross-section
also discussed. area of the wind tunnel at the location of both the first and
The final integration of the four sections into a complete second corners. Due to the stringent limitations on the maxi-
aeroacoustic wind tunnel design is performed iteratively mum opening angle of a diffusor to avoid flow separation,
through the boundary conditions of the individual sections. this is normally achieved by installing a longer diffusor.
Initially, a basic design of the four wind tunnel sections The advantage of this design is twofold. First, it reduces
is completed separately based on preliminary macro flow the local flow velocity in the first and second corners and

13
 J. P. Gomes et al.

consequently the flow-induced noise produced by the flow uniformity, separation, etc.), while minimizing the overall
around the turning vanes. Furthermore, the turning vanes pressure loss. A small pressure loss reduces the required
are located at a greater distance from the test section, which power to drive the tunnel which in turn reduces the noise
increases further the acoustic damping and decreases reflec- generated by the drive unit. The high acoustic transmission
tions between the vanes and the test section. Both effects loss minimizes the noise produced inside the tunnel, in par-
increase the signal-to-noise ratio in the test section. Second, ticular by the fan and the heat exchanger, to reach the test
larger wind tunnel corners give more freedom to the design section.
and optimization of the turning vanes without the penalty The strong interaction between a solid structure with the
of higher pressure loss. In particular, it allows the turning flowing fluid in which it is immersed, as in the case of the
vanes to be thicker and acoustically treated. This type of turning vanes, or by which it is surrounded, as in the case of
vane often has an aerodynamic shape and is made from per- the wind tunnel inner walls, gives rise to a wide spectrum of
forated metal plates with back-filled glass wool used in typi- multi-physical phenomena that results in a strong coupling
cal sound absorbers. Figure 2 depicts the first cross leg of between the aerodynamics and the acoustics. Therefore,
the same wind tunnel test case before and after the acoustic the strategy to optimise only the acoustic properties of the
upgrade. Whereas the sketch on the left-hand side repre- airline by steering it using boundary conditions defined by
sents a pure aerodynamically designed airline, the one on the aerodynamic considerations, or vice versa, does not lead to
right-hand side shows a design optimization based on both a global optimum. This constitutes a closed problem. The
aerodynamic and acoustic considerations. The figure clearly current contribution suggests an iterative optimisation of
shows the size difference between the turning vanes of an both objectives using high-fidelity CFD tools and a pho-
aeroacoustic and a pure aerodynamic wind tunnel. In fact, ton-based ray tracing technique for the acoustic layout. The
the turning vanes in the second corner act as a bent splitter optimization chain starts by generating the airline geometry
silencer and provide a counterpart to the annular acoustic based on the actual designing parameters. After a second
muffler of the same length, situated downstream of the wind step, in which the geometry is prepared and the numerical
tunnel fan. In the example shown in Fig. 1, the acoustic muf- grid is generated, the calculation process is separated into
fler consists of an acoustically treated, very long fan tail cone two trails. The first one deals with the determination of the
surrounded by acoustically treated walls. This design is only aerodynamic quality of the airline circuit. There is a strong
possible after a substantial increase of the airline diameter. aerodynamic and acoustic coupling between the turning
The figures suggest also that it is easier to implement acous- vanes in the first and second corner of the airline because
tically treated turning vanes in large-scale wind tunnels than of the local low velocity of the flow and the short distance
in smaller ones. The enlargement of the airline cross-section between them. To take this into account, the first and the
area near the second corner is also favourable for an opti- second corner must be simulated together in the CFD and
mized positioning of the heat exchanger (see Sect. 3). acoustic code as well. The same is applicable to the third
The turning vanes contribute significantly to the acous- and fourth corner of the wind tunnel airline. The second trail
tic characteristics of a wind tunnel. For that reason, their determines the acoustic quality of the airline circuit.
design must be carried out together with the design of the The evaluation of the aerodynamic quality is performed
internal contour of the wind tunnel airline in such a way as by a CFD solver. Based on the given geometry, which is
to simultaneously maximize both the acoustic transmission discretized by a mesh, a flow solver produces a numerical
loss through the airline and the aerodynamic characteris- solution of the averaged Navier–Stokes equations describ-
tics of the tunnel. The latter means maintaining a high flow ing the flow physics. The numerical simulations are usu-
quality through the entire airline (regarding flow directivity, ally performed for a 2D cut of the airline to reduce the

Fig. 2  Comparison between an
aerodynamic optimized wind
tunnel airline design (left) and
a design optimized for both
acoustic and aerodynamic appli-
cations (right) (for clarity, only
the section of the airline located
between the collector and the
drive unit is represented)

13
Aircraft noise generation and assessment

computational time. This simplification does not consider a pixel into the scene, recursively generating reflected and
the typical 3D boundary layer found in cascade flows. The refracted rays.
aerodynamic objective is to minimize the pressure loss in the Both wave-based and geometric modelling can be mapped
flow between the inlet and outlet section of the numerical to the propagation of acoustic energy [7]. This implies that
domain. It is defined as follows: the propagation of sound waves can be simplified to a ray-
( ) ( ) based approach. For middle and higher frequencies, this
ptot out − ptot in
𝛥ploss = . (1) assumption is true to a certain extent, because diffraction and
ṁ interference processes are neglected. In contrast, wave-based
The total pressure loss and consequently the aerodynamic effects become prominent at lower frequencies and ray-based
objective, must consider the mass flow rate because small methods can no longer be applied. The frequency range in
changes in pressure loss are followed by changes in the which diffraction and other wave-based interferences are
mass flow through the simulation domain. This coupling important is a function of several parameters. However, it is
effect is neutralized when introducing the mass flow rate a common practice to define a limit based only on the com-
into the equation, if a linear dependency between the total parison between the wavelength of the propagating wave and
pressure loss and total mass flow is assumed. If compress- the size of the solid with which it interacts. Some authors
ibility effects are introduced through the Mach number, the state that the diffraction-free limit is lowered to the object of
total energy of the flow becomes the sum of the mechanical the size of the wavelength. Others require an order of mag-
and the thermal energy components. In fact, it can be argued nitude difference between the wavelength and the size of the
that the Mach number squared represents the ratio of the body before geometric optics could become useful. Bertram
mechanical (kinetic) energy and the thermal component of et al. [2] for example, define the lower frequency threshold
the total energy. Following this path, the total pressure can for the applicability of ray tracing techniques as follows:
be written as: c c
) 𝜅
f = ≈ .
𝜆 L (3)
𝜅 − 1 2 𝜅−1
(
ptot =p 1+ Ma (2)
2 Using these criteria, the lowest frequency that can be simu-
lated with the geometric modelling technique is approxi-
for the inlet and outlet sections of the numerical domain.
mately 100 Hz for a turning vane with an arc length equal
A light analogy is implemented to access the acoustic
to 3 m. For a mid-size aeroacoustic wind tunnel focused on
behaviour of the wind tunnel airline circuit including the
tests with scaled models, this is a plausible vane size; the
turning vanes. Sound and light are in general of a different
frequency value is about six times smaller than the lowest
nature and perceived by two distinct senses [27, 28]. How-
frequency of interest. Thus, the simulation of the transmis-
ever, in terms of propagation and interaction with objects,
sion loss along the airline can be performed with a good
they present some similarities. They both propagate as
degree of confidence with the geometric approach.
waves, sound waves being longitudinal waves of varying
In conclusion, to determine the acoustic transmission
pressure. The propagation of light energy and its interac-
loss in an aeroacoustic wind tunnel airline, a global illu-
tion with objects can be measured and described using tech-
mination model is needed. Besides direct reflections of the
niques of radiometry from which global illumination models
sound waves on the turning vanes surface and internal walls,
used in computer graphics are derived. There are two well-
a significant amount of diffuse reflections takes place inside
documented models [13, 28]: wave-based modelling and
the tunnel airline. Ray tracing can only handle specular
geometric modelling. The so-called wave-based modelling
reflections, refractions and direct illumination (local illumi-
is based on the numerical solution of the wave equation, or
nation model). Effects such as caustics and indirect illumi-
Helmholtz–Kirchhoff equation. Therefore, it corresponds
nation (e.g., from diffuse reflections) cannot be computed,
to a global illumination model. The numerical simulation
because it assumes perfect specular materials [13]. To take
based on this model provides very accurate results, but the
diffuse reflections into account, the ray tracing technique has
complexity and the computation time increase drastically
been extended in recent years with Monte Carlo methods
while increasing the highest frequency to be resolved [2].
in which rays are distributed stochastically to account for
The geometric modelling is a local illumination model. It
all light paths. Another viable idea is to combine radiosity
is based on the calculation of the path of waves or particles
(wave-based modelling) with ray tracing into a hybrid tech-
through an environment with regions of varying propaga-
nique. However, this technique needs a fine representation
tion velocity, absorption characteristics and reflecting sur-
of the scene to render, which makes it very computationally
faces. This can be simulated with only a small amount of
demanding. A more modern approach makes use of a photon
computational resources using a ray tracing technique. Ray
map. Photon mapping is a different approach than hybrid
tracing solves the problem by repeatedly emitting rays from
techniques [13, 14]. The idea is to change the representation

13
 J. P. Gomes et al.

of the illumination. Instead of tightly coupling lighting infor- airline and consequently a better acoustic quality. To keep
mation with the geometry, the information is stored in a the method simple, the image plane is discretized by N moni-
separate independent data structure, a so-called photon map. toring points, or microphones in the acoustic analogy. The
The photon map is constructed from photons emitted from brightness histograms of all microphone positions are typi-
the light sources and traced through the domain. It contains cally calculated. This histogram reflects the distribution of
information about all photon hits and this information can the n brightness levels of a picture. For each microphone
be used to efficiently render the object in a similar manner position, the overall brightness is calculated based on the
as radiosity is used in hybrid techniques. The decoupling of logarithmic sum assuming incoherent sound sources using
the photon map from the geometry is a significant advantage the following equation:
that both simplifies the representation and makes it possible
n
to use the structure to represent lighting in very complex ∑ hj

objects. The combination of photon mapping and Monte

Lmi = 10 log 10 10 . (4)
j=1
Carlo ray tracing-based rendering algorithms not only results
in a more efficient algorithm but also represents a global This value, in the acoustic analogy, represents the sound
illumination model. pressure level measured by each microphone. The logarith-
The photon-based ray tracing techniques are normally mic summation takes into account the physical effect that
applied to an airline in such a way that the physical bound- the sound sources with the highest levels make a bigger
ary conditions of the numerical domain coincide with the contribution to the overall sound pressure level. In the end,
geometry of the first or second cross leg of the wind tunnel. the overall sound pressure level is computed adding the con-
The inlet section of the domain is defined in the core of the tribution of all N microphones again using the logarithmic
airline circuit at the wind tunnel drive unit and the outlet summation as follows:
section is defined close to the wind tunnel test section. The
technique uses a planar light source at the inlet plane of N
∑ Lmi

the acoustic domain as a representation of the noise gener- Lp ≡ Lm = 10 log 10 10 . (5)

i=1
ated by the fan. In detail, a number of small, punctual light
sources distributed over the plane are used to create a uni- This value defines the acoustic objective. At the end of both
directional source of light over the complete cross-section trails, the aerodynamic and acoustic results for the current
of the tunnel. This is comparable with the noise radiated by geometry are known as numerical values and constitute a
the fan into the airline. The interaction between light and the design point. After that, a new set of designing parameters
solid boundaries is characterized by the surface properties of is generated and a new calculation run is started.
the walls and by the type of reflection they generate. Thus, it Figure 3 shows a typical two-point design optimization
must be distinguished between hard-walled and soft-walled map for an airline design of an aeroacoustic wind tunnel.
surfaces. A hard-walled surface means the complete sound
energy is reflected and nothing is absorbed. With a soft-
walled surface, part of the sound energy is dissipated into
4
heat in the material and therefore the reflected sound wave is
weakened. Further on, depending on the surface of the wall,
there are two kinds of reflections: specular and diffuse reflec- Design point
Pareto front
tions. Specular reflection is the light reflected from a smooth 3
surface at a definite angle. Diffuse reflection is the reflection
Acoustic objective

of light such that the incident ray on the surface is scattered

at all directions. In the end, the different reflection mecha-
2
nisms are considered by the acoustic ray tracing approach
using two parameters: the fraction of soft- and hard-walled
surface and fraction of diffuse and specular reflection. For
acoustically treated walls, like turning vanes made from 1

perforated metal plates with back-filled glass wool, it is a

non-trivial task to determine both values.
The acoustic quality of the actual configuration is a func- 0
0 1 2 3
tion of the brightness of the image constructed in the outlet Aerodynamic objective
plane of the acoustic domain. For the same inlet light source
intensity, a less bright image at the outlet of the acoustic Fig. 3  Typical two-point design optimization map for an aeroacoustic
domain means a higher acoustic transition loss through the wind tunnel airline design

13
Aircraft noise generation and assessment

The result of each calculation run is represented by a circle mode of the duct. When it manifests, tonal noise arises at
in the plot. The horizontal axis represents the aerodynamic the resonance frequencies.
objective. A lower aerodynamic objective is associated with Any heat exchanger also introduces a significant pres-
a smaller pressure loss and represents a better aerodynamic sure loss in the airline of the wind tunnel. For this reason,
quality of the airline. The acoustic objective is shown on the it is often located in the settling chamber to take advan-
vertical line. As for the aerodynamics, the lower the acoustic tage of the large cross-section area and low flow velocity
objective the higher the acoustic quality of the airline. It in this part of the wind tunnel, despite the decrease in cool-
means a higher acoustic damping through the airline legs and ing efficiency while reducing the flow velocity. Therefore,
better absorption of the internal noises generated in the wind an acoustic interaction between the heat exchanger and the
tunnel. The two-point design optimization map is followed other components that are in a settling chamber, such as the
by the determination of the Pareto front, or optimization line. flow straightener and screens, must be expected and con-
Figure 3 also highlights the strong dependent and competi- sidered along with the self-generated heat exchanger noise.
tive character of an aeroacoustic design. Improvements on Both experimental tests and in situ inflow measurements
the acoustic properties of a wind tunnel airline are associated have shown that the sound pressure level created by the flow
with a reduction of the aerodynamic qualities and vice versa. through the combined heat exchanger, flow straightener and
This is especially valid for mid-size and small wind tunnels, screens is significantly higher than the noise radiated by the
because the thickness of the turning vanes cannot be scaled. flow passing through the individual elements [12]. Depend-
The final design point of the airline is found along the Pareto ing on the frequency range, the sound pressure level due to
front and should correspond to the best compromise solution interaction effects can be 10 dB greater, or more, than the
that fulfils the global aerodynamic and acoustic requirements sound pressure level generated in the screens alone. Due to
of the aeroacoustic wind tunnel design. It should also fit the the proximity of these components to the test section, the
design of the other wind tunnel sections. A good example of negative acoustic consequences that arise from the inter-
this dependency is the interaction between the airline design action between the heat exchanger, flow straightener and
and the design of the wind tunnel drive unit. If fan power is a screens have a direct and relevant impact on the background
capital design constraint, the final airline design point must noise. The same studies also concluded that the combined
prioritise the aerodynamic objective with a penalty to the heat exchanger and flow straightener is primarily responsi-
acoustic quality of the aeroacoustic wind tunnel. The final ble for the background noise spectrum in the mid- to high-
design point moves accordingly to the left along the Pareto frequency range (3 kHz to above 10 kHz). Additionally,
front. In opposition, the final design point can be moved to some tonal noise components were registered at low flow
the right along the Pareto front if the available power in the velocities.
fan allows for a slight increase of the pressure loss in the air- To minimize this problem, the heat exchanger, which in a
line of the wind tunnel. This way, the acoustic characteristics classic aerodynamic wind tunnel is usually placed in the set-
of the wind tunnel airline can be maximized. An example of tling chamber, must be repositioned in an upstream location,
the application of the method suggested in the present test is ideally near the drive unit and far away from the test section.
described in Ref. [16]. To achieve this, the cross-section area of the wind tunnel
airline in between the second corner and drive unit fan must
be of similar size as the settling chamber. With this design,
the heat exchanger is placed close to the noisiest element of
3 Heat exchanger the entire wind tunnel so the acoustic penalty it introduces
is minimal. Additionally, all acoustic measures introduced
Because of their singular geometry and mechanical charac- in the flow path inside the airline, such as the annular muf-
teristics, tubes in heat exchangers are prone to experience fler installed downstream of the fan and splitter attenuators,
fluid-elastic instabilities and/or fluid–acoustic coupling will constitute a damping barrier for the propagation of the
effects [21]. Whenever this is observed, the heat exchanger noise generated by the drive unit and heat exchanger alike.
can be improved to reduce the self-generated noise at the To inhibit the self-generated noise and any acoustic reso-
source. Among all flow-induced sound interaction mecha- nance triggering mechanism in the heat exchanger tubes,
nisms, the so-called Parker resonance is a common one [23, either the periodic vortex shedding or the synchronization
24, 36]. This interaction mechanism is commonly found in with an acoustic mode must be disrupted, or both simul-
gas heat exchangers in power plants [8] and in axial-flow taneously. This is the reason why the heat exchangers of
fans [34]. It describes the flow-excited resonant sound gen- some aeroacoustic wind tunnels are installed with baffles
eration in cascades of flat plates in a rectangular duct, as in the upstream side (see Fig. 4). These elements, gener-
the vortex shedding frequency from the trailing edge of the ally located in a random arrangement, significantly increase
plates approaches the frequency of an acoustic transversal the unsteadiness and the turbulence intensity of the flow

13
 J. P. Gomes et al.

Fig. 4  Baffles (left) and acoustic partitions (right) installed in the heat exchanger of DNW-NWB

upstream from the tubes that constitute the heat exchanger. The funnel collector has its name attributed to its funnel-
As a direct consequence of the new upstream flow condition, shaped geometry. The collectors of this type are commonly
the wake flow created by the tubes is significantly smaller made of straight walls and are designed to ingest the free
and weaker as is the coherence of the vortex shedding devel- flow together with its shear layer. As a result of different
oped behind them. An alternative solution to the baffles is entrainment mechanisms, the diameter of the free flow
to manufacture the heat exchanger with tubes mounted with increases in the stream-wise direction along the test section.
tripping elements or with dimple structures on their surfaces. Because the mass flow must remain constant in the airline
Both solutions will force the flow boundary layer on the circuit, part of the flow that enters the funnel is deflected
surface of the tubes to become turbulent with similar effects backwards close to the funnel walls and is forced to flow out
to the vortex shedding developed behind them. Figure 4 also of the collector. This implies that the flow stagnation line,
shows the section of the airline immediately after the heat i.e., the line at the surface of the collector where the fluid is
exchanger, divided into small, irregular partitions. In oppo- brought to rest, is located at some point inside the collec-
sition to the baffles, this solution acts on the fluid–acoustic tor. If the inlet cross-section area of the funnel collector to
rather than fluid–structure interaction mechanisms. The flow cross-section area of the contraction ratio is small, undesired
partitions have the property to disrupt the mechanism of fluid–acoustic coupling effects arise as the flow stagnation
energy transfer transversal to the flow. More importantly, line moves close to the leading edge of the collector walls.
the irregular partitions breakdown the acoustic transversal For this reason, funnel collectors are only effective if their
resonance mode of the duct, proportional to its diameter or cross-section area is considerably bigger than that of the
width, into a wider range of smaller wave lengths. As soon contraction. In addition to the large space they occupy, fun-
as the principal acoustic transversal resonance modes of the nel collectors also present an aerodynamic drawback. The
duct are not restricted to a single wave length or frequency flow experiences a strong deceleration in front of this type of
and its higher harmonics, the transversal standing waves that collector, because of its larger diameter compared with the
will be triggered perpendicular to the flow will be out of contraction area. Consequently, the static pressure rises in
phase and will destructively interfere with each other. This the stream-wise direction and its distribution can no longer
mechanism will prevent the amplification through resonance be considered constant along the test section.
of an acoustic transversal standing wave across the airline The disadvantages of the funnel collectors are counterbal-
cross-section downstream of the heat exchanger. anced by their superior acoustic quality when installed with
the appropriate acoustic lining. Because of their size and
simple geometry, the collectors can be made with sound-
absorbing walls and/or covered with pile fabrics. The acous-
4 Collector tic advantage is a consequence of the fact that the entire
shear layer flow interacts directly with an acoustically treated
Various types of collectors have been tried in both aerody- surface. Moreover, funnel collectors are not installed with
namic and acoustic wind tunnels. However, just two of these a breather gap between the collector and the diffusor which
types have been selected as a standard solution in modern can be a relevant noise source in the test section. For these
wind tunnel designs. reasons, this type of collector is preferable for pure acoustic

13
Aircraft noise generation and assessment

wind tunnels with the penalty of a minor degradation of the fluid–structure–acoustic interaction mechanisms in the
aerodynamic quality of the flow. wind tunnel. The wind tunnel pulsations they generate, also
Wing collectors, or the so-called Prandtl rings, are wing- known as wind tunnel pumping or buffeting, have a nega-
shaped deflectors that are mounted upstream of the diffusor tive impact in the aerodynamic quality of the flow and thus
such that a breather gap is created between the collector and in the quality of the measured data [37]. As far as acoustics
the diffusor. In this design, the shape of the collector walls is concerned, wind tunnel pulsations not only increase the
is optimized so that the flow stagnation line coincides with background sound pressure level within the low-frequency
the leading edge of the collector. This way, the additional range but also distort the acoustic measurements, because
mass flow that results from the jet growth in the test section the noise generated by the flow around the test model is
is not reverted inside the collector walls but is forced to flow modulated at the frequency of the pulsations. The physical
out of the collector through the breather gap. This gap and mechanisms of the amplification of low-frequency pressure
the flow it generates constitutes a significant noise source in fluctuations in open test sections and their suppression have
the test section especially when it is hit by vortices in the been the subject of a variety of investigations and proposals
flow. The acoustic treatment of this type of collector is more to solve the problem. From a purely aerodynamic point of
challenging because of their smaller size and highly aerody- view, the most widely used method are vanes, tabs and teeth
namic shape. In most of the cases, flocked or pile fabrics are (Seiferth flaps) that protrude into the flow at the contraction
used as wall material for the collector surface to reduce the exit [15, 22, 30]. However, these vortex generators produce
noise generation (see discussion on pile fabrics in Sect. 7). high-frequency, broadband flow noise right in the inlet of the
The advantage of the wing collectors is that their geom- test section and thus are generally considered unacceptable
etry can be optimized to impose a minimum deceleration in for most aeroacoustic wind tunnel applications [25].
the flow. Consequently, the static pressure increase in front More complex alternatives for the attenuation of the pres-
of the collector can be mitigated. Additionally, the collec- sure fluctuations comprehend both active and passive sup-
tor to contraction cross-section area ratio can be minimized pression methods. Active resonance control (ARC) systems
resulting in small collectors. Thus, the complete length of essentially consist of a microphone to measure the pressure
the test section can be also minimized which is advantageous fluctuations in the wind tunnel anechoic plenum and a large
for the overall energetic balance of the wind tunnel. Because speaker driver system to pump sound waves inside the airline
of this, wing collectors are the most appropriate choice for circuit. A phase delay is introduced between the microphone
aerodynamic wind tunnels. signal and the speaker output to generate anti-phase sound
Despite the numerous attempts to solve this problem levels at the resonance frequencies inside the airline to can-
by means of CFD optimization algorithms, an established cel the pressure fluctuations. This system has proved to be
method to design an aerodynamically optimized collector very efficient in eliminating the periodic velocity fluctua-
that is at the same time acoustically very quiet has not yet tions in small wind tunnels and does not introduce additional
been found. The collectors that are installed in the most noise into the wind tunnel. In actual cases where systems of
modern aeroacoustic wind tunnels were designed almost by this kind were applied, reductions of pressure fluctuations
trial and error using a numerical, iterative approach similar of up to 20 dB were registered [37]. However, the high price
to that suggested for the design of the turning vanes inside and complex operation associated with ARC systems tend to
the wind tunnel airline. Thus, the results from this approach define them as not cost effective.
were specific for each particular wind tunnel and have not As far as passive methods are concerned, Helmholtz reso-
yet led to an optimized type of collector to be implemented nators can be implemented to remove the pressure fluctua-
as standard in all aeroacoustic wind tunnels to come. tions that occur at distinct frequencies [35]. The physical
characteristic of a Helmholtz resonator consists of a large
volume communicating with the airline circuit through a
5 Buffeting suppression mechanisms short duct. The main acoustic characteristic of a Helmholtz
resonator is the ability to be tuned to a single frequency.
Another aspect which deserves special attention during the The resonance frequency of a single resonator is defined as:
design of an acoustic wind tunnel is the low-frequency fluc- √
c A
tuations in pressure and velocity that are prone to appear f = , (6)
2𝜋 lef V
at distinct ranges of flow velocity in many open test sec-
tions. The large-scale vortices shed from the trailing edges where V is the fluid volume, A is the duct cross-section area
of the wind tunnel contraction, combined with the devel- and lef is the duct effective length. The latter is given by
opment of shear layers between the wind tunnel core flow
and the quasi-quiescent surrounding region, provide a lef = l + 𝛿, (7)
source of flow unsteadiness that can couple with various

13
 J. P. Gomes et al.

where l is the duct length and 𝛿 = 0.85d , where d is the duct The fan is known for generating different types of noise
diameter. that arise from all sorts of mechanisms of noise generation.
The absorption capability of the Helmholtz resonator can The most relevant include rotor blade and stator vane trailing
be understood as follows: the air in the communicating duct edge noise, rotor– and stator–inflow–turbulence interaction
oscillates back and forward as a single mass and the large noise, and rotor–stator interaction noise. Apart from the har-
volume acts as a spring or restoring force. Friction resistance monic tonal contribution of the rotor–stator interaction, all
is encountered by the alternating fluid inside and around the other noise sources are broadband in nature. Among these,
communicating duct. Hence, acoustic energy is absorbed, the inflow turbulence-induced noise is normally predicted to
mainly in the region around the resonance frequency. The play a role in the lowest frequency range, while rotor blade
larger the mass flow, the larger the damping effect of the trailing edge noise is expected to be dominant at higher
resonator. The Helmholtz resonator is an efficient method frequencies. As far as peak frequencies are concerned, the
of mitigation of the low-frequency fluctuations inside the peaks of the broadband contributions are determined by the
wind tunnel but at the expense of being effective for only a relative flow velocity and turbulence length scale, whereas
short frequency range around the resonating frequency. Mul- the peak frequencies of the tonal noise are centred on the
tiple resonator volumes are needed to broaden the frequency blade passing frequency and its harmonics.
range of the absorber. In addition, it normally requires large The impact of the fan design modifications on the dif-
volumes relative to airline volume to be truly effective at ferent noise levels of a ventilator or an aero-engine fan is
very low frequencies. commonly predicted by means of computational tools based
Another passive method of attenuation of pressure fluc- on semi-analytical acoustic models. The prediction of noise
tuations is the implementation of a sudden contraction in is given by semi-empirical scaling laws that describe the
the cross-section area of the airline circuit. With a sophisti- evolution of the sound pressure level of a noise source as a
cated design, a sudden contraction originates an impedance function of flow and geometry parameters. The aerodynamic
mismatch inside the wind tunnel airline that is capable of impact of those design modifications is later verified using
providing a strong absorption of pressure fluctuations in the turbomachinery-oriented CFD solvers. Examples of both
low-frequency regime. It is not yet completely understood acoustic and aerodynamic solvers for fan applications can
what the real interaction mechanisms between the airline be found elsewhere [1, 10, 17, 18, 33]. This approach allows
discontinuity and the low-pressure waves are. Preliminary for only the prediction of trends. Otherwise, it requires com-
tests tend to prove that the impedance mismatch method plicated calibration methods for producing results on the
constitutes a very efficient method of suppression of the low- absolute sound pressure levels.
frequency pressure fluctuations inside the airline for a wide Recent developments have led to a fully analytical formu-
range of flow velocities. In addition, it does not show the lation of the noise generation without the necessity of cali-
major penalties that are characteristic of the other methods. brating the models to predict absolute levels [19]. With these
The shape of the contraction should be chosen to create a improvements, it is possible to compare and superimpose
sudden increase of the flow velocity with a minimum penalty the absolute sound pressure level produced by several noise
on the local pressure loss in the airline circuit (for example, sources. This is very useful when the turbomachinery design
using sudden contractions with rounded corners). An exam- is supported by an automatic multi-object optimisation pro-
ple of a sudden contraction is shown in Fig. 1. It is located cess. No matter the prediction method, the acoustic design
after the second corner of the wind tunnel, in between the procedure must be based on the variation of noise relevant
heat exchanger and the fan. In this design, the sudden con- parameters and systematic evaluation of their impact on the
traction was also used to adapt the large cross-section area of different noise contributions. The most relevant parameters
the airline after the second corner to the narrower diameter are: blade count of the rotor and stator, rotor velocity, axial
of the fan. distance between rotor and stator, sweep angle of the stator
vanes, and inflow turbulence.
Figure 5 shows the results of the fundamental contri-
bution of the blade passing frequency and also the first
6 Drive unit three harmonic components as a function of the rotor and
stator blade count obtained after the simulation of a typi-
The wind tunnel fan is certainly one of the few components cal wind tunnel fan. The impact of the blade count on the
of a wind tunnel that can be independently improved to broadband noise for the same fan is presented in Fig. 6. Two
reduce noise emission at the source. The main objective of regions presenting a cutoff design can be identified in Fig. 5
the acoustic optimization is the reduction of broadband noise (1-BPF). The triangular domain in the upper right-hand
while keeping the tonal noise at a low level and the aerody- corner of the scalar map corresponds to the standard cutoff
namic performance unchanged. design in which the number of stator vanes is higher than

13
Aircraft noise generation and assessment

Fig. 5  Tonal sound pressure level of a typical wind tunnel fan as a function of both rotor and stator blade count for the first four blade passing
frequencies

Broadband rotor–stator interaction noise increases propor-

tionally with the number of sound emitters, in this case the
stator vanes. Furthermore, broadband rotor–stator interac-
tion noise is proportional to the integral turbulence length
scale in the rotor wake. With the latter being proportional to
the pitch, it is coherent to expect the broadband rotor–stator
interaction noise to decrease while increasing the number of
rotor blades. The combined effect of these two tendencies
is shown in Fig. 6.
Both broadband noise-generating mechanisms,
rotor–stator and stator–inflow–turbulence interaction noise,
are proportional to the number of stator vanes. Similarly,
Fig. 6  Broadband sound pressure level of a typical wind tunnel fan as
rotor–inflow–turbulence broadband interaction noise is
a function of both rotor and stator blade count proportional to the number of rotor blades. Consequently,
and from an acoustic point of view, a small number of rotor
blades are more suitable in the event of high incoming flow
the number of rotor blades. Classic aerodynamic wind tunnel turbulence.
fans often utilize this design. It is optimized for tonal noise The solidity of the blades in both the stator and rotor is a
cancellation, since the first harmonics of the blade passing very important aerodynamic parameter and directly affects
frequency of the rotor–stator interaction are non-propagated. the pressure rise of the fan. To maintain the aerodynamic
For acoustically optimized fans, it is advantageous to characteristics of the fan unchanged, the number of blades
apply the so-called inverse cutoff design. This design makes must be changed while keeping the blade solidity constant.
use of a lower stator to rotor blade count ratio and its domain This means to adjust the blade chord as a function of the
is restricted to a diagonal line in Fig. 5 (1-BPF). In this line, number of blades. This design constrain has a direct impact
the number of stator vanes is smaller than the number of on the trailing edge noise, because it does not depend on the
rotor blades. The new design is selected to reduce primarily blade count at constant solidity. For this reason, no trailing
the broadband noise through the lower count of stator vanes. edge noise variations are expected for the rotor and stator.

13
 J. P. Gomes et al.

However, the inverse cutoff design has an acoustic pen- induces a noise reduction despite the lengthening of the
alty. While the standard cutoff design can be extended to leading edge. Combined with the effect on correlation,
the higher harmonic components of the blade passing fre- this could explain why broadband noise is also reduced
quency, no such opportunity is given by the inverse cutoff by choosing a swept stator [39]. However, special atten-
design method, because only the fundamental frequency is tion must be given to the sweep angle of the stator vanes
cutoff (see Fig. 5). Therefore, it is safe to expect higher tone close to the wind tunnel wall, because they can increase
levels with the inverse design, since the second harmonic the local pressure loss significantly. To keep this effect
of the blade passing frequency (and higher) can propagate bounded, it is interesting to investigate V-shaped or inter-
inside the tunnel airline. To counterbalance the higher tonal rupted sweep vanes. The acoustic efficiency of these types
contributions of the fan-generated noise, the axial distance of vanes is expected to be slightly lower than the original
between the rotor and stator and also the sweep angle of the sweep vanes. Thus, the design must achieve a compromise
stator vanes can be increased. between acoustic and aerodynamic quality. Several experi-
Variations of the axial distance between the rotor and sta- ments have already shown that stopping the sweep angle at
tor have no influence on the aerodynamic performance of the approximately 85% of the blade span leads to a significate
fan. This is true to a certain extent. From an acoustic point acoustic gain with very limited aerodynamic penalty com-
of view, both rotor–stator broadband and tonal noise contri- pared to a radial stator.
butions decreased while increasing the separation between Comparing all advantages and disadvantages of the
the two blade rows. The noise reduction is due to the decay acoustic design of a wind tunnel fan, it can be shown that
of the turbulence intensity as the wakes after the rotor are the configurations with a small number of stator vanes and
convected downstream. Actually, this trend should be even a larger number of rotor blades are more beneficial. This
more pronounced for the tonal contributions, because inter- is true assuming the incoming flow turbulence intensity is
ference effects contribute to the noise reduction as well as not too high and the tonal noise associated with the higher
the wake decay. The radial phase shift of the incoming wakes harmonic components of the blade passing frequency is
(source excitation) is amplified by the inter-stage swirl as mitigated by an optimized geometric definition of the rotor
the wakes develop downstream. Due to this phase shift, the and stator blades.
leading edges are excited at different instants along the span, Finally, one should account for the effects of the veloc-
which induces noise cancellations. It should be emphasised ity of the rotor. It is easy to assume that this parameter has
the rotor–stator separation must be bigger if the number of great importance because the sound pressure level gener-
blades is reduced, because with fewer blades the decay of the ated by a blade increases roughly with the 5th power of
potential field is slower and therefore needs a larger distance the flow velocity measured in the relative frame of refer-
to decrease by the same amount as with the original design. ence. Rotor blade trailing edge noise, rotor–inflow–turbu-
As a rule of thumb, the most efficient distance between lence interaction noise, and rotor–stator interaction noise
the two blade rows is about one blade chord spacing. For decreased while reducing the rotor velocity at constant
larger distances, the decay of turbulence is slower and fan pressure differential. This noise reduction is a direct
reaches the point where the acoustic benefit is too small consequence of the reduction of the local flow velocity. In
when compared with the aerodynamic penalty due to the this case, the solidity must increase to compensate for the
increase of the pressure loss. Trailing edge noise and inflow velocity reduction, which means an increase of the chord
turbulence interaction noise due to the ingested turbulence length. Stator–inflow–turbulence interaction broadband
are unchanged, because those sources should be independent noise slightly increases due to the increase of the relative
from the separation distance between the blade rows. flow velocity at the stator vanes’ leading edge, whereas
The introduction of a sweep angle on the stator vanes stator trailing edge broadband noise remains almost con-
is proposed mainly to reduce the tonal contribution of the stant. The rotor velocity reduction is directly associated
noise produced by the periodical impingement of the rotor with the optimization of the wind tunnel internal pressure
wakes on the stator vanes [9, 38]. The principal mecha- loss and to the early design considerations of the drive
nism responsible for the noise reduction is the radial phase unit. Reducing the rotor velocity is not only beneficial for
shift introduced by the tilted wakes as explained by Envia supersonic rotors (for which the suppression of shock-
[9]. This effect can be strongly amplified by introducing induced buzz-saw noise is a well-known technique) but
a sweep [11]. Presumably, the sweep also has an effect on also for low-speed rotors. The maximum velocity reduc-
the source strength due to the change in the inclination tion at constant fan pressure rise is limited by the tech-
of the leading edge relative to the mean velocity and the nological advances in aerodynamic design and represents
lengthening of the span. This means a reduction of the one of the major challenges for designing both quiet and
relative velocity in the case of swept stator, which usually efficient wind tunnel fans.

13
Aircraft noise generation and assessment

7 Anechoic plenum microphones or between different measurement runs were

observed at up to 10 dB.
Acoustic measurements should be performed in a fully The discrepancy observed in the results of the local sound
anechoic, or at least semi-anechoic, environment to pro- pressure level can only be attributed to the complex acoustic
vide reliable results of the strength and directivity of an interference pattern inside the closed test section created by
acoustic noise source. Such an environment is reproduced multiple reflections on the hard walls. Small flow velocity
with sound-absorbing materials, often in special shapes, variations modify the local interference pattern due to con-
which efficiently suppress the reflections of acoustic waves vection effects, which produces a similar effect to chang-
on the solid walls of the test section. In such an environ- ing the position of the traversing microphone inside the test
ment, the noise source can be investigated in quasi-free- section. In a similar way, small mechanical changes in the
field conditions. But most actual wind tunnel facilities test setup are capable of changing the interference pattern
make use of closed test sections with hard walls for aero- for tonal noise signals. For broadband noise contributions,
dynamic investigations. To provide well-defined boundary like the wind tunnel background and model airframe noise,
conditions, the test section walls are made as flat as possi- smaller differences between the sound pressure levels meas-
ble and from solid materials, like metal or glass for optical ured by adjacent microphones were registered. In general,
access to the inside of the test section. Acoustically, these the results from different test setups and runs also showed a
hard and flat walls act as efficient reflecting surfaces for much better agreement compared with those obtained with
sound waves. tonal noise signals.
To gain insight on the impact of the interference effects Further investigations involving noise measurements in
created by hard-walled test sections in the acoustic meas- closed test sections with microphones placed outside of the
urements, DNW investigated the standard closed test sec- flow, like inside the cabin of a passenger car in the test sec-
tion of the DNW-LLF. This standard test section has a tion, provided similar conclusions. In this case, the effects
width of 8 m, a height of 6 m and an overall length of of reflections on the car were added to the reflections on
20  m. The tests were carried out in both wind-off and the hard walls of the test section. Once again, the measured
wind-on conditions with one traversing and several sta- sound pressure level of tonal signals was much more unreli-
tionary inflow microphones to cover the entire area of the able compared to broadband noise contributions.
test section. All microphones were installed at a height of Based on these investigations, it can be said that it is
about 1.5 m above the test section floor with aerodynamic very difficult to reliably measure absolute sound pressure
forebodies, also known as nose cones. During the investi- levels inside closed test sections with hard walls with the
gation in wind-off conditions, a loudspeaker was used to classic inflow free-field microphone technique. Even delta
generate tonal noise signals at the typical position of the measurements do not provide reliable results for the sound
test model. Sinusoidal signals between a few hundred and pressure level comparison between two test configurations.
several thousand Hertz were tested. The results showed a For measurements of broadband noise contributions, the sce-
complex pattern of interferences for the local sound pres- nario is slightly better but still not acceptable for quantitative
sure level as a function of both the axial and lateral posi- analysis.
tion in the test section. Within a distance variation of a few Differently, noise mapping techniques have been success-
centimetres in the position of the traversing microphone, fully applied in closed test section environments, in special
the local sound pressure level could change up to more high order deconvolution methods such as CLEAN-SC [31]
than 20 dB. When repeating the tests with a white noise or, more recently, high-resolution CLEAN-SC [32]. These
signal, fluctuations of the overall sound pressure level deconvolution methods have proven to be significantly more
were still observable but significantly smaller and limited reliable than the conventional beamforming technique in
to several dB. The investigations in wind-on conditions reverberant, noisier environments. The noise mapping tech-
were concentrated on the characteristic tonal noise of the niques make use of phased microphone arrays to detect the
wind tunnel fan. These tones were clearly observed in all spatial distribution of sound sources. The array microphones
acoustic spectra with a typical distance to the broadband are often flush mounted in the walls of the test section and
noise level of 10 dB and more. During measurements at therefore directly exposed to the boundary layer that devel-
different flow velocities, a clear correlation between the oped along them. As long as the size of the test section is
measured sound pressure levels and the required electric big enough compared with the dimensions of the model and
power to operate the wind tunnel fan was not registered. the acoustic wave length of interest, wall-mounted phased
A low reproducibility between different test configura- microphones arrays are significantly less sensitive to inter-
tions was also observed. For the flow velocity range of the ference effects caused by the hard walls.
tests, the sound pressure level variations between adjacent In recent years, tests have been performed in closed test
sections with acoustic absorbing walls, like perforated walls

13
 J. P. Gomes et al.

with a sufficiently thick layer of acoustic material behind the shape of a rectangular cuboid. Even though this design
them, or acoustically semi-transparent walls, like walls is prone to developing standing waves, the fundamental
made from thin Kevlar fabric [5, 6, 26]. In both cases, the frequency of resonance is usually low enough to be disre-
reflected acoustic waves on the inner walls of the test section garded. In the case of smaller plenums, this effect should not
are minimized. However, these new types of walls have their be neglected and non-regular shapes should be considered.
own problems when it comes to defining the aerodynamic Figure 7 shows an example of an aeroacoustic wind tun-
boundary conditions inside the test section accurately, and nel anechoic plenum. It has a cross-section area equal to
they have not yet proven their efficiency. 16.4 m × 7.8 m (the cross-section area of the contraction is
The best acoustic alternatives to closed test sections are equal to 3.25 m × 2.8 m) and it is symmetrically mounted
the open-jet and 3/4 open test sections surrounded by a with respect to the wind tunnel axis. The length of the ane-
large acoustic plenum with all walls covered with acoustic choic plenum is equal to 14.2 m, about 2.4 times the length
lining. Most modern aeroacoustic wind tunnels offer such of the wind tunnel test section. Acoustic lining made from
kind of test sections for acoustic tests [12]. They provide an sound-absorbing material shaped into wedges, such as those
anechoic environment for the measurements and the pos- shown in the figure, is a well-proven method to optimise the
sibility to reliably measure absolute noise pressure levels. acoustic damping and to achieve a free-field environment
Because the microphones can be placed outside of the flow inside the anechoic plenum. The wedge-shaped geometry
at a bigger distance from the test model, the acoustic far ensures a gradual change in the acoustic impedance of the
field can also be evaluated correctly. The biggest drawback transmission media, ensuring that sound waves are absorbed
imposed by open test sections to acoustic measurements is by the material, rather than reflected at the interface. In
associated with the fact that the sound waves emitted by choosing the dimensions and material of the acoustic lining,
the test model must propagate through the free flow shear the most important decision point is the frequency range in
layer before reaching the microphones. The propagation which the anechoic plenum must be effective. For the most
of the sound waves through the shear layer causes refrac- aeronautical acoustic problems, the sound frequency range
tion, spectral broadening and loss of coherence between the below 4 kHz is of special interest. Because of the specific
signals arriving at different microphones. The latter affects aerodynamic problems found in road vehicles, this value is
mostly the results of the noise mapping techniques. These even smaller for automotive tests. Thus, wind tunnels that
effects depend on the Mach number of the flow, both noise focused on full-scale models have anechoic plenums gener-
source and microphone position relative to the shear layer, ally designed to operate effectively in the frequency range
and sound frequency. Because they increase while increas- from about 100 Hz up to 4 kHz. In smaller wind tunnels,
ing the frequency of the propagating sound, the maximum where test models are scaled down relative to their full size,
sound frequency that can be measured in an open test section the frequency range of interest increases by the same factor.
with acoustic plenum is limited. This can cause problems in Consequently, frequencies above 0.6 ≈ 0.8 kHz and up to
measuring the noise generated by small-scale models. 40 kHz are important during a test with a 1:10 scale model.
The installation of a fully anechoic or a semi-anechoic The effectiveness of the acoustic lining is defined by its
plenum around the wind tunnel test section serves a twofold geometry and construction materials. The lowest frequency
objective. First, to isolate the test section and consequently
the acoustic experimental setup against external noise, so
the resulting internal noise does not overrule the measure-
ments. Second, to reproduce with high fidelity the acoustic
free-field conditions inside the test section. This may require
the use of single- or double-wall construction with proper
vibration isolation to reduce air- and structure-borne noise
transmission. At the same time, the volume of the anechoic
plenum must be as big as possible. The noise pressure level
decays with the square of the distance. Consequently, the
amplitude of the sound pressure waves reaching the walls
of the plenum is smaller in large facilities. Additionally, any
pressure fluctuation and/or secondary flow that might occur
inside the anechoic plenum have less interaction with the
plenum walls. In practice, the dimensions of the plenum
are limited by the host wind tunnel building infrastructure
and by construction and maintenance costs. For the same
reasons, large anechoic plenums are usually constructed in Fig. 7  Open-jet test section with anechoic plenum of DNW-NWB

13
Aircraft noise generation and assessment

at which the absorption is effective (cutoff frequency) is broadband noise reduction of up to 15 dB was observed
inversely proportional to the height of the wedges. For exam- compared with the same untreated hard surface [25]. This
ple, 0.8 m high wedges made from mineral wool covered by finding corresponds well with the results presented by
glass-reinforced fibre fabric have a typical cutoff frequency Nishimura [20].
close to 100 ≈ 200 Hz and a reflection factor for plane waves The acoustic characteristics of an anechoic plenum in
at normal incidence smaller than 0.1 for frequencies above wind-off conditions can be assessed utilizing the methods
the cutoff frequency. At lower frequencies, the reflection fac- described in a series of international standards on room
tor increases rapidly. On the other hand, the cutoff frequency acoustics [42, 44, 45]. In most cases, these standards are
of a 0.2 m thick flat panel made from the same combination the same ones chosen to certify the acoustic quality of the
of materials is approximately 425 Hz. anechoic plenum. When applying these norms to the aero-
When increasing the frequency of the sound, the most dynamic wind tunnel design, it is reasonable to compare the
commonly encountered noise transmission problems are anechoic plenum to an anechoic chamber. This analogy sug-
those resulting from sound reflections and sound leakages. gests removing the collector from the plenum and covering
Leaks through narrow gaps tend to be significant in the fre- both contraction and diffusor openings with acoustic lining
quency range above approximately 2 kHz. Reflections on for the in situ tests of the acoustic insulation and damping
the other hand occur on every hard surface. The minimum characteristics.
area to create a sound reflection decreases while increasing The requirements on the airborne acoustic insulation the
the sound frequency. A generally accepted rule defines it of anechoic plenum walls are defined by weighted sound
to a quarter wavelength of the highest sound frequency of reduction indices [45]. Quantities such as noise absorption
interest. This corresponds to an area of about 2 mm × 2 mm and sound transmission through wall partitions are funda-
for a sound frequency around 40 kHz. Although unrealistic, mentally related to sound power. Therefore, sound intensity,
this rule shows that every small structural surface inside or sound power, is a reliable quantity to be evaluated and
the plenum that must be acoustically protected. If not prop- provides an alternative approach for assessing the airborne
erly acoustically treated, small gaps resulting from moving sound insulation through walls. Measurement of the sound
parts (like doors, moving walls, detachable ceilings, etc.), transmission through wall partitions between two rooms can
auxiliary elements (like ladders, walking platforms, lift- be carried out in the laboratory or in situ. In the source room,
ing cranes, etc.) and interior lightning systems can produce a sufficiently diffuse sound field must be assumed. An easy
undesired sound reflections and leakages that deviates the way to achieve this is to distribute one or more noise sources
acoustic conditions inside the plenum from those registered away from the wall, so the portion of direct sound from the
in a pure acoustic free field. sound source to the wall is not significant in relationship to
However, there are cases in which it is impossible to hide the portion of the diffuse sound.
all solid structures. These cases include all mechanical struc- The sound power passing through a surface can be related
tures that are directly exposed to high flow velocities and to the spatial averaged sound pressure level measured at the
structures that need to keep a defined aerodynamic shape, surface. Thus, the sound power incident on the wall is cal-
like model and inflow microphone supports and the wind culated based on sound pressure level measurements in the
tunnel collector. All these cases share a common point. The source room. The spatial averaging can be realized by either
resulting interaction between the highly turbulent flow and sweep measurements or by discrete point measurements
the solid surfaces constitutes a major source of broadband over the entire wall that separates both source and receiving
noise. room. In the receiving room, the sound power transmitted
Nishimura [20] investigated this phenomenon and dem- through a wall partition is obtained using a similar type of
onstrated that vortex interaction-induced noise from a turbu- measurement over the wall partition area. Finally, the appar-
lent flow that impinges on a solid body can be significantly ent sound reduction index is defined as:
reduced when the latter is covered with pile fabrics. Pile [ ( )]
fabrics are a type of cloth material with fine and high-density R�I = Lp inc − 6 + 10 log
( ) A
fibres and can be found in different thicknesses and struc- A0
tures. The purpose of using pile fabrics is to attenuate the
[
Ap
( )] (8)
( )
near-wall flow vorticity. When vortices collide with pile fab- − Lp trs + 10 log ,
A0
rics, their kinetic energy is gradually reduced by the interac-
tion with the fibres. As a result, the rapid deceleration of the
where Lp inc and Lp trs is the sound pressure level measured
( ) ( )
flow near the wall does not occur and dipole-type noise is
reduced. Another study compared different types of cloths in the source and receiving room, respectively.
and showed that the noise reduction highly depends on the In the case of an aeroacoustic wind tunnel, the rooms
structure of the pile fabric. A maximum sound pressure level adjacent to the anechoic chamber are used as source rooms,

13
 J. P. Gomes et al.

whereas the anechoic chamber is the receiving room. In con- control of the sound power stability of the sound source
clusion, the anechoic plenum must be designed in such a way is made by means of a stationary microphone during the
as to maximize the apparent sound reduction index of all the measurements.
side walls, floor and ceiling of the plenum. This way, the The acoustic measurements must also account for the
acoustic isolation of the test section against external noise atmospheric absorption inside the plenum and consequent
is maximized. sound pressure level attenuation along the propagation paths.
The damping effectiveness of an anechoic plenum pro- For simple( tones,
) the attenuation due to the atmospheric
vides a direct comparison between the acoustic conditions absorption 𝛥Lp atm is given in terms of an attenuation coef-
inside the plenum and those of a perfect acoustic free field. ficient as a function of four variables: the frequency of the
Therefore, it is evaluated comparing acoustic measurements sound, temperature, humidity and pressure of the air [40].
at N positions along predefined acoustic propagation paths Consequently, the acoustic measurements along the propaga-
inside the plenum against the theoretically predicted sound tion paths must be followed by measurements of the ambi-
pressure level for a sound propagation in the free-field. The ent pressure, air temperature and relative air humidity. The
sound pressure level at the measurement distance ri for the acoustic damping effectiveness of the anechoic plenum is
ideal free-field propagation is modelled by the optimal refer- finally defined at each measurement location as follows:
ence method [44] according to the following equation: ( )
( ) 𝛥Lp i = Lp i + 𝛥Lp i − Lp (ri ). (12)
a atm
Lp (ri ) = 20 log , (9)
ri − r0
The lower the value, the closer the acoustic propagation of
where the sound inside the anechoic chamber is to the ideal free-
field condition. It should be stressed that the deviation from
�∑
N
�2 ∑N the ideal free-field condition is evaluated as a function of the
r
i=1 i
−N r2
i=1 i sound frequency, the position of the sound source and the
a = ∑N (10)
∑N −0.05Lp i
−N
∑N � −0.05Lp i
� measurement paths. Regarding the two last parameters, room
i=1 ri i=1 10 i=1 ri 10
acoustics require that the sound source is located in the cen-
tre of the room. Starting from the sound source, four measur-
and
ing paths should run diagonally into the corners of the room.
∑N ∑N � � ∑N ∑N
r i=1 ri 10−0.05Lp i − i=1 ri2 i=1 10−0.05Lp i Additionally, one measuring path can be chosen arbitrarily.
i=1 i
r0 = − ∑N ∑N ∑N � . In a wind tunnel, it is useful to make some modifications to
r i=1 10−0.05Lp i − N i=1 ri 10−0.05Lp i
�
i=1 i the standard procedure. The sound source must be located
(11) at the usual position of the test model regardless of the
Here, it is considered that the acoustic and the geometric geometry of the anechoic chamber. The propagation paths
centre of the sound source may not coincide. The collinear must be defined while considering the most common micro-
offset between the two is estimated based on the measured phone positions (free-field microphones, phased microphone
sound pressure level along the propagation path using the arrays, etc.) during the wind tunnel measurements.
regression (11). This value describes the deviation of the The results of the acoustic damping test only indicate
acoustic centre of the sound source from the geometrically deviations from the ideal free field. It is not possible to
defined one. Reference [44] does not define a limit for the extract information about the reasons for those deviations.
collinear offset, however, if this value exceeds a certain For example, they can be related to insufficient sound
threshold it can be assumed with certainty that the free-field absorption of the acoustic lining or sound reflections on
conditions are not fulfilled inside the anechoic plenum. As both hard and semi-hard surfaces. To identify possible
a first approximation, one can consider the collinear offset sources of reflection, new acoustic methods such as the room
to be limited to 0.2 m or to double wavelength of the tested impulse response method can be implemented. In the room
frequency, whichever is achieved first. A multi-frequency impulse response method, the sound source at the origin of
tonal acoustic source is used as test signal, with frequencies the propagation paths is substituted by a time-pulsed, high-
corresponding to approximately the centre frequency of the pass filtered pink noise acoustic source [43]. The impulse
one-third octave bands within the frequency range of inter- response is evaluated in the time domain for every micro-
est. On one hand, the sound source is required to ensure a phone position in the propagation path. To improve the clar-
minimum signal-to-noise ratio of 10  dB at all microphone ity of the results, the impulse response along one path is
positions. On the other hand, the sound source must radiate a commonly presented in a scalar map plot referred to as a
spherical sound wave as specified in Ref. [44]. Additionally, reflectogram. Figure 8 shows an example of a reflectogram
the sound power level of the sound source must not change measured inside an anechoic plenum along an arbitrary
during the measurement time of one propagation path. The acoustic propagation path. In the figure, the first oblique,

13
Aircraft noise generation and assessment

120

Typical aerodynamic
wind tunnel
100
1st generation of
aeroacoustic wind tunnels

OSPL [dB(A)]
80

Newest
60 aeroacoustic wind tunnels

40

20
0 20 40 60 80 100
flow velocity [m/s]
Fig. 8  Example of a spectrogram obtained by the impulse response
method along an acoustic propagation path inside an anechoic plenum
Fig. 9  Overall background noise pressure level in different aeroacous-
tic wind tunnels in comparison with the same measured in a typical
aerodynamic wind tunnel [29]
brighter line corresponds to the direct sound. This line is
found in all reflectograms and is associated with the direct
propagation of the sound wave along the path. Therefore, results shown in Fig. 9 were measured using a similar meth-
its slope is directly proportional to the propagation velocity, odology as described in subsequent paragraphs. The distance
i.e., to the speed of sound. The second oblique line indi- of the microphone to the tunnel axis was approximately 6 m.
cates a weak reflection from a flat surface. In a reflectogram, The overall background sound pressure level inside the
all impulse responses are normalised by the amplitude of anechoic plenum is measured as a function of the flow veloc-
their individual direct sound to remove the distance-related ity in the absence of any test model and should consider the
attenuation of the sound pressure level. contribution of all acoustic frequencies from approximately
20 Hz up to 20 kHz. The measurements are standardly per-
formed using a free-field microphone outside of the flow at
8 Performance measurements a predefined distance from the wind tunnel axis. The micro-
phone should be positioned perpendicular to the flow and
The overall background sound pressure level and the pulsat- aligned with the most common position of the test model. In
ing coefficient are the most recognised quantities to deter- most cases, the intersection between the wind tunnel model
mine the acoustic performance of an aeroacoustic wind rotating table axis and the flow axis is used to represent
tunnel in wind-on conditions. They are of special relevance this position. The selection of the microphone, the signal
because they provide the background noise level in the test acquisition and post-processing systems must guarantee
section within the entire flow velocity range of the wind an accurate reconstruction of the measured acoustic signal
tunnel. Consequently, they define the signal-to-noise ratio in the form of a one-third octave band spectrum from the
for the measurements described below. one-third octave band centred at 20 Hz up to 20 kHz. It is
Figure 9 shows the overall background sound pressure common practice to apply an A-weighted filter to the output
level inside the anechoic plenum of a group of aeroacoustic signal of the microphone to give more weight to the sound
wind tunnels as a function of the flow velocity. The plot frequencies of interest for human hearing and to attenuate
shows that the background noise level inside the test sec- the influence of both very low and very high acoustic fre-
tion of an acoustically oriented wind tunnel design is typ- quencies [41]. With the inclusion of this filter, the evaluation
ically 30 dB(A) lower, or more, than that measured in a of the background noise becomes very robust and provides
classic aerodynamic wind tunnel. This value continues to very comparable results even when the measured frequency
decrease while the design tools are developed, especially the range has slightly different boundaries. As a consequence of
numerical ones. The most modern aeroacoustic wind tunnels the small scale of the test models, the A-weighted filter must
designed with the highest level of development tools in the only be skipped for very small aeroacoustic wind tunnels
field of aerodynamic and acoustic show a background noise because of the shift of the frequencies of interest towards
reduction of almost 50 dB(A). It should be emphasised, that the high frequencies.
the comparison between the overall background noise of Finally, the overall background sound pressure level is
different wind tunnels should be performed only if both the calculated integrating the sound pressure level over the
size of the wind tunnels and the position, or distance, of the entire frequency spectrum. If the one-third octave band
microphone relative to the test section are comparable. All spectrum is used for that purpose, it corresponds to the

13
 J. P. Gomes et al.

summation of the noise level, in decibels, of all frequency 0.01

bands between the 13th and the 43th [41], inclusive,

according to the following equation: 0.008

43 Lp
∑ i 0.006
(13)

Cp,RMS [-]
Lp,OSPL = 10 log 10 10 .
i=13
0.004

The pulsating coefficient provides a comparison between

the magnitude of the low-frequency pressure fluctuations 0.002
inside the anechoic plenum and the dynamic pressure in the
free flow. When its magnitude exceeds a certain threshold, 0
pressure fluctuations of low frequency lead to pulsating 0 20 40 60
flow velocity [m/s]
80 100

problems inside the wind tunnel with negative mechanical

and acoustic consequences (see also Sect. 5). The frequency Fig. 10  Pulsating coefficient result in a modern aeroacoustic wind
range of interest for the wind tunnel pulsation is located in tunnel (DNW-NWB)
the lower part of the frequency spectrum, typically between
1 Hz and 20 Hz.
The evaluation of the pulsating coefficient starts with coefficients of about 0.02 or more in classic aerodynamic
the measurement of the overall sound pressure level, in wind tunnels.
dB(Z), within the lower frequency range. This should be Conversely, aeroacoustic wind tunnels must be designed
performed with a microphone outside the flow at the same to minimize the pulsating coefficient and consequently any
position as for the background noise measurements inside pumping effect in the test section and airline. Recent expe-
the anechoic chamber. Special attention must be paid to rience shows that low-frequency pressure fluctuations in a
the selection of the microphone and filters to be used in the well-designed aeroacoustic wind tunnel are very small, typi-
conditioning of the signal, because all the frequencies of cally lower than 1 % of the dynamic pressure inside the free
interest during this measurement are smaller than 20 Hz. flow. In terms of the pulsating coefficient, this represents a
The overall sound pressure level can be calculated based value smaller than 0.01. For this reason, this value is com-
on the narrow band spectrum of the measured acoustic monly accepted as the standard requirement for current and
signal after the logarithmic summation of all narrow bands future aeroacoustic wind tunnel designs. Notwithstanding
within the frequency range from 1 up to 20 Hz. In this this requirement, the aeroacoustic wind tunnel design is so
process, one must correct the results for the spectral leak- advanced nowadays that some of the most modern aeroa-
age introduced by the windowing during the fast Fourier coustic wind tunnels have pulsating coefficients smaller than
transformation. For a Hanning window, the noise equiva- 0.002, as shown in Fig. 10.
lent bandwidth is equal to 1.5𝛥f which corresponds to the
worst case amplitude error equal to 1.76 dB.
Finally, the pulsating coefficient is defined as: References
Lp,
p0 10
OSPL
20
1. Becker, K., Heitkamp, K., Kgeler, E.: Recent progress in a hybrid-
Cp,RMS = , (14) grid CFD solver for turbomachinery flows. In: Proceedings of
1
𝜌U 2 the 5th European Conference on Computational Fluid Dynamics
2
ECCOMAS CFD 2010 , Lisbon (2010)
2. Bertram, M., Deines, E., Mohring, J., Jegorovs, J., Hagen, H.:
where p0 is the reference pressure used in the computation Phonon tracing for auralization and visualization of sound. In:
of the sound pressure level (equal to 20 μPa) and 𝜌 is the Proceedings of IEEE Visualization (2005)
density of the air. The latter is calculated based on the mete- 3. Broszat, D., Selic, T., Marn, A.: Verification of the inverse cut-off
orological information at the moment of the measurements. effect in a turbomachinery stage—Part 1—numerical results. In:
18th AIAA/CEAS Aeroacoustics Conference, Colorado Springs,
Pulsating coefficient measurements are rarely reported Colorado (2012)
for classic aerodynamic wind tunnels. This is because very 4. Broszat, D.U., Selic, T., Marn, A.: Verification of the inverse
low-frequency pressure fluctuations and their minimiza- cut-off effect in a turbomachinery stage—Part 2—comparison to
tion are not considered a priority during their design. At experimental results. In: 19th AIAA/CEAS Aeroacoustics Confer-
ence, Berlin (2013)
best, the pumping effects are mitigated after the wind tun- 5. Brown, K., Devenport, W., Borgoltz, A.: Exploiting the character-
nel design using a posteriori solution if the low-frequency istics of Kevlar-Wall wind tunnels for conventional aerodynamic
fluctuations in the test section are registered to be too big. measurements. In: AIAA AVIATION 2014—30th AIAA Aerody-
For this reason, it is not uncommon to register pulsating namic Measurement Technology and Ground Testing Conference,
Atlanta, Georgia (2014)

13
Aircraft noise generation and assessment

6. Camargo, H., Remillieux, M., Burdisso, R., Crede, E., Devenport, Calibration and demonstration of the new Virginia Tech anechoic
W.: The Virginia Tech stability wind tunnel from an aerodynamic wind tunnel. In: 14th AIAA/CEAS Aeroacoustics Conference,
into an aeroacoustic facility. In: 19th International congress on Vancouver, British Columbia Canada (2008)
acoustics, Madrid (2007) 27. Roeber, N., Andres, S., Masuch, M.: HRTF simulations through
7. Cremer, L., Mueller, H.A.: Principles and Applications of Room acoustic Raytracing, Tech. Rep., Fakultaet fuer Informatik, Otto-
Acoustics, vol. 1. Peninsula Publishing, Newport Beach (2016) von-Guericke Universitaet Magdeburg (2006)
(ISBN 0932146732) 28. Roeber, N., Kaminski, U., Masuch, M.: Ray acoustics using com-
8. Eisinger, F.L., Sullivan, R.E.: Unusual acoustic vibration in heat puter graphics technology. In: Proceedings of the 10th International
exchanger and steam generator tube banks possibly caused by Conference on Digital Audio Effects, Bordeaux (2007)
fluid-acoustic instability. J Eng Gas Turbines Power 115(2), 411– 29. Schneider, S., Wiedemann, J., Wickern, G.: Das Audi-Wind-
417 (1993) kanalzentrum, Aerodynamik des Kraftfahrzeugs, Haus der Tech-
9. Envia, E., Nallasamy, M.: Design selection and analysis of swept nik, Essen, Germany (1998)
and leaned stator concept. J. Sound Vib. 228(4), 793–836 (1999) 30. Seiferth, R.: Vorausberechnung und Beseitigung der Schwingun-
10. Gurin, S., Moreau, A., Tapken, U.: Relation between source mod- gen von Freistrahl-Windkanlen, Aerodynamische Versuchsanstalt,
els and acoustics duct modes. In: 16th AIAA/CEAS Aeroacoustics Goettingen, Germany (1946)
Conference, Miami, Florida (2009) 31. Sijtsma, P.: CLEAN based on spatial source coherence. Int. J.
11. Gurin, S., Moreau, A.: Accounting for sweep and lean in the Aeroacoust. 16(4), 357–374 (2007)
design-to-noise of rotor-stator stages. In: Proceedings of the 32. Sijtsma, P., Merino-Martinez, R., Malgoezar, A., Snellen, M.:
DAGA Conference, Berlin (2010) High-resolution CLEAN-SC: theory and experimental validation.
12. Holthusen, H., Bergmann, A.: Investigations and measures to Int. J. Aeroacoust. 16(4–5), 274–298 (2017)
improve the acoustic characteristics of the German-Dutch Wind 33. Voss, C., Becker, K., Lawerenz, M.: Multi-objective optimization
Tunnel DNW-LLF. In: 18th AIAA/CEAS Aeroacoustics Confer- in axial compressor design using a linked CFD-Solver, In: ASME
ence, Colorado Springs, Colorado (2012) Turbo Expo 2008 Conference (2008)
13. Jensen, H.W.: Realistic Image Synthesis Using Photon Mapping. 34. von Heesen, W.: Abnormal high-level tonal noise in axial-flow
A.K. Peters Ltd, Natick (2001) (ISBN 1568811470) fans. In: 4th AIAA/CEAS Aeroacoustics Conference, Toulouse
14. Jensen, H.W., Christensen, N.J.: Photon maps in bidirectional (1998)
Monte Carlo ray tracing of complex objects. Comput Graph 19(2), 35. Waudby-Smith, P., Ramakrishnan, R.: Wind tunnel resonances
215–224 (1995) and Helmholtz resonators. Can. Acoust. 35(1), 3–11 (2007)
15. Martin, R.M., Brooks, T.F., Hoad, D.R.: Reduction of background 36. Welsh, M.C., Stokes, A.N., Parker, R.: Flow-resonant sound inter-
noise induced by wind tunnel jet exit vanes. AIAA J 23(10), action in a duct containing a plate—Part I: semi-circular leading
1631–1632 (1985) edge. J. Sound Vib. 95(3), 305–323 (1984)
16. Melber-Wilkending, S., Bergmann, A.: Aeroacoustic optimiza- 37. Wickern, G., von Heesen, W., Wallmann, S.: Wind tunnel pulsa-
tion of the NWB airline and turning vanes based on high fidelity tions and their active suppression. SAE Technical Paper 2000-01-
CFD and acoustic simulation. In: 18th AIAA/CEAS Aeroacoustics 0869 (2000)
Conference, Colorado Springs, Colorado (2012) 38. Woodward, R., Hughes, C., Berton, J.: Benefits of swept and
17. Moreau, A., Enghardt, L.: A first step towards a parametric model leaned stators for fan noise reduction. In: 37th Aerospace Sciences
for fan broadband and tonal noise. In: Proceedings of the DAGA Meeting, Reno, Nevada (1999)
Conference, Dresden, Germany (2008) 39. Woodward, R., Hughes, C., Jeracki, R., Miller, C.: Fan noise
18. Moreau, A., Gurin, S.: Development and application of a new source diagnostic test: Far-field acoustic results. In: 8th AIAA/
procedure for fan noise prediction. In: 16th AIAA/CEAS Aeroa- CEAS Aeroacoustics Conference, Breckenridge, Colorado (2002)
coustics Conference, Stockholm (2010) International standards:
19. Moreau, A., Gurin, S.: Similarities of the free-field and in-duct
formulations in rotor noise problems. 17th AIAA/CEAS Aeroa-
coustics Conference, Portland, Oregon (2011)
International standards
20. Nishimura, M., Kudo, T., Nishioka, T.: Aerodynamic noise reduc-
ing techniques by using pile-fabrics. In: 5th AIAA/CEAS Aeroa- 40. ISO 9613-1:1993: Attenuation of sound during propagation
coustics Conference, Bellevue, WA (1999) outdoors—Part 1: calculation of the absorption of sound by the
21. Paidoussis, M.P.: Flow-induced vibrations in nuclear reactors and atmosphere. International Organization for Standardization (1993)
heat-exchangers. In: Naudasher, E., Rockwell, D. (eds.) Practical 41. ISO 266:2003: Normal equal-loudness-level contours. Interna-
Experiences with Flow-Induced Vibrations. Springer, New York tional Organization for Standardization (2003)
(1980) 42. ISO 15186-2:2003: Measurement of sound insulation in build-
22. Papenfuss, H.D. et al.: Lowering the turbulence level in an open- ings and of building elements using sound intensity—Part 2: field
jet wind tunnel by novel adjustable jet-exit vanes. In: FVP Sym- measurements. International Organization for Standardization
posium on Aerodynamics of Wind Tunnel Circuits and their Com- (2003)
ponents, CP-585, Moscow (1996) 43. ISO 18233:2006: Application of new measurement methods in
23. Parker, R.: Resonance effects in wake shedding from parallel building and room acoustics. International Organization for Stand-
plates: some experimental observations. J. Sound Vib. 5(2), 330– ardization (2006)
343 (1966) 44. ISO 3745:2012: Determination of sound power levels and sound
24. Parker, R.: Resonance effects in wake shedding from parallel energy levels of noise sources using sound pressure—precision
plates: calculation of resonant frequencies. J. Sound Vib. 5(2) methods for anechoic rooms and hemi-anechoic rooms. Interna-
(1967) tional Organization for Standardization (2012)
25. Pott-Pollenske, M., von Heesen, W., Bergmann, A.: Acoustic pre- 45. ISO 717-1:2013: Rating of sound insulation in buildings and of
examination work and characterization of the low-noise wind tun- building elements—Part 1: airborne sound insulation. Interna-
nel DNW-NWB. In: 18th AIAA/CEAS Aeroacoustics Conference, tional Organization for Standardization (2013)
Colorado Springs, Colorado (2012)
26. Remillieux, M.C., Crede, E.D., Camargo, H.E., Burdisso,
R.A., Devenport, W.J., Rasnick, M., Van Seeters, P., Chou, A.:

13