In Chapters 1–3, wave propagation is considered in unbound media, and boundaries occurred only in the context of scattering, reflection, or radiation. Beams or shells were characterized by their boundaries, but their effect was the creation of different forms of wave propagation. For example, the dimension (thickness) of a plate causes the different wave types, bending, in-plane shear, and dilation, but in principle there was always at least one infinite dimension. In Chapter 4 we will approach the system from Chapter 1, meaning that we consider bounded systems, for example a closed cavity with specific input and output positions.
A one-dimensional subsystem can be a small tube with radius small enough that the wavelength is larger than the diameter for the highest considered frequency (Pierce, 1991). The tube has rigid surfaces and is filled with a fluid, e.g. air with given density and speed of sound .
The solution of the wave Equation (2.42) in the tube is given by the one-dimensional solution of the wave equation for pressure (2.34) and velocity (2.36)
Interaction with the tube fluid takes place at the end and the beginning of the tube. So, we look for a representation similar to a 2DOF system (1.73) but with the velocities and instead of the displacement. We must clearly separate between state and external variables in order to avoid confusion. So, we separate between the state pressure , denoting the pressure at the right side at and at the left side at . A positive pressure at leads to a force in the negative -direction that must be compensated by a positive pressure due to the face normal and -coordinate both pointing in positive directions. At a positive pressure leads to positive blocked force in the -direction that must be compensated by a negative pressure . We must keep this in mind when we introduce external forces and for coping with the transfer matrix method. Consequently the external force for the system description is replaced by the pressures and
Using Equation (4.2) at the related positions of we get for the velocities
The difference (4.4) (4.5) gives
and entering this into (4.1) at we get the following expressions for and depending on the velocities
This provides the following 2DOF system matrix in impedance form
This is the matrix description of the tube as a system with two (external) DOFs. If the velocities are given, the system response can be derived directly. For calculating the response to pressure excitation, the matrix must be inverted to give the mobility form
The pressure and velocity field can be derived from (4.1) and (4.2) and the solutions for and :
In order to interpret the above result we reduce our system to a 1DOF system by setting an impedance boundary condition at . Thus,
Using Equation (4.10) we get for a given velocity amplitude the pressure response at
In a first step we consider a rigid boundary condition with leading to
and from that the radiation impedance assuming a damping according to (2.59)
The tan function is zero at for when we assume a real . This is the case for when half wavelengths fit into the tube. The wave transmitted by the piston at is reflected at and matches perfectly in phase at . So, the pressure would add up to infinity without damping, and this is called the resonance or first mode of the tube.
Solving expression (4.17) for gives for the amplitude of at the resonances
In Figure 4.2 the result is shown for values of . For low frequencies () we see a clear deterministic behavior. The impedance switches between the typical values. Going higher in frequency, the peaks are very close to each other. For high frequencies, there is no reflected wave because of the damping. Consequently, this is like radiation into half space and the impedance equals the characteristic impedance of air.
The power introduced to the subsystem is according to (2.45) the acoustic intensity times the surface area. For this we link the wavenumber to the damping loss and separate the real from the imaginary part
Thus the introduced power is
and for the input power becomes
Due to the absence of reflection by damping the radiation into the tube equals radiation into the free field.
In order to derive the interior pressure field inside the tube, we replace in Equation (4.5) by
After some algebra we get for the two amplitudes
Entering this into (4.1) and (4.2) we get
and for a rigid end this yields
In Figure 4.3 the pressure and velocity amplitudes along the tube axis are shown.
In Chapter 1 it was shown that every dynamic and closed system has natural states of vibration: the modes of the system. For the harmonic oscillator that was the resonance of the system, and for a system of connected springs and masses that was the modal solution (1.104). Continuous and bounded systems also have modes and natural shapes of vibration linked to a certain frequency.
Here, we are looking for the mode shapes of our system with given boundary conditions, eg. rigid conditions such as . So, the solution (4.2) must fulfil these conditions:
The first equation gives . Inserting this in the second equation leads to the following solutions for pressure and velocity
The choice corresponds to a static pressure. These wave numbers correspond to modal frequencies and wave lengths
Mode shapes are natural shapes of the system. That means the system tends to vibrate in this form at the specific frequency. We know that the mode shapes can be used for coordinate transformations such as when the state of the system is given in modal coordinates
Entering this into the one-dimensional inhomogeneous Helmholtz Equation (2.123) gives
For the derivation of the modal coordinates we need the orthogonality relations according to equations (1.111) and (1.112). As we don’t have a discrete system the vector product is replaced by a scalar product of the continuous mode
The complex conjugate does not have any effect, as the mode shape is supposed to be real. But in later application we may have complex modes that require the scalar product in complex form. For practical reasons we normalize the modes with
so that
We make use of the orthogonality by multiplying Equation (4.34) from the left with , and keeping in mind that the source term contains the volume source strength density , so that we have to multiply with the tube cross-section to get the dimensions right
providing an expression similar to (1.28) and (1.126)
An equivalent relationship can be found with the free boundary modes.
When systems are excited by broadband signals, it is relevant how many modes are in the frequency range of interest. The more modes are available, the higher the probability for the exciting force to excite resonators. We define the number of modes until frequency as
is the distance between each modal frequency as shown in Figure 4.4. The modal density is the number of modes per band given by
These quantities are indicators of the dynamic complexity. A high number of modes in a band may denote high dynamic complexity. In addition they are criteria for the probability to excite a mode when excited by signals with a specific bandwidth.
Damping is considered with Equation (2.59) by using a complex wavenumber
With this assumption we end up with a modal damping according to (1.63) and (1.127). So, the damping in the wave propagation is in line with the modal damping expressions from Chapter 1.
The first example with a piston excitation at is taken as first application of the modal method. According to Equation (4.39) the response is
We assume a piston vibrating with at , so the volumetric source strength density would be . Using the sifting property of the delta function the result is
and transforming back into real space with (4.33) gives
Obviously, in numerical implementation the infinite series must be replaced by a certain number of modes. As we now the exact solution from Equation (4.26), we can calculate the error
A modal solution may require a large number of modes to synthesize the exact solution similar to the Fourier transform of a rectangular signal. If the solution is similar to a mode, saying that is chosen as , the approximation is naturally exact if the damping is not too high. In Figure 4.5 six modes () in the sum of Equation (4.45) are shown. On the left hand side the system is excited at resonance frequencies, the upper two are well represented, but the case shows that six modes are not sufficient. The required shape to represent the solution is simply not available in the used mode set. When the system is excited out of resonance as shown on the right hand side, even for low frequencies the representation is not perfect, the different colors show increasing numbers of modes. Investigating the absolute value of the modal coordinates as shown in Figure 4.6 the above considerations are confirmed. The out-of-resonance excitation requires a reasonable number of modes, with decreasing contribution, the more the modal frequency is separated from the frequency of excitation.
Choosing a high value of we see that due to large damping even the result of resonance frequencies leads to visible deviations, but for both curves coincide. Consequently for out-of-resonance frequencies 100 modes are required for reasonable precision. In other words, for systems large compared to the wavelength and with damping, a larger number of modes is required to describe the dynamics correctly.
In this section we deal with large three-dimensional cavities or rooms. We focus on simple geometries to describe the principle behavior over the full frequency range and for being able to work out the details on analytical solutions. Treatments on room acoustics can be found for example in Morse and Ingard (1968), Pierce (1991) and Jacobsen (2008).
We consider a room of rectangular shape as shown in Figure 4.8. We solve the Helmholtz Equation (2.42) in Cartesian coordinates with rigid boundary conditions . With (2.35) this yields
We assume that the solution can be factorized, meaning that the solution of can be expressed as the product of three independent functions of , and such as
Entering this into (2.42) gives
All terms must be independent from the other space variables, e.g. the term with is independent from and equals .
From Equation (4.51a) follows the constraint
The solutions for the equations can be adapted from the tube case
and the factorisation leads to
The rigid boundary condition at and implies that and and leading to a set of cosine functions
The boundary conditions at the opposite walls and require that a natural multiple of half-wavelength fit into the room dimensions
with and . Finally, we denote every mode of the room with
and the solution in modal coordinates is
Here, we have to include static modes in any direction. Thus is defined as
For simplicity the three indexes are combined into one index
where represents the three integers and . Using Equation (4.53) and (4.57) we get an expression for the natural frequencies of the room
The modes are orthogonal as will be shown by entering (4.58) into the Helmholtz Equation (2.42)
Multiplying the first with and the second with leads to the difference
that can be rewritten as
In accordance with the one-dimensional system we integrate over the volume and apply Gauss's theorem on the first term
with denoting the room surface. At rigid walls any normal component of the gradient is zero according to (4.47)–(4.49), so the functions are orthogonal regarding the three-dimensional functional
It can be shown as in (4.35) that the modes can be normalized with the following expression
The number of modes in a three-dimensional system must be naturally higher than for one-dimensional systems. We can count the possible modes below a specific frequency using Equation (4.62). As the mode count becomes important when the system gets complex and thus many modes exist, we need a more appropriate way to estimate the number of modes that occur until frequency or .
A geometrical approach is used to estimate the mode count with some corrections as proposed by Maa (1939). In Figure 4.9 the plane of modal wavenumbers for the tangential modes is shown. Each oblique mode occupies a full rectangle of area in the quarter circle of radius . The number of modes would be the quarter circle area divided by the rectangle area. However, the axial modes with additionally or are not correctly considered. Only half of their occupied area is taken into account by the quarter circle. This is corrected by adding half-rectangles of area and . The static mode is covered by one quarter through the quarter circle, but two quarters are already compensated by rectangular area. So, only a correction by is necessary.
The total area is given by sum of
The total area of all these components divided by the area of one oblique mode area provides an estimate for two-dimensional systems.
For the three-dimensional wavenumber grid in -space a similar compensation is required. See Figure 4.10 for details. The octant of radius is divided by the cube space any oblique mode is occupying, i.e. to estimate the mode count. The oblique modes with all are considered correctly.
The cubes corresponding to the tangential modes are cut in half; this is compensated by three quarter slabs. Only the quarter of the axial modes is covered by the octant, but one half is already considered by the two adjacent quarter slabs. So, we compensate by the columns with length and quarter area. To be fully consequent the static mode is considered by one-eighth from the octant and two times the contribution of the three slabs and columns. Finally, a missing one-eighth must be added to the full volume.
The summarized total mode count volume has the following components:
Thus, the total volume divided by the volume on one mode is
Introducing the quantities perimeter and surface we get the approximate expression for the mode count formulated for basic geometric quantities , , and
and for the modal density
In Figure 4.11 the exact curves derived by counting all modes with and the estimated curves from function (4.70) are shown. At high mode count the estimation is precise enough. The more irregular the cavity the worse the estimation becomes. Irregular cavities mean having a large perimeter and surface compared to the volume, hence flat and column shaped cavities.
Figure 4.12 shows precise and estimated modal density for the same rooms. Here, the derivative leads to more sensitivity to small errors. Irregular cavities lead to inaccurate modal density estimations.
The mode shapes of the room can be used to calculate the solution of the inhomogeneous Helmholtz Equation (2.123) similar to the one-dimensional case in this chapter. For consideration of absorbing boundaries we include the treatment on non-rigid cavity surfaces. We start with
Multiplying from the left with and integrating over the room volume leads to
With the Green identity, the first term on the left hand side can be converted and we get
The second term in the surface integral is zero because of the boundary conditions in the mode shape determination. The pressure gradient can be replaced with (2.35)
and This equation can be used for implementing boundary conditions, for example vibrating surfaces. Here we consider a locally reacting surface with
Using this in (4.73) and the modal sum for the pressure
we get
and by using the orthogonality relationship (4.68) this reads as
The surface integral leads to mode coupling, meaning that all modes contribute to the -th modal coordinate. We define
In rooms with nearly rigid surfaces the off-diagonal components can be neglected and only is considered. Using this leads to
From the source term we would get the modal coordinates of the Green’s function of rectangular rooms:
or the Green’s function as a series
This is the generalized Green’s function of the rectangular room, because it provides the point source response under given boundary conditions. Using acoustic point sources and equations (4.80) and (4.76) we get the modal response of the rectangular room with point source of strength located at
The practical aspect of the above equation is that it considers damping due to surface absorption and propagation simultaneously. For rigid walls vanishes and the equation corresponds again to the usual modal response form.
Whether the Green’s function or the modal response form (4.83) is used we may calculate the response of the system. For the understanding of low- and high-frequency behavior we investigate the source characteristics and wave field of point sources. They can by located in the room volume, at the walls, edges, or corners. In our first considerations we neglect wall absorption and apply an overall modal damping.
A point source at the wall is represented by a small piston with as shown in Figure 4.14. In this case the velocity at the wall leads to the following volume source
Together with (4.84) the response can be calculated. This expression can also be derived by introducing the vibration as a boundary condition applying . As we need the radiation impedance for later applications the mechanical impedance is derived by using the source area .
This impedance can be used when vibrating surfaces are described by discrete meshes. Every vibrating node is related to a specific area and thus experiences a mechanical radiation impedance from the fluid. In Figure 4.13 the mechanical impedance of a point source in the room wall is shown and compared to the impedance of the circular piston. In the range of validity for the point source, i.e. or the real part of the room point impedance equals the piston result.
With the above defined impedance the radiated power is given by
From the considerations in section 2.4.1 we know that the radiation impedance is more appropriate to describe radiation of point sources. For the rectangular room this reads as
for the power radiated into the room we get
The room itself is assumed to be source free, so . Therefore, there is no contribution from the volume term in the Kirchhoff integral (2.137). We assume a rectangular piston in the wall at ; the dimensions are as shown in Figure 4.14.
The surface integral over the room surface is zero except at the wall at and the limit of the rectangular surface
This integral can be analytically solved for the modal Green’s function from Equation (4.82)
Assuming the solution for is
With Equation (4.92) the pressure field due to the vibrating piston can be derived. But, we are interested in the radiation impedance to the piston, in order to get the power radiated into the room.
For this we have to integrate the pressure over the piston as in Equation (2.146)
because we have to validate the same integral again. In Figure 4.15 the impedance is shown. Please note that the radiation impedance approaches in the free field limit.
For most cavity shapes and geometries an analytical expression of a closed form solution is not possible. This is the field of numerical solutions. Due to the homogenity of the fluid medium (or the existance of the Green’s function) there are two options for numerical solutions:
FEM Finite Element Method. Here the fluid is discretized into small fluid elements, of for example tetrahedal, pentahedral, or hexahedral shape. Most solvers discretize the wave equation for pressure.
BEM Boundary Element Method. The second option is, to make use of Equation (2.137) and to integrate over elements of a surface mesh. Thus, the mesh is a surface mesh, which makes the meshing procedure mush easier. But, the resulting matrices are numerically complex.
The theories of FEM and BEM are not part of this book, but it makes sense to create model conceptions about the meaning of acoustic finite elements.
In Chapter 1 we introduced a discrete equation of motion that was created automatically by using discrete elements as springs, dampers and masses. The discrete acoustic wave equation of a fluid looks similar to equations (1.89) or (1.103) but with a different meaning of the matrices.
is the acoustic mass matrix, but has nothing to do with masses. It is called the mass matrix because of the factor. The coefficients correspond to the inverse bulk modulus of the discrete fluid element. So, it is more a stiffness related quantity. The acoustic stiffness matrix corresponds to the inertia of the fluid element. Thus, in contrast to mechanical systems where the discrete equations can be visualized by a set of point masses and springs we have to find a different model.
This is because the pressure is the degree of freedom in the wave equation. When we would have chosen the displacements the discrete equation would be more alike to mechanical systems.
The sum of all three matrices is the acoustic dynamic stiffness matrix . It is frequency dependent and complex. However, it is helpful to develop physical models for the discrete presentation and the components of the discrete acoustic mass and stiffness matrix.
As shown in Figure 4.16 there are two elements to be considered. The fist element is a local compressible volume . For the change in volume per pressure change we have
Expressing the change in volume and pressure by complex amplitudes and using this reads as
and with the source term from (4.95) we get
Thus, the components of the mass matrix in Equation (4.95) correspond to element volume and the compressibility or inverse bulk modulus.
In contrast to this, the second lumped element we must deal with is an acoustical mass. In the simplified approach of the lumped acoustic network, we assume an incompressible fluid of mass . That means the wavelength in the fluid must be larger than the element length, to fulfill that condition. The relationship between the fluid velocity and pressures follow from Newton’s law
This reads as
using . When we consider the external flow into the volumes is considered as positive and thus the system of equations for the mass element is:
The stiffness matrix of the acoustic system is linked to the mass of the connecting fluid pipes.
The exchange of physical meaning results from the exchange of state and excitation variables. The mechanical equation of motion was defined for displacement and the excitation by forces; in acoustics the state variable is pressure (better related to forces) and the excitation given by volume flow (better linked to displacement).
The sum of all volume flows at each node must equal the externally introduced volume flow . The final set of equations for the system shown in Figure 4.16 is as follows
When dealing with one-dimensional fluid systems such as air conditioning, mufflers, and hydraulic pipes the system can be described by frequency dependent system functions without discretisation. We learned above that the appropriate quantities for an acoustic network are pressure and volume flow . Thus Equation (4.11) must be slightly modified to get this relationship for one-dimensional fluid pipes by multiplying both lines by .
The internal pressure and velocity field can be derived from (4.1) and (4.2) now solving and for the pressure, because this is the state variable in our network, or by simply multiplying (4.12) by with consideration of the differing propagation directions of both waves.
When inspecting equations (4.83) and (2.137) one can see that the source and receiver locations can be exchanged. That means that in an acoustic system that relates the relationship of a source strength at location 1 to a pressure at location 2, source and receiver positions can be exchanged; hence
Globally the reciprocity is a general principle for linear systems. Thus, the only restriction to the variables is that they must be conjugate, meaning their product is related to power; thus, we use here and . The same principle holds for all linear networks, structures, etc. One conclusion from this principle is that the stiffness matrix from Equation (4.95) must be symmetric to fulfill the reciprocity relationship.
18.221.140.111