characterization of the non - damped modal response … · characterization of the non - damped...
TRANSCRIPT
1
Characterization of the Non - Damped Modal Response of a Portuguese Guitar Including its Twelve Strings
Jorge A. S. Luis
Dept. Engenharia Mecânica, Instituto Superior Técnico Av. Rovisco Pais, 1049-001 Lisboa, Portugal
Abstract
The Portuguese guitar is a pear-shaped twelve-string musical instrument (six pairs of strings).
This thesis is dedicated to the study of the analysis of the un-dampened vibrational response of a
Portuguese guitar, including its twelve strings. This study is motivated by the interest in predicting, in
the project stage, which geometrical characteristics (dimensions) allow for a certain amount of modal
characteristics, in this case, focusing on the three first frequencies associated with the modes (0,0),
(0,1) and (0,2) in the guitar sound board. Besides the body, part of the arm is included in the
simulation, as well as the twelve strings tuned in the standard Lisbon Portuguese guitar tuning (B, A,
E, B, A, D). As far as the author knowledge is concerned, there is no study which such characteristics
in the current literature, connecting CAD (computer assisted drawing) with a un-dampened modal
analysis with pre-tension in the cables. The strings were tuned to the centesimal and the frequencies
of modes (0,0), (0,1) and (0,2) adjusted to values found in the literature.
Keywords: Portuguese guitar, structure, strings, vibration, acoustics.
1. Introduction
The goal of this thesis is the development of prediction methodologies for the structural and
acoustic response of a Portuguese guitar, through the analysis of the un-dampened modal vibrational
response, including its twelve strings, so as to predict, in the project stage, which geometrical
characteristics mainly dimensions, allow for a certain amount of modal characteristics, and doing so,
furthering the contemporary knowledge of these instruments, so that the results produced may be
used by the industries that build these instruments.
Guitar-makers like Ervin Somogyi [1], whose California produced guitars can reach 31000 dollars,
and Gerald Sheppard [2], compare hand-made guitars to industrially produced guitars in their articles,
where they stress the quality and uniqueness of hand built guitars, while recognizing the quality of
some industrially produced guitars as well as the unique characteristics of some of these guitars.
With the introduction of new materials and the further investigation of previously existing ones, it is
possible, in my opinion, to produce very high quality industriously produced guitars. In [3], “The sound
of a concert guitar will be clean every string and frets" adding that the goal of state-of-the-art
technology nowadays is to replace the subjective quality assessments usually associated with the
concert guitar, with simulations and experimental science, in order to increase quality and lower costs.
It follows that the introduction of models with the ability to produce an effective simulation, within
the possibilities of computer simulation, using the Finite Element Method (FEM), for instance, would
be invaluable for music instruments industry. It would also be an excellent contribution to applied
science.
2
In this work, a modal characterization of the structure with and without strings is reviewed, and
the mode tuning is explored, considering the presence of strings. As far as the author knowledge there
is no publication on FEM in this field that includes the strings.
1.1 Revision of the literature
One of the first studies of the modal response of a Portuguese guitar was actually made by
[4]. The authors obtained the modal characterization of the guitar soundboard in fully mounted Lisbon
and Coimbra guitars. They compared and described the frequency response curves for several
specimens, as well as some significant vibrational modes.
In the experimental setup the guitar was suspended by rubber bands on a rigid structure. The
authors made a detailed modal identification in one of the guitars. They defined a grid of 114 possible
impact points, in the sound board as well as the arm, so that they could identify possible coupled
movements.
The authors reached the following conclusions: 1) frequencies bellow 200 Hz are not
significant for the sound radiation, 2) frequencies between 121 Hz and 160 Hz are due to cavity
resonances (Helmholtz), and 3) between 250 and 450 Hz, there is, at least one resonance or
resonance group that is responsible for a significant part of the radiated spectrum (see figure 1). The
most important is the monopole mode (0,0) (shown in figure 1a), which is the mode that radiates the
sound more efficiently, as opposed to the longitudinal dipole (0,1) (see figure 1b), where the adjacent
antinodes move in anti-phase, eliminating the air movement inside the box. The tripole longitudinal
mode (0,2) (see figure 1c) shows up at 635 Hz. Others appear above these, but are not mentioned
here.
a) b) c)
Figure 1 - Modal forms of three resonances of one of the guitars in study: a) 𝑓 !,! = 275 𝐻𝑧 ; b)
𝑓 !,! = 360 𝐻𝑧; c) 𝑓 !,! = 6355 𝐻𝑧 (fonte: [4]).
In [5], the author modulates the Portuguese guitar, using MEF, for the first time. The modal
analysis of the guitar produces results that are compared with the experimental results in [4]. The
dimensions of the guitar and of the harmonic braces were supplied beforehand. From these
dimensions the author performed the modal analysis and obtained the results, which are the first 20
modes of vibration of the guitar, amongst which she compared the frequencies closest to the ones
obtained by [4].
In [6], the authors use a Portuguese guitar model that is based on the reference geometry [5]
to study the influence of the inclusion of the guitar strings and its respective modal response. The
3
twelve steel strings are stretched in a way that their natural resonances correspond to the frequencies
of the Portuguese guitar notes. For that, they use pre-tensions related to temperature variations in the
strings and so obtain the frequencies corresponding to the right tones of the Portuguese guitar.
2. Methodology
The strategy adopted in this study uses a CAD program to model the guitar’s geometry according
to the dimensions, which were taken from [7], of Álvaro Merceano da Silveira’s Portuguese guitar,
including harmonic braces, gluing belt (linings) and the tail block, that constitute the grid and arm with
thickness variation, see figures 2 and 3.
Simplifications: neither the part of the arm that supports the fan (string holding mechanism), nor
the volute (wood decorative piece) are considered in the adopted geometry, as these elements were
ignored to the modal behavior of the guitar and at this step the box is modeled with shell elements.
Due to the difficulties in finding the properties of some woods, the arm is modeled with mahogany
instead of Brazilian cedar, and the parts that should be modeled with spruce are modeled with Sitka-
Spruce. The side of the box opposed to the arm is projected from a circumference centered in the
exact point i.e. 205mm of radius instead of being bulging with 165mm of distance from the exact point
of the nut, see figure 2. The strings are all considered bulk (actually some are spiral shaped) and
modeled by beam elements. Instead of a continuum string, two are considered, one from nut (opposite
side of de arm) to bridge and another from the bridge to the nut (near head of guitar).
a) b) c) d)
Figure 2 – a) Álvaro Merceano da Silveira’s guitar (source [17]), b) CAD model, c) front view
of the model d) side view of the model.
In the Portuguese guitar, the soundboard, the back and the sides are glued through a gluing
belt that acts as reinforcement as well as being useful in the fixing of the braces. Also in the model
used, a grid is used that unites all the braces and the tail block, figure 3.
Figure 3 – Gluing grid with the braces a) Front view, b) side view c) perspective.
a) b) c)
4
Afterwards the geometry is exported to the finite elements program (EF) and the modal
analysis is performed with the goal of obtaining the natural frequencies and the vibrational modes of
the guitar. The obtained frequencies and modal forms for the three first modes, namely modes (0,0),
(0,1), (0,2) and respective natural frequencies are compared with the frequencies and the three first
vibration modes of the guitar tested experimentally by [4], see figure 1. This process is reproduced
interactively, varying, in each iteration, at least one of the following dimensions: gluing belt, harmonic
braces of the back or harmonic braces of the soundboard or all at once (three pairs of braces) or the
distances between them and the distance to the exact point (position of the bridge) while checking the
effect on the modal forms and frequencies. When the deviations relative to the experimental values of
[4] are considered acceptable, as far as this work in concerned, the bridge is positioned (protuberance
of 17 mm of height by 4 mm of length, in mahogany) on the soundboard, followed by the installation of
just one string, and finally the installation of all strings.
In the case of the strings the method used is based on the one proposed in [6]: insertion of a
pre-tension in the strings by introducing a reference temperature and the variation of the temperature
of each string, depending on the desired frequency, followed by a static analysis with pre-tensions and
modal analysis of the complete model (with strings). Thereby it is possible to create a tension in each
string, so that its natural vibration frequency in the model’s complete structure corresponds to the
string’s natural vibration frequency in the tuned guitar.
An iterative process is also used in the case of the installation of the strings, where the results
obtained with the model are compared with the experimental results of [4]. The installation of the
bridge and the strings produces a downward dislocation of the frequencies and change the vibration
modes. In this case, the variation of the distances and dimensions of the braces is not enough to
adjust the results of the EF analysis to the values of [4], so, the geometry of the guitar is corrected (by
replacing the 205mm radius in the half circumference opposite the arm by an elyptical arch that makes
the guitar bulging with 165mm of the exact point to the nut (opposite side of the arm), see figure 4.
With this new geometry and with all the components (bridge and strings), a new iterative adjustment is
made on the dimensions and distances of the braces, so that it is possible to obtain the frequencies
and modes of vibration adjusted to the values of [4]. To speed up the convergence of the iterative
process, several thicknesses for the soundboard are tried so as to obtain an estimate of the thickness
of the braces.
a) Figure 4 – Modified CAD model. a) perspective of the guitar with strings, b) front-view, c) side-
view.
a) c)b)
5
The final guitar geometry, external dimensions included, are thus determined by this iterative
process.
Special attention was given to work [5], which is the source of the material data, see table 1, with
the exception of the material used in the strings. In this work, the author begins with pre-determined
dimensions, that she keeps constant, and performs the modal analysis, comparing the frequencies
obtained with the frequencies in [4], not considering whether these modal frequencies correspond,
except in the case of the first one. In this case it becomes relatively easy to find all the coincident
frequencies with some modal form.
Table 1- Characteristics of the materials (source [5])
Wood Density (Kg/m3)
E1 (108 N/m2)
E2 (108 N/m2)
E3 (108 N/m2)
G23 (108
N/m2)
G13 (108
N/m2)
G12 (108
N/m2) ν12 ν13 ν23
Sitka-Spruce 390 116 9 5 0.39 7.2 7.5 0.37 0.47 0.25
Indian Rosewood 775 160 22 7.2 3 8.4 11 0.36 0.03 0.26
Ebony 1100 190 21.1 9.5 4 11.2 16.7 0.3 0.03 0.26
Mahogany 450 106.7 5.3 11.8 6.3 2.2 9.4 0.3 0.03 0.26
In [6], the author placed the strings in the model created and analysed by [5], with previously
defined dimensions and the same vibration modes and frequencies found by [5], which he kept fixed,
having afterwards adjusted the strings to the nearest hundredth. In this case the two last modes do not
correspond to the modes obtained by [4]. Nothing is said regarding the effect of the string placement
on the guitar body.
The material used to model the strings was a stainless steel AISI INOX 302, already
employed by [6], whose mechanical properties are presented in Table 2.
Table 2 – Stainless steel (AISI INOX 302) properties [6]. Density [Kg/m^3] 7900 Young's modulus [GPa] 193 Thermal Expansion [1/oC] 0.0000172
3. Dynamic analysis According to the theory of elasticity, the dynamic behaviour of a linear elastic solid, for small deformations, is (in Cauchy’s form) [8]
𝜎!",!!𝑓!!𝜌!𝑢! (1)
where 𝜎!" is the stress tensor, 𝑓! is the sum of the force vectors acting on the body, 𝜌! is the density of the solid, 𝑢! is the displacement vector and I, j = x, y, z.
The weak form can be obtained by the residual method whose function of choice is based on the Galerkin method. In this way, the approximate solution by finite methods in terms of nodal displacements can be written as:
𝑴𝒖 + 𝑲𝒖 = 𝒇 (2)
where 𝑢 is the nodal displacement, M is the global mass matrix, K is the global stiffness matrix and 𝒇
is the force vector. These matrices and force vector can be obtained by FE (Finite Element)
assembling of the following elements:
6
𝒎𝒆 = 𝜌𝑵!𝑵𝑑𝑉! (3)
𝒌𝒆 = 𝑩!𝑪𝑩𝑑𝑉! (4)
𝒇𝒆 = 𝑵!𝑭𝒗𝑑𝑉! + 𝑵!𝑭𝒔𝑑𝑆! (5)
were N is the matrix of the element’s shap function, B is the matrix of the extensions – nodal displacements, C is the constitutive law matrix, 𝑭𝒗 is the vector of the volume forces and 𝑭𝒔 is the vector of the surface nodal forces. Equation 2 restrained to static analysis gives
𝑲𝒖 = 𝒇 (6)
Considering thermal expansion due to temperature change and despised (for being too small) the variation of the elastic constants can apply the principle of superposition.
𝜎!" = 𝐶 !"#$ 𝜀!" − 𝛼!"∆𝑇 (7)
The tensions can be obtained through Hooke’s law. Rewriting (2) in the frequency domain for free vibration conditions we get
𝑲 + 𝒌𝒈𝒆𝒐𝒎 + 𝝀𝑴 𝒖 = 𝟎, (8)
where 𝜆 is a diagonal matrix of the frequencies squared and u is the nodal displacements vector
matrix (one per column) of the corresponding vibration modes and 𝒌𝒈𝒆𝒐𝒎 is the geometric matrix.
4. Results for the model without strings The problem analysed in this section corresponds to the determination of the first frequencies
of the guitar without the strings, with the boundary conditions of guitar free in space, as was
considered in the experimental model of [4]. Table 3 presents the final dimensions (the ones which
better approximate the modes and frequencies obtained by [4]), attained after several Finite Elements
(FE) analyses.
Table 3 – Dimensions and brace distances of the analysed guitar model.
(lxh mm) Top Back Distance to the exact point Between braces Gluing Belt
Brace 1 12x15 12x15 6x12 Brace 2 8x16 8x16 52 Brace 3 8x18 8x18 16
4.1 Mesh Convergence Table 4 presents the results obtained with four meshes, corresponding to the size parameters
of the finite elements which vary from 4,49𝑥10!! 𝑚 to 8,98𝑥10!! 𝑚.
Table 4 – Frequencies obtained with different meshes.
Modes
Mesh/numberofelements1stMesh8146
2ndMesh24077
3rdMesh95558
4thMesh108409
Frequency(Hz)1 306,59 291,27 283,93 283,042 409,73 376,13 371,02 369,623 605,95 552,7 553,85 549,6
Figure 5 presents three meshes, 1st, 2nd and 4th of Table 4, respectively.
7
Figure 5 - a) 1st mesh; b) 2nd mesh; c) 4th mesh.
In Table 4 we see that the variation between the third and the fourth meshes is not significant
anymore, and, therefore, it can be validated as a fairly accurate approximation for the objectives of this
work. It is true that the number of elements increases from mesh to mesh, which also corresponds to
an increase in computation time; however, since the computation time is not too high (130,8s of CPU
time), mesh 4 was chosen for this analysis.
4.2 Modal analysis Now we present the three vibration modes of the guitar, obtained through FE analysis, which
better approximate the modes and frequencies obtained by [4].
Figure 6 - Modal forms and frequencies. a) Mode (0,0); b) Mode (0,1); Mode (0,2)
Table 5 presents the results and the deviations of the frequencies obtained in the FE analysis,
in comparison with the results obtained by [4].
Table 5 – Frequencies of modes 1,2 and 3, experimental and obtained through FE.
4.3 Guitar with strings
At this point the objective is to place and tune the strings of the FE model.
The guitar is assumed to be free in space as in the experimental model of [4]. FE analyses are
performed with two different models using the mesh with the characteristics indicated in section 4.1.
The modal analysis is first performed in the CAD model, presented in the previous section, and
afterwards in the modified CAD model (see Figure 4).
Mode Shapes Frequency (Hz) Experimental FE Analysis Deviation (%)
1 275 283,04 2,84 2 360 369,62 2,6 3 635 549,6 13,45
a) b) c)
a) b) c)
8
4.4 Mesh
When the strings are placed and tuned, the CPU time increases drastically. It is important to
note that the strings are tuned iteratively and, therefore, the software has to be run several times in
order to tune each string and the whole set. Based on the results presented in Table 6, a mesh with
48855 elements was chosen.
Table 6 – Different meshes (with strings)
Modes
Mesh/numberofelements1stMesh21191
2ndMesh48855
3rdMesh77423
4thMesh118685
Frequency(Hz)1 283,78 268,83 266,89 267,612 395,81 380,76 378,57 377,593 536,34 511,06 507,67 506,1
Details of the strings and bridge are represented in Figure 7.
Figure 7 – Detail of the strings. a) Whole set of strings; b) Strings and bridge.
4.5 Modal analysis of the initial model with bridge and of the initial model with one string
Successive FE analysis of the model were undertaken: with bridge and without strings, with
one string, and with the whole set of strings, using in the model the dimensions of Table 3. The results
are presented in Table 7.
Table 7 – Frequencies of modes 1,2 and 3, experimental and obtained through FE analysis (initial
model).
Mode Shapes
Frequency (Hz) – Initial model
Experimental With bridge With bridge and one
string With all strings
FE Deviation (%) FE Deviation
(%) FE Deviation (%)
1º 275 233 18,03 227,96 17,45 193,05 29,8 2º 360 306,7 14,81 268,58 25,39 307,74 14,5 3º 635 449,27 29,25 442,07 30,38 426,56 32,8
In the modified model, the following FE analysis were carried out: with bridge and without
strings, with a 5 mm thick top and with the whole set of strings. As a result of an iterative
approximation to the results of [4], the braces 2 (see table 3) of the top and back were changed to
8x18 mm instead of 16x18 mm. In the case of the last column, which corresponds to the guitar with
the whole set of strings, the dimensions of braces 2 and 3 (see table 3 and figure 4) of the top and
back were changed to 8x25 mm. The obtained results are shown in Table 8.
a) b)
9
Table 8 – Frequencies of modes 1,2 and 3, experimental and obtained through FE analysis (modified model)
Mode Shapes
Frequency (Hz) – Modified Model
Experimental
With bridge (without strings)
With bridge and 5 mm of thickness of the top (without
strings) With all strings
EF Deviation (%) EF Deviation (%) EF Deviation
(%) 1 275 266,29 3,17 276,14 0,41 268,34 2,4 2 360 368,4 2,33 366,12 1,67 380,76 5,45 3 635 521,23 17,9 629,04 0,94 511,06 19,5
Figure 7a) shows string 11 in the fundamental form and the respective tuning frequency
rounded to the nearest hundredth. Figure 7b) shows the 3rd mode of vibration with all the strings
attached.
The fundamental frequencies of the remaining strings and the two remaining modes of
vibration were found in a similar way, see table 9.
Figure 7 – a) Frequency and mode shape of the eleventh string. b) All strings on guitar and third
modal shape of vibration
Table 9 – String diameters for the Portuguese guitar and experimental frequencies (source [6]).
Frequencies obtained through FEM
String
Diameter[mm]
Frequency values [Hz] – FEM
Frequency values [Hz] - from [9] B (b4) 0,24 493,880 493,880 A (a4)
0,25
440,000
440,000
E (e4)
0,32
329,630
329,630 B bordão (b3) 0,50 246,940 246,940 A bordão (a3)
0,64
220,000
220,000
D (d4)
0,44
293,660
293,660 D bordão (d3) 0,79 146,830 146,830
5. Results and conclusions.
For the initial guitar model without the strings, the modes (0,0), (0,1) and (0,2) were obtained
in the correct form. The frequencies estimated through FE analysis had deviations of 2,84%; 2,6% and
13,4%, respectively, when compared with the experimental values of [4].
From the dimensions of the braces and the relative distances between them and between the
braces and the exact point (where the bridge is placed, see Figure 2 c), only one of the dimensions
was changed: the middle braces (close to the hole, Figure 2 c) with 18x8 mm (height x thickness)
changed to 18x16 mm.
a) b)
10
In the case of the guitar (modified model) with strings, the correct modes were also obtained
and the deviations in the frequencies were of 2,4%, 5,45% e 19,5%, respectively, when compared to
the experimental values of [4]. Regarding the geometry, two pairs of central braces were changed,
from 18x8 mm to 25x8; furthermore, the guitar was redesigned in order to better approximate the
reference model from [7], thus improving the results. The strings were tuned to the nearest hundredth
through the application of a pre-tension to each string. The effect of the bridge placement on the two
previous cases was also verified.
These results support the hypothesis that the behaviour of the guitar can be reasonably well
predicted during the design stages, demonstrating that it is possible to quantify the effects caused by
changes in the guitar structure, bridge and string placement and string tuning.
Acknowledgments to supervisor
I would like to thank my supervisor, Professor Miguel de Matos Neves, for the support and
advice provided during the writing of this thesis, as well as for the availability to make our schedules
compatible for the orientation meetings.
References
[1] Ervin Somogyi, “Some thoughts on the difference between handmade and factory made guitars”,
http://www.esomogyi.com/handmade.html.
[2] Gerald Sheppard ,The Question of Handcrafted Versus, Mass-Produced Instruments, 2015,
(http://www.sheppardguitars.com/handbuilt_vs_manufactured.htm)
[3] Torres T. Jesús Alejandro, “Análise modal de la tapa armónica de una guitarra clássica mediante
la aplicación de Ansys”, Instituto Tecnológico de Querétaro. p ( 2).
[4] O. Inácio, F. Santiago and P. Caldeira Cabral, “The Portuguese Guitar Acoustics: Part 1 –
Vibroacoustic Measurements”, paper ID: 187 p (2). 2004
[5] Raquel Vaz, “Modelo de elementos finitos da Guitarra Portuguesa”, Universidade Técnica de
Lisboa, Instituto Superior Técnico, Mestrado em Engenharia Aeroespacial 2011/12
[6] João P. B. Lourenço and Miguel M. Neves “Modal Analysis of a Portuguese Guitar using a 3D
Finite Elemente Model with string tension” Acoustics Conference 2013 and Sound and Music
Computing Conference, pp.1-8, 2013.
[7] http://www.jose-lucio.com/Construir_2008/Constuir_01.htm
[8] Claus Jurgen Bathe, Finite Element Procedures, Prentice Hall, New Jersey, 1982.
[9] José Maria Campos dos Santos e Guilherme Orelli Paiva, “Vibroacoustic Numerical Analisys of a
Brazilian Guitar Resonance Box”, 2013 ESSS Conference & Ansys Users Meeting, Atibaia, Brasil,
Abril 2013.