0% ont trouvé ce document utile (0 vote)

69 vues135 pages

Mems

Transféré par

Nous prenons très au sérieux les droits relatifs au contenu. Si vous pensez qu’il s’agit de votre contenu, signalez une atteinte au droit d’auteur ici.

Formats disponibles

Téléchargez aux formats PDF, TXT ou lisez en ligne sur Scribd

0% ont trouvé ce document utile (0 vote)

69 vues135 pages

Mems

Transféré par

Nikola Vidaković

Nous prenons très au sérieux les droits relatifs au contenu. Si vous pensez qu’il s’agit de votre contenu, signalez une atteinte au droit d’auteur ici.

Formats disponibles

Téléchargez aux formats PDF, TXT ou lisez en ligne sur Scribd

Conception, fabrication and characterization of a MEMS

microphone
Jaroslaw Czarny

To cite this version:

Jaroslaw Czarny. Conception, fabrication and characterization of a MEMS microphone. Acoustics
[physics.class-ph]. INSA de Lyon, 2015. English. �NNT : 2015ISAL0003�. �tel-01247487�

HAL Id: tel-01247487

https://theses.hal.science/tel-01247487
Submitted on 4 Jan 2016

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
N◦ d’ordre : 2015 ISAL 0003 Année 2015

Thèse

Conception, fabrication et
caractérisation d’un microphone
MEMS
(Conception, fabrication and characterization of a MEMS
microphone)

Présentée devant
L’institut national des sciences appliquées de Lyon
pour obtenir
Le grade de docteur
Formation doctorale : Acoustique
École doctorale : MEGA - Mécanique, Énergétique, Génie Civil, Acoustique
Préparée au CEA Grenoble en collaboration avec le LVA de l’INSA LYON

Par
Jaroslaw Czarny
(Master of Science, Engineer)

Soutenue le 27 janvier devant la Commission d’examen

Jury MM.

Directeurs de thèse: J-L. Guyader Professeur (INSA de Lyon)

E. Redon Maı̂tre de Conférences (Université de Bourgogne)
Rapporteurs: B. Dubus Directeur de Recherche (CNRS)(IEMN)
N. Joly Maı̂tre de Conférences HDR (Université du Maine)
Examinateurs: M. Collet Directeur de Recherche (CNRS)(E. C. de Lyon)
M. Ouisse Professeur (ENSMM)
T. Verdot Docteur Ingénieur (CEA-LETI)
Invité: A. Walther Docteur Ingénieur (Infineon Technologies AG)

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Electret microphones dedicated to consumer electronics and medical applications (hear-

ing aids) have reached the miniaturization limits. Since the release of the first micro-
phone based on Silicon micromachining, electret microphones are constantly replaced
by MEMS microphones. MEMS (Micro-Electro-Mechanical Systems) microphones use
Silicon that provides exceptional mechanical characteristics along with good electric
properties and mature fabrication technology. Regardless of the transduction principle
(capacitive, piezoresistive, piezoelectric, optical), all of the MEMS microphones reported
in the state of the art literature are based on a membrane deflecting out of the plane of
the base wafer. Most of the reported microphones and all of the commercially available
MEMS use capacitive transduction. Downscaling of capacitive microphones is problem-
atic, since the sensitivity depends on capacitance value. Moreover capacitive sensors
suffer of high sensitivity to parasitic capacitance and nonlinearity.
The drawbacks of capacitive detection may be overcome with use of piezoresistive prop-
erties of Silicon nanowires. Unlike the classical piezoresistors integrated into silicon
membrane, suspended nanowires do not suffer of leakage current. Further improvement
of piezoresistive detection is possible since the longitudinal piezoresistive coefficient rises
inversely proportional to nanowire section.
This thesis presents the considerations of novel MEMS microphone architecture that
uses microbeams which deflect in the plane of the base wafer. Signal transduction
is achieved by piezoresistive nanogauges integrated in the microsystem and attached to
the microbeams. Acoustic pressure fluctuations lead to the deflection of the microbeams
which produces a stress concentration in the nanogauges. Accurate simulations of the
discussed transducer couple acoustic, mechanical and electric behavior of the system.
Due to micrometric dimensions of the MEMS acoustic system, thermal and viscous
dissipative effects have to be taken into account. To reliably predict the sensor behavior
two acoustic models are prepared: the complete Finite Element Model based on the
full set of linearized Navier-Stokes equations and the approximative model based on the
Lumped Elements (Equivalent Cirtuit Representation). Both models are complementary
in the design process to finally retrieve the frequency response and the noise budget of
the sensor.
The work is completed by the description of the technological process and the chal-
lenges related to the prototype microfabrication. Then the approach to the MEMS
microphone characterization in pressure-field and free-field is presented.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Les microphones à électret dédiés à l’électronique grand public et les applications

médicales (les audioprothèses) ont atteint les limites de la miniaturisation. Depuis
la sortie du premier microphone basé sur une technologie microsystème sur silicium
(MEMS: Micro-Electro-Mechanical Systems), les microphones à électret sont progres-
sivement remplacés par les microphones MEMS. Les MEMS utilisent le silicium car il of-
fre des caractéristiques mécaniques exceptionnelles avec de bonnes propriétés électriques
et la technologie de fabrication est maintenant bien maı̂trisée. La plupart des micro-
phones MEMS qui sont décrits dans la littérature sont constitués d’une membrane qui
vibre en dehors du plan du capteur, et utilisent la transduction capacitive. La minia-
turisation de tels microphones est limitée car leur sensibilité est liée à la valeur de la
capacité qui dépend de la taille de la membrane. En outre, les capteurs capacitifs sont
très sensibles aux capacités parasites et aux non-linéarités.
Cette thèse présente une nouvelle architecture de microphone MEMS qui utilise
des micro-poutres qui vibrent dans le plan capteur. La transduction du signal est
réalisée par des nanojauges piézorésistives intégrées dans le microsystème et attachées
aux micro-poutres. Ce système de détection original ne présente pas les inconvénients
de la détection capacitive et à la différence des piézorésistors classiques intégrés dans la
membrane de silicium, les nanofils suspendus permettent d’éliminer les courants de fuite.
De plus, l’amélioration de la détection est possible puisque le coefficient piézo-résistif
longitudinal est inversement proportionnel à la section du nanofil.
Les fluctuations de pression acoustique entraı̂nent les déviations des micro-poutres
qui produisent une concentration de contraintes dans les nanogauges. Le comporte-
ment du capteur, que l’on cherche à modéliser, est lié à des phénomènes mécaniques,
acoustiques et électriques qui sont couplés. En raison des dimensions micrométriques
du MEMS, les effets des dissipations thermique et visqueuse doivent être pris en compte
dans le comportement acoustique. Pour prédire de façon fiable le comportement du
capteur, deux modèles vibroacoustiques sont utilisés: un modèle éléments finis basé sur
l’ensemble des équations de Navier-Stokes linéarisées et un modèle approché basé sur
un schéma à constantes localisées (représentation par circuit électrique équivalent). Les
deux modèles sont complémentaires dans le processus de conception pour déterminer la
réponse en fréquence et le taux de bruit du capteur.
Le travail est complété par la description des processus technologiques et les défis liés
à la fabrication du prototype. Puis deux approches pour la caractérisation fonctionnelle
du microphone MEMS sont présentées, la première en tube d’impédance, la seconde en
champ libre.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

First of all I would like to show my gratitude to all the people that I had a pleasure
to work with during my PhD. I am very grateful that during this time I was never left
alone and every arising diﬃculties could have been discussed with more experienced
co-workers.
J’aimerais tout d’abord remercier mon directeur de thèse, Jean Louis Guyader d’avoir
accepté ma candidature. Les remerciements vont aussi à Emmanuel Redon and Thierry
Verdot, qui ont bien eu la gentillesse de m’avoir suporté durant ces trois années. J’ai
particulièrement apprécié toutes les discussions que nous avons pu avoir et tous les
conseils que je pu en tirer. Je remercie aussi Arnaud Walther, d’avoir eu conﬁance en
moi et de m’avoir donné l’opportunité de travailler sur les microphones MEMS et d’avoir
encadré mon travail durant les deux premières année de ma thèse. Il m’est important de
souligner le travail du jury et de rapporteurs qui ont lu mon mémoire et ont contribué
à l’évaluation de mon travail.
Je voudrais remercier à toutes les personnes travaillant à la division M&NEMS du
Laboratoire Composants Micro Capteurs, pour leur travail en amont et qui a sensible-
ment contribué à la fabrication des microphones MEMS. Mes remerciements vont aussi
à Hélène Lhermet, Brigitte Desloges et Audrey Berthelot, impliquées dans le projet
MADNEMS, pour leur travail de gestion et de fabrication de microphones.
La conception des microphones n’aurait pas été possible sans le travail des person-
nes du Laboratoire Vibrations Acoustique à Lyon. Je leurs remercie aussi tout partic-
ulièrement, je remercie aussi Cécile Guianvarc’h, Thierry Verdot, Emmanuel Redon et
Kerem Ege de m’avoir présenté les aspects d’acoustique.
Je remercie aussi tous mes collègues d’avoir mis une ambiance très positives au
bureau et pour toutes les discussions qui s’avéraient très productives.
Dzie˛ kuje˛ ”polskiej rodzince” za wspólne wieczory, pikniki w parku i wycieczki.
Na koniec dzie˛ kuje˛ najbliższym mi osobom. Bez waszego nieustaja˛ cego wsparcia i do-
brego slowa napisanie niniejszej pracy nie bylo by możliwe. Moje myśli wedrowaly ku
was przez cale trzy lata spe˛ dzone we Francji.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

1 Introduction 1
1.1 General introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 International context of this work . . . . . . . . . . . . . . . . . . . . . . 6
1.3 The “MADNEMS” project . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4 Organization of the manuscript . . . . . . . . . . . . . . . . . . . . . . . 13

2 New architecture of a MEMS microphone 15

2.1 Concept presentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Possible transduction mechanisms . . . . . . . . . . . . . . . . . . . . . 16
2.2.1 Capacitive transduction . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.2 Piezoresistive transduction . . . . . . . . . . . . . . . . . . . . . 19
2.2.3 Transduction mechanisms comparison . . . . . . . . . . . . . . . 22
2.3 The M&NEMS technology . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.1 Technological platform . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.2 M&NEMS sensors review . . . . . . . . . . . . . . . . . . . . . . 24
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

3 Models for simulations of acoustic phenomena in MEMS 27

3.1 Propagation of sound wave in dissipative fluid . . . . . . . . . . . . . . . 27
3.2 Scaling effects in acoustics . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3 Full Linearized Navier-Stokes model (FLNS) . . . . . . . . . . . . . . . . 32
3.3.1 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3.2 Problem formulation . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3.3 Solving method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3.4 Means of FLNS model enhancement . . . . . . . . . . . . . . . . 36
3.4 Low Reduced Frequency model (LRF) . . . . . . . . . . . . . . . . . . . 37
3.4.1 Problem formulation . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.5 Benchmarking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.5.1 Finite Element Models based on Linearized Navier-Stokes equa-
tions (LNS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.5.2 Low Reduced Frequency model . . . . . . . . . . . . . . . . . . . 43
3.5.3 Results discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.6 Towards Lumped Element model . . . . . . . . . . . . . . . . . . . . . . 48
3.6.1 Validation of Lumped Element model . . . . . . . . . . . . . . . 48
3.6.2 Acoustic resistance of rectangular guide . . . . . . . . . . . . . . 49
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
4 Sensor design considerations 52
4.1 Architecture properties of considered MEMS . . . . . . . . . . . . . . . 52
4.2 Transduction chain - assemble of the microphone subsystems . . . . . . 56
4.3 2D finite element analysis of MEMS with fluid-structure interaction . . 57
4.3.1 Structure of the model . . . . . . . . . . . . . . . . . . . . . . . . 57
4.3.2 Discussion on the MEMS chip acoustic properties . . . . . . . . . 59
4.3.3 Influence of MEMS chip-scale packaging on the microphone response 63
4.4 Detailed discussion on the mechanical system . . . . . . . . . . . . . . . 66
4.4.1 Analytical model . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.4.2 FEM model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.5 Truss structured beam - from 3D to 2D simulations . . . . . . . . . . . . 72
4.6 Electrical system transfer function . . . . . . . . . . . . . . . . . . . . . 76
4.7 Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.8 Noise budget . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.8.1 Origins of the noises . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.8.2 Equivalent circuit representation of the microphone for the noise
budgeting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.8.3 Total transduction chain noise . . . . . . . . . . . . . . . . . . . 83
4.9 Influence of overetch on the microphone response . . . . . . . . . . . . . 85
4.10 Tolerance to mechanical shocks and sound pressure overload . . . . . . . 87
Summary of sensor specifications and discussion . . . . . . . . . . . . . . . . 89

5 Technological implementation 92
5.1 Process flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.2 Advances and difficulties encountered in microfabrication . . . . . . . . 95
5.3 Out-of-plane deflection of MEMS . . . . . . . . . . . . . . . . . . . . . . 96
5.4 Chip scale packaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.4.1 Technology overview . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.4.2 Proposed solution . . . . . . . . . . . . . . . . . . . . . . . . . . 102
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

6 Approach to MEMS microphone calibration 105

6.1 Selection of suitable calibration methods . . . . . . . . . . . . . . . . . . 105
6.2 Specimen description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.3 Influence of experimental setup on microphone response . . . . . . . . . 108
6.4 Pressure-field investigations with use of impedance tube . . . . . . . . . 110
6.4.1 Pressure-field experimental setup and measurements . . . . . . . 112
6.5 Free-field investigations in anechoic chamber . . . . . . . . . . . . . . . . 115
Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

Conclusion and outlook 119

A How COMSOL implements FLNS 121

Bibliography 122

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Introduction

1.1 General introduction

Origins of microphones dates XIX century [1], research in this area was accelerated
by need of convenient and fast means of long distance communication. If we consider
microphone as a electroacoustic transducer, its development dates down to 1874 when
patent of E.W. Siemens on magneto-electric apparatus [2] was granted. However the
invention of Siemens was not applied to microphone until late 1920’s and in a meanwhile,
in 1876 A.G. Bell has used liquid transmitter microphone in his telephone. Further
designs of telephones used carbon microphone presented in 1877 by T.A. Edison [3].
Condenser microphone (which is the most common type in daily use) was invented by
E.C. Wente [4] in 1916 and improved by G. Sessler and J. West [5] in 1960 by integration
of electret material. The first commercial MEMS1 microphone was presented in 2002 by
Knowles - nearly 20 years after the report of first working condenser microphone based
on silicon micromachining by G. Sessler and D. Hohm in 1983 [6].

• 1874 - E.W. Siemens patent on magneto-electric apparatus was granted,

• 1876 - A.G. Bell patent entitled “Improvement in Telegraphy” [7],
• 1877 - T.A. Edison invents carbon microphone,
• 1916 - E.C. Wente invents first stable and commercially viable condenser micro-
phone,
• 1960 - Invention of Electret Condenser Microphone (ECM) by G. Sessler and
J. West of Bell Labs,
• 1983 - G. Sessler and D. Hohm fabricates the first working silicon condenser mi-
crophone based on bulk machining of silicon,
• 2002 - Knowles releases the SiSonic - first commercial MEMS microphone.

Principle of operation
Microphone is an instrument that transduces acoustic waves into electrical signal, it
operates within audible range, that is from 20 Hz to 20 kHz. Depending on the applica-
tion we may distinguish microphones dedicated to sound acquisition (consumer devices
and studio recording equipment) and microphones dedicated to sound measurements.
1
MEMS stands for Micro-Electro-Mechanical System

1
Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf
© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 1. Introduction 2

Measurement microphones are characterized by higher performance and proportionally

higher price.
Regardless of the transduction means, microphone consists of a diaphragm that
is usually a membrane (fig. 1.1) anchored within a rigid housing. In order to ensure
the equal static pressure level between the backvolume and surrounding, the housing
is supplied with pressure equalization vent. On the occurrence of sound pressure the
membrane deflection is transduced into alternative electrical signal by one of the fol-
lowing transduction mechanisms: electromagnetic induction, capacitance or resistance
variation or piezoelectric effect.
Rigid housing
Pressure equalization vent
Flexible diaphragm

Sound waves Signal transduction mechanisms

Backvolume

Figure 1.1: General structure of a microphone.

Means of sound pressure transduction

As we already presented, microphones use variety of mechanisms in signal transduc-
tion, microphones with conversion based on electromagnetic induction (dynamic micro-
phones) are widely used on the stages, while measurement and MEMS microphones are
mostly based on capacitive variation measurements. Moreover the MEMS technology
enables integration of piezoelectric and piezoresistive materials, therefore the MEMS
microphones with detection based on this materials are also reported.
Dynamic microphone consists of a flexible membrane which is connected with a
coil (see fig. 1.2). On the occurrence of a sound pressure the membrane deflects and
moves the coil relative to the fixed magnet. As a result, electric current proportional
to the sound pressure variation flows through the coil. This kind of microphone does
not need an external polarization, however the output signal has to be amplified by
dedicated readout electronics.
Flexible diaphragm
Magnet

Moving coil Output signal

Magnet

Figure 1.2: General structure of a dynamic microphone.

Another type is a condenser microphone that uses capacitance variations to detect

a sound pressure (see fig. 1.3). In this kind of a microphone the membrane serves as an
electrode which moves relative to the fixed electrode (called the backplate). With air
as a dielectrical material between two electrodes one manage to measure the value of
the capacitance that varies as a function of a distance between the membrane and the
backplate. Initially the main drawback of capacitive detection was a need of external
polarization, this problem has been solved by G. Sessler and J. West [5] by use of

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

permanently charged material (electret2 ) integrated into the membrane. Prepolarized

condenser microphones are called the electret microphones (ECM).
Flexible diaphragm

Rigid backplate Output signal

Figure 1.3: General structure of a condenser microphone.

Other popular microphones are those that use piezoelectric and piezoresistive
effects. Regarding the fact that booth effects concerns materials that change their
properties when subjected to stress (tensile or compressive), the construction of the
microphone is presented on the common figure (fig. 1.4). Stress may be applied to the
piezoelectric or the piezoresistive material directly by integration of the material into
the membrane. The other mean to apply the stress is to transfer membrane deflection to
stress-sensitive material by simple mechanisms. Since piezoelectric material generates
electrical signal while stressed, microphones based on this material do not need external
polarization, however the piezoresistor has to be externally polarized in order to read
the resistance variations.
Flexible diaphragm

Stress-sensitive Output signal

material

Figure 1.4: General structure of a microphone that uses the stress-sensitive materials.

Microphone specifications
We will now introduce several terms that are used in description of microphones prop-
erties. Brief introduction to the most important terminology used in microphones de-
scription can be found in papers of J. Lewis [8, 9], we recommend also the handbook
prepared by Brüel & Kjær [10].

Sensitivity is a quantity that describes the output electrical signal of a microphone

as a function of applied acoustic pressure. Datasheets provide the value of sensitivity at
1 kHz and sound pressure level of 94 dB (1 Pa). Sensitivity of a microphone with analog
output signal is expressed in mV/Pa or in decibels referring to 1 V (dBV). Values in
mV/Pa and dBV can be easily compared by use of the following formula:

SensitivitydBV = 20 log10 (Sensitivity/1[V /P a]). (1.1)

2
Electret materials are fabricated by melting of dielectric material with charged particles and solidify
it in electrostatic field.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Sensitivity of a microphone with digital output is expressed in dBFS (decibels referring

to the full scale). To obtain value in dBFS we suppress reference pressure (94 dB) from
the maximum acoustic load of a microphone.
Ideal response of a microphone is flat, however due to mechanical (solid and acous-
tical) properties the sensitivity may vary at the extremities of a frequency range. Varia-
tions of sensitivity with the frequency are represented by the frequency response curves
normalized to 1 kHz (fig. 1.5).

Figure 1.5: Frequency response curve of Knowles SPU0409LE5H MEMS micro-

phone [11].

Bandwidth describes microphone capabilities to detect acoustic signal. Each micro-

phone is characterized by its frequency range which is limited by low and high cutoff
frequencies. We will show later that the low cutoff frequency is determined by the prop-
erties of acoustic vents and the housing of a microphone while the high cutoff frequency
is set by a mechanical resonance of a diaphragm (membrane, beam etc.). Auditory re-
gion covers frequencies from 20 Hz up to 20 kHz, where the sound generated by speech
is within 170 Hz - 4 kHz and music 50 Hz - 8.5 kHz (fig. 1.6); therefore depending on
the application, microphones are designed in order to operate within full auditory region
(bandwidth of 20 kHz) or a reduced bandwidth.

Signal to Noise Ratio (SNR) expressed in dBA (A-weighted3 decibels) is a ratio

between standard reference signal (94 dB at 1 kHz) and the equivalent input noise
(noise floor) of a microphone. Equivalent input noise sets the lower detection limits of
a microphone, it is measured in a quiet environment and it consists of sensing element
noise and readout electronic noise.

Dynamic range expressed in dB is a domain of operation of the microphone within

the noise floor and acoustic overload point. Acoustic overload point (also called MaxSPL
or maximal acoustic load) is defined by the sound pressure level where the response of
the microphone becomes nonlinear. Long-lasting exposition of microphone to the SPL
3
A-weighting curve corresponds to the sensitivity of human ear that varies over the frequency range.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 1.6: Auditory region according to F.A. Everest [12].

greater than maximal indicated value can lead to mechanical failure.

Acoustic overload point of a microphone is usually adjusted to be close to the threshold
of pain which varies slightly over frequency range and is approximately 120 dB, however
due to the growing demands of end users (concerts recording, aeroacoustic applications)
this value tends to increase.

Directionality is a microphone property that describes the capability of detecting

sound waves incidents from different directions relative to the sound port. It is repre-
sented on a polar grid plot called the “polar pattern”, where the sensitivity is normalized
to 0◦ incidence. Typically the sensitivity of a MEMS microphone does not depend on
the sound source position (fig. 1.7), such response is called omnidirectional.

Figure 1.7: Polar pattern of omnidirectional microphone, on the example of Analog

Devices MEMS microphone [9].

Total Harmonic Distortion (THD) is a ratio between total power of a signal and
power of the fundamental frequency, it is expressed in percents:

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

P total − P fundamental
T HD = × 100. (1.2)
P fundamental
THD is most commonly measured along with noise (THD+N). In this measurement
microphone is excitated with sine wave signal. Power of harmonics is then measured by
applying notch filter on the fundamental frequency (fig. 1.8). Comprehensive discussion
on audio distortion measurements can be found in paper of S. Temme [13].

Figure 1.8: THD+N measurement with use of notch filter. From the paper of
S. Temme [13].

1.2 International context of this work

Overview of a MEMS microphones
Commercial products
Micro-Electro-Mechanical Systems (MEMS) have their origins in well established semi-
conductors technology. It profits of electronic properties of Silicon and combines it with
its exceptional mechanical characteristics in order to miniaturize wide range of trans-
ducers. It has been a long time since MEMS followed our daily life unnoticed - hidden
in the ink-jet print heads, saving lives in such automotive applications as Airbags, tyre
pressure monitoring systems (TPMS) and Electronics Stability Program (ESP) systems.
Established MEMS devices enabled engineers to work on new designs and invent new,
dedicated technological processes (example the Bosh process [14, 15]). Improvement in
MEMS sensors technology enabled designers to aim consumer electronic market. After
integration of accelerometers into consumer devices, the next natural step was to develop
gyroscopes, magnetometers and microphones.
Despite the fact that the first working silicon microphone has been reported by
G. Sessler and D. Hohm in 1983 [6] it has taken nearly 20 years when on 2002 Knowles
has released SiSonic - the first commercial MEMS microphone. If we review the evolution
of the international MEMS microphones market until 2013, we can define the following
milestones:

• 2002 - Knowles releases the first commercial MEMS microphone,

• 2006 - Infineon enters the MEMS microphone market,

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

• 2006 - Akustica releases the first digital MEMS microphone which is additionally
the first single chip microphone (MEMS and ASIC on the same die),
• 2007 - AAC Technologies begins mass production of MEMS microphones,
• 2009 - Bosch enters MEMS microphone market by acquiring Akustica,
• 2010 - Apple employs MEMS microphones in iPhone,
• 2011 - STMicroelectronics enters MEMS microphone market with focus on digital
microphones,
• 2012 - Knowles claims to ship over 3 billions of chips since the release of its first
SiSonic series,
• 2013 - InvenSense acquires MEMS microphone business line from Analog Devices.

Since the year 2002, U.S. - based Knowles has maintained the position of the leader
in MEMS microphones market (fig. 1.9), however its domination has shrunk from 75%
in 2011 to 50% in 2012 in favor of the other key players such as Chinese AAC Tech-
nologies, U.S. Analog devices, another Chinese company - GoerTek and French-Italian
STMicroelectronics.

Figure 1.9: 2012 MEMS microphone market suppliers according to IHS iSuppli MEMS
& Sensors Special Report.

Figure 1.10 shows MEMS microphone shipment forecast according to IHS iSuppli
Research, February 2013. In addition data from the previous years starting from the year
2006 were included. On this figure we can observe the irregularity in shipment growth
between the years 2010 and 2011 caused by the introduction of MEMS microphones
into the iPhone 4 by Apple which has previously used exclusively ECMs and it now
became the biggest buyer of MEMS microphones. Current version of Apple flag product
- iPhone 5 uses 3 microphones, with two dedicated to speech acquisition and noise
canceling while the third (placed at the back of the phone) is used for high definition
video recording. Integration of MEMS microphones has set cellphones as the largest area
of its employment (54% of microphones shipped in 2012 has been bought by two biggest

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 1.10: MEMS microphone shipment forecast according to IHS iSuppli Research,
February 2013.

players that are Apple and Samsung). Cellphone application is followed by notebooks,
headsets, media tablets and so on.
All of the commercially available MEMS microphones use capacitive means of de-
tection. Such detection is realized by two electrodes micromachined in Silicon. One of
the electrodes is rigid while the other deflects on the presence of sound pressure caus-
ing by that variation of capacitance. Packaged microphones have dimensions of single
millimeters (fig. 1.11(a)), the package consist of a MEMS chip and dedicated readout
electronics (ASIC) with analog or digital output (fig. 1.11(b)).

(a) Overall view (b) Top view

Figure 1.11: Knowles SPU0409LE5H microphone presented in reverse costing analysis

by R. Fraux [16].

We have gathered the specifications of microphones of the key market players in

table 1.1.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 1. Introduction
Table 1.1: Specifications of commercial MEMS microphones with the digital output.

Specifications
Manufacturer Model Sensitivity Bandwidth SNR MAX SPL Current supply Size
[dBFS/1Pa] [kHz] [dBA] [dB] [µA] [mm3 ]
Knowles SPK0833LM4HB -26 0.1-10 63 122 500 4x3x1
AAC SDM0401B-263-M02 -26 no data 60.5 no data 550 4x3x1
Analog Devices ADMP521 -26 0.1-16 65 120 900 4x3x1
GoerTek SD04OT263-01/02 -26 0.1-8 58 no data 600 4.7 x 3.8 x 1.3
STMicroelectronics MP34DB01 -26 0.02-20 62.6 120 650 4x3x1

9
Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf
© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 1. Introduction 10

State of the art

Micromachined microphones are based on the rigid backplate and flexible membrane
that deflects out of the wafer plane (fig. 1.12). Numerous surface micromachined mem-
branes have been reported, among them those made of metalized silicon nitride [17–19],
the combination of silicon nitride and polysilicon [20], polysilicon [19, 21–26] and bulk
micromachined membrane reported by Y. Iguchi [27]. Regardless the design, membrane
surface is in range of single square millimeters: from 0.38 mm2 reported by A. Dehe [26]
to 4.41 mm2 reported by Y. Iguchi [27].

Figure 1.12: Example of state of the art condenser MEMS microphone [23].

Use of surface micromachining requires precautions in order to minimize the residual

stress which preloads membrane and affects the response of a microphone. Residual
stress in thin films can be caused by the mismatch of coefficient of thermal expansion,
differences in technological process parameters (intrinsic stress), or in case of monocrys-
tals - mismatch of crystal lattice. Some of the authors adjust the membrane stress
by use of interlayers [20], the others by adding additional annealing steps [21], while
J.W. Weigold [22] from Analog Devices, A. Dehe [23,26] from Infineon Technologies and
C. Chan [25] have presented spring-suspended rigid membranes (fig. 1.13). A. Dehe has
reported that employment of springs or corrugation rings increases the sensitivity from
two to six times in comparison to standard clamped membrane.

(a) By J.W. Weigold (b) By A. Dehe

Figure 1.13: Different types of spring-suspended membranes.

Measured performance and characteristics of state of the art microphones have been
gathered in table 1.2. Pressure response of the reported microphones was character-
ized by comparison method in anechoic environment or by use of Kundt’s tube. It is
problematic to adequately compare these works since many crucial elements affecting
microphone performance varies (membrane surface, bias voltage, acoustic port config-
uration - backvolume). However on figure 1.14 we have compared the SNR of some of
the condenser microphones with regard to the surface of the membrane and the year
of publication. SNR is proportional to the surface of the membrane, this tendency is

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

easily seen if we focus on the microphone reported by P. Rombach [20] where surface of
the membrane equals 4 mm2 and tree microphones reported by A. Dehe [26] where the
surface of the membrane decreases from 0.95, 0.63 to 0.38 mm2 .

Figure 1.14: SNR with regard to the surface of the membrane and the year of publication.

It seems that the condenser MEMS microphones have reached their miniaturization
limits. The limitations comes mainly from the fact that the sensitivity of the microphone
decreases together with the miniaturization of the membrane. In this work we will
propose the novel architecture of the microphone which improves the yield between
the surface of the diaphragm and the resulting sensitivity. To investigate this new
architecture the consortium of several partners has started the project MADNEMS
which is described in the next subsection.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 1. Introduction
Table 1.2: Specifications of state of the art MEMS microphones

Specifications
Author Year Type Sensitivity [dBV] Bandwidth [kHz] SNR [dBA] Remarks
Y. B. Ning et al. [17] 1996 Condenser -43 0.10 - 20 no data
A. Torkkeli et al. [21] 2000 Condenser -48 0.01 - 12 60
M. Mullenborn et al. [28] 2001 Condenser -46 0.01 - 10 69
W. Kronast et al. [18] 2001 Condenser -47 0.05 - 20 56
P. Rombach [20] 2002 Condenser -38 0.02 - 20 71
J.J. Neumann et al. [29] 2003 Condenser no data 0.02 - 10 48
G.W. Elko et al. [30] 2005 Condenser no data 0.10 - 20 34 dB unweighted SNR, direc-
tional
J.W. Weigold et al. [22] 2006 Condenser -47 no data no data
A. Dehe [23] 2007 Condenser -39 0.10 - 10 65
Y. Iguchi et al. [27] 2007 Condenser -52 0.03 - 20 47
T. Kasai et al. [24] 2011 Condenser -25 dBFS 0.02 - 20 63 digital output
C. Chiang et al. [31] 2011 Condenser -42 0.10 - 20 56 integrated ASIC
C. Chan et al. [25] 2011 Condenser -38 0.20 - 20 33
A. Dehe et al. [26] 2013 Condenser -38 0.08 - 10 66
M. Sheplak et al. [32] 1999 Piezoresistive -93 0.20 - 6 1 aeroacoustic applica-
tion, MaxSPL 155 dB
M.L. Kuntzman et al. [19] 2011 Optical no data 0.04 - 10 66
M.D. Williams [33] 2012 Piezoelectric -88 0.07 - 20 54 MaxSPL >172 dB

12
Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf
© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 1. Introduction 13

1.3 The “MADNEMS” project

My thesis was realized simultaneously to the industrial research project funded by the
French National Research Agency and called the MADNEMS project (MADNEMS is
abbreviation from the french phrase “Microphone A Détection par jauge NEMS” - micro-
phone with detection realized with nanogauge). The target of the project was to develop
novel MEMS microphone dedicated to the hearing aids and to the consumer electronic
(especially portable devices). To prepare the prototype a consortium of several partners
has been appointed:

CEA-Leti - “Laboratoire d’électronique des technologies de l’information” based in

Grenoble. CEA-Leti has an extensive experience in field of MEMS, it has proposed
the new concept of a MEMS microphone.

LVA - “Laboratoire Vibrations Acoustique”. The laboratory of vibration and acoustics

is based in Lyon and it is affiliated to the National Institute of Applied Science
(INSA). It introduces the knowledge on acoustic phenomena and on characteriza-
tion methods of the microphones.

IM2NP - “Institut Matériaux Microélectronique Nanosciences de Provence” based in

Marseille is in charge of conception and fabrication of ASIC4 with ultra-low power
consumption.

Neurelec/Oticon Medical - Neurelec (acquired by Oticon Medical in 2013) is an

industrial partner based in Sophia Antipolis technology park. An expert in field
of cochlear implants that introduces its knowledge in signal processing and is in
charge of final integration of the prototype into the BTE5 .

Motivation for the MADNEMS project comes from the innovative technological platform
developped at CEA-Leti and called the M&NEMS. This platform based on piezoresistive
detection realized with nanowires has already been proven to be relevant for fabrication
of the accelerometers, gyroscopes, magnetometers and pressure sensor. High sensitivity
of strain gauges in form of nanowires - called the nanogauges enabled the engineers from
CEA-Leti to decrease the footprint of these sensors. Now we would like to prepare novel
MEMS microphone based on the M&NEMS. Use of single technological platform for the
microphone, pressure and inertial sensors may play a key role in the future integration
of several sensors on one chip.

1.4 Organization of the manuscript

This thesis was prepared at CEA-Leti in collaboration with the LVA. It is the first
thesis on MEMS microphone prepared at the Laboratory of Microsensors at CEA-Leti
and its main target was to prepare the first prototype of the MEMS microphone to
complete the M&NEMS sensors line. To initiate the research on acoustic sensors in the
laboratory and to prove the new concept of the microphone there was a need to prepare
the extended bibliographical research on several subjects. This bibliographical research
covered the following issues:

• nomenclature used in microphone specifications,

4
Application Specific Integrated Circuit
5
BTE stands for Behind The Ear, it is a sound processor unit used in cochlear implant systems.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

• evolution of the microphones and state of the art of the MEMS microphones,
• transduction mechanism suitable for the new architecture,
• proper tolls for simulation of the acoustic phenomena at the microscale,
• investigations on the chip-scale packaging of MEMS microphones,
• functional characterization methods.

This bibliographical research was done at the different development stages and it had
a key role to competently assemble the subsystems of the microphone and prepare the
first prototype that is in fabrication. The manuscript is divided into six chapters, to
facilitate the reading, we present the organization of the chapters:

Chapter 1 is a general introduction to this work and it places it among the state of
the art and the commercial market microphones.

Chapter 2 introduces the new concept of MEMS microphone patented by CEA-Leti

and compares it to the state of the art MEMS microphones. Then we present
the architecture and possible transduction mechanisms. Finally the technological
platform used for sensor fabrication and the other sensors based on that technology
are presented.

Chapter 3 discusses the approaches to model the acoustic phenomena in MEMS. We

present models based on full set of linearized Navier-Stokes equations (state of the
art and commercial solutions) and compare it with approximative Low Reduced
Frequency model and Lumped Element model. This chapter is an introduction to
the design considerations of proposed new MEMS microphone.

Chapter 4 consists of the microphone design considerations including the investiga-

tions of its acoustical, mechanical and electrical behavior. We designate the key
factors that formulate the microphone characteristics. This considerations are
completed by the discussion of the noise sources typical for MEMS microphones.

Chapter 5 gives the closer view on the process flow used in the designed microphone
fabrication. Afterwards we review the chip-scale packaging used in commercial
products and the in-house solution is proposed. At the end we discuss the advances
in fabrication and the encountered difficulties.

Chapter 6 illustrates the work that has been done on the preparation for the func-
tional characterization of the MEMS microphone in pressure-field and free-field
conditions. While waiting for the prototype of designed microphone, the charac-
terization methods have been validated with use of commercial MEMS microphone.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

New architecture of a MEMS

microphone

The following chapter will present the concept of a Silicon MEMS microphone with a
diaphragm that deflects in the plane parallel to a Silicon wafer (we call it the in-plane
deflection). Amplitude of the diaphragm vibration is proportional to sound pressure
fluctuations and can be detected by a capacitance or a resistance variation. Relevance
of these detection means to the new architecture will be discussed and more suitable
mechanism will be indicated. Finally the technological platform will be described and
the existing sensors that are based on this technology (accelerometer, gyroscope, mag-
netometer and pressure sensor) will be reviewed.

2.1 Concept presentation

Presented concept is based on the invention by Ph. Robert and A. Walther of the MEMS
Sensors Laboratory at CEA-Leti [34]. Microphone in this invention is described as a
diaphragm enclosed between a bottom and a top wafer (fig. 2.1), where the one of the
wafers is used as a base for MEMS fabrication and the second one is a lid. Acoustic
vents that ensure sound pressure propagation and static pressure equalization are etched
in both wafers.
Outlet vent Top wafer

e1
Bottom wafer Diaphragm Inlet vent

Figure 2.1: Cross-sectional view of the new MEMS microphone architecture (not to
scale).

On the occurrence of sound, the pressure fluctuations propagates through the inlet

15
Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf
© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 2. New architecture of a MEMS microphone 16

vent and reach the diaphragm which deflects proportionally to the pressure difference
between inlet and outlet vents (fig. 2.2). Given concept can be modified by multiplication
of the diaphragms and acoustic vents that leads to sophisticated system with higher
performance.
Diaphragm deflection

p2 p1

Sound waves

Figure 2.2: Diaphragm deflection mechanism (not to scale).

2.2 Possible transduction mechanisms

Regarding specific microphone concept and available technology of silicon micromachin-
ing two possible transduction mechanism have been identified: capacitive and piezore-
sistive transduction. Lets now review both of this mechanisms with emphasis on the
electro-mechanical coupling and the sensitivity designation.

2.2.1 Capacitive transduction

For the considerations of electro-mechanical coupling in capacitive microphone we will
use simplified model represented on figure 2.3. The model consists of the fixed electrode

p2 R i(t)
g0
p1

C(u)
e2 E(t) ΔV
e1

Sound waves

(a) Overall view (b) Electrical schematic

Figure 2.3: Capacitive transduction mechanism.

and the rigid diaphragm suspended on the springs with the stiffness k; g0 is the initial
distance between the electrodes. The diaphragm moves in-plane and it serves as an
electrode with lateral surface Sd . The diaphragm displacement is denoted with u.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

General governing equations

To begin the considerations we present the general governing equations, for the sake of
simplicity we omit the damping:
mü + ku = Fa + Fes , (2.1)
d E(t) − ∆V
(C(u)∆V ) = . (2.2)
dt R
where m is a mass of the diaphragm and k is total stiffness of suspension springs.
External forces in the system are Fa - acoustic force and Fes - electrostatic force:
1 ∂C(u) 1 0 Sd
Fes = ∆V 2 = − ∆V 2 (2.3)
2 ∂u 2 (g0 + u)2
Governing electrical equation is expressed by the conservation of electric charge in the
system. Capacitance of the sensor C(u) depends on the actual position of the diaphragm
according to:

0 S d
,
C(u) = (2.4)
g0 + u
where 0 and Sd are permittivity of the vacuum and surface of the electrodes respectively
and the displacement u is generated by the acoustic or electrostatic force.
Replacing Fes and C(u) in equations 2.1 and 2.2 results in the following set of coupled
nonlinear equations:
1 ∂C(u)
mü + ku = ∆V 2 + Fa , (2.5)
2 ∂u
d∆V ∂C du E(t) − ∆V
C(u) + ∆V = . (2.6)
dt ∂u dt R

Static equilibrium and the pull-in voltage

We will now appoint the equilibrium points under static conditions (E = const,
∆V = E). At the static equilibrium u is denoted U and it is a solution of:
1 0 Sd
kU + E 2 = 0. (2.7)
2 (g0 + U )2
We may deduce that when the static polarization exceeds certain value, the electro-
static force may exceed the mechanical stiffness and the moving electrode will be pulled
towards the fixed electrode. Static polarization for which the system becomes non stable
and the electrodes get in contact is called the static pull-in voltage Epi . The critical
displacement of the electrode that corresponds to Epi is denoted Upi :
s 3
k 2 g0
Epi = g0 and Upi = − . (2.8)
0 S d 3 3
The problem of the pull-in was thoroughly studied by V. Rochus [35], where this
phenomenon has been introduced by use of a graph similar to the one presented on
fig. 2.4: for the pull-in voltage only one solution is found (marked with doted line)
whereas for E < Epi we found two solutions: stable and unstable one. Green area on
fig. 2.4 represents the static polarization for which the system is stable - the sensor is
operational.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 2.4: Illustration of the pull-in effect.

Electro-mechanical coupling governing equations

To study electro-mechanical coupling of the presented system, we will express the dis-
placement u and voltage across the capacitor ∆V as a functions of time:
u(t) = U + ũ(t), (2.9)

∆V (t) = E + ṽ(t), (2.10)

where U is the diaphragm displacement caused by the static voltage E; ũ(t) and ṽ(t)
are the small perturbations around U and E.
To study time-dependence of a system, the governing equations have to be linearized
around the equilibrium points with use of Taylor series. Linearized governing mechanical
equation become:
d2 ũ 0 Sd 0 S d
m + kũ = F˜a + E 2 ũ − Eṽ, (2.11)
dt 2 (g0 + U ) 3
(g0 + U )2
We realize that the resulting stiffness k ∗ of a system consists of the mechanical stiffness
and the electrostatic stiffness terms:

∗ 0 S d 2
k ũ = k − E ũ. (2.12)
(g0 + U )3
Linearization of the electrical equation governs the voltage fluctuations:
ṽ C(U )
C(U )ṽ˙ + = ˙
E ũ, (2.13)
R (g0 + U )

Sensitivity of capacitive sensor

Considering small harmonic pressure fluctuations p̃ and knowing the corresponding gap
and voltage variations we can express the sensitivity of a capacitive microphone:
ω02

ṽ Sd E RC(U )jω
S(ω) = = ∗ · · , (2.14)
p̃ k (g0 + U ) ω02 − ω 2 1 + RC(U )jω

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

where ω0 denotes the resonance of the mechanical structure.

2.2.2 Piezoresistive transduction

Piezorestive effect involves the change of the conductor resistance as a function of ap-
plied stress. Piezoresistive effect exists in doped crystalline silicon and it is commonly
used as a detection mechanism in MEMS sensors. One way of adoption of piezoresistive
effect for sound pressure detection is integration of piezoresistor into silicon membranes.
Such architecture has been already used in static pressure sensors (figure 2.5). Integra-

(a) MEMS piezoresistive pressure (b) Stress generation inside

sensor [36]. the piezoresistors [37].

Figure 2.5: Example of MEMS piezoresistive pressure sensor.

tion of piezoresistors in this case can be achieved by selective deposition of doped silicon
layer on the membrane in epitaxial process or by selective doping of silicon membrane
in diffusion or ion implantation process [38]. Since the stress is distributed over the
whole surface of the membrane, sensor optimization covers the stress maximization in-
side the piezoresistors. It can be achieved by placing piezoresistor in the maximal stress
regions [39] and by keeping high ratio between the membrane and piezoresistors thick-
ness [37]. Maximal stress in rectangular membrane occurs in the middle point of the
membrane edge (tensile stress) and in the center of a membrane (compressive stress).
Stress concentration regions can be additionally introduced by the special membrane
features including boss [40] and trenches [41]. Integration of piezoresistors into the
membrane creates p-n junction at the membrane-piezoresistors interfaces. This junc-
tion may lead to generation of leakage current. Leakage current values are negligible
corresponding to piezoresistors bias current at room temperature, however it increases
and may impact response of a sensor at elevated temperatures [42, 43].
Alternative design of piezoresistive sensor involves use of suspended Silicon nanowires
(fig. 2.6) which serves as a piezoresistors. Piezoresistive effect in p-type nanowires fabri-
cated on silicon-on-insulator (SOI) wafers has been studied by R. He and P. Yang [44].
They have reported that piezoresistive effect at nanoscale significantly rises compared
with bulk Silicon and the nanowires seems to be a good direction towards more efficient
MEMS sensors.

Electro-mechanical coupling governing equations

We may consider microphone architecture with piezoresistive nanogauge attached and
stressed by diaphragm moving in-plane of SOI wafer (fig. 2.7). Mechanical governing

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 2.6: Silicon nanowire [44]. The scale bar is 2 µm.

p2 p1

e1 Suspended nanowire
Sound waves

Figure 2.7: MEMS microphone with nanowire piezoresistor (not to scale).

equation for presented architecture will be as follow:

mü + (k + Kg )u = Fa , (2.15)

where total stiffness will consist of diaphragm stiffness k and longitudinal stiffness of a
nanogauge Kg .
Variation of nanogauge resistance ∆R depends proportionally on diaphragm displace-
ment:
Kg u
∆R = πpzr σg R = πpzr R, (2.16)
Sg
where πpzr , σg , R and Sg are piezoresistive coefficient, nanogauge longitudinal stress,
nominal resistance of the nanogauge and surface of its longitudinal cross-section.

Wheatstone bridge architecture

Accurate measurements of ∆R can be carried out by use of electrical architecture called
the Wheatstone bridge. It consists of two voltage dividers connected in parallel (fig. 2.8).
If we feed both voltage dividers of Wheatstone bridge with current Ib and we assume
that the values of resistance R are perfectly matched, then the output voltage ∆V will
be null.
At the time when the resistors are not matched, the divided voltages are not equal which
results in output voltage ∆V proportional to the resistance difference. Depending on
the application, one to four piezoresistors can be introduced into Wheatstone bridge
architecture (see fig. 2.9). Output signal for different bridge configurations is then

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

R R

ΔV

R R

Figure 2.8: Resistors arranged into Wheatstone bridge architecture.

Ib Ib Ib

R R R R-ΔR R+ΔR R-ΔR

ΔV ΔV ΔV

R R+ΔR R R+ΔR R-ΔR R+ΔR

Figure 2.9: Possible Wheatstone bridge architectures for piezoresistive sensor [45].

related to the number of piezoresistors by the formulas 2.17-2.19.

Ib R ∆R Ib
1 piezoresistor: ∆V = ∆R
≈ ∆R, (2.17)
4 R+ 4 4
Ib
2 piezoresistors: ∆V = ∆R, (2.18)
2
4 piezoresistors: ∆V = Ib ∆R. (2.19)

Further considerations will be based on 4 piezoresistors configuration since it provides

the highest output signal. If we now introduce ∆R into the equation 2.19, the governing
equations of one-way coupled system will be given by:
" #
k + Kg

Fa mü
= Kg u+ (2.20)
∆V Ib πpzr Sg R 0

Sensitivity of piezoresistive sensor

Sensitivity of piezoresistive sensor can be formulated analytically to capacitive sensor
sensitivity. We start with expression of ∆V :
Kg
∆V = Ib πpzr Ru, (2.21)
Sg

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

by development of u we obtain:
˜
∆V Sd Kg

ω02

S(ω) = = Ib πpzr R , (2.22)
p̃ Sg k + Kg ω02 − ω 2
where ω0 denotes the resonance of the mechanical structure.

2.2.3 Transduction mechanisms comparison

In the previous sections we have proposed two distinct planar microphone architectures:
capacitive and piezoresistive based on suspended nanowires. Both architectures are
comparable in terms of technological implementation - sensors can be fabricated basing
on SOI wafers and standard microelectronic process; the main differences come from
their mechanical and electric properties.

Comparison of mechanical properties

For the purpose of comparison we will imagine the diaphragm as a microbeam suspended
at both extremities by the springs. Beam is considered to be rigid with stiffness k com-
ing from the springs. Mechanical resonance frequency will be defined by the beam mass
and by the total stiffness of a sensor which is stiffness k modified by introduction of neg-
ative electrostatic stiffness and longitudinal stiffness of a nanogauge for capacitive and
piezoresistive detection relatively. While the longitudinal stiffness of nanogauge has a
constant value defined by geometry, the electrostatic stiffness depends of static polariza-
tion of capacitor and may be used for active control of sensor resonant frequency. To set
the static polarization of capacitive sensor one has to take precautions and do not exceed
the pull-in voltage which will stuck the two electrodes. Since the capacitive system is
two-way coupled, we may drive the displacement of a diaphragm with electrostatic force
Fes . This possibility of actuation may be used for a self-test of a sensor which is not
possible in case of piezoresistive MEMS that need integration of additional self-test elec-
trode. Finally the capacitive sensor introduces considerably larger squeeze-film damping
because of the electrodes (for example in case of comb-structures).

Sensitivity optimization
Analysis of sensitivity formulas - for capacitive (equation 2.14) and for piezoresistive
(equation 2.22) sensor enable us to reveal the common rules that lead to sensitivity
optimization. Therefore in both cases we need to increase the diaphragm surface and
supply power. Moreover the sensitivity is amplified in the vicinity of the mechanical
resonance. To profit of this mechanical amplification the value of the first mechanical
resonance has to be set close to the sensor frequency range, however in such case the
sensor sensitivity will vary across the bandwidth.
In case of capacitive sensor sensitivity optimization, we may simplify sensitivity formula
on the assumption that electrostatic stiffness value is in general much lower than me-
chanical stiffness. Therefore we need to decrease the mechanical stiffness and the gap
between the electrodes. Furthermore capacitive sensor output signal will be influenced
over the frequency range by high-pass filter with corner frequency at 1/(2πRC(U )).
If we now consider the sensitivity of piezoresistive sensor, we need to pay attention on
proper design of piezoresistive nanogauge. In perfect case the stress in the nanogauge
can be maximized by decrease of nanogauge section Sg and by optimization of the dis-
tribution of strain energy between the gauge and the diaphragm Kg /(k + Kg ).

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Miscellaneous aspects and conclusion

Thanks to research on Silicon nanowires which can be fabricated on SOI wafers, piezore-
sistive sensors became more competitive to capacitive sensors. Unlike the classical
piezoresistors integrated into silicon layer, suspended nanowires do not suffer of leakage
current. Moreover further improvement of piezoresistive detection is possible since the
longitudinal piezoresistive coefficient rises inversely proportional to nanowire section.
Nanowire detection has an advantage on capacitive sensors in terms of high linear-
ity of response and facility of sensor downscaling. Main features of both transduction
mechanism has been summarized in table 2.1, adopted from work of D. Ettelt and
Y. Deimerly [46, 47].

Table 2.1: Transduction mechanisms comparison.

Transduction mechanism
Feature Condenser Piezoresistive
Linearity 7 4
Downscaling 7 4
Technological complexity 4 4
Possibility of differential measurement 4 4
Possibility of actuation 4 7
Sensitivity to chip-package stress high low
Sensitivity to parasitic capacitance high low
Influence of squeeze-film damping high low
Temperature sensitivity no yes
Humidity dependent yes no
Sensitivity to electromagnetic interface yes no

2.3 The M&NEMS technology

Fundamentals of so-called M&NEMS technology have been initially presented in 2009
by Ph. Robert [48]. Name of the technology comes from combination of MEMS and
NEMS structures as a components of final functional sensor. Ph. Robert has shown the
application of M&NEMS concept to 3D accelerometer design. Figure 2.10 illustrates

(a) Architecture of device. (b) Example of realization.

Figure 2.10: M&NEMS technology concept [48].

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

structure of in-plane accelerometer with inertial mass suspended on a hinge and a sus-
pended nanogauge. Nanogauge is situated along the line between the rotation axis of
the hinge and the center of inertia of the seismic mass. It is suspended between the
seismic mass and the anchor in silicon substrate. On the occurrence of acceleration,
the seismic mass will move rotationally with respect to the rotational axis, the inertial
force will be exerted on the section of a nanogauge causing compressional or tensional
longitudinal stress. Stress in the nanogauge generates resistance variation (piezoresistive
effect) proportional to inertial force and amplified by factor coming from a lever effect
(distance between rotation axis and center of seismic mass divided by distance between
rotation axis and position of the nanogauge).

2.3.1 Technological platform

Technological platform developed at CEA-LETI, supports M&NEMS sensors fabrication
on 200 mm SOI wafers. The simplified technological process (fig.2.11) begins with
lithography and etching of NEMS nanowires (b), next the nanowires are protected by
the oxide layer (c). Due to use of DUV lithography process, dimensions of nanogauges
are limited to 250 nm. Afterward epitaxial layer of Silicon is grown (d) to be the base for
DRIE of MEMS component (e). Electrical contacts are then realized by metalization of
doped Silicon (e). The process is terminated with HF-vapor etching of oxide to release
the MEMS structure and the nanogauges (f). Further steps may be realized for the
specific designs such as magnetometers and pressure sensors. Additional steps required
for M&NEMS microphone fabrication will be described later.

(a)

(b)
(e)

(c)

(f)
e2
(d)
e1

Figure 2.11: The M&NEMS technological platform.

2.3.2 M&NEMS sensors review

Since the first presentation, M&NEMS has shown the capabilities for multi-sensor in-
tegration on chip scale. Concept was proved by functional tests of 3D accelerometer.
As a consequence of market demand on integrated IMU (Inertial Measurement Unit),
research on magnetometer and gyroscope has been started simultaneously.
M&NEMS magnetometer [46, 49, 50] use three MEMS structures that assure magnetic
field detection in 3 spacial directions (figure 2.12). In contrary to previous realizations
of MEMS magnetometers based on Lorentz force (increased power consumption), it uses
a permanent magnet. Permanent magnet was fabricate by sequential layer deposition
of ferro- and anti-ferromagnetic materials on top of the MEMS rotating mass. Magnetic

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 2.12: SEM image of 3D magnetometer with zoom on the nanogauges on the right
side [50].

torque is then transduced into resistance variations in the way similar to acceleration
sensor. It is worth to note that although both sensors are sensitive in 3 axes, the sensi-
tivity of z-component is always noticeably lower than x- and y-component sensitivities.
It is explained by higher stiffness of the hinge in case of z-direction, that in consequence
reduces the stress in piezoresistive nanogauge.
MEMS gyroscope was presented on 2012 [51], it is based on dual mass structure vibrat-
ing at resonance (fig. 2.13). Actuation of structure is realized by use of comb-drive while

Figure 2.13: SEM image of gyroscope with zoom on nanogauge on the right side [52].

the signal coming from Coriolis force is detected by use of nanogauges. Actuation and
detection may be done at the same (matched) or at different (unmatched) frequency
depending on the sensitivity and noise contributions. Accelerometer works in ambient
pressure, while the gyroscope as a vibrating structure requires vacuum environment.
Problem of accelerometer and gyroscope chip scale integration has been investigated
by Y. Deimerly [47]. Two potential solutions that have been selected and preliminary
tested use resonating accelerometer or accelerometer with integrated electrostatic damp-
ing source.
Finally, the absolute pressure sensor concept has been demonstrated in 2013 [53]. Signal
transduction in this sensor is based on deformations of thin Silicon membrane. These
deformations are then transferred to the nanogauges by use of rotating, rigid lever arm
(fig. 2.14).

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

(a) Architecture of device. (b) SEM image.

Figure 2.14: M&NEMS pressure sensor [53].

Summary
Considering the fact that sensitivity of capacitive sensors decreases along with capaci-
tance value, the further downscaling of those sensors that use relatively big comb struc-
tures became problematic. On the other hand presented M&NEMS technology seems to
manage the downscaling issues while preserving the sensitivity. Fabrication and valida-
tion of individual M&NEMS sensors has been already pursued by first attempts towards
multi-sensor chips. Under these circumstances use of M&NEMS technology in new ar-
chitecture of MEMS microphone seems to be a natural step. So far we have briefly
presented the new microphone architecture. This architecture will be investigated in
chapter 4, however at the beginning we need to designate the proper tools to model the
acoustic behavior of the system. This tools are discussed in the next chapter.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Models for simulations of acoustic

phenomena in MEMS

Comprehensive study of the new microphone architecture that was proposed in previ-
ous chapter requires proper tools to investigate the acoustic behavior. Classical acoustic
approach is irrelevant in case of MEMS architecture where the scaling effects have to
be considered. We start with presentation of full set of linearized Navier-Stokes equa-
tions and their implementation into the COMSOL Multiphysics finite element solver.
We believe that this model approach is essential and it is the first step to understand
the acoustic phenomena at the microscale. Afterward we compare the Full Linearized
Navier-Stokes model (FLNS) with the approximative models: Low Reduced Frequency
model (LRF) and Lumped Element model to designate the most efficient tool for MEMS
microphone design.

3.1 Propagation of sound wave in dissipative fluid

As we will show later, in case of microsensor design the dissipative effects may severely
impact the resultant performance. Depending on the dimensions of the acoustic duct,
boundary and bulk regions may contribute to the medium properties equitably or - in
the extreme cases, the boundary effects may completely prevail.
We start our discussion by introduction of the governing equations for the sound prop-
agation in dissipative fluid.

Momentum equation
M. Bruneau [54] shows that conservation of the momentum in the dissipative fluid
is governed by the linearized Navier-Stokes equations. The left side of general form
represents the motion of a fluid where ρ0 is quiscent fluid density and v is the particle
¯ and f̄ , which are the Cauchy
velocity vector; while the right side consist of sum of σ̄
stress tensor and the external body forces:
∂v
ρ0 ¯ + f̄ .
= ∇ · σ̄ (3.1)
∂t
We assume that the fluid is homogeneous and isotropic (this conditions were presented
by M. Bruneau) and the Cauchy stress tensor comprise the acoustic pressure fluctuations
and the shear stress generated by the viscous effects:
¯ = −pI + σ̄
σ̄ ¯v , (3.2)

27
Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf
© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 28

¯v is a viscous shear stress. To understand the

where p denotes the acoustic pressure and σ̄
viscous shear we need to imagine the elementary volume of the fluid (fig. 3.1). If we divide
the fluid domain into thin layers, where each layer have a different velocity starting from
zero and limited by the bulk velocity, then the portion of a fluid situated between two
different layers undergoes the shear force in the manner showed on figure 3.1. Viscous
original strained
element element

δv

Figure 3.1: Elementary fluid element undergoing the viscous shear force.

stress is comprehensively discussed in book of M. Bruneau [54], where the final formula
¯v is given by:
for σ̄
T 2
¯v = µ ∇v + ∇v
σ̄ − µ − η (∇ · v) I, (3.3)
3
where µ is the shear viscosity (dynamic viscosity) and η is the coefficient of the bulk
viscosity (η ≈ 0.60µ after A.D. Pierce [55] and R. Kampinga [56]).
In the further considerations the momentum equation for dissipative problem is pre-
sented in the following manner:
∂v
ρ0 ¯v + f̄ .
= −∇p + ∇ · σ̄ (3.4)
∂t

State equation
In the dissipative process we need to investigate the heat introduced to the system by
the viscous heat flux. At this point we may also introduce the eventual external heat
source that may perturb the sound propagation. Therefore the entropy of the system
varies and the state equation reads:
∂τ ∂p
ρ0 c P − = −∇ · q + Q̄, (3.5)
∂t ∂t
where cP and τ are the heat capacity ratio at constant pressure and the temperature
fluctuations respectively, Q̄ denotes the external heat source inside the system and q
denotes the heat flux density defined by the Fourier heat law:
q = −κ∇τ, (3.6)
where κ is the thermal conductivity.

Continuity equation
We present the conservation of mass in the system knowing that the motion of fluid
induces the density fluctuations (ρ):
∂ρ
= −ρ0 ∇ · v, (3.7)
∂t

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 29

To present this term with use of pressure and temperature variations R. Kampinga [56]
used the identity coming from the equation of state of the perfect gas:
ρ p τ
= − , (3.8)
ρ0 P0 T 0
where P0 and T0 are the quiescent pressure and quiescent temperature respectively. By
introducing equation 3.8 into equation 3.7 we relate the divergence of the velocity field
with the pressure and temperature variations:

1 ∂τ 1 ∂p
− + = −∇ · v. (3.9)
T0 ∂t P0 ∂t

Final set of general equations

We present the propagation of sound in dissipative fluid with momentum 3.10, state 3.11
and continuity 3.12 equations:

∂v
ρ0 ¯v + f̄ ,
= −∇p + ∇ · σ̄ (3.10)
∂t
∂τ ∂p
ρ0 c P − = −∇ · q + Q̄, (3.11)
∂t ∂t
1 ∂τ 1 ∂p
− + = −∇ · v. (3.12)
T0 ∂t P0 ∂t

3.2 Scaling effects in acoustics

Relevance of viscous boundary layer
In classical acoustic problems, where the fluid is assumed to be non-dissipative, the
particle velocity is proportional to the gradient of pressure. If we consider the sound
wave propagation in large cavity we can indeed rely on that relation and assume that the
velocity profile (that is parallel to the rigid boundary) is constant across the waveguide
and the velocity field corresponds to the one in perfect fluid (fig. 3.2(a)).
However the real velocity profile at the microscale (fig. 3.2(b)) differs from the perfect one
in the vicinity of the rigid boundary. Moreover the velocity of a fluid on the boundary
is null (no-slip conditions) and it increases with the distance from the boundary and
obtain the bulk velocity at the distance δv (thickness of the viscous boundary layer).
v v

bulk

δv boundary layer

(a) Perfect fluid. (b) Real conditions.

Figure 3.2: Velocity field in the vicinity of rigid boundary, based on work of A. Lalle-
mand [57].

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 30

Relevance of thermal boundary layer

The origins of thermal dissipation effects in sound propagation are related to uneven
temperature of fluid and adjacent material. Therefore fluid-solid interface suffers from
temperature gradient that generates thermal conduction. This gradient is appointed by
difference between the temperature on the boundary (imposed by solid material) and
the temperature of the bulk.
Similar to the viscous boundary layer, the thickness of thermal boundary layer (δt ) is
determined by the distance between the fluid-solid interface and the point where the
fluid temperature is comparable to the one of the bulk material

Thicknesses of the boundary layers

The exact formulas for viscous (eq. 3.13) and thermal (eq. 3.14) boundaries thicknesses
presented by N. Joly [58] are as follows:
r
2µ
δv = (3.13)
ρ0 ω
r
2κ
δt = (3.14)
ρ0 cP ω
Both viscous and thermal boundary layers thicknesses are inversely proportional to
the square root of the frequency, where ω is an angular frequency (ω = 2πf ).
Dependence of the boundary layers on the frequency is presented on figure 3.3. Dashed
rectangle indicates the region of the exploitation of thermoviscous effects in this thesis.
Frequency range is limited by the audible bandwidth (20 Hz - 20 kHz), while the thick-
ness axis limit the rectangle by the characteristic dimensions of MEMS devices (from
single micrometers to single millimeters).
Within the audible range the thicknesses of boundary layers are as follows:
• 0.5 mm - 15 µm for the viscous effects,
• 0.6 mm - 18 µm for the thermal effects.
Although the listed dimensions are in the same order of magnitude and the thermal
layer is slightly thicker, the viscous effects are believed to be of higher significance.
Comparison between this effects is summarized by Prandtl number presented in work
of M. Bruneau [54] and R. Kampinga [56]:
δv2 µcP
P rN = = . (3.15)
δt2 κ
For air at ambient pressure and room temperature the Prandtl number is about 0.75.
Knowing the thickness of viscous boundary layer for 20 Hz and 20 kHz, its evolution
with the frequency may be additionally illustrated with use of simple geometry as on
fig. 3.4. The dimensions of the geometry are 50 × 200 µm2 . We impose gradient of
pressure 1 Pa between the shorter sides and no-slip conditions on the longer sides. When
the sound propagates at frequency of 20 Hz, δv is much thicker than the characteristic
dimension of the guide (0.5 mm 50 µm). The boundary layer is not distinguishable
from the bulk zone and the velocity profile is parabolic. For the propagation frequency
of 20 kHz δv becomes 15 µm and we can distinguish two boundary zones close to the
walls and the bulk zone which appears in the center of the guide and where the velocity
over the section is constant.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 31

Figure 3.3: Viscous and Thermal boundary layers thicknesses as a function of frequency.

Figure 3.4: Evolution of viscous boundary layer thickness in audible range.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 32

3.3 Full Linearized Navier-Stokes model (FLNS)

The most straightforward way of modeling the sound propagation in dissipative fluid is to
implement the full set of linearized Navier-Stokes equations into FEA solver. It is making
this model the most complete one: with velocity, pressure and temperature fluctuations
as degrees of freedom. Such modeling approach has been chosen by R. Kampinga [56,59]
that has presented comprehensive studies on FLNS model. The following procedure is
then mainly based on work of R. Kampinga and some aspects of the FEA theory is
based on book of O.C. Zienkiewicz [60]. Moreover it is a following of work on MEMS
microphone model presented already by Laboratoire Vibrations et Acoustique (LVA-
INSA de Lyon) [61, 62].

Assumptions
We assume that the fluid is Newtonian1 and use the ideal gas assumption.
Linearization assumes that the pressure and temperature variations are small relatively
to quiescent values and the velocity variations are small relative to the speed of sound
in the medium:
|p/p0 | 1, |τ /T0 | 1, |v/c0 | 1. (3.16)

3.3.1 Boundary conditions

The boundary conditions used in viscothermal models by N. Joly [58] and
R. Kampinga [59] are of Dirichlet or Neumann type. Dirichlet type is used when we
want to impose the fixed value of the variable on the boundary, while the Neumann type
impose the fixed value for the gradient of variable in normal direction. Depending on
the circumstances, the boundary conditions for the temperature are:

τ =0 ⇔ isothermal wall, (3.17)

τ = τ̄ ⇔ prescribed temperature fluctuations, (3.18)
∂τ /∂n = q · (−n) = 0 ⇔ adiabatic wall, (3.19)
q · (−n) = h̄ ⇔ heat flux, (3.20)

where h̄ denotes the external heat rate. The mechanical boundary conditions are as
follows:

v=0 ⇔ no-slip conditions, (3.21)

v = v̄ ⇔ prescribed velocity with no-slip conditions, (3.22)
¯·n=0
σ̄ ⇔ no normal stress, (3.23)
¯ · n = t̄
σ̄ ⇔ prescribed normal stress, (3.24)

where t̄ denotes the surface force density coming from external forces.
1
Under assumptions of Newtonian fluid the viscous stress at every point is proportional to the rate
of strain.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 33

3.3.2 Problem formulation

It is very difficult to resolve the FLNS model analytically, however it is possible to
implement the full set of governing equations into the FEA solver to obtain the estimated
solution. To do so we use the procedure presented in book of O.C. Zienkiewicz [60]. In
this procedure the problem is implemented into FEA in form of the weak equations.
Weak formulation approach enables one to avoid the second-order and higher-order
derivatives, to say colloquially - it weakens the general equations.

Domain
We know that the exact solution of the problem in the domain is obtained for:
∂v
0 = f̄ − ρ0 ¯v ,
− ∇p + ∇ · σ̄ (3.25)
∂t
∂τ ∂p
0 = Q̄ − ρ0 cP + − ∇ · q, (3.26)
∂t ∂t
1 ∂τ 1 ∂p
0= − − ∇ · v. (3.27)
T0 ∂t P0 ∂t
This equations are then weakened by integration over the domain D and can be presented
in the following manner:
Z
∗ ∂v
v · −ρ0 ¯v + f̄ dV
− ∇p + ∇ · σ̄
D ∂t
τ∗
Z
∂τ ∂p
+ · −ρ0 cP + − ∇ · q + Q̄ dV (3.28)
D T0 ∂t ∂t
Z
∗ 1 ∂τ 1 ∂p
+ p · − − ∇ · v dV = 0,
D T0 ∂t P0 ∂t

where each term is multiplied by v∗ , τ ∗ and p∗ that are the test functions.

Boundaries
Dirichlet boundary conditions constraint the value of the variable on the boundary.
Procedure of setting this condition as an addition to Neumann boundary conditions is
well explained in Comsol Multiphysics Reference Manual [63]. Dirichlet boundary terms
for the problem solution are as follows:

0 = v̄ − cm v, (3.29)
0 = τ̄ − cτ τ, (3.30)

where cm denotes the row vector that designates the spatial direction of applied velocity
and cτ is a scalar (1 × 1 row vector) representing value 1 on the boundaries where the
temperature fluctuations are imposed. This particular boundary conditions impose the
constraints but in practice the imposed values vary and to maintain the allowed error we
need to establish the reaction forces. Comsol Multiphysics use for this purpose variables
defined only at the boundaries and called the Lagrangian multipliers. Lagrangian multi-
pliers for our problem are: λm and λτ for mechanical and thermal boundary conditions

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 34

respectively. Lagrangian multipliers, which store the reaction forces are then introduced
into the Neumann boundary conditions that become:

¯ · n − cT
0 = t̄ − σ̄ m λm , (3.31)

0 = h̄ − q · (−n) − cT
τ λτ (3.32)

We may now present the weak form of the terms that have to be respected on the
boundary ∂D, starting with Neumann conditions for the velocity and the temperature
fluctuations (first two lines) and completing with Dirichlet conditions (third line):
Z
v∗ · t̄ − σ̄
¯ · n − cT

m λm dS
∂D

τ∗
Z
+ · h̄ − q · (−n) − cT
τ λτ dS (3.33)
∂D T0
λ∗τ
Z
+ λ∗m · {v̄ − cm v} + · {τ̄ − cτ τ } dS = 0.
∂D T0

Complete problem
We can arrange the weak form so the consecutive lines corresponds to the momentum,
state and continuity terms, then the last line corresponds to the Dirichlet boundary
conditions:
Z Z
∗ ∂v
v∗ · t̄ − σ̄
¯ · n − cT

v · −ρ0 ¯
− ∇p + ∇ · σ̄v + f̄ dV + m λm dS
D ∂t ∂D

τ∗ τ∗
Z Z
∂τ ∂p
+ · −ρ0 cP + − ∇ · q + Q̄ dV + · h̄ − q · (−n) − cT
τ λτ dS
D T0 ∂t ∂t ∂D T 0
Z
∗ 1 ∂τ 1 ∂p
+ p · − − ∇ · v dV
D T0 ∂t P0 ∂t
λ∗
Z
+ λ∗m · {v̄ − cm v} + τ · {τ̄ − cτ τ } dS = 0.
∂D T0
(3.34)

Foregoing equation contains the second order derivatives which can be reduced with use
of Green’s theorem as in work of R. Kampinga [56]. As a result we obtain:
Z Z Z
∗ ∂v ∗ ∗
v∗ · t̄ − cT

v · −ρ0 ∗ ¯
+ (∇ · v ) p + ∇v : σ̄v dV + v · f̄ dV + m λm dS
D ∂t D ∂D

τ∗ ∇τ ∗ · q τ∗ τ∗
Z Z Z
∂τ ∂p
+ · −ρ0 cP + + dV + · Q̄ dV + · h̄ − cT
τ λτ dS
D T0 ∂t ∂t T0 D T0 ∂D T0
Z
1 ∂τ 1 ∂p
+ p∗ · − − p∗ · (∇ · v) dV
D T0 ∂t P 0 ∂t
λ∗
Z
+ λ∗m · {v̄ − cm v} + τ · {τ̄ − cτ τ } dS = 0.
∂D T0
(3.35)

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 35

3.3.3 Solving method

Weighted residuals
Exact solution of the model is presented in eq. 3.34 or 3.35, however to find this solu-
tion FEA starts with replacement of variables with its approximations. Problem solved
with approximate variables gives intermediate solution that suffers of inaccuracies. In
weighted residual approximation method described by O.C. Zienkiewicz [60] this in-
accuracies are represented with the functions called the residuals. Residuals in our
document are denoted R and A, which are the functions of the same variables as the
general equations defined in the domains:

(D) ∂ ṽ
Rv (ṽ, τ̃ , p̃) = −ρ0 ¯v + f̄ ,
− ∇p̃ + ∇ · σ̄
∂t
∂ τ̃ ∂ p̃
R(D)
τ (ṽ, τ̃ , p̃) = −ρ0 cP + − ∇ · q + Q̄, (3.36)
∂t ∂t
1 ∂ τ̃ 1 ∂ p̃
R(D)
p (ṽ, τ̃ , p̃) = − − ∇ · ṽ,
T0 ∂t P0 ∂t
where ṽ, τ̃ and p̃ are the arbitrary variables. Residuals for the boundaries yields:
(∂D)
Rv ¯ · n − cT
(ṽ) = t̄ − σ̄ m λ̃m ,
(3.37)
R(∂D)
τ (τ̃ ) = h̄ − q · (−n) − cT
τ λ̃τ ,

(∂D)
Av (ṽ) = v̄ − cm ṽ,
(3.38)
A(∂D)
τ (τ̃ ) = τ̄ − cτ τ̃ ,

where λ̃m and λ̃τ are the arbitrary Lagrangian multipliers. Until the solver reaches
the most precise solution, subsequent arbitrary variables and Lagrangian multipliers are
used. Therefore until the exact solution is found the system response is non-null and
corresponds to the solution inaccuracy:

L̃(v˜∗ , τ˜∗ , p˜∗ , λ˜∗m , λ˜∗τ ,ṽ, τ̃ , p̃, λ˜m , λ˜τ ) = . . .

Z Z
(D) (∂D)
v˜∗ · Rv (ṽ, τ̃ , p̃) dV + v˜∗ · Rv (ṽ) dS
D ∂D

τ˜∗ τ˜∗
Z Z
+ · R(D)
τ (ṽ, τ̃ , p̃) dV + · R(∂D)
τ (τ̃ ) dS
D T0 ∂D T 0 (3.39)
Z
+ p˜∗ · R(D)
p (ṽ, τ̃ , p̃) dV
D

λ˜∗
Z
(∂D)
+ λ˜∗m · Av (ṽ) + τ · A(∂D)
τ (τ̃ ) dS,
∂D T0

where v˜∗ , τ˜∗ , p˜∗ , λ˜∗m , λ˜∗τ are the arbitrary test functions.

Discretization and convergence

R. Kampinga has studied the influence of the shape functions (used in the discretiza-
tion) on the stability of the solver and on the convergence. Combinations of the shape

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 36

functions for different degrees of freedom has shown that choosing the same shape func-
tions for the velocity and the pressure result in problems with stability. The stability
problems occurs mostly at low frequencies when the problem becomes incompressible.
The combinations of shape functions presented by Kampinga are gathered in table 3.1,
where:
• P1 - piecewise linear,
• P2 - quadratic,
• Q1 - bilinear,
• Q2 - biquadratic,
• P1+ - shape function space enriched by cubic bubble functions,
• P−1 - piecewise linear functions discontinuous over the element boundaries,
• P2+ - shape function space enriched by cubic bubble functions.

Table 3.1: Combinations of shape functions investigated by R. Kampinga [59].

Stokes element shape v, v ∗ p, p∗ τ, τ ∗

Taylor hood 4 P2 P1 P2
Taylor hood Q2 Q1 Q2
Crouzeix Raviart 4 P2+ P−1 P2
MINI 4 P1+ P1 P1

R. Kampinga correlated the performance of each shape function for simple 2D ge-
ometry with different mesh sequences (triangular and quadrilateral with different refine-
ment). He indicated the Taylor Hood and Crouzeix Raviart to be the best choices for
FLNS. Moreover both velocity and temperature elements are the most efficient if we use
quadratic functions.

3.3.4 Means of FLNS model enhancement

Another approach for modeling acoustic effects in dissipative fluid has been presented by
N. Joly [64]. This model is dedicated for numerical studies and the full set of linearized
Navier-Stokes equations is modified in order to present full acoustic problem only with
use of velocity and temperature fluctuations. It means that the model employs only two
out of initial four degrees of freedom and as a consequence it reduces the computation
cost. Equations have been derived under assumptions of Stokesian fluid2 which is ho-
mogenous and it is at rest. Additionally the linear behavior for the differential form of
fluid state equation has been assumed.
Starting with the momentum, continuity and state equations N. Joly introduces the
complementary parameters, where first two are the viscous characteristic lengths and
the third one is the thermal characteristic length:
lv = (η + 4/3µ)/ρ0 c0 , (3.40)

lv0 = µ/ρ0 c0 , (3.41)

lh = κ/(ρ0 c0 cP ). (3.42)
2
Assumption of Stokesian fluid means that the viscous stress is proportional to the rate of strain and
the heat flux is proportional to the temperature gradient.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 37

Additionally fluctuations of the volumetric mass density and the entropy are:

γ P0
ρ= 2 p− τ , (3.43)
c0 T0

cP T0 γ − 1
s= τ− p , (3.44)
T0 P0 γ
where γ denotes the specific heat ratio. Combination of the foregoing parameters and
the expressions of density and entropy with general set of equations gives:

∂2v
2
c0 ∂ ∂ P0 ∂
− 2 + + c0 lv ∇ (∇ · v) − c0 lv0 ∇ × (∇ × v) − ∇τ = 0, (3.45)
∂t γ ∂t ∂t ρ0 T0 ∂t
∂τ T0 γ − 1 2
− γlh c0 ∇ · (∇τ ) + ρ0 c0 ∇ · v = 0. (3.46)
∂t P0 γ

The foregoing set of equations appoint the temperature and velocity perturbations how-
ever it may be completed by the expression of acoustic pressure perturbations and den-
sity perturbations related to the temperature and velocity by the following identities:

∂p P0 ∂τ ρ0 c20
= − ∇ · v, (3.47)
∂t T0 ∂t γ
∂ρ
= −ρ0 ∇ · v. (3.48)
∂t
Model of N. Joly has been successfully applied for simulations of several acoustic el-
ements [58, 65]. However, regarding the problematic of this thesis the best boundary
condition for acoustic simulations is to prescribe the pressure (prescribed stress bound-
ary condition) at the microphone inlet. This kind of boundary condition is not evident
in case of the model presented by N. Joly which provides boundary conditions for tem-
perature and velocity [58].

3.4 Low Reduced Frequency model (LRF)

In 1975 H. Tijdeman presented so-called Low Reduced Frequency model which solves
the propagation of sound in viscothermal fluid enclosed by rigid cylindrical tube [66].
Later the problem was studied by W.M. Beltman [67], while R. Kampinga [56] applied
LRF to solve the problem for geometry that consists of two parallel plates.

Assumptions
LRF is a simplified viscothermal model for small tubes and slits. In this model the
pressure is assumed to be constant across the section and varies along the sound wave
propagation direction (axial direction). On the opposite hand, the velocity and the
temperature variations are considered only across the section of the duct. All three
authors make the similar assumptions for the problem which can be summarized:
• homogenious medium,
• no mean flow,
• laminar flow,

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 38

• small amplitude, sinusoidal perturbations,

• cross-section much smaller than the acoustic wavelength,
• cross-section is uniform or slowly varies,
• geometry length must be large to neglect the end effects,
• no internal heat generation.

Lets rewrite the general set of equations. We consider that direction x corresponds to
the propagation direction, then the v1 corresponds to the particle velocity fluctuations
in propagation direction, while the v2 and v3 components are considered null. We omit
the divergence of the velocity, which becomes ∇ · v = ∂v1 /∂x. Then the momentum
equation is limited to the propagation direction and it gives:
2
∂ v1 ∂ 2 v1

∂p
ρ0 jωv1 − µ 2
+ 2
=− . (3.49)
∂y ∂z ∂x

We then neglige the temperature fluctuations in axial direction to obtain the modified
state equation:
2
∂2τ

∂τ ∂ τ ∂p
ρ0 cP −κ 2
+ 2 = . (3.50)
∂t ∂y ∂z ∂t

Finally the continuity equation is not modified:

1 ∂τ 1 ∂p ∂v1
− + =− . (3.51)
T0 ∂t P0 ∂t ∂x

3.4.1 Problem formulation

Velocity and temperature fields
Further procedure is based on the approach of R. Kampinga [56] where the approximate
viscothermal solutions enable one to express the velocity and temperature fluctuations
with use of so-called shape functions for velocity field Ψv and temperature field Ψτ :

1 ∂p
v(x, y, z, ω) = Ψv (y, z, ω) − (x, ω) , (3.52)
ρ0 jω ∂x

p
τ (x, y, z, ω) = Ψτ (y, z, ω) (x, ω) . (3.53)
ρ0 cP

On first sight we may have impression that the expressions in brackets (eq. 3.52 and 3.53)
are not complete and that we describe the isentropic problem, however the dissipative
effects are included in the shape functions as shown below. The shape functions are the
solutions of the following differential equations:
2
∂ 2 Ψv

−2 ∂ Ψv
Ψv + kv + = 1, (3.54)
∂y 2 ∂z 2
2
∂ 2 Ψτ

−2 ∂ Ψτ
Ψτ + kτ + = 1, (3.55)
∂y 2 ∂z 2

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 39

where kv and kτ are the viscothermal wave numbers defined by R. Kampinga as:
kv2 = −jωρ0 /µ and kτ2 = −jωρ0 cP /κ. This definitions correspond in fact to the thick-
nesses of boundary layers (see section 3.2):

1−j 1−j
kv = and kτ = . (3.56)
δv (ω) δτ (ω)

Analytical solutions of shape functions for several cross-sections are presented by

R. Kampinga [56]. Moreover the solution of shape function for any cross-section may be
computed with use of FEA. Figures 3.5 and 3.6 show the solution of velocity shape func-
tion for circular cross-section. The results obtained with COMSOL reveal the evolution
of shape function with the frequency.

Figure 3.5: Evolution of the velocity shape function for circular vent with radius of
150 µm.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 40

(a) f = 500 Hz (b) f = 1 kHz

(c) f = 10 kHz (d) f = 100 kHz

Figure 3.6: Evolution of the velocity shape function for circular vent with radius of
150 µm

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 41

Towards the pressure field

Further considerations which lead to the designation of pressure fluctuations are based
on the mean values of shape functions over the section. With these mean values of shape
functions for velocity and temperature (Yv , Yτ ) and with modified mean temperature
0
field Yτ = γ − (γ − 1)Yτ , R. Kampinga defines mean (lumped) velocity, temperature
and density fluctuations:
0
Yv ∂p Yτ p Yτ p
< v >s = − , < τ >s = , < ρ >s = . (3.57)
ρ0 jω ∂x ρ0 cP c20

If we consider no mean flow (as in model assumptions), the pressure field may be ex-
pressed in form of modified Helmholtz equation:

∂2p
+ kl2 p = 0, (3.58)
∂x2
where kl is so-called LRF wave number defined as:
0
Y
kl2 = k02 τ and k0 = ω/c0 . (3.59)
Yv
Moreover with use of air acoustic impedance Z0 R. Kampinga introduces LRF charac-
teristic impedance Zl :

Z02
Zl2 = and Z0 = ρ0 c0 . (3.60)
Yτ0 Yv
Knowing the LRF wave number we may determine the pressure field inside the
domain:
p(x, ω) = pi (x, ω) + pr (x, ω) = Aejkl x + Be−jkl x , (3.61)
where the indexes i and r denote the incident and reflected waves, A and B are the
complex amplitudes. To continue we will assume that the pressure at the inlet (p(x1 , ω)
denoted p1 ) and the outlet (p(x2 , ω) denoted p2 ) of the domain is known. We introduce
the lenght of the domain Lt = x2 − x1 . Then the complex amplitudes are identified:

1/2j −e−jkl x2 e−jkl x1 p1

A
= . (3.62)
B sin kl Lt ejkl x2 −ejkl x1 p2

With the values of complex amplitudes we may describe the pressure field inside the
domain:
sin kl (x2 − x) sin kl (x − x1 )
p(x, ω) = p1 + p2 (3.63)
sin kl Lt sin kl Lt
and the pressure gradient:

∂p cos kl (x2 − x) cos kl (x − x1 )
− (x, ω) = kl p1 − p2 . (3.64)
∂x sin kl Lt sin kl Lt

3.5 Benchmarking
Low Reduced Frequency model seems to be a good alternative that may reduce the
computation cost in comparison to the FEM based on Linearized Navier-Stokes equa-
tions. To validate this model we present the benchmark where we examine the acoustic

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 42

phenomena inside the cylindrical tube (see fig. 3.7) with radius Rt = 150 µm and length
Lt = 0.5 mm. Knowing the pressure p1 and p2 imposed at the inlet and outlet of the
tube we estimate the flows to finally compare the transfer functions calculated by the
subsequent models. We emphasize this geometry since R. Kampinga provided the reso-
lution of shape functions for LRF, while FEM of LNS (Linearized Navier-Stokes) models
profits on the symmetry of the tube (we can model quarter of the geometry).
Lt

p1 p2
Rt q1 q2

Figure 3.7: Cylindrical geometry used in acoustic models benchmark.

The constants adopted for the benchmark are gathered in table 3.2. The similar
constants are used throughout this work.

Table 3.2: Constants used in acoustic simulations.

Symbol Value Unit Description

T0 298.15 [K] Quiescent temperature
P0 101325 [P a] Quiescent pressure
Mmol 28.8 [g/mol] Molar mass of air
R 8.3144621 [J/mol/K] Ideal gas constant
rair = R/Mmol 288.6966 [J/kg/K] Air gas constant
ρ0 = P0 /rair /T0 1.1772 [kg/m3 ] Quiescent density
γ = cP /cV 1.4 [1] Ratio of specific heats
cP = γ/(γ − 1)rair 1010.4 [J/kg/K] Specific heat at constant pressure
cV = 1/(γ − 1)rair 721.7415 [J/kg/K] Specific heat at constant volume
κ 25.4341 [mW/K/m] Thermal conductivity
µ 18 [µP a · s] Shear viscosity
c0 = γP0 /ρ0 347.1381 [m/s] Speed of sound in air
Z0 = ρ0 c0 408.6414 [P a · s/m] Acoustic impedance of air

3.5.1 Finite Element Models based on Linearized Navier-Stokes equa-

tions (LNS)
In section 3.3 formulation of FLNS (Full Linearized Navier-Stokes) model has been pre-
sented. This complete model describes the acoustic phenomena by taking into account
the dissipation of energy coming from the air viscosity and the thermal conductivity
of air. Such model is challenging in terms of computation cost in case of the complex
geometries.
In order to decrease the number of DOF (Degrees Of Freedom) we would like to compare
FLNS model to the simplified one, where the energy dissipation coming from the thermal
effects is neglected and the process becomes adiabatic. To obtain the simplified model
all of the temperature terms from equation 3.35 are removed. In further considerations

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 43

this model is called the ALNS (Adiabatic Linearized Navier-Stokes).

Finally the models list is completed by the COMSOL Thermoacoustic module, which
with minor differences (see appendix A for the details) corresponds to the FLNS.

Geometry and meshing

For the benchmark we use 3D geometry of cylindrical tube. The FEM take advantage
of the geometry symmetry, thus we calculate the solution using only quarter of the
cylinder (see fig. 3.8(b)). Figure 3.8 presents the mesh applied on examined geometry.
For the increase of computational efficiency the swept mesh was chosen. It was created
by triangular meshing of one cross-sectional face and then the reproduction of triangular
mesh over the 3rd axis. For proper modelisation of boundary layers, the element size on
the cross-section outer edge is limited to 1.5 µm (thickness of viscous boundary layer
at 1 MHz is 2.18 µm) and the maximum element growth rate is set to 1.1. Complete
mesh consists of 16345 domain elements and it is applicable to the simulations where
the maximal frequency equals 680 kHz (limited by 7 elements per wavelength in swept
mesh). As shown later, our benchmark includes the frequency up to 1 MHz, however
the results for the frequencies above 680 kHz are strictly comparative (all finite elements
models in the benchmark suffer of the same error).

(a) Triangular mesh on geometry cross-section. (b) Sweept mesh on 3D geometry.

Figure 3.8: Meshing of 3D geometry used in acoustic models benchmark.

3.5.2 Low Reduced Frequency model

Key elements of LRF model presented in section 3.4 are the solutions of viscothermal
fields Ψv and Ψτ on the geometry cross-section. For this benchmark we use the analytical
solution of viscothermal fields and their mean values Yv and Yτ given for cylindrical

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 44

cross-section by R. Kampinga [56]:

J0 (kv r) J2 (kv Rt )
Ψv (r, ω) = 1 − , Yv = − , (3.65)
J0 (kv Rt ) J0 (kv Rt )
J0 (kτ r) J2 (kτ Rt )
Ψv (r, ω) = 1 − , Yτ = − , (3.66)
J0 (kτ Rt ) J0 (kτ Rt )

where J0 , J2 are the Bessel functions of the first kind where the subscript denotes their
order; r varies between 0 and R.
The flow rate q which equals the integral of velocity over the surface is designated from
the combination of equation 3.57 and 3.64:
Scs Yv kl
q(x, ω) = − [cos kl (x2 − x)p1 − cos kl (x − x1 )p2 ] , (3.67)
ρ0 jω sin kl Lt
where Scs is the surface of the tube cross-section. We modify equation 3.67 with use of
the identity ρ0 ω = k0 Z0 and we denote the flow outgoing the tube on the input (q1 ) and
on the output (q2 ):

Scs Yv kl
q1 = [cos kl Lt p1 − p2 , ] , (3.68)
jk0 Z0 sin kl Lt
Scs Yv kl
q2 = − [p1 − cos kl Lt p2 ] . (3.69)
jk0 Z0 sin kl Lt
We may represent equation 3.68 and 3.69 as a system of equations:

q1 Scs Yv kl cos kl Lt −1 p1
= (3.70)
q2 jk0 Z0 sin kl Lt −1 cos k L
l t p2

and the input admittance of the system is:

q1 Scs Yv kl
= . (3.71)
p1 jk0 Z0 tan kl Lt

Following the example of A.D. Pierce [55] we represent the circular tube as a two-port
element with acoustic impedance Zd and acoustic compliance Yc (see figure 3.9). To
Zd

p1 Yc /2 Yc /2 p2

Figure 3.9: Simplified representation of circular tube as a two-port element.

identify acoustic impedance and acoustic compliance that appear under two different
circumstances A.D. Pierce [55] investigates two cases:

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 45

First case: Yc → ∞
The length of the tube is much smaller than the acoustic wavelength, the flow in-
side the tube is nearly incompressible. In this case we demonstrate the impedance
behavior of the tube and since the flow is nearly incompressible the acoustic com-
pliance goes to infinity and it is replaced by open circuit (fig. 3.10(a)).

Second case: Zd → 0
In the second case the air is compressed inside the tube and the value of air
flowing outside the tube is null. The acoustic impedance in this case approaches
zero, therefore it is treated as short-circuit and Yc is the only element describing
the acoustic behavior of the tube (fig. 3.10(b)).

p1 p2 p1 Yc p2

P0 P0

(a) First case: Yc → ∞ (b) Second case: Zd → 0

Figure 3.10: Two different representation of circular tube regarding its acoustic behavior.

In practice, to demonstrate the first case with use of equation 3.70 we need to estab-
lish the flow passing through the tube - excite the enclosed air in differential mode
(p1 = −p2 ). On the other hand to demonstrate the second case we need to compress the
air inside the tube - thus the air must be excited in common mode (p1 = 1, p2 = 0). If we
separate the differential and the common mode excitation, the equation 3.70 becomes:

q1 Scs Yv kl 1 1 −1 p1 Scs Yv kl 1 − cos kl Lt 1 0 p1
= + − ,
q2 jk0 Z0 sin kl Lt −1 1 p2 jk0 Z0 sin kl Lt 0 1 p2
(3.72)
where the expressions that precede the matrices corresponds to the acoustic admittance
1/Zd and the acoustic compliance Yc :
jk0 Z0
Zd = sin kl Lt , (3.73)
Scs Yv kl
Scs Yv kl 1 − cos kl Lt
Yc = − . (3.74)
jk0 Z0 sin kl Lt

3.5.3 Results discussion

Table 3.3 summarizes the cost of the compared models. Under this terms the LRF
model is clearly the best choice for the users with limited resources that at the same
time would like to keep the high resolution (number of frequencies). The FEM solutions
are significantly more expensive. Finite element analysis were made with use of machine
with 10 CPUs and they require 9 - 13 gigabytes of RAM. The number of frequencies
was limited to the minimal value that enabled us to compare the models based on FEA
with LRF. Nonetheless the simulations have taken from 30 minutes in case of the ALNS

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 46

model (omit of temperature terms reduces the number of DOF) to 60 minutes in case of
FLNS and commercial COMSOL Thermoacustic models which use full set of linearized
Navier-Stokes equations.

Table 3.3: Comparison of model computation cost.

Model
Feature Unit LRF ALNS FLNS COMSOL Thermoacoustic
DOF [1] - 229588 302728 302728
Number of frequencies [1] 1200 61 61 61
Physical memory [GB] 0.01 9.44 13.26 13.14
Virtual memory [GB] - 14.76 18.26 18.76
Solution time [min] 0.05 33 58 60

Bode plot of the input admittance (fig. 3.11) gathers the transfer functions for the
frequencies from 10 Hz to 1 MHz calculated with subsequent models. The shapes of
magnitude plots are similar in all the cases and only the COMSOL Thermoacoustic
model shows minor nonconformity on the phase plot. However the magnitude of LRF
model is slightly higher (10 dB) for all the frequencies. To understand this shift of LRF
magnitude one need to remember that in case of LRF model we neglect the fluid velocity
in cross-sectional directions. This simplification reduces the energy dissipation through
viscous effects and results in overestimation of the tube admittance. On the contrary

Figure 3.11: Bode plot of the input admittance (q1 /p1 ).

the 3D FEM models evaluate the velocity in every dimension and the admittance given
by those models is more precise. Figure 3.12 presents the velocity field distribution
in the tube. The module of velocity in propagation direction (longitudinal velocity) is
constant throughout the tube length (fig. 3.12(a)) while the velocities in cross-sectional
plane (radial velocity) have their maximum in the vicinity of the inlet and outlet of the
tube (fig. 3.12(b)).

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 47

(a) Module of longitudinal fluid velocity. (b) Module of radial fluid velocity.

Figure 3.12: Velocity field distribution in the tube, f = 10 kHz.

Figure 3.13 presents the same results as figure 3.11, however it focuses on the acoustic
resonances. Difference in magnitude between ALNS, FLNS and COMSOL Thermoa-
coustic models is displayed. ALNS presents slightly higher magnitude in the vicinity of
acoustic resonances since the energy is not dissipated by thermal effects.

Figure 3.13: Bode plot of the input admittance (q1 /p1 ) focused on acoustic resonances.

It has been proven that Low Reduced Frequency model is reliable replacement of
Finite Element Analysis for fast analysis of acoustic phenomena. It is true that this ap-
proach becomes problematic for the geometries with complicated cross-sections where
the analytical resolutions of shape functions are not available. However for those ge-
ometries the shape functions Ψv and Ψτ might be calculated numerically.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 48

It is then possible to evaluate the acoustic system with use of LRF model to adjust the
critical parameters of the system. At the end of design process the LRF-based simula-
tions can be enriched with use of finite element analysis to obtain the precise values of
the admittance.

3.6 Towards Lumped Element model

Lumped element modeling is a common technique used in modelisation of variety of
physical properties with use of analogy to simple mechanical system or discrete electronic
elements (if we use electronic elements the model is commonly called the Equivalent
Circuit model). The theoretical background of Lumped Element model in acoustic field
can be found in books of A.D. Pierce and M. Bruneau [54, 55]. Analogies of general
physical properties to their equivalents in fluidics, mechanics and discrete electronics
are gathered in table 3.4. We have decided to represent the acoustic elements with
discrete elements of electrical system. With this approach we may simulate the system
analytically or - in case of complex systems - with use of MATLAB or SPICE software.

Table 3.4: Lumped analogies of the physical properties [68]

General Fluidic Mechanical Electrical

Effort Pressure Force Voltage
Flow Vol. flow rate Velocity Current
Displacement Volume Displacement Charge
. Momentum Pressure Momentum Momentum -
Resistance Fluid resistance Damper Resistor
Capacitance Fluid compliance Spring Capacitor
Inertance Inertance Mass Inductor

3.6.1 Validation of Lumped Element model

At the end of subsection 3.5.2 we have introduced the acoustic impedance and com-
pliance of circular tube. Basing on this two elements we may identify the electrical
representation of two-port element (fig. 3.14) analogous to fig. 3.9. To obtain the ex-
Rac Lac

p1 Cac /2 Cac /2 p2

Figure 3.14: Electrical representation of two-port acoustic element.

pressions of lumped resistance Rac , inductance Lac and capacitance Cac we search for

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 49

approximation of impedance (eq. 3.73) and compliance (eq. 3.74) about ω = 0. The
simplified expressions of lumped parameters for circular tube becomes:
8µLt 4ρ0 Lt Lt πRt2
Rac = , Lac = , Cac = (3.75)
πRt4 3πRt2 2P0 γ
and with use of lumped parameters the input admittance of the tube is expressed with
the following formula:
q1 1
= + jωCac . (3.76)
p1 Rac + jωLac
To validate the identified lumped elements we trace again the input admittance of the
circular tube (fig. 3.15). We use the similar geometry to the one from the benchmark.
The admittance of both - LRF and Lumped Element model is coherent nearly till the
frequency of 100 kHz. Keeping in mind that the microphone bandwidth is 20 - 20 kHz,
the Lumped Element model is applicable to the simulations of this kind of sensors.

Figure 3.15: Comparison of the admittance obtained with the LRF and with Lumped
Element model.

3.6.2 Acoustic resistance of rectangular guide

As far we have shown how to identify the lumped elements for the circular tube, however
the new architecture of the MEMS microphone investigated in this thesis consists of the
acoustic elements that have a rectangular cross-section. If one of the rectangle sides is
much larger than the other one (A B - for illustration of A and B please see fig. 3.16)
the geometry can be approximated as a slit (layer). In this case we can analytically
describe the shape functions and their means values, then the acoustic resistance can be
calculated in the similar way as in case of the circular tube and it equals [54]:
12µLr
Rac = , (3.77)
BA3

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 50

Figure 3.16: Cross-section of the rectangular acoustic elelemnt.

where Lr denotes the length of the guide in the sound propagation direction. However
what can we do to identify the lumped elements when A and B are of comparable
dimensions? This problem can be resolved with use of LRF model, for example the
viscous resistance Rac may be identified with equation 3.73 for the Ψv and Ψτ solved in
FEA solver at low frequency.
To begin we have verified this approach by comparison of equation 3.73 and 3.75 on the
example of the circular tube. The results on figure 3.17 shows that the viscous resistance
identified with FEA solution is correct and we can apply it to other types of sections.

Figure 3.17: Acoustic resistance of the cylindrical tube as a function of radius R. The
length Lt of the tube is 500 µm.

We have prepared the same type of simulations for the rectangular guide where the
side lengths A and B were in rage of 0.5 - 50 µm. As a result we have obtained the
color map of the acoustic resistance (fig. 3.18). The figure highlights the importance
of the thermal and viscous boundary layers especially for the geometries smaller than
10x10 µm where the viscous resistance rises drastically.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 3. Models for simulations of acoustic phenomena in MEMS 51

Figure 3.18: Acoustic resistance of the rectangular guide as a function its side length A
and B. Color legend indicates the acoustic resistance with the units [Pa/(m3 /s)]. The
length Lr is set to 500 µm.

Summary
We have gone through the set of models that may be used to investigate the acoustic
behavior of the MEMS microphone. Despite the high computational costs of the models
based on Finite Elements Method, these models seems to be the wisest and quickest way
to understand the acoustic phenomena at the microscale. At the end of this chapter
we have reached the representation of the acoustic phenomena in form of the equivalent
electrical circuit. In the next chapter we apply these two models to the design consider-
ations of the new MEMS microphone architecture. We start with the simulations of the
microphone with use of the FLNS model, then - with the knowledge acquired thanks
to FEA, we construct the model of investigated microphone in form of the electronic
circuit.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Sensor design considerations

In the following chapter we focus on the design of the new MEMS microphone. We now
use FLNS model with fluid-structure interaction to predict the microphone frequency
response and to detect the crucial factors that have an influence on it. Afterward we
discuss the mechanical structure and demonstrate how to represent complicated truss
beam in 2D. Coupling between the microphone subsystems (acoustical, mechanical and
electrical systems) enabled us to present the microphone sensitivity with three different
models: 2D, 3D FLNS models and with Lumped Element model. The last model is
subsequently used to analyze the noise sources that restrict the lower limit of signal
detection. Finally we shortly discuss the influence of technological dispersion (overetch)
and the mechanical shocks on the microphone functioning.

4.1 Architecture properties of considered MEMS

Overall view
Considered MEMS (fig. 4.1(a)) has overall dimensions of 1.5x1.5x0.6 mm3 , it consists of
4 micro-beams placed between the inlet vents and the outlet vents (fig. 4.1(b)). These
beams move rotationally in plane of a Silicon wafer - the axis of rotation coincides with
the vector normal to the surface of a wafer.
outlet vents

p2
p2
electrical connections
p2 p2 p2

p1
p1 e2 e1
p1
e3
e2 e1
Sound waves
(b) Location of two out of four mi-
e3 crobeams
inlet vents
(a) Overall view

Figure 4.1: Cross-sectional view of MEMS dice.

MEMS dice is then packaged; typical chip-scale package showed on fig. 4.2 includes two
separated chips which are MEMS and ASIC (Application-Specific Integrated Circuit).

52
Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf
© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 4. Sensor design considerations 53

Integrated circuit amplifies sensor signal and furnishes it in analog or digital output form.

ASIC

e2
Sound waves
e1

Figure 4.2: Bottom port configuration of MEMS microphone package (not to scale).

Standard PCB (Printed Circuit Board) is used as a support and metal, cup-shaped lid
is used for package sealing. Depending on a sound port position we distinguish two
package configurations called the top port and the bottom port. For the M&NEMS
microphone considerations the bottom port configuration has been chosen. The package
introduces the acoustic cavity into the microphone. The volume of this cavity called
the backvolume is obtained by deducting the volumes of the MEMS and the ASIC chips
from the internal volume of the package. The influence of the backvolume is considered
in the microphone design.

Acoustic system
The cross-section of the MEMS dice acoustic system along with the symbols of each
dimension are presented on fig. 4.3. Microbeams are placed between two silicon wafers
wio

hio
Str(1) Str(2) Str(3) Str(4)

hio

e2
wio
e1

Figure 4.3: The cross-section of the MEMS dice with view on the acoustic system.
One can see two inlets and three outlets. All of the inlets and outlets have the same
dimensions.

(fig. 4.4). This architecture introduces the viscous damping that is related to the gaps a
and b localized above and below the beam. The dimensions of the elements of acoustic
system are listed in the table 4.1.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

wc
a

e2 h

b
e1

Figure 4.4: Cross-section of the beam and the surrounding wafers (so-called ”coupler”).

Table 4.1: Dimensions of the acoustic system

Symbol Value Unit Description

lio 740 [µm] Length of the inlet and outlet vents along the e3 axis
(out of plane dimension of fig. 4.3)
hio 313 [µm] Height of the inlet and outlet vents
wio 100 [µm] Width of the inlet and outlet vents
a 1 [µm] Height of the gap between the beam and SOI wafer
b 2 [µm] Height of the gap between the beam and sealing wafer
h 10 [µm] Height of the beam (defined by epitaxial layer)
wc 30 [µm] Width of the couplers
Var 7.65 [mm3 ] Backvolume

Mechanical system
Our investigations concern two types of the diaphragms: the first type is the simple
beam (fig. 4.5) while the second one (fig. 4.6) it is a beam constructed with use of truss
elements. Both beams are suspended in the same manner, moreover the truss beam
architecture maintains the same mass that the one of the simple beam. Piezoresistive
gauge is attached to the movable element at one extremity while the other side is an-
chored. Nanogauge is situated along the line between the center of rotation and center
of inertia of the beam. The dimensions of the beams used in the project are summarized
in table 4.2.
cdh
L
leq
lh

d
lg
wg

Figure 4.5: Top view of the simple beam.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

cdh
L
θ
lh l

d
lg
cdm
wg

Figure 4.6: Top view of the truss beam.

Table 4.2: Dimensions of the beams

Symbol Value Unit Description

L 740 [µm] Length of the beam
leq 3.52 [µm] Width of the simple beam
l 10 [µm] Overall width of the truss beam
h 10 [µm] Height of the beam (defined by epitaxial layer)
d 4 [µm] Distance between the center of rotation and the center
of the nanogauge
lg 2 [µm] Length of the nanogauge
wg 250 [nm] Width of the nanogauge
hg 250 [nm] Height of the nanogauge
lh 30 [µm] Length of the hinge member
cdh =cdm 1 [µm] Width of the hinge and beam member respectively

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Electrical system
Electrical system of the discussed sensor consists of four piezoresistive nanogauges. They
are arranged into the Wheatstone bridge architecture that was discussed in subsec-
tion 2.2.2. The bridge may be biased with constant current or constant voltage (see
fig. 4.7).
Vb
Ib

R+ΔR R-ΔR R+ΔR R-ΔR

ΔV ΔV

R-ΔR R+ΔR R-ΔR R+ΔR

Figure 4.7: Wheatstone bridge architecture used in MADNEMS sensor.

The electrical specifications of the considered electrical system are gathered in ta-
ble 4.3.
Table 4.3: Electrical specifications of designed sensor.

Symbol Value Unit Description

R 4.46 [kΩ] Nominal resistance of the nanogauges
Ib 100 [µA] Bias current of the Wheatstone bridge
Vb 0.446 [V ] Bias voltage of the Wheatstone bridge
Material properties
N 5 · 1018 [cm−3 ] Impurity concentration (p-Si)
π pzr 0.4 [1/GP a] Piezoresistive coefficient of Silicon

4.2 Transduction chain - assemble of the microphone sub-

systems
The previous section introduced the elements of the considered microphone - the acous-
tic system of the MEMS dice, the chip-scale package, the mechanical structure and
the electrical system. This elements have to be assembled in order to estimate the
microphone transfer function. Transduction chain presented on figure 4.8 includes the
relations between the systems:

• acoustic system of the MEMS dice have three inputs (p1 , p2 and qm ) and three
outputs (q1 , q2 and ∆p).
• The influence of the backvolume is coupled to the acoustic system by the feedback,
the pressure p2 is estimated with use of the flow rate q2 that enters the backvolume.
• The mechanical system is coupled to the acoustic system with use of the pressure
gradient ∆p generated across the microbeams. Mechanical system introduces to

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

the acoustic system the flow rate qm generated by the movement of the beams.
Mechanical system is coupled to the electrical system by the longitudinal stress σg
generated inside the nanogauges.
• Internal stress of a nanogauge results in the variation of the nanogauge resistance
that is specified by the piezoresistive effect. Finally, the resistance variations may
be interpret by Wheatstone bridge that is biased with current Ib and provides an
output voltage ∆V .

p1 q1

Δp ΔV
MEMS Mechanical σg Electrical
acoustic system qm structure system Ib

p2 q2

Chip-scale
package

Figure 4.8: Transduction chain of the M&NEMS microphone.

4.3 2D finite element analysis of MEMS with fluid-

structure interaction
Dimensions of designed microphone ranges from 1 mm down to 1 µm. The smallest
acoustic elements are the slits situated above and below the beam (fig. 4.4). Their
dimensions a = 1 µm and b = 2 µm are much smaller than the thicknesses of thermal
and viscous boundary layers (illustrated on fig. 3.3) for the targeted sensor bandwidth
(20Hz to 20kHz). Detailed analytical and FEM models that investigate thermoviscous
effects have been prepared within the MADNEMS microphone project and presented
by C. Guianvarc’h and T. Verdot [61, 62]. Using a simplified model for moving beams,
these models handle fluid-structure interactions.

4.3.1 Structure of the model

Acoustic and thermal boundary conditions
We present the cross-section of the MEMS chip on fig. 4.9 with the boundary conditions
that are required to meet the acoustic and mechanical behavior of the device. Therefore
most of the surfaces of the acoustic system are simulated as the isothermal wall (non-
slip conditions and no temperature variations); we impose the pressure p1 at the MEMS
inlet and p2 at the MEMS outlet.

Fluid-Structure Interaction (FSI)

Closer look on the couplers, where the fluid-structure interaction is realized is presented
on the figure 4.10. The velocity of the fluid and the beams on the fluid-structure interface

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

isothermal wall

e1 p1

Figure 4.9: Cross-sectional view of MEMS with boundary conditions indicated.

isothermal wall
e2
FSI(1) FSI(2) FSI(3) FSI(4)

Figure 4.10: Closer view on the ”couplers”.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

(the countour of the beam) is set to be equal:

L
θ̇ − v = 0 (4.1)
2
and it is a resultant of the beam movement driven by the acoustic pressure fluctua-
tions where both mechanical and acoustical problems are solved independently. The
mechanical behavior of the beams is governed by the analytical model (presented in
subsection 4.4.1) and introduced into the FEM with use of ordinary differential equa-
tions. The displacement of the beams in 2D model is approximated by the average
displacement of the rotating beam (fig 4.11).
e1
e3
rotational translational
movement movement

L e2

θ θ
e1

Figure 4.11: Representation of rotationall movement of the beam in 2D.

Meshing
The mechanical behavior of the beams is model analytically, thus the meshing is limited
to the acoustic domains and the beams remain not meshed. We employ anisotropic
mesh, which is adjusted on the boundaries to simulate reliably the boundary effects.
The smallest thickness of boundary layers over the audible bandwidth equals 15 µm for
20 kHz. Therefore the maximal size of the mesh elements on the boundaries is restrained
to 5 µm so that the boundary layers are covered by three elements in the worst case
(fig. 4.12). The smallest elements of acoustic system are located above and below the
beams. This part of the acoustic domain is meshed much more finely, the maximum
dimension of elements is restrained to 0.5 µm (fig. 4.13). Comprehensive studies on the
strategy of meshing for the viscothermal models were presented by N. Joly [58].

4.3.2 Discussion on the MEMS chip acoustic properties

Initial verification of the fluid-structure interaction is possible by investigation of the
input admittance Y11 (ω) = q1 (ω)/p1 (ω) of the MEMS chip (fig. 4.14). We have investi-
gated two cases:

first case: the beams are fixed and their boundaries are simulated as a wall;

second case: the beams are moving and the fluid-structure interation is introduced.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Maximal size
restrained to 5 μm

Figure 4.12: Overall view on the mesh.

Maximal size restrained to 0.5 μm

Figure 4.13: Mesh in vicinity of the beams.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

The comparison of the admittances shows that the characteristics of the systems are
the same at low frequency (f <2 kHz), where the viscous resistance prevails and at the
high frequency (f >100 kHz) where we observe the fluid compression inside the MEMS
followed by the acoustic resonances. What distinguish the admittance of the model with
moving beams is the increase of acoustic admittance between 10 and 20 kHz. Over this
section the admittance is dominated by the flow generated by the mechanical structures
that resonate at 16 kHz.

Figure 4.14: Comparison of input admittance with moving and fixed beams.

Graphic representation of the pressure and velocity fields gives another information on
the system. Figure 4.15 focuses on harmonic solutions (pressure fluctuations and velocity
field) computed at 1 kHz and 16 kHz. These solutions highlight two types of behavior:

• at low frequency when velocity of a beam remains weak air velocity in the slits is
mainly governed by the prescribed difference of pressure (velocity profile is similar
to the Poiseuille parabolic flow).

• In the vicinity of mechanical resonance of a structure (16 kHz), beam velocity

affects air velocity in the slits and the velocity field become similar to the Couette
flow (whose velocity profile is linear through the thickness of the slit).

These studies presented in papers by C. Guianvarc’h and T. Verdot [61, 62] prove that
viscous effects occurring in the small slits generate and maintain a pressure drop across
the beams.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

(a) Surface graph of pressure field with the ar- (b) Surface graph of pressure field with the ar-
rows indicating velocity field at 1 kHz. rows indicating velocity field at 16 kHz.

Figure 4.15: Manifest of fluid-structure interaction: for low frequency (1 kHz) the air
flow is governed by the pressure gradient while at the mechanical resonance (16 kHz)
the beam generates the air flow.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Mismatch between the input and output admittance

On figure 4.16 we present the input (Y11 ) and output (Y22 ) admittance of the system.
The traces are initially even, however the output admittance tends to be higher than the
input admittance starting from 1 kHz. Such behavior is motivated by the asymmetry
of the acoustical system (two inlets/three outlets). The dimensions of the inlets and
outlets are the same however the overall volume of the inlets is lower than the overall
volume of the outlets.

Figure 4.16: Comparison of input and output admittance of MEMS chip.

If we inquire further the asymmetry of the acoustic system we may recognize that the
beams located in the center of the MEMS (Str (2), Str (3) on fig. 4.3) are subject to the
lower acoustic compliance than the beams at the periphery of the MEMS (Str (1), Str (4)).
This problem was not investigated more precisely throughout this thesis however one
can expect that the phase of the center beams will be shifted regarding the external
beams.

4.3.3 Influence of MEMS chip-scale packaging on the microphone re-

sponse
To calculate the frequency response of the microphone, we must obviously take into
account the volume called the backvolume (obtained by deducting the volumes of the
MEMS and the ASIC chips from the internal volume of a package). The backvolume
(Var ) introduces acoustic compliance Car :

Var
Car = , (4.2)
γP0
where γ denotes ratio of specific heats for air and P0 is the static pressure. Regarding
the closed cavity of a package with dimensions larger than thermal and viscous boundary

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

layers we can use the law of mass conservation:

Z Z
dV ∂par
= −∇ · v dV, (4.3)
Var γP0 ∂t Var

where par is a pressure in backvolume and it equals p2 . By solving the integrals from
equation 4.3 we obtain the time derivative of pressure inside the backvolume that can
be implemented in FEM as an ordinary differential equation:
∂par qar
= , (4.4)
∂t Car
where qar is a total volumetric flow entering the backvolume and it equals q2 .
Comparison of input admittance of the MEMS with and without the package
(fig. 4.17) reveals the dissimilarities for the low frequencies. At first sight it is diffi-

Figure 4.17: Comparison of input admittance of MEMS chip (infinite backvolume) and
packaged MEMS (backvolume limited by the package).

cult to explain intuitively the elevated input flow for low frequencies in case of packaged
MEMS. To understand this phenomenon we investigated the response of the system in
time - more precisely the time that is required for equilibration of the pressure p1 and
the pressure inside the package par . Figure 4.18 shows that if we impose the pressure of
1 Pa at the input of the microphone, it takes around 20 ms for the pressure inside the
backvolume to equalize. This characteristic time - τRC - is set by the viscous resistance
of the MEMS and the acoustic compliance of the backvolume.
Let us now examine the response to the harmonic input signal p1 applied at three
different frequencies: at the frequency of 60.42 Hz that corresponds to the characteristic
time and at the frequencies that are in range of one decade (fig. 4.19). The signal at
the frequency 6 Hz corresponds to the period of 167 ms. It is longer period than the
characteristic time τRC of MEMS, therefore the pressure inside the backvolume manages
to equilibrate to the input pressure and the pressure gradient across the microbeam

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 4.18: Temporal response to the static pressure of 1 Pa imposed at the input.

(p1 −par ) is negligible. For the frequency of 60.42 Hz we see that the signal par decreased,
and at the frequency of 600 Hz the gradient of pressure across the microbeam equals p1 .
This considerations underline the role of the backvolume and the viscous resistance. At
the frequencies with period lower than the characteristic time τRC the gradient across
the beams is to low to obtain the reliable response of the microphone.

(a) f = 6 Hz.

(b) f = 60.42 Hz.

(c) f = 600 Hz.

Figure 4.19: Temporal response to the harmonic pressure imposed at the input.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

4.4 Detailed discussion on the mechanical system

4.4.1 Analytical model
Mechanical properties
Mechanical governing equation of the system with inertia J, damping DV , stiffness
C and the torque generated by the acoustic pressure ΓA is presented in the following
manner:
J θ̈ + DV θ̇ + Cθ = ΓA . (4.5)
The damping DV of the system and the torque ΓA are the consequence of fluid-structure
interaction and are governed by the viscous effects (section 4.3). Inertia of the beam is
denoted by its mass and its length:

L2
J = mb . (4.6)
3
For the analytical considerations we assume that the beam is rigid and that the to-
tal stiffness of the system assembles the rotational stiffness of the hinge Ch and the
longitudinal stiffness of the nanogauge Kg :

C = Ch + dKg d. (4.7)

The rotational stiffness of the hinge has been calculated by parallel connection of two
hinge members. The Young modulus (Eh ) of the hinge members that are rotated by
30 degrees relative to <110> crystallographic direction is approximately 140 GPa.

h · cd3h

Ch = 2 Eh . (4.8)
3lh

Formula for longitudinal stiffness of the gauge uses the Silicon Young modulus in <110>
direction (E<110> = 169 GPa) and the dimensions of the gauge:
sg
Kg = E<110> , (4.9)
lg

where sg is the cross sectional surface of the gauge (sg = wg hg ).

Knowing the inertia of the beam and the total stiffness of the system we may evaluate
the in-plane resonance: r
1 Ch + dKg d
f0 = . (4.10)
2π J
For the dimensions given in table 4.2 the resonant frequency calculated from the ana-
lytical formula is 16 kHz bor both - the simple and the truss-structured beam.

Discussion on mechanical resonance

The above analysis assumed that the beam is totally rigid. In fact the beam will be
subjected to small scale deformations proportional to its overall stiffness.
Analytical designation of mechanical resonance frequency becomes in this case problem-
atic. We may presume that the exact value of the resonance frequency is a combination
of two extreme behaviors:

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

• the beam is rigid, it is free to move at one extremity and suspended by the micro-
hinge and the nanogauge at the second extremity,
• the beam is non-rigid, it is deprived of microhinge and nanogauge - its behavior
resembles the case of clamped-free beam.
For now we accept the introduced error, however to provide the final specification we
endorse the resonance frequency designated by FEA (subsection 4.4.2), since it is a much
more reliable method in this circumstances.

Mechanical transfer function

Mechanical transmissibility of M&NEMS microphone is expressed by the value of lon-
gitudinal stress in a nanogauge σg generated by pressure difference ∆p. We propose to
start by calculating the quasi-static value of longitudinal stress as a function of rotation
of a beam expressed by θ:
L Sd∗ ∆p
θ= , (4.11)
2 Ch + dKg d
where Sd∗ denotes the lateral surface of the beam Sd that is modified due to the viscous
effects governed by the 2D finite element model. Later we express the quasi-static value
of longitudinal stress:
Kg d
σg = θ. (4.12)
sg
In order to introduce the inertial effects we consider fluctuation of pressure at angular
frequency ω and mechanical resonance of a beam at ω0 . Knowing that the beam is an
oscillator with quality factor QV given by viscous shear damping DV , the mechanical
transfer function becomes:
 
∗
σg Sd L/2 dKg d  1
TM (ω) = =  . (4.13)

2
∆p Sg d Ch + dKg d

ω 1 ω
1 − ω0 + QV j ω0

Mechanical system optimization

Thanks to the mechanical transfer function expression (equation 4.13), we may reveal
the procedure to enhance the “mechanical sensitivity”. Microphone based on M&NEMS
technology profits of two mechanical amplification mechanisms. First amplification
mechanism comes from the ratio between the modified lateral surface of a beam Sd∗
and section of a nanowire Sg while the second comes from the ratio of the lever arm
(L/2d).
Enhancement of mechanical sensitivity with use of these factors seems to be straight-
forward, however the design has to follow the microfabrication restrictions (resolution
of lithography, aspect ratio of DRIE, internal stress).
If we continue the study of mechanical transfer, we may see that the stress exerted
on the section of the nanogauge is related to the distance between the center of rotation
and the center of the nanogauge d. We may simply trace the stress as a function of d
with use of analytical expression and numerical model (illustrated on fig. 4.20).
Optimal value of d may be also obtained by analysis of equation 4.13. We first isolate
the part of mechanical transfer function that depends on d:

∗ L/2 dKg d
f (d) = . (4.14)
d Ch + dKg d

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 4.20: Optimal value of d parameter.

We know that the maximum of f ∗ may be found when ∂f ∗ /∂d = 0. It give us optimal
distance of d defined as: q
d∗ = Ch /Kg . (4.15)
To illustrate how the error in position of the gauge d influence the overall mechanical
transfer function we introduce d∗ into the equation 4.13:
 
∗
∗ σg 1 Sd L/2  1
TM (ω) = =  , (4.16)

d∗
2
∆p 2 Sg

1 − ωω0 + Q1V j ωω0

knowing that Ch /Kg = (d∗ )2 , the ratio TM /TM

∗ can be presented in the following form:

TM 2 (d/d∗ )
∗ = . (4.17)
TM 1 + (d/d∗ )2

Relation of TM and TM ∗ from equation 4.17 can be traced as a function of d/d∗ . The

resulting graph (fig. 4.21) reveals that the misalignment of d introduced either in design
process or during the fabrication may reduce the mechanical sensitivity of MEMS. It is
important to notice that the horizontal axis is presented in logarithmic scale, therefore
the performance drops more drastically if d is lower than its optimal value d∗ (d < d∗ )
rather than in the opposite direction (d > d∗ ).

4.4.2 FEM model

3D Comsol model has been prepared in order to analyze the dynamic properties of
the microbeams. The model use single-crystal Silicon. The Young modulus for the
used material is gathered in elasticity matrix. Microbeam is aligned according to the
crystallographic direction of the wafers used in microfabrication of MEMS.
Microbeam is damped resonator with the viscous damping that is introduced by shear

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 4.21: Influence of nanogauge misalignment on the resultant mechanical transfer.

stress generated inside the gaps below and above the beam. This damping has been
brought into the model by use of Rayleigh damping hypothesis1 :

[DVF EM ] = α [M ] + β [K] , (4.18)

We have considered that the damping is proportional to mass matrix, thus β = 0 while
α equals:
µ 1 1 1
α= + , (4.19)
ρSi h a b
where µ and ρSi are the air shear viscosity and the Silicon volumetric mass density
respectively.

Geometry and mesh

Geometry of the hinge and the beam is build with use of the work plane that is
subsequently extruded to the desirable thickness. The hinge and the nanogauge handle
elements are created as a polygons while the beam is a rectangle. Finally the union of
the foregoing elements with no internal boundaries is generated and the work plane is
extruded. The nanogauge is attached to the resulting beam geometry as a 3D block.
Specific construction of the long beam together with the nanogauge and the microhinge
results in the high scale factors. Regarding the scale factors:

• h/leq ≈ 3,
• h/hg = 40,

different sizes of the mesh elements have to be employed. Therefore we identify four
geometry elements which needs different meshing strategies:
1
Rayleigh damping use an assumption that the damping matrix is proportional to the mass and the
stiffness matrices. Therefore it introduces the damping to the system with use of mass and stiffness
matrices multiplied by coefficients α and β. [69, 70]

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

• microhinge where the maximal element size is cdh /2,

• nanogauge handle where the maximal element size is 2cdm and the mesh is refined
along the edges to the size cdh /6,
• nanogauge where the maximal element size is wg /2,
• the beam where the maximal element size is leq /3.

The resulting mesh is illustrated on figures 4.22 and 4.23.

Figure 4.22: Tetrahedral mesh used in simple beam for mechanical studies.

Figure 4.23: Tetrahedral mesh used for discretization of the hinge and the nanogauge.

Eigenvalue analysis

Eigenvalue analysis were used to determine in-plane mechanical resonance of the

beam, furthermore this analysis may be used to extract the modal damping of the beam
used to estimation of the Brownian noise (see subsection 4.8.1). Unfortunately the
in-plane resonance occurs at the frequency much lower than the 16 kHz approximated
with analytical formula in subsection 4.4.1. The shape of this resonant mode which
occurs at 7.7 kHz is shown on fig. 4.24.
Under such circumstances we have decided to build the beam with use of truss struc-
ture (see fig. 4.6) which is commonly used in structural mechanics. The truss-structured

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 4.24: First mechanical resonance.

beam is build in such manner that it preserves the same mass as the simple beam. How-
ever its stiffness is higher since the truss members are subjected to tensile and compres-
sive forces rather than bending. The 3D model of the beam is prepared with the same
procedure as the simple beam and the truss is build as an array of the truss members.
The mesh of the truss (fig. 4.25) is prepared individually considering the aspect ratio
between the width of the members and the height of the beam (h/cdm = h/cdh = 10).

Figure 4.25: Tetrahedral mesh used in truss-structured beam for mechanical studies.

The results of the simulations (fig. 4.26) confirm that the beam based on truss structure
represents higher stiffness than the simple beam (in-plane resonance occurs at 12.8 kHz
which is much higher than 7.7 kHz given by simple beam). Therefore for the further
design considerations we will use the truss beam.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 4.26: First mechanical resonance.

Frequency domain analysis

Frequency domain analysis were used for sensitivity value extraction. Sensitivity
for FEM model has been evaluated basing on the average longitudinal stress generated
inside the nanogauge by the reference pressure of 1 Pa. The average longitudinal stress
was then introduced to the equation 2.16 to obtain the resulting resistance variation.
The final sensitivity value was obtained from the electrical transfer function for the
full Wheatstone bridge architecture (eq. 2.19). The simulations results are shown and
compared to analytical sensitivity in section 4.7.

4.5 Truss structured beam - from 3D to 2D simulations

On the contrary to the simple beam (fig. 4.5) where the cross-section is constant across
the beam length, the truss beam is much more sophisticated and there is no straight-
forward 2D representation of its geometry. Since we have preserved the mass of the
beam, the adjustment of resonant frequency coming from equation 4.10 is not required,
however we face the acoustic problem which is the adjustment of viscous resistance in-
troduced by the gaps above and below the beam. To investigate this problem we have
prepared three models with different geometries of beam cross-sections:

first: truss is represented with rectangle where the width corresponds to the equivalent
width of truss members which in fact equals leq (fig. 4.27),

second: truss is represented with rectangle where the width equals the overall with of
truss beam l (fig. 4.28),

third: we represent the beam with three rectangles: two external where width equals the
width of truss members (cdm ) and the central one which is slightly wider (1.2 µm)
to take into account the angle of the central member of the truss (fig. 4.29).

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 4.27: Truss beam represented with use the equivalent width.

Figure 4.28: Truss beam represented with use the overall width.

Figure 4.29: Truss beam represented with use of 2D ”truss”.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

The aim of these models is to investigate only the acoustic properties, therefore to
simplify the model: we omit fluid-structure interaction and the beams outlines are simu-
lated as the isothermal walls. Moreover in order to save the computation cost we model
just one coupler, where the pressure at the inlet and outlet is set to 1 and 0 Pa. The
resulting pressure and velocity fields for 1 kHz presented of figures 4.27 - 4.29 are insuf-
ficient to compare the 2D modeling approaches and appoint the most precise representa-
tion of 3D structure. Therefore to validate our choice we have prepared 3D model where
the 3D beam geometry was subtracted from the coupler geometry (no fluid-structure
interaction). Even the 3D model of the coupler (without the fluid-structure interaction)
that was used to extract the results showed on figure 4.30 is extremely expensive in
terms of computation. To visualize the problematic of 3D simulations: properly meshed
model that use COMSOL Thermoacoustic module takes 6 minutes to solve one fre-
quency using 10 CPU’s. During that time it solves the problem for over 5 millions DOF
while using 100 GB of RAM. This extreme computation cost makes 3D model inefficient
for engineering problem which is the design of MEMS microphone where the numerous
geometry parameters have to be revised.

Figure 4.30: Pressure and velocity field distribution in 3D truss beam.

The quantitative comparison of beam representation was made basing on the input
admittance for the audible bandwidth (fig. 4.31). We may determine two extreme cases:

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

geometry with equivalent width that has the lowest viscous resistance and the geometry
with overall width that has the highest viscous resistance. We lean towards the interme-

Figure 4.31: Comparison of input admittance for different 2D representations of truss

beam.

diary solution which is the 2D ”truss”. This 2D representation slightly underestimates

the input admittance of coupler with truss beam. This underestimation of the input
admittance will generate a proportional error in estimation of the frequency response
of the microphone (sensitivity - section 4.7). We accept this error and in the further
design procedure we use the 2D model with the beam represented as the 2D ”truss”.
However to estimate the error introduced by the 2D representation we have prepared
the fully-coupled 3D model in which we have limited the microphone geometry to one
coupler and the beam. We have set the gradient of pressure across the coupler to 1 Pa
at the frequency of 1 kHz (fig. 4.32(a)). The estimation of the average stress gener-
ated inside the nanogauge (fig. 4.32(b)) lead us to the value of the sensitivity which is
compared to the one obtained with 2D model in section 4.7.

(a) Pressure field distribution. (b) Stress accumulation inside the nanogauge.

Figure 4.32: 3D model of one coupler and the beam with fluid-structure interaction.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

4.6 Electrical system transfer function

MADNEMS microphone consists of 4 active gauges, thus the resulting ∆V is propor-
tional to the longitudinal stress in the nanogauge in the following manner (as presented
in subsection 2.2.2):
∆R
∆V = Vb = Vb πpzr σg (4.20)
R
and the electrical transfer function becomes:
∆V
TE = = Vb πpzr . (4.21)
σg

4.7 Sensitivity
Total sensitivity of a microphone is the assembly of its acoustical, mechanical and elec-
trical transfer functions. The sensitivity response (fig. 4.33) given in this section is
designated with use of three different models:
• 2D FEM model,
• simplified 3D FEM model,
• Lumped Element Model.
FEM models were already presented while the Lumped Element Model is introduced
later in section 4.8. 3D model is not an efficient tool for the design of M&NEMS
microphone and the value of the sensitivity has been calculated only for 1 kHz. Two
remaining models give nearly similar response for the whole bandwidth. In these models
the mechanical behavior of the beam is governed with the same analytical approach,
however the acoustic phenomena model differs and it is more precise in the case of the
FEM. This difference is visible at low frequency and at the resonant frequency. It is
mainly caused by different way of viscous resistance estimation.
For the purpose of sensitivity analysis we have prepared the general equation that
describes the sensitivity, however we need to remember that such elements as: Sd∗ , QV
and τRC are designated in different manner for each model. The components that do
not depend on the frequency are gathered and expressed as S0 , the nominal sensitivity
of a microphone: ∗
Sd L/2 dKg d
S0 = [πpzr Vb ] . (4.22)
Sg d Ch + dKg d
Then, to obtain the total sensitivity we add the terms that depends on the frequency:
 

∆V 1 τRC jω
S(ω) = = S0  . (4.23)
 
2
p1 1 + τRC jω

ω 1 ω
1 − ω0 + QV j ω0

Lower limit of a microphone bandwidth is fixed by the inverse of time constant 1/τRC
coming from viscous resistance of the slits and the acoustic compliance of the back
volume. High frequency limit is set by the first mechanical resonance frequency of
M&NEMS structure.
Nominal sensitivity at f = 1 kHz and bias voltage of Wheatstone bridge Vb = 0.46 V is
estimated with 2D FEM simulations to -56.4 dBV (1.5 mV/Pa). This value is an output
voltage of Wheatstone bridge, the final sensitivity of the microphone will be enlarged
by amplification of readout electronics (ASIC).

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 4.33: Frequency response of the MADNEMS microphone.

Influence of chip-scale packaging

In subsection 4.3.3 the influence of chip-scale packaging on the MEMS microphone re-
sponse was introduced. We have identified that the major factors are the viscous resis-
tance of the gaps (proportional to the gaps height b) and the acoustic compliance of the
backvolume (Var ). Indeed the figures 4.34 and 4.35 shows how easy is to affect the low
frequency response of the microphone by varying this two parameters.

Figure 4.34: Comparison of normalized frequency response S(ω)/S0 of the microphone

for different b parameters.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 4.35: Comparison of normalized frequency response S(ω)/S0 of the microphone

for different backvolumes.

4.8 Noise budget

Detection threshold of the sensor is limited by the noises included in the transduction
chain. We will now discuss the origins of this noises and their contribution to the noise
budget of the microphone. The considerations are based on the comprehensive report
of T. Gabrielson [71] which is recommended as an introduction to this section.

4.8.1 Origins of the noises

Depending on the origin, the noises are introduced into the transduction chain in the
different manner, they are then subjected to the same conversion mechanisms (for ex-
ample amplification and filtering) as the functional signal. Therefore the noise budget
considerations that are of high importance, are presented in order of the appearance of
the noise in the transduction chain (fig. 4.36).
Four main noise sources were identified in the designed microphone:
• thermoacoustic noise which generates the pressure fluctuations as a result of energy
dissipation by the viscous resistance of the acoustic system,
• the Brownian noise which induces the force on the MEMS surface similarly to the
force generated by the gradient of pressure.
• Johnson noise introduced by the thermal noise of the strain gauge,
• Flicker noise introduced into the strain gauge.
Some of the mentioned noises have mutual origins. These are the thermoacoustic noise,
the Brownian noise and the Johnson noise, which are the thermal noises that accompany
every dissipative effect (every dissipative mechanism cause the fluctuations). Thermal
noises are constant across the spectrum (white noise), we can express them in the fol-
lowing way:
Fn = 4kb T R∗ , (4.24)

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Flicker
noise

Acoustic Δp σg Electrical ΔV
Mechanical
system structure system

Thermoacoustic Brownian Johnson Thermal

noise noise noise noises

Figure 4.36: Noise model of the microphone.

where kB denotes the Boltzmann constant, T is temperature in Kelvins and R∗ denotes

the dissipation mechanism (in case of the microphone it is the viscous resistance of the
acoustic system, damping of the mechanical structure and the electric resistance of the
nanogauges).

Thermoacoustic noise
This noise is related to the viscous resistance of the acoustic system. Dissipation of
acoustic energy inside the MEMS will then generate the pressure fluctuations according
to the thermal noise formula:

FVn = 4kb T Rviscous . (4.25)

Brownian noise
Mechanical-thermal noise (Brownian noise) is a white noise caused by the Brownian
motion of the fluid particles. Those particles randomly interact with the surface of the
microsystem causing the beam displacement in the same manner as in the case of sound
pressure load. The resulting beam displacement is then transduced into variations of
output voltage.
Problematic of the Brownian noise might be omitted at large scale, however not in case
of the microsensors. The most comprehensive studies on Mechanical-thermal noise were
prepared by T.B. Gabrieson [71, 72]. He have presented the method of Mechanical-
thermal noise evaluation basing on the Equipartition Theorem and Nyquist Relation.
Spectral density of force generated by mechanical-thermal noise is then:

FBn = 4kb T Rmech , (4.26)

kB denotes the Boltzmann constant, T is temperature in Kelvins and Rmech denotes the
mechanical resistance (damping).

Analytical model examination. Knowing the viscous shear damping DV we may

estimate the torque ΓBn exerted on the beam by the Brownian noise. To do so, we

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

replace mechanical resistance Rmech in equation 4.26 by the viscous shear damping:
p
ΓBn = 4kb T DV . (4.27)

Knowing the torque generated by Mechanical-Thermal noise we may modify and com-
bine equations 4.11 and 4.12 to obtain:
 

Kg d ΓBn 1
σgBn =  . (4.28)
 
2
sg Ch + dKg d

ω 1 ω
1 − ω0 + QV j ω0
h √ i
Finally, the contribution of this noise expressed in V / Hz over the sensor bandwidth
is:
VBn = Vb πpzr σgB . (4.29)

FEM examination. Finite element model allow us to estimate the value of damping
by use of modal analysis presented in section 4.4.2. It may be extracted from the modal
mass µj in the following manner:

[DVF EM ] = α t φj [M ] φj = α [µj ] , (4.30)

where φj is a mode shape matrix and µj is a modal mass. Knowing thehmodal damping,
√ i
we give the Mechanical-Thermal noise as an input noise expressed in P a/ Hz with
the modified formula given by T.B. Gabrielson [72]:
s s
Rmech DVF EM
Bn = 4kb T 2 = 4kb T . (4.31)
(Lh) (Lh)2

Flicker noise
Flicker noise is also called Hooge noise or 1/f noise because its spectrum is inversely
proportional to the frequency. It is an intrinsic noise of each conductor, it is introduced
into the sensor transduction chain at the level of silicon nanogauges.
The origins of this noise are still under study, however they are linked to the fluctuations
of conductivity that can be the results of fluctuation of carriers density and fluctuation
of their mobility. The behavior of the noise has been empirically described in 1969 by
F.N. Hooge [73] and the comprehensive discussion on 1/f noise origins may be found in
his more recent work [74]: r
αH
VFn = Vb , (4.32)
cn f
where αH is called the Hooge parameter and it is a material constant while the cn
denotes the number of carriers inside the resistor. The Hooge parameter depends on
the material (quality of the crystal). It can be optimized by following the doping of the
Silicon by annealing process.
The number of carriers cn inside the gauge is calculated with:

cn = lg sg N · 106 , (4.33)

where N is an impurity concentration (see table 4.3).

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Johnson noise
Johnson noise (Johnson-Nyquist noise) is another intrinsic noise that is related to the
nanogauge resistance and contributes to the total noise of the nanogauge. It has a white
frequency spectrum and it is represented by the following formulation:

VJn = 4kb T R. (4.34)

4.8.2 Equivalent circuit representation of the microphone for the noise

budgeting
Circuit representation and designation of lumped elements
Equivalent circuit representation is a popular technique of MEMS microphones simu-
lations. It has been applied by several authors [19, 26, 29, 32, 33] to model MEMS mi-
crophones response and the noise budget. Moreover it seems to be the most convenient
method to estimate the output noise generated by the system. For sake of simplicity, we
present the lumped elements of the designed microphone on the example of one beam
(fig. 4.37).

Car
Rvent
Rslit
Cvent /2 Cvent /2
qm

Rvent

e2
p1
e1

Figure 4.37: Lumped elements of the microphone presented on single coupler.

We may then identify the following elements:

• p1 is the acoustic pressure at the input of the microphone,

• Rvent is the viscous resistance of the inlet and outlet vents,
• Cvent is the acoustic compliance of the inlet and outlet vents,
• Rslit is the combined viscous resistance of the gaps above and below the beam,
• qm is the air flow generated by the mechanical structure,
• Car is the acoustic compliance of the backvolume.

If we want to represent the full architecture of designed microphone, we need to take four
beams, two inlet vents and three outlet vents. The equivalent circuit is then presented

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

on fig. 4.38. The microphone simulations based on such circuit may be done in Spice
software or by use of Linear Time-Invariant (LTI) models in MATLAB.

Car

Rvent Cvent /2 Rvent Cvent /2 Rvent

Rslit Rslit Rslit Rslit

qm1 qm2 qm3 qm4

Cvent /2 Rvent Cvent /2 Rvent Cvent /2

Figure 4.38: Lumped model of the microphone.

Our considerations assume that the air is non compressible in the slits above and
below the beams. Moreover for audible bandwidth inertial effects can be neglected.
Including viscous shear stress in the air, Navier-Stokes momentum balance equation
provides relationship between the pressure p and velocity v fields [75]:
∂p ∂2v
= µ 2, (4.35)
∂x ∂y
Equation 4.35 is integrated considering difference of pressure across the coupler ∆p
and no-slip conditions at fluid-structure interfaces. As a result we obtain formula for
volumetric air flow inside the coupler qc :
1 L
qc = ∆p + Sd∗ θ̇, (4.36)
Rslit 2
where θ̇ is the angular velocity of the beam (see fig. 4.6), Rslit denotes the total viscous
resistance of gaps that is calculated with LRF model and given later and Sd∗ is modified
lateral surface of a beam driven by viscous effects:

a+b
Sd∗ = (Lh) 1 + . (4.37)
2h
The second part of equation 4.36 is the air flow generated by the mechanical structure:
L ∗
S θ̇. qm = (4.38)
2 d
Torque applied to rotating beam is defined by acoustic force reduced by viscous shear
damping:
L
ΓA = Sd∗ ∆p − DV θ̇, (4.39)
2

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

where viscous shear damping lumped coefficient DV is:

µL2

1 1
DV = (lL) + . (4.40)
3 a b
The acoustic compliance of the backvolume is calculated with use of the following for-
mula:
Var
Car = . (4.41)
γP0
In this model the values of some elements are designated with use of the LRF model
approach described in section 3.6.2. With this approach we have determined the value
of Rslit , Rvent and Cvent :

Table 4.4: Lumped elements determined with use of LRF model.

Symbol Value Unit

Rslit 3.2444e11 [P a/(m3 s)]
Rvent 9.987e7 [P a/(m3 s)]
Cvent 3.59e-19 [m3 /P a]

Introduction of noise sources

Resistors presented on figures 4.38 and 4.37 are the noiseless resistors, in fact both
schematic have to be modified to introduce the noises coming from the viscous resis-
tances. For the noise analysis each noiseless resistor is complemented by the voltage
noise source as shown on fig. 4.39.

noiseless
noiseless resistor
resistor
voltage noise
source

Figure 4.39: Introduction of noise sources.

4.8.3 Total transduction chain noise

Total noise of the microphone corresponds to signal (voltage) level measured at the
output of microphone located in a perfectly quiet environment (no pressure fluctuations).
Knowing the sensor sensitivity, noise can also be represented as an equivalent pressure
noise at the entrance of the microphone. This second form gives an order of magnitude
of the signal level detectable by the sensor.

Contributions of noise sources

We present the global view of noise spectral density on the figure 4.40 it reveals the
three noises that dominates: the Flicker noise, the Johnson noise and the Brownian
noise. Flicker noise (which decreases along with frequency) predominates the total noise
of the sensor at low frequencies and it is negligible in the vicinity of the mechanical

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

resonance of the sensor. Johnson noise dominate the noise budget between 1.3 and
6.5 kHz, then in the vicinity of the mechanical resonance the performance is limited by
the mechanical Brownian noise which is amplified mechanically in the same manner as
the functional input signal. The thermoacoustic noises are of the smaller importance in
the architecture of presented microphones, however the noise of the couplers may disturb
the microphone performance as its value is relatively important at low frequencies.

Figure 4.40: Overall contributions of noises.

On figure 4.41 we present once again the importance of the backvolume for the
microphone performance. The scale of this figure has been adjusted to the scale of fig-
ure 4.40 to facilitate the comparison. To trace this noise contributions we have assumed
that the backvolume is infinite. We see that the backvolume has an impact only on the
thermoacoustic noises. It is important information for the future optimization of the
sensor and shows that in the miniaturization of the sensor, the reduction of backvolume
is one of the most challenging limitations.

Noise floor
Typically the noise floor of every sensor is given at the output of the readout electronics.
The value of the noise floor is given in V and more common in dBV which is a voltage
relative to 1 V. Therefore to obtain the noise floor of the sensor we need to integrate
the value of the noises over the bandwidth:
v
u fZmax
u
Vtotn = t VV2n + VB2n + VF2n + VJ2n df . (4.42)
u

fmin

Equivalent input noise (EIN)

The value of EIN provides the information on the detection threshold of the microphone.
Namely the signal with magnitude below the EIN will not be detected since it will be

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 4.41: Overall contributions of noises under assumption of infinite backvolume.

hidden below the intrinsic noise of the sensor. The equivalent input noise is presented
in dBSPL which is the sound pressure level referenced to the threshold of hearing. In
order to obtain EIN, first we obtain its value in Pa:
s
fmax
VV2n + VB2n + VF2n + VJ2n df
R
fmin
EINP a = , (4.43)
S(ω)
then we convert it to dBSP L by the following formula:

EINP a
EIN = 20 log10 , (4.44)
pref

where pref , the threshold of hearing equals 20 µPa.

4.9 Influence of overetch on the microphone response

MEMS structure is obtained in dry etching process (see section 5.1). The dry etching is
a very precise process, however depending on the specification of the geometry and the
process parameters the dimensions of final structure slightly vary [76]. This derogation
(example of overetch effect is shown on fig. 4.42) is introduced into the transduction
chain at the level of acoustic system (overetch of acoustic vents) and at the level of
mechanical structure.
Observations of cleaved wafer with SEM suggests the overetch of 9 µm in case of
acoustic vents and 125 - 200 nm for the mechanical structure. However this measure-
ments certainly suffer of error and give rough approximation. According to the author
experience, the overetch value for micromechanical structure may reach 300 nm and
the final value has to be measured by use of built-in structures used for technological
process characterization. Overetch of the mechanical structure appears in the shift of

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 4.42: Overetch issue in DRIE. The 2 µm intended width varies from 2.25 to
2.40 µm over the thickness of epitaxial layer.

mechanical resonance (MEMS mass will vary) and in critical cases it may affect the
integrity of mechanical structure (in case of truss and microhinge the initial sizes goes
down to 1 µm).
We have investigated the the consequences of overetch on acoustic transfer function.
The results are compared to the nominal sensitivity response on fig. 4.43 and 4.44. Size
dispersion of inlets and outlets affects the sensitivity value at the mechanical resonance
while the overetch of the truss beam marginally shifts the low cut-off frequency of the
microphone (overetched truss beam has lower viscous resistance).

Figure 4.43: Influence of inlets and outlets overetch on the response of the microphone.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 4.44: Influence of the truss beam overetch on the response of the microphone.

4.10 Tolerance to mechanical shocks and sound pressure

overload
Mechanical shock
MEMS microphones are most commonly used in portable devices. This type of ap-
plications impose the challenging requirements in terms of mechanical shock reliability
since the hand held devices frequently suffer of drop impacts. Typical failures in MEMS
devices are:

• fracture,
• stiction,
• short-circuit,
• package failure.

Maximum ratings usually consider the acceleration of 10 000 g. MEMS that are already
in advanced phase of development are simulated and simultaneously undergoes the drop
tests together with their assembly (chip-scale package). Mechanical shock studies [77,78]
of commercial MEMS microphones shows the highest fragility (membrane rupture) in
Z-axis. It is not surprising for the out-of plane moving membrane.
Our investigations are limited to the fracture inside the MEMS chip. Furthermore
we may already assume that the crucial element that suffers the highest stress is the
nanogauge. Lets now consider the type of numerical studies that simulates the drop
impact. The device is dropped on the hard surface and the generated stress will be
studied, thus we need to perform transient studies with acceleration in form of the
pulse [79]. We have chosen the half-sine pulse (fig. 4.45) expressed as:
π
Gs (t) = Gs0 sin t, (4.45)
T
where Gs0 and T are the peak acceleration and the width of the pulse. The shock is

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 4.45: Characteristic of the mechanical shock applied to the MEMS.

applied as a body load to the whole mechanical structure sequentially in each spatial
direction. The highest stress corresponds to the e1 direction (fig. 4.46) which is relevant
since it is the sensor working orientation. Regarding the fig. 4.46 one might be concerned
of the high von Mises stress. In e1 direction stress spans to nearly 10 GPa while the
crystalline Silicon fracture stress adopted by [46] is 2.8 GPa.
If we trace the spatial displacement generated at he beam extremity by the acceleration
(fig. 4.47) it oscillates with amplitude of 20 µm in case of e1 direction and 10 µm in case
of e2 direction. In fact the MEMS has embedded stops that restrain the displacement to
1 µm in case of e1 and e3 directions. For the e2 direction the displacement is restrained
by the gaps above (2 µm) and below (1 µm) the beam. Keeping in mind the stops we

Figure 4.46: Maximal von Mises stress in the nanogauge.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 4.47: Spatial displacements of the beam extremity.

may say that the shock-generated stress in real structure will be largely reduced.

Maximum acoustic input

Maximum acoustic input it is a maximum sound pressure level that can be transduced
by the M&NEMS microphone within the linear range. After [46] the maximal stress
in the gauge that assures the linear behavior equals 100 MPa. From 100 MPa we can
designate the maximum input pressure (∆pmax ) by combining equations 4.11 and 4.12:

sg 2 Ch + dKg d
∆pmax = σgmax . (4.46)
Kg d L Sd∗

If we express the maximum input pressure in dB, we obtain the maximum acoustic input
of 115 dB.

Summary of sensor specifications and discussion

Sensor specifications
In this chapter we have investigated the new architecture of the microphone. Those
investigations resulted in the prevision of the sensor specifications that are gathered in
table 4.5.
We can not squarely compare this specifications to the commercial products because
we give the specifications of the MEMS dice without the ASIC. Normally the ASIC will
amplify the output signal and the sensitivity of designed microphone will reach the one
given by commercial products (typically -38 dBV). The given SNR is much lower than
the best results of the commercial and state of the art sensors. We can accept this value
for the prototype, however for the future designs this value has to be optimized. It can
be done by splitting the noise sources and investigating the problem in two fields:

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Table 4.5: Evaluated specifications of MADNEMS microphone (without signal condi-

tioning).

Parameter Value Conditions

Performance
Sensitivity -56.4 dBV 1 kHz, 94 dB SPL
Signal-to-Noise Ratio (SNR) 44 dB
Equivalent Input Noise (EIN) 50 dB SPL bandwidth 20 Hz - 20 kHz
Frequency response 0.04 to 20 kHz limited by -3 dB point
Maximum acoustic input 115 dB SPL limited by the linearity

Power supply
Supply voltage 0.46 V supply current 100 µA

Output characteristics
Maximum output voltage 16.9 mV 115 dB SPL input

Electrical noises consist of the intrinsic noises of the nanogauge - the Flicker noise
and the Johnson noise. First of this noises can be decreased by the optimization
of technological process and improving the crystal lattice quality. The second one
is proportional to the electric resistance.
Acousto-mechanical noises are placed together since both: thermoacoustic noise
and Brownian noise depends on the architecture of the acoustic system. We have
shown that the viscous effects allow to establish the pressure gradient across the
beam and enable the functioning of the microphone. However the same effects in-
troduce the viscous resistance which generates the thermoacoustic and Brownian
noise. To decrease these noises we need to optimize the architecture of the acoustic
system (cavities in MEMS and the backvolume). This type of noises seems to be
even more challenging than the electrical noises since the noises optimization and
further miniaturization do not come along.

As for the rest of the specifications, it corresponds to the ones of the commercial
products. We have been aware that matching all of the specifications to the commercial
products is challenging in terms of the sensor design, however we want our microphone
to be competitive. Therefore our design corresponds to the commercial products in such
critical aspects as the overall chip-package size and the power consumption.

Tools for design of a MEMS microphone

Another aspect of this chapter was the application of FLNS and Lumped Element model
to the sensor design. Both of this models are valuable in the design process. Initial de-
sign and especially the dimensioning of the microphone were prepared basing on the
FLNS model with fluid-structure interaction. This model is efficient if prepared in 2D
and it is essential to understand the acoustic phenomena at the microscale. Afterward
we have used the Lumped Element model for the analysis of the noises in the micro-
phone. Moreover the frequency response of the microphone obtained with both of the
models shows a good match. Despite the high computational cost we have prepared
simplified 3D model to verify if 2D representation of the microphone is correct. The

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

sensitivity value obtained with 3D model is in slightly lower than the one obtained with
approximative models, however it is in the same scale.
These tools have to be validated with the measurements of the fabricated device.
Unfortunately the technological problems have postponed the fabrication of the pro-
totype, however the fabrication process has been already optimized and the prototype
should be available shortly. Anyway the presented models are proven to be a powerful
tools for the MEMS microphone designer.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Technological implementation

Basic M&NEMS process flow is presented at the end of chapter 2. This basic process
flow has been adapted to fulfill the requirements of the designed microphone architecture
and it is presented in the first section of this chapter. Afterward we discuss the MEMS
microphone packaging approaches based on the commercial products review to finally
propose our in-house solution. Due to the problems that have been encountered, the
MEMS fabrication is still in progress, however at the end of the chapter we present
the microscope images of the MEMS elements and the measurements of MEMS vertical
deflection.

5.1 Process flow

Basic M&NEMS technological platform has been described in chapter 2. The same
platform is used in the first steps of MEMS wafer fabrication, however the further
process is different from the one described in chapter 2 - the electrical connections are
not realized by direct metalization of the MEMS wafer but they are realized on the
sealing wafer. Both wafers are fabricated simultaneously, their process flow presented
on fig. 5.1 consists of the following steps:

MEMS wafer :

(a) we start with the SOI wafer,

(b) nanogauges are defined,
(c) nanogauges are covered with protective Silicon Oxide layer,
(d) Silicon epitaxial layer is grown,
(e) mechanical structures are defined in the epitaxial layer.

Sealing wafer :

(f) we start with a standard silicon wafer which is oxidized and the metal layer
for the electrical network is deposited,
(g) electrical network is realized in the metal layer,
(h) the electrical network is isolated with the oxide layer and the vias for electric
connections are realized,
(i) finally the metal which is used in eutectic bonding is deposited.

92
Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf
© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 5. Technological implementation 93

MEMS wafer
(a)

Sealing wafer
(f)
(b)

(g)
(c)

(h)
(d)

(i)

(e)

e2 Si SiO2 metal

Figure 5.1: Side view of MEMS wafer process flow (not to scale).

Metal patterns on the sealing wafer that are used in eutectic bonding have two functions:
they provide electric connections to MEMS and they form the sealing ring around the
MEMS structures to provide the mechanical attachment and hermeticity. With both
MEMS and sealing wafers prepared, we pursue the process (fig. 5.2):

Sealing and process finish :

(a) surfaces of both wafers are prepared and the eutectic bonding in vacuum is
realized,
(b) both wafers, that are initially 725 µm thick are thinned down to 300 µm
with use of grinding. High quality surface is then obtained with use of CMP
(Chemical Mechanical Polishing) process,
(c) the inlet and outlet acoustical vents are etched in DRIE process, the oxide is
used as an etch-stop layer. At the same time we etch the cavity that enables
the access to the electric pads (left side of the chip),
(d) The last step involves the vapor HF etching of the oxide to open the acoustic
vents and then to release the MEMS structures.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

(a)

(b)

(c)

(d)
e2
Si SiO2 metal
e1

Figure 5.2: Side view of wafer bonding and process finish (not to scale).

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

5.2 Advances and difficulties encountered in microfabrica-

tion
Fabrication process is carried out at CEA-LETI. Figure 5.3 illustrates the final shape of
MEMS part with zoom on nanogauge and ellipses indicating the future positions of the
inlets (red) and the outlets (blue). One of the MEMS wafers has been withdrawn from
the process at this point, then the mechanical structures have been released with HF to
investigate the out-of-plane deflection of the beams (see section 5.3).

p2 Piezoresistive nanogauge
p1

Rotation axis

Figure 5.3: SEM image: top view of the microbeam and the nanogauge.

On figure 5.4 we present the patterns on finished sealing wafer: the sealing ring that
assures hermeticity, the electrical pads for wire bonding and the paths that distribute
the signal (Wheatstone bridge wiring: power supply and signal acquisition; wiring of
test electrode for MEMS electrostatic activation test). The electrical and sealing metal
layers are isolated with use of the oxide layer and the vias are prepared to connect the
paths with the nanogauges.

Sealing ring

Vias

Electrical connections

Figure 5.4: Optical microscope image: top view of the sealing wafer with the sealing
ring, the electrical connections and the vias used for Wheatstone bridge wiring.

Both MEMS and sealing wafer have been bonded, the infrared inspection (fig. 5.5)
showed that the electrical paths are well aligned with the corresponding patterns on
MEMS wafer.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 5.5: Infrared microscope image of the bonded wafers. Zoom reveals the points
where the electrical connection is provided to the nanogauges.

Encountered difficulties
Designed MEMS microphone requires the modification of standard M&NEMS process
flow. Some of processes prepared for previous M&NEMS sensors contain the wafer
eutectic bonding, however it is the last step in fabrication process. On the contrary
the process flow of the microphone continues after the eutectic bonding and includes
grinding and essential step: etching of acoustic vents.
Several of the assembled wafers were damaged in the steps following the bonding. The
possible causes are associated to the poor bonding strength combined with the fact
that the wafers are weakened because of the grinding. This process have been already
optimized and the prototype should be ready shortly.

5.3 Out-of-plane deflection of MEMS

Mechanical architecture of discussed microphone includes relatively long structures an-
chored at one of the extremities. Because of that one may suspect that the released
structures tend to deflect in vertical (out-of-plane) direction (fig. 5.6). Vertical deflec-
tion is a known problem in MEMS, the most common origin of long structures deflections
comes from the mismatch between the thermal expansion coefficients of the various ma-
terials of MEMS stack. It is true for the multilayered MEMS, however in our case the
MEMS beam is fabricated entirely from the crystalline Silicon and the foregoing inter-
pretation is no longer valid.
However if we study the process flow (see section 5.1) we see that the beam is in fact
composed of two crystalline Silicon layers: first one which is 250 nm thick in which the
nanogauges are defined and the second layer which is 10 µm thick. The first layer is
the one of SOI wafer whereas the second layer is grown in epitaxial process. We do not
know if the both process were done under the same conditions (temperature, pressure),
therefore we presume that the interface of both layers is not free of the crystal lattice
defects. This defects may lead to the internal stress and finally (after the release of the
structure) the stress relaxation leads to the vertical deflection of the beam.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

out-of-plane deflection

e2
e1 e3

Figure 5.6: Schematic of out-of-plane deflection of the microphone beam.

Examination of the internal stress is not in the scope of the presented thesis, however
there is a need of such statistical studies (with focus on different geometries and process
parameters) for further development of the M&NEMS microphone. The considerations
of stress-induced deflection in this thesis are restricted to the designation of the over-
all trends. The measurements (fig. 5.7) were made with use of optical profilometer
Veeco Wyko NT9100 (white light interferometer) on the MEMS wafer withdrawn from
the fabrication process before the wafer level packaging and released with vapor HF.

Figure 5.7: Optical interferometry measurements sketch.

We have tested the beams with length of 500, 740 and 1000 µm. The results gathered
on the fig. 5.8 shows that the deflection does not depend on the width of the hinge (cdh )
and beam (cdm ) members (for the description of geometric parameters see section 4.1),
however we see clearly that the deflection increases with the beam length and it is within
the range 0.50 - 2.75 µm. The designs where the deflection exceeds 2 µm (distance be-
tween the beam and the sealing wafer) should be avoided since the beams will get in
contact with the sealing wafer and the microphones with such beams will not work.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 5.8: Deflection of the beam extremity depending on the beam length.

5.4 Chip scale packaging

As shown in section 4.3, the MEMS dice housing remains an equal element of the
microphone that influences its final performance. Therefore for the final functional test,
the MEMS dice should be packaged in the same manner as the existing commercial
products. Such solution has been foreseen for the presented M&NEMS microphone,
however our package will be introduced after the overview of the popular packages used
in the industry.

5.4.1 Technology overview

MEMS microphones are typically assembled in LGA (land grid array) package that is
compatible with surface mount technology. It can be soldered under the same conditions
as the other SMD elements. Depending on the sound port position we distinguish two
package types called the top and the bottom port. Although the bottom port architec-
ture is more problematic (inlet vent must be drilled in the PCB laminate - fig. 5.9), it
is more common since it enables increase of the package backvolume.

Knowles SPU0410LR5H
Most of the MEMS microphones packages resemble this of Knowles SPU0409LE5H mi-
crophone. The packages are in form of thin rectangular cuboids (typical dimensions are
around 4 x 3 x 1 mm3 ) where the electric pads are located on the surface of the PCB
(see fig. 5.9). The surface with electric pads that after soldering comes into contact
with PCB indicates the bottom of the package. The embodiment consists of PCB card
equipped with soldering pads and sound port. MEMS and ASIC dies are covered with
metal lid which is attached to the PCB with the conductive adhesive. Thanks to Sys-
tem Plus Consulting report [16] we may have a look on the cross-section of the Knowles
microphone package (fig. 5.10). MEMS die is aligned with the PCB sound port so the
sound propagation path is reduced to the minimum. The connections between MEMS
and ASIC (that is placed aside) are realized with wire-bonding (fig. 5.11), then the

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 5.9: Overview of Knowles SPU0409LE5H bottom port package [16].

connections and ASIC are covered with protective resin. Cross-sectional view reveals
additionally the vias in PCB that provide the integrated circuit output signal to the
PCB soldering pads.

Figure 5.10: Cross-section of Knowles SPU0409LE5H package [16].

Epcos T4000 and T4060

Epcos presents another approach for MEMS microphone packaging (fig. 5.12). It is
still bottom port package, however the MEMS and ASIC are upside-down (microphone
diaphragm is facing the substrate while the back volume cavity is created only with
internal MEMS cavity). The dies are placed side by side and the electrical connections
are provided with golden bumps. Epcos uses the ceramic substrate and the sealing is
first realized with the elastic films and then it is finished with metal layer (fig. 5.13).
This approach reduces drastically the volume of the package, which in case of T4060
model is 2.65 x 1.81 x 0.95 mm3 .

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 5.11: Wire-bond connections in Knowles SPU0409LE5H package [16].

Figure 5.12: Overview of Epcos T4060 bottom port package [80].

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 5.13: Cross-section of Epcos T4000 package [81].

Akustica AKU230
Akustica is the only company that managed to integrate the readout electronics di-
rectly into the MEMS microphone dice (fig. 5.14:Die Overview). This asset facilitates

CMOS/MEMS die

Figure 5.14: Overview of Akustica AKU230 top port package [82].

the chip-scale packaging and (as shown by Akustica) broaden the packaging possibil-
ities. The problem of top port packages is a small backvolume which is problematic

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

for low frequency response. This issue has been resolved by Akustica with the so-
phisticated package that contains the acoustic system shown on package cross-section
(fig. 5.14:Cross-section Overview). This advanced inlet vent assures that MEMS is
working roughly under the same acoustic conditions as in bottom port package.

5.4.2 Proposed solution

While selecting the most suitable packaging for the considered microphone we must take
into account that it is a device under development and it is currently fabricated on a
small-scale. Under such circumstances the prepared package must fulfill the following
requirements:

• package back volume have to correspond to the one used in commercial MEMS
microphones,
• for the first tests the MEMS and ASIC dies wont be integrated into one package,
• during first tests the readout is realized by use of standard laboratory equipment
or the dedicated discrete electronics,
• package has to provide simple interface for the tests instruments (preferred coaxial
connections),
• to facilitate testing of several chips, MEMS has to be integrated with one of the
standard packages (for example DIP - Dual In-line Package).

Regarding the above requirements we have decided to prepare the interface that consists
of three elements: the PCB card (support) with coaxial connections and socket for the
DIP 16 (fig. 5.15), the DIP 16, and the microphone package that is similar to the one
used by Knowles (fig. 5.16). The Knowles package made of miniature PCB and the
metal lid was chosen as a model because it is a package relatively easy to prepare for
the small-scale fabrication.

Figure 5.15: Support board designed and fabricated by IM2NP.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Foreseen assembly
in DIP 16

Bottom port package

elements

Figure 5.16: Elements of package and interface for M&NEMS microphone tests.

Two types of PCB were prepared: dedicated for top and bottom port configurations,
however we will focus on bottom port type (fig. 5.17). PCB is prepared basing on double
sided, 400 µm thick FR 4 substrate with electric pads surface finished accordingly to
the requirements of wire bonding process.
Acoustic sound port of 1 mm diameter is situated below the MEMS die. The “cross”
and “L” alignment marks are for MEMS and metal lid assembly. This marks along with
adjacent rectangular fields are not covered with solder mask in order to ground MEMS
die and the lid. Finally PCB has two sets of electric pads - one for MEMS-PCB wire
bonding and the second for PCB-DIP 16 wire bonding connections.
Electric pads for
PCB-DIP bonding
Outline of metal lid
Outline of MEMS die
Electric pads for
MEMS-PCB bonding
MEMS

Vias

Figure 5.17: PCB substrate prepared for M&NEMS microphone chip-scale packaging
(bottom port). Left side presents the PCB layout while the right side indicates the
specific elements of the PCB.

Miniature substrates prepared for the MEMS packaging are problematic in terms of
PCB depaneling (fig. 5.18). The overall size of the individual PCB is 5 x 7 mm2 and
the spacing between the copper paths and the cut line is around 200 µm, while the
standard methods of depaneling such as V-Scoring and break-routing [83,84] impose the
minimum spacing between the board and cut path of 200-500 µm and the minimum
PCB size of 5 mm.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 5.18: Panel with 300 miniature substrates for M&NEMS microphone. Panel
overall size is 10.0 x 10.5 cm2 .

Depaneling with popular CO2 laser is not suitable either since it damages the PCB due
to generation of high temperature. Finally the depaneling was realized with use of UV
laser that reduces the cutting path and the generated heat (“cold ablation”) [85].

Summary
In this chapter we have presented the technology prepared for the fabrication of the new
type of MEMS microphone. The fabrication revealed couple difficulties, however the
process has been already optimized.
The specification of the microphone makes it problematic in terms of packaging. There
are no standard packages available for MEMS microphones, therefore the in-house so-
lution has been prepared. Our packaging solution ensures that the designed MEMS
prototype will work in the similar conditions to the ones of the commercial products.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Approach to MEMS microphone

calibration

While the designed microphone is under fabrication, the functional test approach has
been developed and validated on a commercial MEMS microphone. The chapter begins
with discussion on a suitable methods for MEMS microphone sensitivity determination.
Impedance tube and anechoic chamber are appointed to characterize the microphone
under pressure-field and free-field conditions. Moreover it is shown that the anechoic
chamber experimental setup is suitable for microphone noise measurements.

6.1 Selection of suitable calibration methods

By term of calibration we mean the investigations of the microphone frequency response
(typically from single hertz to 100 kHz) and the sensitivity at 1 kHz. Numerous calibra-
tion methods are standardized by the International Electrotechnical Commission (IEC).
The method depends on the microphone which can have the laboratory (LS) or working
(WS) standard and on the working environment (see the work of R. Barham, V. Nedzel-
nitsky or E. Frederiksen for further details [86–88]). The most precise microphones
(laboratory standard), which are then used as a standards for the most important mea-
surements and for the calibration of the less precise microphones (working standard) are
calibrated with primary calibration techniques. However we need to remember that the
goal of this thesis is to work on the microphone that is in early stage of development.
That is why we will not discuss the primary calibration methods and we will go directly
to the secondary calibration methods.
Secondary calibration is usually made by comparison of tested microphone to the ref-
erence microphone that have higher precision. It can be done simultaneously with
reference microphone placed next to the tested microphone or the measurements of two
microphones may be done one after another. At this point we need to consider the work-
ing environment which has an influence on the frequency response of the microphone at
higher frequencies (see fig. 6.1). Working environment (after L.L. Beranek [89]) corre-
sponds to the type of sound field in the vicinity of the diaphragm:
• pressure-field - sound pressure is distributed uniformly over the diaphragm (mi-
crophone is introduced into the coupler),
• free-field - this type of field is obtained inside the anechoic chamber where the
sound waves reflected from the walls are negligible in comparison to the sound
waves coming directly from the source,

105
Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf
© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 6. Approach to MEMS microphone calibration 106

• diﬀuse-ﬁeld - where the sound wave of similar amplitude may arrive from the
arbitrary direction.

G%

)UHHILHOG5HVSRQVH

'LIIXVHILHOG
5HVSRQVH
–4 3UHVVXUHILHOG
5HVSRQVH
'LIIXVHILHOG0LFURSKRQH
–8

–12
+]

Figure 6.1: Example of pressure-field, free-field and diffuse-field frequency responses of

diﬀuse-ﬁeld microphone given by E. Frederiksen [88].

6.2 Specimen description

Approach to functional characterization of MEMS microphone is presented with use of
commercial INMP504ACEZ microphone. It is an analog MEMS microphone in bottom
port conﬁguration. It was designed by Analog Devices as a part of MEMS microphone
line that was lately purchased by InvenSense (Q4 2013).

Microphone speciﬁcation
Selection of the microphone was done based on its features (table 6.1, ﬁg. 6.2) and the
asset of a wide range of associated application notes.

Acquisition chain prerequisites and architecture

The acquisition chain starts with MEMS microphone which is a capacitive sensor with
integrated impedance converter and output ampliﬁer. Eventually the signal is digitized
by OROS analyzer, however the microphone has to be connected to the analyzer through
the additional discrete electronics which provide the signal conditioning.
Requirements for the discrete electronics are as following:

• provide relevant gain of the output signal for eﬃcient use of the ADC,
• cut-oﬀ the redundant signal (signal below and above sensor bandwidth, unwanted
signal coming from impedance tube setup),
• self noise of discrete electronics has to be lower than the microphone noise.

We have chosen to feed the signal coming from the microphone to the amplifier
prepared with use of operational amplifier in non-inverting configuration (see fig. 6.3).
Preliminary measurements in impedance tube done with use of laboratory reference

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Table 6.1: Specifications of INMP504 MEMS microphone [90]

Parameter Value Conditions

Performance
Sensitivity -38 dBV 1 kHz, 94 dB SPL
Signal-to-Noise Ratio (SNR) 65 dBA
Equivalent Input Noise (EIN) 29 dBA SPL
Frequency response 0.1 to 20 kHz limited by -3 dB point
Maximum acoustic input 120 dB SPL

Power supply
Supply voltage 1.6 - 3.3 V

Output characteristics
Maximum output voltage 0.35 V 120 dB SPL input
Noise floor -103 dBV 20 Hz - 20 kHz, A-weighted,
rms

Figure 6.2: Typical frequency response of INMP504 MEMS microphone [90].

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

microphone (B&K) have shown that the sound pressure level inside the tube is in range
of 60 - 120 dB. Therefore the amplification has been set by the ratio of the resistors
R1 and R2 and adjusted for the expected SPL. All of the components are situated on

Figure 6.3: Readout electronics selected for INMP504 microphone [90].

the same side of PCB while the sound port is on the opposite side (see fig. 6.4). This

(a) Top view - this side of PCB is contains (b) Bottom view - this side of PCB is exposed
discrete electronics to sound pressure

Figure 6.4: PCB prepared for INMP504 MEMS microphone.

configuration was chosen to eliminate the perturbation of sound field in the vicinity of
the microphone sound port. SMB connectors have been chosen for the pcb interface.
The bias voltage 5 V which is used for operational amplifier is dropped to 1.8 V by the
voltage regulator to power the MEMS microphone.

6.3 Influence of experimental setup on microphone re-

sponse
Before starting the characterization of the microphone let us examine the measurements
chain of the experimental setup (fig. 6.5). It consists of:

• the generator that introduces the signal driving the loudspeaker uexc ,

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

• the loudspeaker that transduces the voltage signal into the acoustic pressure ac-
cording to its transfer function HL (s),
• the acoustic pressure that propagates inside the acoustic domain and introduces
transfer function HA (s),
• the MEMS microphone (marked dut - device under test) and the reference micro-
phone(s) that capture(s) the pressure fluctuations according to their sensitivities
Sdut and Sref .

uexc udut
Loudspeaker Acoustic domain Microphone
transfer function transfer function sensitivity
uref

Figure 6.5: Block diagram of experimental setup.

Figure 6.6 represent the experimental setup used for MEMS microphone characteriza-
tion. This representation is valid regardless of the sound field type (pressure-field and

Multi-channel analyzer

Output channel Input channels

OUT IN1 IN2 IN3 IN4

uexc uexc uref1 uref2 udut

Acoustic domain
reference
microphone(s)
Sref

HL(s) HA(s)

Sdut
MEMS
microphone

Figure 6.6: Graphic representation of experimental setup.

free-field) and the only distinction is the amount of reference microphones. Pressure-field
investigations described in section 6.4 need two reference microphones - microphone x1
with signal uref1 connected to the second input channel of the analyzer and microphone
x2 with signal uref2 connected to the third input channel of the analyzer. On the con-
trary the free-field investigations (section 6.5) require only one reference microphone
with signal identified as uref and connected to the second input channel of the analyzer.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

If we consider all of the elements of of the experimental setup, the signal coming
from MEMS microphone becomes:
udut = uexc [HL HA Sdut ] . (6.1)
To extract the sensitivity of the microphone we need to suppress the excitation signal
along with the loudspeaker and the acoustic domain transfer functions. This is feasible
if we use one reference microphone placed next to the investigated microphone. In such
case the responses of both microphones are linked to the same transfer functions HL
and HA . The reference microphone signal uref is:
uref = uexc [HL HA Sref ] . (6.2)
Knowing the sensitivity Sref of reference microphone (that is given for the frequency of
1 kHz and is assumed to be equal over the range given by the manufacturer), we may
appoint the sensitivity of characterized microphone:
udut
Sdut = Sref . (6.3)
uref
Measurements described in section 6.4 employ two reference microphones located on
different positions than the investigated microphone. The sound pressure applied to
the MEMS microphone is then calculated analytically. This calculations suffer from the
error introduced by the different acoustic transfer functions (HAdut , HAref 1 and HAref 2 )
measured by all three microphones:
udut = uexc [HL HAdut Sdut ] , (6.4)
uref 1 = uexc [HL HAref 1 Sref ] , (6.5)
uref 2 = uexc [HL HAref 1 Sref ] . (6.6)
Suppression of transfer function mismatch in case of response from three microphones is
not trivial. Therefore in our considerations we assume that the resulting error is negligi-
ble and we proceed with measurements. The method for pressure-field characterization
is validated subsequently in subsection 6.4.1.

6.4 Pressure-field investigations with use of impedance

tube
Presented technique of microphone sensitivity investigation is based on the impedance
tube (fig. 6.7). It is measuring instrument build of a cylindrical waveguide with acoustic
pressure source placed at one of the extremities. For the purpose of a measurements, the
opposite extremity is considered to be rigid with investigated microphone mounted at
the x = 0. Impedance tube considerations are true only in case of planar propagation of
sound waves inside the tube; to ensure planar wave propagation the following condition
must be fulfilled:
c0
fmax < . (6.7)
2DIT
We will now show, that with use of two well known, calibrated microphones (x1 and x2 )
integrated to the border of a waveguide we can express the pressure field p(x) over the
entire length of a impedance tube:
p(x) = pi (x) + pr (x) = Aejkx + Be−jkx , (6.8)

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

microphone x1 microphone x2

loudspeaker
investigated
microphone
DIT

0 x1 x2 LIT

Figure 6.7: Sketch of impedance tube test set.

where pi (x) and pr (x) are the pressure of incident and reflected waves, A and B are the
complex amplitudes, and k is a wave number (k = 2πf c0 ). In order to determine p(x),
unknown variables A and B has to be found; for that purpose we prepare set of two
equations with use of microphones x1 and x2 :
e−jkx2 A
jkx
p(x2 ) e 2
= jkx1 . (6.9)
p(x1 ) e e−jkx1 B
Please note that A and B can be found from equation 6.9 if the square matrix standing
before A and B is invertible:
ejkx2 e−jkx2
6= 0. (6.10)
ejkx1 e−jkx1

To meet the foregoing condition sin k(x2 − x1 ) 6= 0, thus ∆x 6= 0, it means that the
frequency has to in the range:
c0
0<f < . (6.11)
2∆x
Reproduction of pressure field inside the tube with use of two microphones is possible if
∆x is shorter than the shortest wavelength; according to Shannon sampling theorem [91]:
λmin 1 c0
∆x < = . (6.12)
2 2 fmax
With the assumptions from equations 6.11 and 6.12, A and B are described by the
following relation:
−jkx
−e−jkx2 p(x2 )

A 1/2j e 1
= , (6.13)
B sin k(x2 − x1 ) −ejkx1 ejkx2 p(x1 )
For further consideration we will introduce velocity field with use of Euler equation:
∂v(x) ∂p(x)
ρ0 =− , (6.14)
∂t ∂x
thus:
Z0 v(x) = −Aejkx + Be−jkx . (6.15)
We may now describe pressure and velocity fields inside the tube:
e−jkx A
jkx
p(x) e
= , (6.16)
Z0 v(x) −ejkx e−jkx B

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

finally pressure and velocity at x = 0 equals:

p(0) 1 − sin kx1 − sin kx2
= p2 − p1 . (6.17)
Z0 v(0) sin k(x2 − x1 ) j cos x1 j cos x2
x1 +x2 x2 −x1
By introducing xm = 2 and d = 2 into equation 6.16, we obtain:

p(x) 1 sin k(x − xm ) p2 − p1 1 cos k(x − xm ) p2 + p1
= +
Z0 v(x) sin kd j cos k(x − xm ) 2 cos kd −j sin k(x − xm ) 2
(6.18)
p(0) 1 − sin k(xm ) p2 − p1 1 cos k(xm ) p2 + p1
= + (6.19)
Z0 v(0) sin kd j cos k(xm ) 2 cos kd j sin k(xm ) 2

Alternative measurement set

The foregoing measurement set and the presented analytical sound pressure designation
may suffer of some incertitude’s. This incertitude’s, identified in equation 6.19 are:

• value of speed of sound c0 ,

• misalignment of the reference microphones (x1 , x2 ) ,

• frequency response of the x1 and x2 microphones (sensitivity is calibrated only for

f = 1 kHz).

Moreover the reposes of reference microphones placed at different positions inside the
tube manifest different transfer functions of the tube (see section 6.3). Possible solution
of this inaccuracies is to use of single reference microphone placed at the end of a tube
together with investigated microphone (see fig. 6.8).

investigated loudspeaker
microphone

DIT

reference
microphone

0 LIT

Figure 6.8: Alternative configuration of measurement set.

6.4.1 Pressure-field experimental setup and measurements

The experimental setup for pressure-field investigations is presented on figure 6.9. We
may see the impedance tube build of PVC with the loudspeaker attached on the right
side and the tested microphone (MEMS or the reference microphone) at the opposite
extremity of the tube. Two calibrated reference microphones x1 and x2 are positioned
at the side of the tube.
The aim of the first type of measurements was to validate the method presented at
the beginning of the section. For that purpose we have used three well known, calibrated
microphones. Two of them were used to identify the pressure at the end of the tube with

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 6.9: Experimental setup used for pressure-field measurements.

use of equation 6.18 while the third one measured the pressure at the end of the tube
directly. Comparison of experimental data and identified data is shown on fig. 6.10.
Both of this results show a good match and confirm the utility of prepared method.

Figure 6.10: Comparison of experimental and identified data for impedance tube setup.

However the data start do diverge from the 2 kHz. In fact the application of impedance
tube for frequency response measurements is limited by its diameter (DIT ) and the
distance between microphones x1 and x2 :
c0
fmax < , (6.20)
2DIT
1 c0
x2 − x1 < . (6.21)
2 fmax
The first limitation concerns planar wave propagation criteria (on fig 6.10 we observe
the disorders coming from the radial mode starting from 2 kHz) while the second one is
set by the Shannon sampling theorem. Nevertheless we will continue the investigations
of this method keeping in mind that the high frequency performance can be improved
by use of the tube with smaller diameter. The second results show the comparison

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

between the response of the reference and the MEMS microphones (fig. 6.11). Both
microphones were placed at the end of the tube and the measurements were made by
the exchange, with equal sound pressure level. The responses of both microphones shows
a good match.

Figure 6.11: Comparison of normalized responses of the reference microphone and the
MEMS.

Finally we have prepared the normalized pressure-field response of the MEMS mi-
crophone. Figure 6.12 shows the results obtained by the rapport of signal coming from
the MEMS and the laboratory microphone placed at the end of the tube (consecutive
measurements). This figure corresponds to the low frequency response of the specimen
showed on figure 6.2. We may observe the influence of the packaging and approximate
the low cut-off by giving the frequency value for -3 dB. We estimate it to be around
70 Hz while the MEMS data sheet gives 100 Hz. Finally the sensitivity of the MEMS
microphone has been estimated to -37.4 dBV/Pa which agrees with the datasheet value
where the limit values are -41 to -35 dBV/Pa.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 6.12: Normalized pressure-field response of INMP504ACEZ microphone.

6.5 Free-field investigations in anechoic chamber

We use the anechoic chamber to provide the conditions that are close to the free-field.
The experimental set is presented on the figure 6.13; it consists of the loudspeaker, the
reference microphone and the MEMS microphone. All the elements are placed on the
stands to adjust their vertical positions, then the loudspeaker is placed at a distance of
70 cm from the tested microphones.

Figure 6.13: Experimental setup used in free-field investigations.

We have compared the frequency response of the reference and the MEMS micro-
phone on figure 6.14. It shows that the MEMS microphone if fully operational. We have
estimated the MEMS sensitivity at 1 kHz to -34.8 dBV/Pa. The difference between the
sensitivity -37.4 dBV/Pa obtained in impedance tube may come from the experimental
setup used in anechoic chamber. To verify it one should prepare the measurements where
the reference and the MEMS microphones are tested one after another while placed ex-
actly in the same distance from the loudspeaker to assure the equal sound pressure level.

It is problematic to retrieve the sensitivity of the MEMS microphone over the whole
frequency band. Normally it can be done with use of formula 6.3. Nevertheless the

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 6.14: Normalized frequency response measured in anechoic chamber.

sensitivity of the reference microphone is calibrated only at 1 kHz and we see that the
response of reference microphone is not ideally uniform over the bandwidth. Therefore
use of proposed formula can introduce large error in the MEMS microphone sensitivity
response, especially between 10 and 20 kHz where the reference microphone response
decreases.
Calm environment of anechoic chamber enabled us to analyze the noise of the mi-
crophones. In this measurements the loudspeaker was removed from the chamber and
the signals presented on the fig. 6.15 results from the inherit noise sources of the mi-
crophones. The MEMS noise is higher however we need to remember that this noise
is amplified by the discrete electronics. If we suppress the gain of the electronics, the
noise decreases (fig. 6.16(a)). Both of the figures (6.16(a) and 6.16(b)) present the noise
of MEMS microphones with acoustic signal capacitive detection based on perforated
membrane. According to the paper of A. Dehe [26] at low frequencies the noise level is
defined by the Flicker noise coming from the ASIC, then the noise level is set by two
thermal noises: thermomechanical noise of the membrane and thermoacoustic noise of
the membrane perforations.
Moreover the noise measurements unveil the mechanical resonance of the MEMS mi-
crophone that for measured microphone occurs close to 7 kHz. It is slightly lower than
the value that may be read from the typical frequency response given in the data sheet
where the resonance is announced above 10 kHz (fig. 6.2).

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Figure 6.15: Noise spectral density of reference microphone and the MEMS microphone
with discrete readout circuit.

(b) Noise measurements od the Infineon MEMS

(a) Noise of INMP504ACEZ microphone.
microphone [26].

Figure 6.16: Comparison of the identified noise with the state of the art example.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Summary
We have selected and validated two methods to extract the crucial characteristics of the
MEMS microphone. First method use the impedance tube and it allowed us to trace the
low frequency response of the microphone which is shaped by the viscous resistance and
the backvolume. The backvolume of used commercial microphone is unknown, however
for the future tests of the new prototype we will be able to deduct the viscous resistance
of the whole acoustic system of the MEMS and compare it to the one obtained with
the FEM and Lumped Elements models. Moreover both pressure-field and free-field
characterization techniques allowed us to find the sensitivity of characterized MEMS.
Both values: -37.4 and -34.8 dBV/Pa are comparable, however the sensitivity obtained
in pressure-field seems to be more reliable and shows good match with -38 dBV/Pa
given in the datasheet.
We have managed to measure the noise floor of the MEMS in the anechoic chamber.
This measurement gave us additional information on the mechanical resonance of the
membrane which is slightly lower comparing to the frequency response given by the
manufacturer (measured value is around 7 kHz while in the datasheet the resonance is
announced above 10 kHz).

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

This work aimed the fabrication of the MEMS microphone for the hearing aids and
consumer electronics. Designed microphone is supposed to complement the existing
sensor line at the MEMS sensors laboratory at CEA-LETI with particular interest in
future fabrication of multi-sensor chips. Realization of this project called the MAD-
NEMS was possible with support of the French National Research Agency. Since it
was the first microphone realized in the laboratory, this work covered the search for
suitable solutions that may be now applied in future work. The work included: the
concept considerations, the technological platform, the simulations with acoustical, me-
chanical and electrical behavior and finally the approaches to the MEMS microphone
characterization.
As a result of this work it has been demonstrated that FEA and equivalent circuit
representation are the powerful tools for comprehensive design considerations of the
MEMS microphone. Both models are complementary and were used to understand
the acoustic phenomena at the microscale and the evaluation of noise sources in the
sensor respectively. Initial design and the dimensioning of the microphone were prepared
basing on the FLNS model with fluid-structure interaction. Afterward we have used the
Lumped Element model for the analysis of the noises in the microphone. The frequency
response of the microphone obtained with both of the models shows a good match.
Despite the high computational cost we have prepared simplified 3D model to verify if
2D representation of the microphone is correct. The sensitivity value obtained with 3D
model is in slightly lower than the one obtained with approximative models, however it
is in the same scale.
The evaluated performance of the first M&NEMS Microphone prototype can not
be squarely compared to the commercial products because we give the specifications
of the MEMS dice without the ASIC. Normally the ASIC will amplify the output sig-
nal and the sensitivity of designed microphone (-56.4 dBV) will reach the one given by
commercial products (typically -38 dBV). The given SNR is much lower than the best
results of the commercial and state of the art sensors. We can accept this value for the
prototype, however for the future designs this value has to be optimized. M&NEMS
technology offers high sensitivity, however the optimization of SNR is challenging in
terms of electrical noises of the piezoresistive nanogauge (1/f and Johnson noises) and
the thermoacoustic noise generated in small acoustic cavities.
As for the rest of the specifications, it corresponds to the ones of the commercial prod-
ucts. We have been aware that matching all of the specifications to the commercial
products is challenging in terms of the sensor design, however our design corresponds
to the commercial products in such critical aspects as the power consumption and the
overall chip-package size. It is worth mentioning that there are no standard packages
available for MEMS microphones and the in-house solution has been prepared.

119
Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf
© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Chapter 6. Approach to MEMS microphone calibration 120

While the designed microphone is under fabrication, the functional test approach
has been developed and validated on a commercial MEMS microphone. Two methods
to extract the crucial characteristics of the MEMS microphone have been selected and
validated. First method use the impedance tube and it allowed us to trace the low
frequency response of the microphone which is shaped by the viscous resistance and the
backvolume. The backvolume of used commercial microphone is unknown, however for
the future tests of the new prototype we will be able to deduct the viscous resistance
of the whole acoustic system of the MEMS and compare it to the one obtained with
the FEM and Lumped Elements models. Moreover both pressure-field and free-field
characterization techniques allowed us to find the sensitivity of characterized MEMS.
Both values: -37.4 and -34.8 dBV/Pa are comparable, however the sensitivity obtained
in pressure-field seems to be more reliable and shows good match with -38 dBV/Pa
given in the datasheet.

M&NEMS microphone improvement

In order to enable prompt fabrication of the microphone, the prototype which is pre-
sented in this work was prepared during the first six months of this thesis. This early
design which is not free of errors may be now improved basing on the experience that
the author has gathered during a later period. The microfabrication of the designed
microphone did not finish due to technological problems, however the overall concept
of the microphone seems to be promising and the first prototype that will prove the
concept of this new architecture is highly expected.

Design optimization:
• the acoustic system of the MEMS dice should be rearranged in order to be sym-
metrical (equal overall volume of the inlets and outlets).
• The truss-structured beam has to be very rigid in the vicinity of the microhinge
and the nanogauge, however its stiffness might be decreased on the other extremity.
Therefore the mass of truss-structured beam may still be decreased by use of the
truss members with variable dimensions.
• The angle between the microhinge members may be adjusted to obtain lower
angular stiffness of the hinge. In the first design the angle of 30 degrees has been
chosen, while for the crystalline Silicon the minimal Young’s modulus is obtained
for the angle of 45 degrees relative to the <110> crystallographic direction.

Technology optimization:
• the Flicker noise of the nanogauges may be decreased by the optimization of the
fabrication process (improvement of Silicon crystal lattice quality).
• The piezoresistive coefficient πpzr might be increased by reduction of nanogauge
section or reduction of the doping level.
• Improvement of the DRIE ratio from 20 to 40 can lead to the fabrication of the
beam with doubled lateral surface while maintaining low mass.
• Optimization of the epitaxial growth of MEMS layer can lead to lower stress
between the NEMS and MEMS layer. With lower stress the out-of-plane deflection
will be decreased and longer beams with larger lateral surface might be used.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

How COMSOL implements FLNS

Documentation of COMSOL Acoustic module and especially documentation of Ther-

moacoustic branch [92] do not present the actual implementation of the viscothermal
problem. However one can conclude that Comsol use the set of full linearized Navier-
Stokes equations (see chapter 3).
Inquisitive COMSOL user may access the exact weak formulation of the problem through
the equation view. Therefore we find the set of three weak formulations used in Comsol
Thermoacoustic branch (for the sake of simplicity we consider 2D problem):

0 =((−2 ∗ ta.mu ∗ ux + (2 ∗ ta.mu/3 − ta.muB) ∗ (ux + vy) + p) ∗ test(ux)

− ta.mu ∗ (uy + vx) ∗ test(uy) − ta.mu ∗ (uy + vx) ∗ test(vx)
+ (−2 ∗ ta.mu ∗ vy + (2 ∗ ta.mu/3 − ta.muB) ∗ (ux + vy) + p) ∗ test(vy) (A.1)
− ta.iomega ∗ ta.rho0 ∗ u ∗ test(u)
− ta.iomega ∗ ta.rho0 ∗ v ∗ test(v)) ∗ ta.delta ta,

0 = − (ta.iomega ∗ ta.rho + ta.rho0 ∗ (ux + vy)) ∗ test(p) ∗ ta.delta ta, (A.2)

0 =(−ta.kcond ∗ (T x ∗ test(T x) + T y ∗ test(T y))

(A.3)
+ ta.iomega ∗ (ta.alpha0 ∗ ta.T 0 ∗ p − ta.rho0 ∗ ta.Cp ∗ T ) ∗ test(T )) ∗ ta.delta ta.

Lets now change the notation to match the one in section 3.3. We see that the foregoing
equations correspond respectively to the velocity, pressure and temperature fields:
Z
∗ ∂v
0= ∗ ¯
∇v : σ̄v + p(∇ · v ) − ρ0 v∗ · ta.delta ta dV, (A.4)
D ∂t
Z
∂ρ
0= − + ρ0 (∇ · v) p∗ · ta.delta ta dV, (A.5)
D ∂t
Z
∂p ∂τ
0= −κ∇τ + − ρ0 cP τ ∗ · ta.delta ta dV, (A.6)
D ∂t ∂t

the term ta.delta ta has been introduced by COMSOL as a scaling factor and it denotes
1/(jω).

121
Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf
© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Bibliography

[1] G.W. Elko and K.P. Harney. A history of consumer microphones: The electret
condenser microphone meets micro-electro-mechanical-systems. Acoustic Today,
5:4–13, 2009.
[2] E.W. Siemens. Improvement in magneto-electric aparatus, 1874.
[3] T.A. Edison. Speaking telegraph, 1879.
[4] E.C. Wente. A condenser transmitter as a uniformly sensitive instrument for the
absolute measurement of sound intensity. Physical Review, 10(1):39–63, 1917.
[5] G.M. Sessler and J.E. West. Self-biased condenser microphone with high capaci-
tance. The Journal of the Acoustical Society of America, 34:1787–1788, 1962.
[6] D. Hohm and G. Sessler. Kapazitive wandler auf siliziumbasis mit siliziumdioxid-
elektret (silicon-based capacitive transducer with silicon dioxide electret), 1983.
[7] A.G. Bell. Improvement in telegraphy, 1876.
[8] J. Lewis. Understanding microphone sensitivity. Analog Dialogue, 46(2):14–16,
2012.
[9] J. Lewis. Microphone specifications explained. Analog Devices. Application Note
AN-1112.
[10] Microphone handbook vol.1. Technical report, Brüel & Kjær, 1996.
[11] Knowles Electronics. Zero-Height SiSonicTM Microphone With Enhanced RF Pro-
tection.
[12] F.A. Everest and K.C. Pohlmann. Master Handbook of Acoustic. McGraw-Hill,
2009.
[13] S. Temme. Audio Distortion Measurements. Brüel & Kjær. Application Note.
[14] F. Laermer and A. Schilp. Method of anisotropically etching silicon, 03 1996.
[15] F. Laermer and A. Schilp. Method of anisotropically etching of silicon, 03 2003.
[16] R. Fraux. Knowles SPU0410LR5H MEMS microphone - reverse costing analysis.
Technical report, System Plus Consulting, 2011.
[17] Y.B. Ning, A.W. Mitchell, and R.N. Tait. Fabrication of a silicon micromachined
capacitive microphone using a dry-etch process. Sensors and Actuators A: Physical,
53(1–3):237 – 242, 1996. Proceedings of The 8th International Conference on Solid-
State Sensors and Actuators.

122
Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf
© [J. Czarny], [2015], INSA de Lyon, tous droits réservés
Bibliography 123

[18] W. Kronast, B. Muller, W. Siedel, and A. Stoffel. Single-chip condenser microphone

using porous silicon as sacrificial layer for the air gap. Sensors and Actuators A:
Physical, 87(3):188 – 193, 2001.

[19] M.L. Kuntzman, C.T. Garcia, A.G. Onaran, B. Avenson, K.D. Kirk, and N.A. Hall.
Performance and modeling of a fully packaged micromachined optical microphone.
Microelectromechanical Systems, Journal of, 20(4):828–833, 2011.

[20] P. Rombach, M. Mullenborn, U. Klein, and K. Rasmussen. The first low voltage,
low noise differential silicon microphone, technology development and measurement
results. Sensors and Actuators A: Physical, 95(2–3):196 – 201, 2002. Papers from
the Proceedings of the 14th IEEE Internat. Conf. on MicroElectroMechanical Sys-
tems.

[21] A. Torkkeli, O. Rusanen, J. Saarilahti, H. Seppä, H. Sipola, and J. Hietanen. Capac-

itive microphone with low-stress polysilicon membrane and high-stress polysilicon
backplate. Sensors and Actuators A: Physical, 85(1–3):116 – 123, 2000.

[22] J.W. Weigold, T.J. Brosnihan, J. Bergeron, and X. Zhang. A MEMS condenser
microphone for consumer applications. In Micro Electro Mechanical Systems, 2006.
MEMS 2006 Istanbul. 19th IEEE International Conference on, pages 86–89, 2006.

[23] A. Dehe. Silicon microphone development and application. Sensors and Actuators
A: Physical, 133(2):283 – 287, 2007. Selected Papers from the 9th International
Conference on Materials for Advanced Technologies.

[24] T. Kasai, S. Sato, S. Conti, I. Padovani, F. David, Y. Uchida, T. Takahashi, and

H. Nishio. Novel concept for a MEMS microphone with dual channels for an ultra-
wide dynamic range. In Micro Electro Mechanical Systems (MEMS), 2011 IEEE
24th International Conference on, pages 605–608, 2011.

[25] Chun-Kai Chan, Wei-Cheng Lai, Mingching Wu, Ming-Yung Wang, and Weileun
Fang. Design and implementation of a capacitive-type microphone with rigid di-
aphragm and flexible spring using the two poly silicon micromachining processes.
Sensors Journal, IEEE, 11(10):2365–2371, 2011.

[26] A. Dehe, M. Wurzer, M. Fuldner, and U. Krumbein. The Infineon silicon MEMS
microphone. In Proceedings SENSOR 2013, pages 95–99, 2013.

[27] Yoshinori Iguchi, Masahide Goto, Masakazu Iwaki, Akio Ando, Kenkichi Tanioka,
Toshifumi Tajima, Futoshi Takeshi, Susumu Matsunaga, and Yoshinobu Yasuno.
Silicon microphone with wide frequency range and high linearity. Sensors and
Actuators A: Physical, 135(2):420 – 425, 2007.

[28] M. Mullenborn, P. Rombach, U. Klein, K. Rasmussen, J.F. Kuhmann, M. Heschel,

M.Amskov Gravad, J. Janting, J. Branebjerg, A.C. Hoogerwerf, and S. Bouw-
stra. Chip-size-packaged silicon microphones. Sensors and Actuators A: Physical,
92(1–3):23 – 29, 2001. Selected Papers for Eurosensors XIV.

[29] J.J. Neumann Jr. and K.J. Gabriel. A fully-integrated CMOS-MEMS audio mi-
crophone. In TRANSDUCERS, Solid-State Sensors, Actuators and Microsystems,
12th International Conference on, 2003, volume 1, pages 230–233, 2003.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

[30] G.W. Elko, F. Pardo, D. Lopez, D. Bishop, and P. Gammel. Capacitive MEMS
microphones. Bell Labs Technical Journal, 10(3):187–198, 2005.

[31] Cheng-Ta Chiang and Chia-Yu Wu. A CMOS digitized silicon condenser micro-
phone for acoustic applications. Sensors Journal, IEEE, 11(2):296–304, 2011.

[32] M. Sheplak, J.M. Seiner, K.S. Breuer, and M.A. Schmidt. A MEMS microphone for
aeroacoustics measurements. In 37th AIAA Aerospace Sciences Meeting & Exhibit,
AIAA Paper 99-0606, Reno, NV, January 1999.

[33] M.D. Williams, B.A. Griffin, T.N. Reagan, J.R. Underbrink, and M. Sheplak. An
AlN MEMS piezoelectric microphone for aeroacoustic applications. Microelectrome-
chanical Systems, Journal of, 21(2):270–283, 2012.

[34] P. Robert and A. Walther. MEMS dynamic pressure sensor, in particular for ap-
plications to microphone production, 06 2011.

[35] V. Rochus. Finite Element Modelling of Strong Electro-Mechanical Coupling in

MEMS. PhD thesis, The University of Liege, Faculty of Applied Sciences, 2006.

[36] Chien-Hung Wu, C.A. Zorman, and M. Mehregany. Fabrication and testing of bulk
micromachined silicon carbide piezoresistive pressure sensors for high temperature
applications. Sensors Journal, IEEE, 6(2):316–324, April 2006.

[37] C. Liu. Foundations of MEMS. Pearson Prentice Hall, 2006.

[38] A.A. Barlian, Woo-Tae Park, J.R. Mallon, A.J. Rastegar, and B.L. Pruitt. Re-
view: Semiconductor piezoresistance for microsystems. Proceedings of the IEEE,
97(3):513–552, March 2009.

[39] G. Blazquez, P. Pons, and A. Boukabache. Capabilities and limits of silicon pressure
sensors. Sensors and Actuators, 17(3–4):387 – 403, 1989.

[40] R.H. Johnson, S. Karbassi, U. Sridhar, and B. Speldrich. A high-sensitivity ribbed

and bossed pressure transducer. Sensors and Actuators A: Physical, 35(2):93 – 99,
1992.

[41] Ahmed A S Mohammed, Walied A Moussa, and Edmond Lou. Optimization of

geometric characteristics to improve sensing performance of mems piezoresistive
strain sensors. Journal of Micromechanics and Microengineering, 20(1):015015,
2010.

[42] Sea-Chung Kim and K.D. Wise. Temperature sensitivity in silicon piezoresistive
pressure transducers. Electron Devices, IEEE Transactions on, 30(7):802–810, Jul
1983.

[43] Yi Cai Sun, Zhenbin Gao, Li Qiang Tian, and Yi Zhang. Modelling of the reverse
current and its effects on the thermal drift of the offset voltage for piezoresistive
pressure sensors. Sensors and Actuators A: Physical, 116(1):125 – 132, 2004.

[44] Rongrui He and Peidong Yang. Giant piezoresistance effect in silicon nanowires.
Nature Nanotechnology, 1:42–46, Oct 2006.

[45] W. Kester. Practical design techniques for sensor signal conditioning. Analog
Devices.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

[46] D. Ettelt. Conception et fabrication d’un magnétomètre à jauge de contrainte. PhD

thesis, Université de Grenoble, 2012.

[47] Y. Deimerly. Vers des centrales inertielles compactes basées sur des nanojauges
piezorésistives : problématique de co-intégration. PhD thesis, Université Paris-Est,
2013.

[48] P. Robert, V. Nguyen, S. Hentz, L. Duraffourg, G. Jourdan, J. Arcamone, and

S. Harrisson. M&NEMS: A new approach for ultra-low cost 3d inertial sensor. In
Sensors, 2009 IEEE, pages 963–966, Oct 2009.

[49] D. Ettelt, P. Rey, M. Savoye, C. Coutier, M. Cartier, O. Redon, M. Audoin,

A. Walther, P. Robert, Y. Zhang, F. Dumas-Bouchiat, N. M. Dempsey, and J. Dela-
mare. A new low consumption 3d compass using integrated magnets and piezoresis-
tive nano-gauges. In Solid-State Sensors, Actuators and Microsystems Conference
(TRANSDUCERS), 2011 16th International, pages 40–43, June 2011.

[50] D. Ettelt, P. Rey, G. Jourdan, A. Walther, P. Robert, and J. Delamare. 3d magnetic

field sensor concept for use in inertial measurement units (imus). Microelectrome-
chanical Systems, Journal of, PP(99):1–1, 2013.

[51] A. Walther, M. Savoye, G. Jourdan, P. Renaux, F. Souchon, P. Robert, C. Le Blanc,

N. Delorme, O. Gigan, and C. Lejuste. 3-axis gyroscope with si nanogage piezo-
resistive detection. In Micro Electro Mechanical Systems (MEMS), 2012 IEEE 25th
International Conference on, pages 480–483, Jan 2012.

[52] A. Walther, C. Le Blanc, N. Delorme, Y. Deimerly, R. Anciant, and J. Willemin.

Bias contributions in a mems tuning fork gyroscope. Microelectromechanical Sys-
tems, Journal of, 22(2):303–308, April 2013.

[53] Y. Deimerly, P. Robert, G. Jourdan, P. Rey, and T. Bourouina. Ultra-compact

absolute pressure sensor based on mechanical amplification coupled to a sus-
pended piezoresistive nanogauge. In Solid-State Sensors, Actuators and Microsys-
tems (TRANSDUCERS EUROSENSORS XXVII), 2013 Transducers Eurosensors
XXVII: The 17th International Conference on, pages 1767–1770, June 2013.

[54] M. Bruneau. Manuel d’acoustique fondamentale. Hermes, 1998.

[55] A.D. Pierce. Acoustics: An Introduction to Its Physical Principles and Applications.
Acoustical Society of America, 1989.

[56] W.R. Kampinga. Viscothermal acoustics using finite elements : analysis tools for
engineers. PhD thesis, Enschede, June 2010.

[57] A. Lallemand. Écoulement des fluides dynamique des fluides réels. Techniques de
l’ingénieur Fluides et combustion, base documentaire : TIB213DUO.(ref. article :
be8157), 2014. fre.

[58] N. Joly. Finite element modeling of thermoviscous acoustics on adapted anisotropic

meshes: Implementation of the particle velocity and temperature variation formu-
lation. Acta Acustica united with Acustica, 96(1):102–114, January 2010.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

[59] W.R. Kampinga, Y.H. Wijnant, and A. de Boer. Performance of several viscother-
mal acoustic finite elements. Acta Acustica united with Acustica, 96:115 – 124,
2010.

[60] O.C. Zienkiewicz, R.L. Taylor, and P. Nithiarasu. The Finite Element Method for
Fluid Dynamics (Sixth Edition). Butterworth-Heinemann, 2005.

[61] C. Guianvarc’h, T. Verdot, J. Czarny, E. Redon, K. Ege, J.-L. Guyader, A. Walther,

and P. Robert. New planar nano-gauge detection microphone: analytical and nu-
merical acoustic modeling. Proceedings of International Congress on Acoustics,
19(1):–, 2013.

[62] T. Verdot, C. Guianvarc’h, J. Czarny, E. Redon, K. Ege, and J.L. Guyader.

Modélisation d’une architecture innovante de microphone MEMS à détection par
nano jauges. In Actes du XXIe Congrès Français de Mécanique, CFM2013, Bor-
deaux, France, 2013.

[63] Comsol. Comsol Multiphysics, Reference Manual: Equation-Based Modeling.

[64] N. Joly, M. Bruneau, and R. Bossart. Coupled equations for particle velocity and
temperature variation as the fundamental formulation of linear acoustics in thermo-
viscous fluids at rest. Acta Acustica united with Acustica, 92:202–209, 2006.

[65] N. Joly and Durand S. Modélisation numérique de l’acoustique en fluides ther-

movisqueux: application à l’étude de micro-systèmes. 18ème Congrès Français de
Mécanique, Grenoble, 27-31 août 2007, 2007.

[66] H. Tijdeman. On the propagation of sound waves in cylindrical tubes. Journal of

Sound and Vibration, 39(1):1 – 33, 1975.

[67] W.M. Beltman. Viscothermal wave propagation including acousto-elastic interac-

tion, part I: Theory. Journal of Sound and Vibration, 227(3):555 – 586, 1999.

[68] Joel Voldman. Course materials for 6.777J / 2.372J Design and Fabrica-
tion of Microelectromechanical Devices, Spring 2007. MIT OpenCourseWare
(http://ocw.mit.edu/).

[69] S. Adhikari. Damping Models for Structural Vibration. PhD thesis, Cambridge
University, Engineering Department, 2000.

[70] A. Preumont. Vibration Control of Active Structures. Springer, 2011.

[71] T.B. Gabrielson. Fundamental noise limits in miniature acoustic and vibration
sensors. Report, Naval Air Development Center, Dec 1991.

[72] T.B. Gabrielson. Mechanical-thermal noise in micromachined acoustic and vibra-

tion sensors. Electron Devices, IEEE Transactions on, 40(5):903–909, May 1993.

[73] F.N. Hooge. 1/f noise is no surface effect. Physics Letters A, 29(3):139 – 140, 1969.

[74] F. N. Hooge. 1/f noise sources. Electron Devices, IEEE Transactions on,
41(11):1926–1935, Nov 1994.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

[75] P.Y. Kwok, M.S. Weinberg, and K.S. Breuer. Fluid effects in vibrating microma-
chined structures. Microelectromechanical Systems, Journal of, 14(4):770–781, Aug
2005.

[76] Y.B. Gianchandani, O. Tabata, and H.P. Zappe. Comprehensive Microsystems,

volume 1 of Comprehensive Microsystems. Elsevier Science & Technology Books,
2007.

[77] J. Meng, T. Mattila, A Dasgupta, M. Sillanpaa, R. Jaakkola, K. Andersson, and

E. Hussa. Testing and multi-scale modeling of drop and impact loading of complex
mems microphone assemblies. In Thermal, Mechanical and Multi-Physics Simu-
lation and Experiments in Microelectronics and Microsystems (EuroSimE), 2012
13th International Conference on, pages 1/8–8/8, April 2012.

[78] J. Li, J. Makkonen, M. Broas, J. Hokka, T.T. Mattila, M. Paulasto-Krockel,

J. Meng, and A DasGupta. Reliability assessment of a mems microphone under
shock impact loading. In Thermal, Mechanical and Multi-Physics Simulation and
Experiments in Microelectronics and Microsystems (EuroSimE), 2013 14th Inter-
national Conference on, pages 1–6, April 2013.

[79] Jing en Luan and Tong Yan Tee. Effect of impact pulse parameters on consistency
of board level drop test and dynamic responses. In Electronic Components and
Technology Conference, 2005. Proceedings. 55th, pages 665–673 Vol. 1, May 2005.

[80] M. Sahmimi. Epcos T4060 MEMS microphone - reverse costing analysis. Technical
report, System Plus Consulting, 2012.

[81] S. Hallereau. Epcos T4000 MEMS microphone - reverse costing analysis. Technical
report, System Plus Consulting, 2011.

[82] R. Fraux. Akustica AKU230 MEMS microphone - reverse costing analysis. Tech-
nical report, System Plus Consulting, 2011.

[83] http://www.eurocircuits.com/index.php/technology-guidelines/pcb-panel-
guidelines/94-how-to-panelise-small-pcbs.

[84] http://www.multi-circuit-boards.eu/en/pcb-design-aid/design-parameters.html.

[85] S. Stafford. Avoid overlooking depaneling step in the PCB manufacturing process.
Electronic Products and Technology, Dec 2013.

[86] R. Barham, S. Barrera-Figueroa, and J.E.M. Avison. Secondary pressure calibration

of measurement microphones. Metrologia, 51(3):129, 2014.

[87] V. Nedzelnitsky. Calibration of pressure and gradient microphones. In Encyclopedia

of Acoustics, pages 1869–1879. John Wiley & Sons, Inc., 2007.

[88] E. Frederiksen. Microphone calibration. In Handbook of Signal Processing in Acous-

tics, pages 1293–1312. Springer New York, 2008.

[89] L.L. Beranek. Acoustic measurements. John Wiley & Sons, Inc., 1949.

[90] Ultra-Low Noise Microphone with Bottom Port and Analog Output. INMP504
MEMS microphone datasheet.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

[91] C.E. Shannon. Communication in the presence of noise. Proceedings of the IEEE,
86(2):447–457, Feb 1998.

[92] Comsol. Acoustics Module, User’s Guide: Theory Background for the Thermoa-
coustic Branch.

Cette thèse est accessible à l'adresse : http://theses.insa-lyon.fr/publication/2015ISAL0003/these.pdf

Vous aimerez peut-être aussi

These A ABBAD Ahmed 2018
Pas encore d'évaluation
These A ABBAD Ahmed 2018
207 pages
Microphones
Pas encore d'évaluation
Microphones
53 pages
Microphone
Pas encore d'évaluation
Microphone
160 pages
Les Microphones Complet
100% (1)
Les Microphones Complet
17 pages
CMT Chapitre3 Exemple de Capteurs Univ Medea
Pas encore d'évaluation
CMT Chapitre3 Exemple de Capteurs Univ Medea
14 pages
Chap 3
Pas encore d'évaluation
Chap 3
5 pages
Tech Audio 1 - Les Microphones
Pas encore d'évaluation
Tech Audio 1 - Les Microphones
17 pages
CR Capteur Anglade Lpgpi
Pas encore d'évaluation
CR Capteur Anglade Lpgpi
12 pages
Microphone Elctrostatique
Pas encore d'évaluation
Microphone Elctrostatique
41 pages
BENOIT Daniel Les Microphones 1997
Pas encore d'évaluation
BENOIT Daniel Les Microphones 1997
25 pages
Dori
Pas encore d'évaluation
Dori
36 pages
Nano Sonde Active Intelligente Pour Mesures de Bruit Et de Puissance Dans La Bande de Frã©quence 130-260 GHZ
Pas encore d'évaluation
Nano Sonde Active Intelligente Pour Mesures de Bruit Et de Puissance Dans La Bande de Frã©quence 130-260 GHZ
203 pages
Principe du Microphone Piézoélectrique
Pas encore d'évaluation
Principe du Microphone Piézoélectrique
4 pages
Les Microphone
Pas encore d'évaluation
Les Microphone
26 pages
17 Microphony
Pas encore d'évaluation
17 Microphony
37 pages
LE Microphone
Pas encore d'évaluation
LE Microphone
7 pages
Chap 1 Microtechnologie
Pas encore d'évaluation
Chap 1 Microtechnologie
26 pages
Introduction aux MEMS : Concepts et Applications
Pas encore d'évaluation
Introduction aux MEMS : Concepts et Applications
9 pages
Un Capteur de Nouvelle Génération Pour La Mesure D'impédance Acoustique en Contexte Industriel
Pas encore d'évaluation
Un Capteur de Nouvelle Génération Pour La Mesure D'impédance Acoustique en Contexte Industriel
7 pages
Microphones 1 Et 2
100% (1)
Microphones 1 Et 2
38 pages
Introduction aux Technologies MEMS
Pas encore d'évaluation
Introduction aux Technologies MEMS
8 pages
Thermoacoustique Principe Et Perspectives en Microfluidique
Pas encore d'évaluation
Thermoacoustique Principe Et Perspectives en Microfluidique
7 pages
Precis Delectro Acoustique
Pas encore d'évaluation
Precis Delectro Acoustique
155 pages
Caractéristiques des cellules microphone ELNO
Pas encore d'évaluation
Caractéristiques des cellules microphone ELNO
92 pages
Comment Utiliser Les Microphones À Condensateur Électret
Pas encore d'évaluation
Comment Utiliser Les Microphones À Condensateur Électret
88 pages
2017-École Polytechnique-Physique Et SI-MP-enonce
Pas encore d'évaluation
2017-École Polytechnique-Physique Et SI-MP-enonce
14 pages
Chapitre3 MEMS 2
100% (1)
Chapitre3 MEMS 2
24 pages
Types de microphones et leur fonctionnement
Pas encore d'évaluation
Types de microphones et leur fonctionnement
29 pages
Réalisation d'un Sonomètre Arduino
Pas encore d'évaluation
Réalisation d'un Sonomètre Arduino
20 pages
electroniqueMicrocapteurPart2 PDF
Pas encore d'évaluation
electroniqueMicrocapteurPart2 PDF
66 pages
Les Micros
Pas encore d'évaluation
Les Micros
17 pages
Capture D'écran . 2024-05-18 À 07.41.33
Pas encore d'évaluation
Capture D'écran . 2024-05-18 À 07.41.33
17 pages
Évolution et Applications du Microphone
Pas encore d'évaluation
Évolution et Applications du Microphone
17 pages
ThA Se Sylvain Godet
Pas encore d'évaluation
ThA Se Sylvain Godet
171 pages
Étude Par Acoustique Picoseconde Des Vibrations
Pas encore d'évaluation
Étude Par Acoustique Picoseconde Des Vibrations
155 pages
Caractérisation Acoustique Avancée
Pas encore d'évaluation
Caractérisation Acoustique Avancée
198 pages
Circuits WLAN 5 GHz à Résonateurs
Pas encore d'évaluation
Circuits WLAN 5 GHz à Résonateurs
159 pages
CFM2007 0502
Pas encore d'évaluation
CFM2007 0502
7 pages
2001LEMA1010
Pas encore d'évaluation
2001LEMA1010
152 pages
Bulk Acoustic Wave Resonators and Oscillators at L PDF
Pas encore d'évaluation
Bulk Acoustic Wave Resonators and Oscillators at L PDF
220 pages
Résonateur Haute Fréquence en MEMS - Thèse
Pas encore d'évaluation
Résonateur Haute Fréquence en MEMS - Thèse
221 pages
Fonctionnement et Construction des Microphones
Pas encore d'évaluation
Fonctionnement et Construction des Microphones
18 pages
MEMS Piézoélectriques en Temps-Fréquence
Pas encore d'évaluation
MEMS Piézoélectriques en Temps-Fréquence
223 pages
Thèse Transduction
Pas encore d'évaluation
Thèse Transduction
270 pages
Introduction aux systèmes MEMS
Pas encore d'évaluation
Introduction aux systèmes MEMS
56 pages
Mémoire d'ingénieur : Amplificateur SiGe
Pas encore d'évaluation
Mémoire d'ingénieur : Amplificateur SiGe
118 pages
Electro A Cous Tique
Pas encore d'évaluation
Electro A Cous Tique
13 pages
05 Ultrasons
Pas encore d'évaluation
05 Ultrasons
48 pages
Introduction aux MEMS et Capteurs
Pas encore d'évaluation
Introduction aux MEMS et Capteurs
43 pages
Capteurs et Technologie MEMS
Pas encore d'évaluation
Capteurs et Technologie MEMS
13 pages
M09 Technologie de Captation de Son-AV-TSAS
Pas encore d'évaluation
M09 Technologie de Captation de Son-AV-TSAS
72 pages
La Technologie de Captation de Son
Pas encore d'évaluation
La Technologie de Captation de Son
34 pages
Types de microphones et leur fonctionnement
Pas encore d'évaluation
Types de microphones et leur fonctionnement
101 pages
TPcapteurultrasons
Pas encore d'évaluation
TPcapteurultrasons
5 pages
Benahmed
Pas encore d'évaluation
Benahmed
14 pages
Canon sonore pour désescalade des violences
Pas encore d'évaluation
Canon sonore pour désescalade des violences
29 pages
L'audit Interne
Pas encore d'évaluation
L'audit Interne
376 pages
Cours TOPO 102 (LTN Morand)
Pas encore d'évaluation
Cours TOPO 102 (LTN Morand)
50 pages
Découverte des Éléments Chimiques
100% (1)
Découverte des Éléments Chimiques
2 pages
Concours Physique II : Loi de Wiedemann-Franz
Pas encore d'évaluation
Concours Physique II : Loi de Wiedemann-Franz
8 pages
Randonnée Belvédère de las Buitreras
Pas encore d'évaluation
Randonnée Belvédère de las Buitreras
4 pages
Chapitre 2 Analyse Fonctionnelle
Pas encore d'évaluation
Chapitre 2 Analyse Fonctionnelle
6 pages
L'activité Ludique Comme Support Pédagogique Motivant Dans Le Développement de La Compétence Orale
Pas encore d'évaluation
L'activité Ludique Comme Support Pédagogique Motivant Dans Le Développement de La Compétence Orale
61 pages
Guide Pedagogique de La Student Academy
Pas encore d'évaluation
Guide Pedagogique de La Student Academy
54 pages
Dimensionnement d'une dalle en béton armé
Pas encore d'évaluation
Dimensionnement d'une dalle en béton armé
21 pages
Test de 3ème Année de Base - Caractéristiques de La Lumière
Pas encore d'évaluation
Test de 3ème Année de Base - Caractéristiques de La Lumière
5 pages
2022 NelleCaledo J1 Exo1 Correction Ingenuity 10pts
Pas encore d'évaluation
2022 NelleCaledo J1 Exo1 Correction Ingenuity 10pts
4 pages
ACV Fairphone 3 : Étude avec SimaPro
Pas encore d'évaluation
ACV Fairphone 3 : Étude avec SimaPro
7 pages
Les Ondes Avec Correction
Pas encore d'évaluation
Les Ondes Avec Correction
11 pages
Logement Moderne et Famille Urbaine
Pas encore d'évaluation
Logement Moderne et Famille Urbaine
25 pages
Echelle Algoplus
Pas encore d'évaluation
Echelle Algoplus
2 pages
La Fraude Fiscale Au Maroc
Pas encore d'évaluation
La Fraude Fiscale Au Maroc
32 pages
Linguistique Descriptive
Pas encore d'évaluation
Linguistique Descriptive
19 pages
Lot 01 Plan Bibliotheque
Pas encore d'évaluation
Lot 01 Plan Bibliotheque
52 pages
Créez un Cahier de Charges Efficace
100% (1)
Créez un Cahier de Charges Efficace
2 pages
Methodes Et Astuces Et Remarques Et Conseils Equation Du Second Degre
Pas encore d'évaluation
Methodes Et Astuces Et Remarques Et Conseils Equation Du Second Degre
1 page
Examen Final - Semaine 8 Mio
Pas encore d'évaluation
Examen Final - Semaine 8 Mio
12 pages
Cons Meca s3 1er GR 2017 PDF
100% (1)
Cons Meca s3 1er GR 2017 PDF
7 pages
Champ de pesanteur terrestre et gravitation
Pas encore d'évaluation
Champ de pesanteur terrestre et gravitation
12 pages
Horaire Cours Fage Du 05 Au 10 Aout 2024
Pas encore d'évaluation
Horaire Cours Fage Du 05 Au 10 Aout 2024
3 pages
Maîtriser la Dynamique Visuelle
Pas encore d'évaluation
Maîtriser la Dynamique Visuelle
20 pages
Estimation des paramètres avec l'algorithme EM
Pas encore d'évaluation
Estimation des paramètres avec l'algorithme EM
16 pages
Avant Propos Manuel PC - 1
Pas encore d'évaluation
Avant Propos Manuel PC - 1
7 pages
Edt SJP 19 Au 24 Mai 2025
Pas encore d'évaluation
Edt SJP 19 Au 24 Mai 2025
11 pages
Bupsy 0007-4403 2000 Num 53 448 14939 t1 0513 0000 3
Pas encore d'évaluation
Bupsy 0007-4403 2000 Num 53 448 14939 t1 0513 0000 3
3 pages