walser 2016

8/17/2019 Walser 2016

1/19

Accepted Manuscript

Title: MEMS Microphones with Narrow SensitivityDistribution

Author: S. Walser C. Siegel M. Winter G. Feiertag M. Loibl

A. Leidl

PII: S0924-4247(16)30196-0

DOI: http://dx.doi.org/doi:10.1016/j.sna.2016.04.051

Reference: SNA 9635

To appear in: Sensors and Actuators A

Received date: 30-11-2015

Revised date: 8-4-2016

Accepted date: 22-4-2016

Please cite this article as: S.Walser, C.Siegel, M.Winter, G.Feiertag, M.Loibl, A.Leidl,

MEMS Microphones with Narrow Sensitivity Distribution, Sensors and Actuators: A

Physical http://dx.doi.org/10.1016/j.sna.2016.04.051

This is a PDF file of an unedited manuscript that has been accepted for publication.

As a service to our customers we are providing this early version of the manuscript.

The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process

errors may be discovered which could affect the content, and all legal disclaimers that

apply to the journal pertain.

http://dx.doi.org/doi:10.1016/j.sna.2016.04.051http://dx.doi.org/10.1016/j.sna.2016.04.051http://dx.doi.org/10.1016/j.sna.2016.04.051http://dx.doi.org/doi:10.1016/j.sna.2016.04.051

8/17/2019 Walser 2016

2/19

MEMS Microphones with Narrow Sensitivity Distribution

S. Walser a* , C. Siegel b, M. Winter b, G. Feiertaga, M. Loibla, A. Leidl b

aMunich University of Applied Sciences, Munich, Germany bEPCOS AG a TDK group company, Munich, Germany

* Corresponding author: Sebastian Walser, Email: [email protected], Tel: +49 089 54020 3160

mailto:[email protected]:[email protected]

8/17/2019 Walser 2016

3/19

Abstract

This paper presents a method for calibrating MEMS microphones after the fabrication process. This allows compensating process

tolerances and packaging stress of the capacitive membrane-backplate system. A programmable MEMS microphone with

variable bias voltage and variable gain was developed and fabricated. The sensitivity of this microphone can be adjusted within a

range of 11 dB. The adjustability was used to trim the sensitivity towards a specified value and to reduce the variation of the

sensitivity. In mass production a tight sensitivity of ±1 dB could be reached, for a specified sensitivity of -38 dBV/Pa @ 1 kHz.

In addition the microphone signal to noise ratio was increased to values above 66 dB(A) by choosing high bias voltages and

matching gain calibration. The relationship between bias supply voltage and the non-linear dynamics of a double backplate

sensor is explained. The influence of programming on the sensitivity, noise, SNR, THD and frequency response was investigated.

Keywords: MEMS, microphone, OTP, programming, high SNR, sensitivity, distribution

1.

Introduction

Today microelectromechanical systems (MEMS) are commonly used as sensors in electronic consumer products,

because of the accurately controlled silicon micromachining technology. MEMS microphones, which convert an

acoustic signal into an electrical signal, reach a high degree of miniaturization [1]. In the last decades electret

condenser microphones (ECM) have been replaced more and more by silicon MEMS microphones, because of theirdurable and reliable performance characteristics at high temperatures or high humidity [2]. For consumer audio

applications, e.g. mobile phone communication, silicon MEMS microphones have become state of the art [3]. High

signal to noise ratios (SNR) of approximately 66 dB(A), sensitivities of -38 dBV/Pa and component sizes of

approximately 3.5 x 2.5 x 1.0 mm3 are state of the art [4, 5].

In addition to the continuous reduction in size, a new demand of the market is a narrow distribution of the

sensitivities down to ±1 dB or beyond for analogue MEMS microphones [6]. The reasons for this demand are new

audio signal processing applications like noise canceling. For noise canceling it is necessary to have microphones

with similar electroacoustic characteristics assembled within one mobile phone [7].

Sensitivity variations are caused by tolerances in the front-end processing of the sensor chip [8] and the

application-specific integrated circuit (ASIC). Another source of variations is packaging stress on the sensor chip [9,

10]. All these influences lead to wide sensitivity variation of MEMS microphones. An example of a measuredsensitivity distribution of MEMS microphones is given in [6]. A sensitivity span of 4 dB was measured. In case of a

specified sensitivity tolerance of ±1 dB, this would result in low production yield.

The simplest way to reach the specified sensitivity tolerances is to select the microphone modules after their

fabrication. This can be done by production test results. A significant drawback of this method would be a low yield.

Another possibility to reduce the sensitivity variations is to pair suitable ASIC and sensor chips. This pairing could

reduce the sensitivity variations by compensating both chip tolerances. A drawback of pairing is that variations

resulting from assembly cannot be compensated. An in situ way to reduce the variations by tuning is described in

[11]. There the acoustic sensitivity tuning was done by using silver metallic electrodeposited nanostructures on the

membrane. This technique offers a step toward reducing sensitivity tolerances after fabrication with the drawback of

a low tuning range of 0.6 dB and the need of an additional external power source for growing mechanism.

8/17/2019 Walser 2016

4/19

Sensors and Actuators A 3

Another way to compensate the tolerances by trimming after completing the fabrication of the microphone

modules is described in [6] and discussed more detailed in this paper. This method also allows compensating the

variations caused by packaging stress. The principle and the benefits of programming are shown. For experimental

characterization a new programmable, capacitive MEMS microphone has been developed. In addition to tight

sensitivity specifications, high signal to noise ratios are also required. By using an appropriate programming

algorithm, it is possible to produce microphones with tight sensitivity deviations and high SNR.

2. Programmable MEMS microphones

Condenser silicon MEMS microphones consist in most cases of two chips, a sensor and an ASIC chip, which are

both integrated in a surface-mount device (SMD) package. During the last few years, two main packaging variants

have been applied for bottom port MEMS microphones [12]. An example for a wire-bonding-type package is given

in [13]. An example for a flip-chip package is given in [14].

The sensor chip converts the incoming acoustic wave into a capacitive change. The ASIC chip transforms the

capacitive into an electrical voltage. During the last few years a trend towards sensor chips with two backplates and

a membrane in between, has become apparent [14, 15, and 16]. The main benefit in comparison to a single-

backplate transducer is the differential output signal. This leads to a 6 dB higher sensitivity. The noise of two

identical transducers is normally distributed and uncorrelated [2]. This results in a 3 dB(A) higher noise level of the

double-backplate system in comparison to a single-backplate system [2]. As a result, this leads to a 3 dB(A) higherSNR in comparison to a single-backplate transducer [2]. For high performance recording the SNR is the most

relevant parameter of a microphone. A detailed theoretical description of differential microphones with two

perforated backplates and a membrane in between was published by Martin et al. [17 and 18].

2.1. Programmable interface

The development of programmable microphones started around 2007 [19]. Figure 1 shows the schematic of the

novel condenser MEMS microphone with a programmable interface for adjusting the acoustic microphone

characteristics.

For calibrating, a new non-volatile one-time-programmable (OTP) memory is integrated in the ASIC. The

principle of the OTP memory is based on a very high cell resistance, which decreases significantly after programming and allows a single write [20]. The advantage of this technology is that an OTP logic can be integrated

in a CMOS process technology without additional process steps or masking layers [20].

A constant preload of the capacitive membrane-backplate system is provided by a programmable bias voltage

generator module, for example by a charge pump. In case of a sound pressure, the oscillating membrane causes a

differential capacitive change between the membrane and the rigid perforated backplates. Transformed into an

electrical signal it is transmitted via a programmable amplifier stage to the differential microphone outputs. The

output stage adapts the impedance and gain of the microphone output signal.

For the programming sequence an external clock and a communication path are necessary. Disadvantages of

programmable microphones are larger ASIC chip sizes and longer test times. The size of a programmable ASIC is

larger, because of the additional area of the OTP module and the digital interface.

With the OTP memory, it is possible to program the bias supply of the membrane-backplate system and the gainof the amplifier. Four bits of the memory cell are reserved for adjusting the amplifiers. The amplifier is designed

with a programmable range of 7.5 dB. The amplifier is linear, so the sensitivity can be adjusted without changing the

other acoustical microphone parameters. Five bits of the memory cell are reserved for adjusting the bias supply

voltage. The choice of the bias working area depends on the sensor design. For a better understanding figure 2

shows the membrane displacement over the bias voltage supply and the forces acting on the membrane for an

incoming sound pressure.

The membrane deflection x and the resulting forces are caused by incoming sound pressure. The mechanical

force Fmech is described by a simplified linear spring model. In case of a double backplate system a capacitive

preload voltage V bias causes two electrostatic forces Fel1 and Fel2.

At the pull-in point, the deflection of the membrane has reached a value where the electrostatic attraction of one

backplate is larger than the forces, which act in the other direction. For a system with only one backplate this point

8/17/2019 Walser 2016

5/19

is roughly reached at x = 2/3 · x0. [21, 22] For a double backplate system, this point is somewhat higher due to the

second electrostatic force (Fel2). In literature [16] this benefit is given with a 30% higher bias field than for a single-

ended sensor. The non-linearity of the system is small as long as the membrane deflection is far from the pull-in

point. By increasing the bias voltage the pull-in point moves to lower sound pressure levels. In fact a pull-in can be

reached even with a double backplate system by increasing bias voltage alone in case of an unstable balance point.

Increasing bias supply voltage results in an increase in sensitivity. With bias voltages close to the pull-in point the

system becomes non-linear. The drawback of this is an increase in the total harmonic distortions (THD). Some

detailed investigations about the non-linear dynamics of double backplate capacitive MEMS microphones can also be found in [23]. So the adjustable bias supply voltage should be chosen within the easier to control linear area.

Therefore, the bias supply was designed with a programmable range of 2.4 V and an upper corner of 12.3 V.

2.2. Fabrication of the programmable MEMS microphone

For this research, a differential sensor chip is used. The sensor chip has a size of 1.45 x 1.45 x 0.45 mm3. A

picture of this sensor chip is shown in figure 3.

The differential sensor chip consists of two perforated backplates with a membrane in between. The diameter of

the membrane and backplates is 1.2 mm. The membrane is coated with an anti-stiction layer and has a thickness ofaround 0.4 μm. The backplates have a thickness of 3 μm. The gap distance between backplate and membrane is

around 2 μm. The venthole is necessary for the pressure exchange between frontvolume and backvolume.

The package of the new programmable MEMS microphone is based on a flip-chip bottom port MEMS

microphone package technology, which was presented in [5]. The microphone with a package size of

3.35 x 2.5 x 1.0 mm3 consists of two chips, the sensor and the ASIC chip. The programmable ASIC chip has a size

of 1.45 x 1.00 x 0.20 mm3. Figure 3a shows a sectional view of the fabricated programmable microphone. The

sensor chip is flip-chip bonded on a ceramic substrate above its sound hole as shown in figure 4b. The separation of

frontvolume and backvolume is done by a polymer foil as shown in figure 4c. The package is closed by a metal lid

as shown in figure 4d. The package combines advantages of a small front-volume to avoid resonances in the

acoustic frequency range and a large back-volume to increase the SNR. The detailed MEMS microphone fabrication

process and the advantages of the package are explained in more detail in [5] and [24].

3. Measurement results

All measurements were carried out in an acoustic pressure chamber, calibrated to 1 Pa by a reference

microphone. For some measurements, a trial mode in the ASIC design was used. This mode allows multiple

programming, by using a volatile memory for debugging applications.

3.1. Characterization of the programmable microphone

To achieve the goal of a small sensitivity variation after MEMS microphone fabrication, the programmable

interface is used. As explained in section 2.1, the microphone sensitivity can be adjusted by setting the gain factor of

the amplifier and the bias supply voltage of the capacitive sensor.

The amplifier with variable gain is part of the output stage of the ASIC. As long as the output stage does not

contribute to non-linearity the signal and its harmonics are amplified linearly. This results in a simple linear shift of

the sensitivity with a slope factor of 1.0 dBV/dB as shown in figure 5a. The noise floor has a slope factor of

0.9 dBV/dB as shown in figure 5b. As a result, an increase in the SNR with increasing gain was measured as shown

in figure 5c. The measured amplification of the noise is smaller than the amplification of the signal because of the

self-noise of the amplifier output stage and the self-noise of the analog measurement input stage. As a consequence,

the gain has no relevant influence the SNR.

The THD influence is shown in figure 5d for a measured sound pressure level (SPL) of 110 dBSPL. Depending

on the amplifier design the THD increases significantly for gain factors above 5 dB. The reason for this is clipping

of the output signal of the microphone. Figure 5e shows the influence of the gain factor on the frequency response.

8/17/2019 Walser 2016

6/19


Increasing the gain factor results in a linear shift to higher sensitivities, without changing the frequency response. A

detailed view of the frequency response of flip-chip MEMS microphone packages is given in [5] and [24].

However, adjusting the bias voltage changes the operating point of the MEMS sensor and thus has an influence

on the overall performance including SNR and THD. Programming bias results in a linear shift of the sensitivity as

shown in figure 5f, as well as of the noise floor as shown in figure 5g. For this sensor design, the sensitivity reacts to

a change of the bias voltage with a slope of 1.5 dBV/V. The microphone noise floor is caused by different sources,

in general the noise of the sensor, the noise of the ASIC and noise caused by the package. The noise reacts to achange of the bias voltage with a slope of 1.0 dBV/V.

The different sensitivity and noise slopes result in a large influence of the bias voltage on the SNR as shown in

figure 5h. With a slope of 0.5 dBV/V the SNR can be modified by the bias voltage. The drawback of high bias

voltages is the THD increase shown in figure 5i. Higher bias voltages shift the sensor operating point closer to the

pull-in point. The non-linear sensor behavior in this region leads to higher THD. Figure 5j shows the influence of

the bias factor on the frequency response. Increasing the bias voltage results in a linear shift to higher sensitivities,

without changing the frequency response.

3.2. Narrow sensitivity distribution

Figure 6 shows the sensitivity distribution of one production lot directly after fabrication. The sensitivities weremeasured in the production end test. The mean value of the sensitivities is -38 dBV/Pa; the span has a value of

3.4 dB. The standard deviation is 0.57 dB. With a specified sensitivity tolerance of ±1 dB the yield would be below

90% for a well-centered process.

To achieve the goal of reducing the sensitivity variation after MEMS microphone fabrication, the programmable

interface is applied. By using 5 bits for bias and 4 bits for gain adjustment, a sensitivity calibration matrix follows

for all possible programming combinations. Figure 7 shows the measured sensitivity for all possible gain and bias

combinations measured for one programmable microphone. With the designed step sizes, the calibration matrix has

a wide sensitivity range of 10.8 dBV. The gain adjustment for rough sensitivity approximation is shown in the

sixteen steps between the different diagonal bias lines. The bias adjustment for precise sensitivity setting is shown in

the diagonal lines with a step count of 32.

The programming is done after completion of the fabrication process within a production test. First, ameasurement characterizes the sensitivity after fabrication and determines the calibration settings with regard to the

required sensitivity specifications, e.g. -38 dBV/Pa. In the next step, the microphone is calibrated by gain and bias

factors. After the programming sequence a complete electroacoustic production test is done. Figure 8 shows the

sensitivity distribution, of the same production lot as in figure 6, before and after programming.

The standard deviation of 0.57 dB before programming was reduced to a standard deviation of 0.06 dB by

programming. With this value, all sensitivities are well within ±1 dB specification limits. The reason for the

remaining variation after programming is that the sensitivity change caused by a bias shift is not well determined.

The change depends on MEMS sensor parameters, like thickness, stress, membrane diameter and the distance

between membrane and backplate.

An even narrower sensitivity deviation could be reached by a precisely characterized bias slope factor. One

solution for this problem could be the characterization of the bias slope factor for each individual microphone deviceand adapting these parameters to the calibration algorithm. The drawback of this characterization is a longer test

time. Around 100 ms are necessary for a single programming cycle in the production test. Two power-ups for

programming and for the normal modus and the first sensitivity measurement are included in this time. For a

characterization of bias slope two additional measurements would be necessary. Additional 100 ms would be

required for this characterization. For a specified sensitivity tolerance of ±1 dB, a calibration without individual bias

slope characterization is sufficient.

A further advantage of microphone calibration is that improvements of other electroacoustic microphone

parameters, e.g. SNR are possible. By choosing the optimal bias and gain combination, the microphone SNR can be

improved. Figure 9 shows SNR and sensitivity for all different settings of bias and gain for one microphone.

8/17/2019 Walser 2016

7/19

Compared to figure 8 the graph is more scattered due to the statistic character of a short noise measurement.

High bias voltages lead to high SNR because the influence of the bias voltage on the sensitivity is higher than the

influence on the noise. Therefore, by using a high bias voltage and a matching gain calibration the microphone can

be trimmed to a high SNR. For a specified sensitivity of -38 dBV/PA a high SNR of up to 66.5 dB(A) was reached.

Figure 10 shows the distribution of the SNRs measured at one hundred microphones. Both groups were

programmed to a specified mean sensitivity of -38 dBV/Pa. After normal programming, the mean SNR was

66.1 dB(A). With an optimized algorithm, a mean SNR of 66.4 dB(A) was achieved.

4. Results and Discussion

The calibration method for microphones after fabrication presented in this paper can be used to compensate

process tolerances and packaging stress. The new programmable MEMS microphone uses an OTP module

integrated in the ASIC to program the bias supply voltage and the gain factor. Trimming bias and gain results in a

parallel shift of the frequency response. By using the optimal bias and gain combination a sensitivity adjustment

with a wide range of approximately 11 dBV is can be achieved.

The programming method allows compensating the sensitivity variations caused by process tolerances of the

fabrication processes. Specification limits of ±1 dB can be fulfilled with a high production yield. In addition, MEMS

microphones can be calibrated towards a higher SNR. High SNR above 66 dB(A) could be reached in mass production. High bias supply voltages shift the microphones closer towards the nonlinear region. Therefore, a

drawback of programming towards high SNR are higher total harmonic distortions.

An even narrower sensitivity deviation would be possible if the influence of the bias setting on the sensitivity was

characterized for each single device. This would, of course, result in longer testing times.

8/17/2019 Walser 2016

8/19


Literature

[1] P. R. Scheeper, A. G.H. Donk, W. Olthuis and P. Bergveld, A review of silicon microphones, Journal of Sensors and Actuators A 44,

1994, pp. 1-11.

[2] B.-H. Kim, H.-S. Lee, Acoustical-Thermal Noise in a Capacitive MEMS Microphone, IEEE Sensors Journal Vol 15, 2015, no 12 pp.

6853-6860.

[3] J.W. Weigold, T.J. Brosnihan, J. Bergeron, X. Zhang, A MEMS Condenser Microphone For Consumer Applications, Proceedings of

Micro Electro Mechanical Systems, Istanbul, Turkey, Jan 22-26, 2006, pp. 86-86.[4] A. Dehe, M. Wurzer, M. Fueldner, U. Krumbein, Design of a Poly Silicon MEMS Microphone for High Signal-to-Noise Ratio,

Proceedings of European Solid-State Device Research Conference, 2013, pp. 292-295.

[5] S. Walser, C. Siegel, M. Winter, M. Loibl, W. Pahl, A. Leidl, G. Feiertag, Flip-Chip MEMS Microphone Package With Small Front-

Volume and Large Back-Volume, Proceedings of European Microelectronics and Packaging Conference, 2015.

[6] S. Walser, C. Siegel, M. Winter, A. Leidl, G. Feiertag, A Novel Method for Reducing the Sensitivity Deviation of MEMS

Microphones by Programming, Procedia Engineering 120 (2015) pp. 206-209.

[7] Z.-H. Fu, F. Fan, J.-D. Huang, Dual-microphone noise reduction for mobile phone application, Proceedings of International Acoustic,

Speech and Signal Processing, Vancouver, Canada, May 26-31, 2013, pp. 7239-7243.

[8] M-C. Yew, C-W. Huang, W-J. Lin, C-H. Wang, P. Chang, A Study of Residual Stress Effects on CMOS-MEMS Microphone

Technology, Proceedings of Microsystems, Packaging, Assembly and Circuits Technology, Taipei, China, Oct 22-23, 2009, pp. 323-

326.

[9] C-H. Chuang, W-H. Li, Parametric study of flip-chip packaging for an MEMS device with diaphragm, Proceedings of Design Test

Integration and Packaging of MEMS/MOEMS, Seville, Spain, May 05-07, 2010, pp. 362-365.

[10] O. Rusanen, A. Torkkeli, Adhesives as a Thermomechanical Stress Source - Comparing Silicones to Epoxies, Proceedings ofAdhesive Joining and Coating Technology in Electronics Manufacturing, Binghamton, USA, Sep 28-30, 1998, pp. 288-294.

[11] S.S Je, J.C. Harrison, M.N Kozicki, B. Bakkaloglu, S. Kiaei, J. Chae, In situ tuning of a MEMS microphone using electrodeposited

nanostructures, Journal of Micromechanics and Microengineering, 2009, vol. 19.

[12] G. Feiertag, M. Winter, A. Leidl, Packaging of MEMS microphones, Proceedings of Smart Sensors, Actuators and MEMS, Dresden,

Germany, May 04, 2009, vol. 7362.

[13] A. Dehe, M. Wurzer, M. Füldner, U. Krumbein, The Infineon Silicon MEMS Microphone, Proceeding of AMA Conference,

Nürnberg, Germany, May 14-16, 2013, pp. 95-99.

[14] G. Feiertag, M. Winter, A. Leidl, Flip chip packaging for MEMS microphones, Journal of Microsystem Technologies Vol 16, 2010,

pp. 817-823.

[15] J. Bay, O. Hansen, S. Bouwstra, Micromachined double backplate differential capacitive microphone, Journal of Micromechanics and

Microengineering, Vol 9, 1999, pp. 30-33.

[16] P. Rombach, M. Müllenborn, U. Klein, K. Rasmussen, The first low voltage, low noise differential silicon microphone, technology

development and measurement results, Proceedings of International Conference on Micro Electro Mechanical Systems, 2001, pp. 42-

45.

[17] D.T. Martin, K. Kardirvel, J. Liu, R.M. Fox, M. Sheplak, T. Nishida, Surface and Bulk Micromachined Dual Back-Plate Condenser

Microphone, Proceeding of Micro Electro Mechanical Systems Conference, Jan 30 – Feb 03, 2005, pp. 319-322.

[18] D.T. Martin, Design, fabrication and characterization of a MEMS dual-backplate capacitive microphone, Dissertation, 2007,

University of Florida, USA.

[19] C. Fallesen, J.K. Poulsen, L.J. Stenberg, J.J.G. Bosch, A calibrated microelectromechanical microphone, European patent application,

EP 1 906 704 A1, filed 05.09.2007.

[20] R.S.C. Wang, R.S.J. Shen, C.C.H. Hsu, Neobit – high reliable logic non-volatile memory (NVM), Proceedings 11th International

Symposium on the Physical and Failure Analysis of Integrated Circuits, 2004, pp. 111-114.

[21] M. Füldner, Modellierung und Herstellung kapazitiver Mikrofone in BiCMOS-Technologie, Dissertation, 2004, University of

Erlangen-Nürnberg, Germany.

[22] M. Winter, Integration kapazitiver Silizium-Mikrofone in ein Chip Scale Package, Dissertation, 2011, University of Saarland,

Germany.

[23] J. Liu, Nonlinear Dynamics of a Dual-Backplate Capacitive MEMS Microphone, Dissertation, 2007, University of Florida, USA.[24] M. Winter, G. Feiertag, A. Leidl, H. Seidl, Influence of a chip scale package on the frequency response of a MEMS microphone,

Microsystem Technologies Vol 16, 2010, pp. 809-815.

8/17/2019 Walser 2016

9/19

M. Sc. Sebastian Walser

Sebastian Walser received his diploma degree in electrical and information technology in 2011

and his M.Sc. degree in electrical engineering in 2013 from the University of Applied Sciences

Munich. Since 2013 he is with the University of Applied Sciences Munich and is working in a

research project, called “twinMikro”, in cooperation with the EPCOS AG and the Fraunhofer

IIS. Since 2014 he is also a Ph.D. student of the “Universität der Bundeswehr München”. His

research interests include MEMS microphones, MEMS packaging, acoustic modeling.

Dr. Christian Siegel

Christian Siegel received his diploma degree in electrical engineering from the Technical

University of Munich and his Ph.D. degree from the University of Ulm. His Ph.D. research wason the development of a new reliable RF-MEMS technology in cooperation with the EADS. He

is currently working in the product development and engineering for MEMS microphones at

EPCOS AG.

Dr. Matthias Winter

Matthias Winter received his diploma degree in physics in 2006 from the Technical University

of Munich and his Ph.D. degree in 2011 from the University of Saarland. His Ph.D. research

was on a new MEMS microphone chip-scale package in cooperation with the EPCOS AG. He

is currently working in the product development and engineering for MEMS microphones at

EPCOS AG.

Prof. Dr. Gregor Feiertag

Gregor Feiertag investigated the precision of the structure transfer of x-ray lithography at theInstitute for Micro Technology in Mainz (IMM) as part of his Ph.D. research work. From 1999

to 2009 he developed packaging technologies for Surface Acoustic Wave Components (SAW),

MEMS microphones and pressure sensors at EPCOS in Munich. He is now at the Munich

University of Applied Sciences as a professor for electronic packaging and sensors in the

department of electrical engineering. His main research subject is packaging of sensors and

SAW-components.

B. Eng. Michael Loibl

Michael Loibl received his B.Eng. degree in electrical and information technology from the

University of Applied Sciences Munich. Since 2013 he is with the University of Applied

Sciences Munich and is working in a research project, called “twinMikro”, in cooperation withthe EPCOS AG and the Fraunhofer IIS.

Dr. Anton Leidl

Anton Leidl received his diploma degree in physics in 1992 from the Technical University in

Munich and his Ph.D. degree in Electrical Engineering from the 'Universität der Bundeswehr'

in 1998. From 1993 till 2000 he was responsible for the team 'Sensors for Liquids' at the

Fraunhofer Insitute for Solid State Technology. Since 2001 Dr. Leidl is with EPCOS AG,

8/17/2019 Walser 2016

10/19


Munich and is heading the Business Unit ‘MEMS Microphones and Pressure Sensors’ since 2010.

Fig. 1: Schematic of a programmable differential capacitive MEMS microphone: As transducer, a double-backplate sensor chip is used. By

programming the one-time-programmable (OTP) module in the ASIC the gain and the bias voltage can be adjusted. Five bits are reserved for bias

and four bits for gain setting.

membrane

backplate

Sensor ASIC

OUT+

MEMS microphone

backplateamplifier

Bias voltage

generator

5-bits 4-bits

4-bits

OUT-

VDD

GND

CLK

DATA

amplifier

one-time-programmable

(OTP) memory

8/17/2019 Walser 2016

11/19

Fig. 2: Top left: Illustration of a double backplate sensor with the membrane deflection x, the mechanical force Fmech and the two electrical forces

Fel1 and Fel2 in case of a sound pressure. Lower graph: Membrane displacement as a function of the bias voltage for a constant sound pressure is.

M e m b r a n e d i s p l a c e m e n t

Bias voltage supply0

Membrane

BP2

BP1

sound

pressure

2·x0

0

x0Fel2

Fel1

x

Fmech

backvolume

V bias

V bias

non-linear

increasing

to pull-in

Fel2 ≈ ε·A· (V bias)2 / (2·(2x0 – x)

2)

Fel1 ≈ - ε·A· (V bias)2 / (2·x2)

Fmech ≈ k membrane · (x0 – x)

8/17/2019 Walser 2016

12/19


Fig. 3: Left: Double-backplate sensor chip with a size of 1.45 x 1.45 mm2; Bottom right: Backplate with perforation; Top right: Backplate with

venthole for pressure exchange between frontvolume and backvolume.

venthole

backplate

Sensor chip

1.45 mm

1 . 4

5 m m

8/17/2019 Walser 2016

13/19

Fig. 4: Sectional view of: a) Complete fabricated microphone [5]; b) Sensor chip flip-chip process; c) Separation of front- and back-volume by

polymer foil lamination process; d) Closure of the microphone back-volume by a metal lid.

Sensor Chip ASIC Chip

Solder Balls

Backplate

Backplate

Membrane

Sound Hole

Metal Cap

Polymer Foil

Ceramic Substrate

(HTTC)

Adhesive

a)

b) c) d)

8/17/2019 Walser 2016

14/19


Fig. 5: Measured influence of programming on the electroacoustic microphone characteristics: a) Sensitivity as a function of the gain factor; b)

Noise as a function of the gain factor; c) SNR as a function of the gain factor; d) THD as a function of the gain factor; e) frequency response for

max and min gain factors; f) Sensitivity as a function of the bias voltage; g) Noise as a function of the bias voltage; h) SNR as a function of the

bias voltage; i) THD as a function of the bias voltage; e) frequency response for max and min bias voltage.

y = 1.0121x - 43.392

y = 1.0093x - 39.525

-45-43

-41

-39

-37

-35

-33

-31

S e

n s i t i v i t y / d B V / P a

Minimum bias voltage Maximum bias voltagea)

y = 1.6231x - 59.504

y = 1.5877x - 51.48

-45-43

-41

-39

-37

-35

-33

-31

S e

n s i t i v i t y / d B V / P a

Minimum gain factor Maximum gain factor f)

y = 0.9188x - 108.18

y = 0.9792x - 105.78

-110

-108

-106

-104

-102

-100

-98

-96

N o i s e / d B V ( A )

b)

y = 1.0116x - 118.19

y = 1.1472x - 112.61

-110

-108

-106

-104

-102

-100

-98

-96

N o i s e / d B V ( A )

g)

y = 0.0931x + 64.792

y = 0.0299x + 66.257

64.5

65.0

65.5

66.0

66.5

67.0

S N R / d B ( A )

c)

y = 0.6119x + 58.686

y = 0.4409x + 61.124

64.5

65.0

65.5

66.0

66.5

67.0

S N R / d B ( A )

h)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0 2 4 6 8 T H D @ 1

1 0

d B S P L / %

Programmed gain factor / dB

d)amplifier influence

y = 0.0258x - 0.2055

y = 0.0454x - 0.3121

0.00

0.05

0.10

0.15

0.20

0.25

0.30

9.5 10.0 10.5 11.0 11.5 12.0 12.5 T H D @ 1

1 0

d B S P L

/ %

Programmed bias voltage / V

i)

-46-44

-42

-40

-38

-36

-34

-32

100 1000 10000

S e n s i t i v i t y / d B V / P a

Frequency / Hz

e)

shift

▲ Minimum gain factor

■ Maximum gain factor

-46-44

-42

-40

-38

-36

-34

-32

100 1000 10000

S e n

s i t i v i t y / d B V / P a

Frequency / Hz

j)

shift

■ Maximum bias voltage

▲ Minimum bias voltage

8/17/2019 Walser 2016

15/19

Fig. 6: Measured sensitivity distribution of one MEMS microphone production lot after fabrication: For a designed mean sensitivity value of -

38 dBV/Pa a standard deviation of 0.57 dB was reached. For a specified sensitivity tolerance of ±1 dB, the yield would be below 90 %.

0

20

40

60

80

100

- 4 0 . 0

- 3 9 . 8

- 3 9 . 6

- 3 9 . 4

- 3 9 . 2

- 3 9 . 0

- 3 8 . 8

- 3 8 . 6

- 3 8 . 4

- 3 8 . 2

- 3 8 . 0

- 3 7 . 8

- 3 7 . 6

- 3 7 . 4

- 3 7 . 2

- 3 7 . 0

- 3 6 . 8

- 3 6 . 6

- 3 6 . 4

- 3 6 . 2

- 3 6 . 0

N u m b e r o f m e a s u r e d

m i c r o p h o n e s

Sensitivity (@ 1kHz) / dBV/Pa

8/17/2019 Walser 2016

16/19


Fig. 7: Sensitivity for all gain and bias combinations, exemplarily measured at one microphone device: gain adjustment shown in 16 steps

between the 32 different diagonal bias lines.

-44

-42

-40

-38

-36

-34

-32

-30

S e n s i t i v i t y ( @ 1 k

H z ) / d B V / P a

Different programmed settings

all prog. values MIN gain MIN bias MAX bias MAX gain

gain steps

bias ↑

gain ↑

8/17/2019 Walser 2016

17/19

Fig. 8. Sensitivity distribution of one production lot after fabrication (“before programming”) and after trimming (“after programming”): The

standard deviation of the sensitivity was reduced from 0.57 dB to 0.06 dB by programming.

0

20

40

60

80

100

120

140

160

180

- 4 0 . 0

- 3 9 . 8

- 3 9 . 6

- 3 9 . 4

- 3 9 . 2

- 3 9 . 0

- 3 8 . 8

- 3 8 . 6

- 3 8 . 4

- 3 8 . 2

- 3 8 . 0

- 3 7 . 8

- 3 7 . 6

- 3 7 . 4

- 3 7 . 2

- 3 7 . 0

- 3 6 . 8

- 3 6 . 6

- 3 6 . 4

- 3 6 . 2

- 3 6 . 0

N u m b e r o f m e a s u r e d m i c r o p h o n e s


before programming (standard deviation = 0.57 dB) after programming (standard deviation = 0.06 dB)

8/17/2019 Walser 2016

18/19


Fig. 9. SNR and sensitivity for all different settings of bias and gain for one microphone: By increasing bias voltages the SNR increases

significantly, Gain can be used for programming sensitivity without significant influence on SNR.

64.5

65.0

65.5

66.0

66.5

67.0

-44 -42 -40 -38 -36 -34 -32 -30

S N R / d B ( A )


min GAIN max GAIN min BIAS max BIAS

high SNR

for a specified

sensitivity

8/17/2019 Walser 2016

19/19

Fig. 10. SNR distribution of one hundred microphones programmed to a mean sensitivity of -38 dBV: with an optimized algorithm, the SNR was

trimmed to a higher mean SNR value of 66.4 dB(A).

0

5

10

15

20

25

65.5 65.6 65.7 65.8 65.9 66.0 66.1 66.2 66.3 66.4 66.5 66.6 66.7 66.8 66.9 67.0

N u m b e r o f m e a s u r e d m i c r o

p h o n e s a m p l e s

Signal to noise ratio / dB(A)

Prog. (-38dBV) Prog. (-38dBV and high SNR)

walser 2016

Documents