348

IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 3, NO. 5, OCTOBER 2009

A Compact and Low-Cost MEMS Loudspeaker for Digital Hearing Aids Sang-Soo Je, Fernando Rivas, Rodolfo E. Diaz, Jiuk Kwon, Jeonghwan Kim, Bertan Bakkaloglu, Senior Member, IEEE, Sayfe Kiaei, Fellow, IEEE, and Junseok Chae

Abstract—A microelectromechanical-systems (MEMS)-based electromagnetically actuated loudspeaker to reduce form factor, cost, and power consumption, and increase energy efficiency in hearing-aid applications is presented. The MEMS loudspeaker has multilayer copper coils, an NiFe soft magnet on a thin polyimide diaphragm, and an NdFeB permanent magnet on the perimeter. The coil impedance is measured at 1.5 , and the resonant frequency of the diaphragm is located far from the audio frequency range. The device is driven by a power-scalable, 0.25- m complementary metal–oxide semiconductor class-D amplifier stage. The class-D amplifier is formed by a differential H-bridge driven by a single bit, pulse-density-modulated bitstream at a 1.2-MHz clock rate. The fabricated MEMS loudspeaker generates more than 0.8- m displacement, equivalent to 106-dB sound pressure level (SPL), with 0.13-mW power consumption. Driven by the class-D amplifier, the MEMS loudspeaker achieves measured 65-dB total harmonic distortion (THD) with a measurement uncertainty of less than 10%. Energy-efficient and cost-effective advanced hearing aids would benefit from further miniaturization via MEMS technology. The results from this study appear very promising for developing a compact, mass-producible, low-power loudspeaker with sufficient sound generation for hearing-aid applications.



61 61

61

61

Index Terms—Acoustic, actuator, class-D amplifier, hearing aids, loudspeaker, microelectromechanical systems (MEMS), microspeaker.

I. INTRODUCTION

P

REVIOUS studies of anatomy, physiology, and psychophysics have contributed to a growing curiosity about normal versus impaired auditory systems [1]–[3], attracting the benefits of hearing aids for the hearing impaired [4], [5]. Approximately 28 million Americans have hearing impairments. Hearing loss affects nearly 17 in 1 000 children under the age of 18, and the incidence increases with age: nearly 314 in 1 000 people over age 65 have hearing loss. According to recent statistics, however, 80% of those who could benefit from a hearing aid chose not to use one. The reasons for the number of untreated cases include reluctance to recognize hearing loss and Manuscript received November 13, 2008; revised February 27, 2009. Current version published September 25, 2009. This work was supported in part by the U.S. National Science Foundation under Grants 0627777 and 0652136. This paper was recommended by Associate Editor Robert Rieger. S.-S. Je, R. E. Diaz, J. Kwon, J. Kim, B. Bakkaloglu, S. Kiaei, and J. Chae are with Arizona State University, Tempe, AZ 85287 USA (e-mail: [email protected]). F. Rivas is with the University of Jaén, Jaén 23071, Spain (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TBCAS.2009.2026429

common misconceptions about hearing aids, such as a social stigma to wearing them. People find hearing aids inconvenient, and some accept that losing hearing capability is a part of aging, known as presbycusis [6], [7]. Since the first digital hearing aid appeared in 1996, hearing aids have advanced substantially [8]. The biggest improvements include directional microphones and associated signal-processing algorithms [9], [10]. Despite these improvements, however, recent statistics suggests that patients’ satisfaction with hearing instruments has not improved [11], and return rates on some hearing aids exceed 40% [12]. Further improvements demanded by patients include long-lasting battery life, high directionality, low cost, low background noise, and low acoustic feedback [13]–[17]. Generally, there are two types of hearing impairment: conductive hearing loss (sound-level reduction due to a blockage) and sensorineural hearing loss (damage to the auditory nerve). The conductive hearing loss can be treated by hearing aids and the sensorineural hearing loss by cochlear implants. Hearing aids that include a microphone-amplifier-loudspeaker system target the conductive hearing loss, ranging from mild (up to 40-dB sound lost) to profound (more than 90-dB sound lost) losses. Depending upon the coverage of hearing impair levels, different types of hearing aids are currently on the market: behind-the-ear (BTE), in-the-ear (ITE), in-the-canal (ITC), and completely-in-the-canal (CIC) types of hearing aids [18]. Among all hearing aid types, the CIC-type hearing aid is designed to be almost invisible and, thus, a good candidate for helping minimize social stigma [19], [20]. However, improving CIC-type hearing-aid performance is challenging. Highly miniaturized components, including microphones, circuitry, battery, and loudspeaker are all packaged in a small enclosure that fits into an average-sized adult ear canal: a space with approximately 0.7 in in diameter and 1 in long. The loudspeaker is one of the most crucial components of a hearing aid, and it consumes 50%–95% of the hearing aid’s overall power [21]. In this paper, a technique to reduce loudspeaker form-factor and cost in hearing aids, using energy efficient MEMS technology is presented. The speaker has a thin diaphragm on a silicon substrate actuated by the Lorentz force. The actuation mechanism and diaphragm deflection are modeled and simulated by using SPICE and ANSYS. A complementary metal–oxide semiconductor (CMOS) driver amplifier is designed in a 0.25- m CMOS process. The hearing-aid digital noise-shaper signal processor (DSP) drives a second-order with a single-bit output. The single-bit output controls the class-D H-bridge driver stage. Characterization results show

1932-4545/$26.00 © 2009 IEEE Authorized licensed use limited to: Arizona State University. Downloaded on October 8, 2009 at 19:55 from IEEE Xplore. Restrictions apply.

JE et al.: COMPACT AND LOW-COST MEMS LOUDSPEAKER

349

that the prototype MEMS loudspeaker can generate 7- m displacement, equivalent to 120-dB SPL, at 1 kHz in hearing aids to cover the profound hearing-impaired patient requirements. The energy-efficient and inexpensive MEMS loudspeaker could further advance miniaturizing CIC-type hearing aids. The paper is organized as follows: Section II discusses MEMS loudspeaker design, fabrication, and experimental setup. Its operating principle, energy efficiency, microfabrication process, and an optical measurement setup to detect displacement of the diaphragm is presented. Section III presents a power-scalable H-bridge class-D driver. Section IV explains the simulation setup of the designed device and discusses simulation results. Section V follows with measurement results and future work. The diaphragm displacement as a function of currents, measurement uncertainties in the setup, and SPL related to input power over the audio frequency range is presented. Future directions to improve the actuation mechanism and different deposition techniques to achieve more compact device size are also presented. Section VI ends with concluding remarks. II. DEVICE DESIGN, FABRICATION, AND EXPERIMENTAL SETUP A. Device Design Many actuation mechanisms have been deployed for loudspeaker applications, including piezoelectric, electrostatic, and electromagnetic (EM) actuations [22]–[24]. Among them, piezoelectric and electrostatic mechanisms are less attractive for hearing-aid applications because of their low conversion factor and the use of high voltage, respectively. EM actuation is the most efficient mechanism and generates the most powerful sound pressure for hearing aids, producing a 106-dB SPL at a distance of 13 mm from the eardrum as required for the profoundly hearing impaired [25]. Given the size limit of CIC hearing aids, EM speakers have the most compatible manufacturing processes for microfabrication technology. Fig. 1 illustrates an electromagnetically actuated MEMS-based loudspeaker with a suspended polyimide diaphragm, soft magnet (1-mm diameter and 10 20- m thick), permanent magnet (20-mm outer diameter, 4-mm inner diameter, and 0.5-mm thickness), and copper coils (10- m coil thickness, 100- m coil width, and 100- m gap between wound coils, and 10- m spacing between multiple coil layers, filled by polyimide). Multiturn and multilayered copper coils surround the soft in the coils magnet core on the diaphragm. Flowing current allow the diaphragm to and an external magnetic field actuate vertically by the Lorentz force. The soft magnet core focuses on the magnetic field. The external magnetic field is provided by a custom-made rare-earth NdFeB magnet that has been manually glued to the chip perimeter. The diaphragm displacement generated by the Lorentz force is a function of the air pressure and can be written as (1) is the air pressure, is the diaphragm diameter, where is the Young’s modules of polyimide, is diaphragm thickness, and is the Poisson ratio [26].

Fig. 1. Schematic of an electromagnetically actuated MEMS loudspeaker for hearing-aid applications: (a) Top view. (b) Side view.

An MEMS loudspeaker is modeled by an analog equivalent circuit model which consists of current to actuate the speaker, voice-coil, and diaphragm [Fig. 2(a)] [27], [28]. The voice coil includes Cu coils and the NiFe soft magnet on the diaphragm. V1 models the driving voltage to supply currents to actuate and are dc resisthe diaphragm via a source resistance. tance and inductance of the voice coil, respectively. and are lumped parameters from the Cu coil and NiFe soft magnet. , , and , which are The diaphragm is modeled by damping constant, compliance, and mass of the diaphragm, respectively. The diaphragm displacement is described by [27] displacement

(2)

is nominal input power, is static (dc) displacement sensitivity of a unenclosed loudspeaker, is the system displacement is the normalized system displacement constant, and function, showing low-pass filter characteristics. Fig. 2(b) shows the frequency response of the displacement using actual parameters used in the MEMS loudspeaker. The displacement is linear as input current increases: 0.168 m for 8-mA and 1.68 m for 80-mA input current. The resonant frequency occurs at 22 kHz. where

B. Fabrication Process Fabricating the MEMS loudspeaker involves a six-mask process that uses multilevel electroplated copper coils on a poly-

Authorized licensed use limited to: Arizona State University. Downloaded on October 8, 2009 at 19:55 from IEEE Xplore. Restrictions apply.

350

IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 3, NO. 5, OCTOBER 2009

Fig. 2. (a) Loudspeaker analog equivalent circuit model, including coils/soft magnet on a diaphragm and (b) expected diaphragm displacement over an audio frequency range.

imide diaphragm suspended on a silicon substrate, described in Fig. 3(a): 1) A first layer of 10- m-thick polyimide is spun, patterned on a silicon wafer, and cured in an oven at 350 C for one hour. 2) Thin seed layers are deposited by using a dc sputter, and 10- m-thick copper coils are electroplated on the patterned polyimide film. 3) SU-8 is spun and patterned for via holes to connect the first- and second-layer coils. 4) A second layer of 10- m-thick polyimide is spun, patterned, and cured in an oven at 350 C for one hour. Thin seed layers are deposited using a dc sputter, and the second 10- m-thick copper coils are electroplated. 5) Again, thin seed layers are deposited by using a dc sputter 20- m-thick permalloy (80% Ni and and patterned. A 10 20% Fe combination) is electroplated at the center of the diaphragm. 6) An AZ 4620 photoresistor is used to pattern the etching holes. The diaphragm is released by etching the back side of the silicon wafer by using deep reactive ion etching 4-mm device (DRIE). Fig. 3(b) shows a fabricated 4-mm pictured next to a U.S. dime. After completing the microfabrication, a permanent magnet with an inner and outer diameter of 4 mm and 20 mm, respectively, is glued on the chip. The magnet is made of NdFeB (from the Quadrant Magnetics group) with a high flux density of 1 T.

Fig. 3. (a) Multilayered MEMS loudspeaker fabrication process. (b) Photograph of the fabricated MEMS loudspeaker next to a U.S. dime.

C. Experimental Setup The diaphragm displacement of the MEMS loudspeaker is characterized by an optical interferometer (NanoGage 100 from OPTRA, Inc.). The interferometer is controlled by LabVIEW, which automatically collects displacement data as shown in Fig. 4(a).

Authorized licensed use limited to: Arizona State University. Downloaded on October 8, 2009 at 19:55 from IEEE Xplore. Restrictions apply.

JE et al.: COMPACT AND LOW-COST MEMS LOUDSPEAKER

Fig. 4. (a) Schematic of optical experimental setup. DUT is on an xyz micromanipulator. (b) Power-scalable H-bridge driver with associated circuitry.

The optical interferometer allows noncontact measurements with 15-nm resolution, up to a frequency of 100 kHz and amplitude of 100 m [29]. The device under test (DUT) is mounted on a Newport Corporation micromanipulator and placed 3 mm away from the optical interferometer. This setup measures the diaphragm displacement and allows 1- m resolution in the x, y, and z directions. A function generator (DSO6104A, Agilent) generates a sinusoidal signal and the diaphragm displacement data are collected via LabVIEW interface. Fig. 4(b) shows the die micrograph of the CMOS driver and MEMS loudspeaker on the printed-circuit board (PCB). III. CMOS DRIVER FOR THE MEMS LOUDSPEAKER A CMOS class-D drives the MEMS loudspeaker. It is a digital-to-analog converter (DAC), where the single- bit discrete-time to continuous-time (DT/CT) interface is implemented by an H-bridge class-D amplifier fabricated in a double poly 0.25- m CMOS process. The control input to the H-bridge is generated from a 16-kHz input sampling rate, 18-b-wide noise shaper. data path, 1.2-MHz output rate single-bit For low-power receiver channels in digital hearing aids, loudspeaker drivers consume the majority of quiescent power. Most state-of-the-art hearing-aid receiver channels use class AB-type amplifiers due to their high linearity, high efficiency, and improved power-supply rejection ratio (PSRR). Recently, class D amplifier topologies are gaining importance due to their high power efficiency, especially in high-power, hands-free-mode applications [30]. Analog-input-based class D stages commonly

351

use pulse-width modulation (PWM), trading off efficiency with tonal content at integer multiples of switching frequency. However, analog-input class D stages require a low-noise digital-to-analog channel before the amplifier, reducing the efficiency of the audio receiver channel. By contrast, digital-input modulator-driven one, class-D stages, such as the digital shown in Fig. 5(a), directly convert digital input signals to output without a dedicated digital-to-analog channel. This architecture is easier to implement in deep-submicron integrated circuits (ICs), achieving a higher spurious-free dynamic range (SFDR) with their quantization noise-shaping characteristics. Power consumption under low audio signal levels should be -modulated class D amplifier used in considered for any -based audio amplidigital hearing-aid applications. The fier switches the power transistors at a higher frequency in the class D output stage compared to a PWM-based amplifier; therefore, it has the potential to dissipate higher switching power. class D amplifiers, the pulse density of In digital input modulator output is proportional to input signal amplithe tude. Therefore, if the input signal amplitude is low, switching -driven audio amplifier are dominant at the power losses in a stage. If the input signal amplitude is high, output-pulse density increases and conduction losses become dominant. Since the transistor’s switching and gate drive losses are proportional to the gate capacitance, to reduce losses, transistor size should be decreased. Conduction loss is proportional to the power transistor on resistance, and increasing the transistor (W/L) ratio reduces conduction losses. In the proposed architecture, switching losses at low amplitude inputs are reduced by decreasing the transistor (W/L) ratio at the output stage; and conduction losses at high input amplitudes are reduced by increasing the transistor (W/L) ratio. Fig. 5(b) shows how connecting transistors in parallel can change transistor size; turning them on separately increases size, and turning them off separately decreases size. This approach enables optimal efficiency for the MEMS loudspeaker driver across all signal-level conditions. Fig. 5(c) shows the total power loss at several output SPLs. Transistor switching is performed based on ambient sound levels and does not have any noticeable audio artifacts. The total harmonic distortion (THD) is measured with the actual loudspeaker load connected to the class D driver circuitry. The class D loudspeaker driver is connected to a dynamic signal analyzer (DSA), and input tone is generated via a 16-b digitally encoded, single-bit bitstream at 3 dB below the sampled, full-scale ( 3 dBFs) level. The input frequency is adjusted to be around 1.015 kHz, a frequency commonly used to achieve coherent detection. After measuring the electrical output on the DSA, the THD integrated up to the 12th harmonic is calculated in the equation (3) where represents the rms value of the th harmonic component and is the rms power of the 1.015 kHz fundamental. The class D amplifier is also characterized across a wide range of audio frequencies for total harmonic distortion (THD) and noise performance. Fig. 5(d) shows the details of

Authorized licensed use limited to: Arizona State University. Downloaded on October 8, 2009 at 19:55 from IEEE Xplore. Restrictions apply.

352

IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 3, NO. 5, OCTOBER 2009

Fig. 6(a) shows a typical loudspeaker amplifier, which is either a built-in linear mode with a push-pull (class AB) configuration, or in a switching (class D) configuration. As the voltage in the push-pull swing comes closer to the voltage rail configuration, power losses tend to increase. The power conin class AB amplifiers is defined as the ratio of sumption for a given operating fresignal swing to the power supply quency, as shown

(4) As seen in (4), losses are a function of the average signal swing, which can be significant in ultra-low power hearing aids. By contrast, power consumption in class D amplifiers is , made up of three major components: 1) switching losses 2) conduction losses , and 3) gate drive losses . Overall power losses for a class D amplifier are a sum of these losses, which can be represented as (5) The conduction loss is related to the ratio between the switching amplifier’s on resistance, the load resistance, and total output power provided to the load (6) Switching losses are characterized by power losses from current losses during the switching transients of the fall time and the discharged power of the output capacitance at the power train output (7) is the switching frequency of the amplifier, and where is the drain-source voltage of the power devices. Finally, the gate drive losses at the class D stage are caused by power needed for switching the power stage. This is represented as (8)

Fig. 5. (a) Schematic of the integrated class D driver for the MEMS loudspeaker. (b) Schematic of the integrated class D driver for the MEMS loudspeaker with efficiency scalable Class D drivers. (c) Different efficiency loss lines for several output sound levels for a fixed driver stage versus minimum loss line with variable switch sizing. (d) Total harmonic distortion plus noise (THD+N) for the powerscalable class D amplifier at 1 Vp-p output swing values.

this characterization. At 0.5-Vrms signal levels, the proposed driver achieves 90% efficiency and 65-dB THD.

and represent the gate charge required during where switching transients. The factor of 2 represents the PMOS and NMOS switching pairs. These first-order loss models show that output losses for a class D amplifier are almost independent of the output voltage swing (or power delivered to the load), as shown in Fig. 6(b). However, the efficiency depends on device parameters ( and ). In this design, since the load resistance from the MEMSbased loudspeaker can have wide variations, and audio signals have varying peak-to-average ratios, especially during daily operation, the device size in the power stage is dynamically adjusted, moving the optimization curve as shown in Fig. 6(b)(ii) to achieve minimum loss across all operating conditions. As shown in Fig. 6(b)(ii), during typical daily use, audio signals

Authorized licensed use limited to: Arizona State University. Downloaded on October 8, 2009 at 19:55 from IEEE Xplore. Restrictions apply.

JE et al.: COMPACT AND LOW-COST MEMS LOUDSPEAKER

353

Fig. 6. (a) Comparison of (i) class AB versus (ii) class D amplifiers in terms of their loss characteristics with respect to output swings and (b) high loss areas shown in (i) typical histogram with daily use of hearing aids and (ii) different signal levels with respect to switching device sizing.

with typical rms amplitude occur around 20 mVrms, and those around silent areas peak at 0 Vrms [31]. By achieving dynamic scaling at the power stage, an optimum loss line associated with the ambient audio signal levels has been achieved. IV. SIMULATION OF THE ELECTROMAGNETICALLY ACTUATED MEMS LOUDSPEAKER Fig. 7(a) shows the simulation geometry and schematics for simulated models (Lorentz force and dipole method models). and are the permanent magnet’s inner and outer diameters, respectively, and and are the coils’ inner and outer diameters, respectively. The diaphragm and coils are located at the bottom of the hollow cylindrical permanent magnet. The cylinder is modeled as the equivalent of an assembly of upper and lower planes of the cylinder. In the upper plane, is an equivalent magnetic charge density; in the lower plane, . The permanent magnet induces a magthis equivalent is netic flux density on the diaphragm plane. The Lorentz force vibrates the diaphragm by flowing current through the coils. Since we are interested in the movement along the z-axis (out is considered. The of plane), only the radial component of magnitude of the Lorentz force is a function of this radial com-

ponent and the radius of the coils. When the resulting force is

turns are considered,

(9) where is the current of coils, is the radius of each turn, and is the magnitude of the radial component of in the plane of the diaphragm at . Fig. 7(b) shows a side view of the MEMS loudspeaker. The loudspeaker is 0.5 mm in height and the permanent magnet has 20-mm/4-mm outer and inner diameters, respectively. The di. aphragm is 3 mm in diameter and the coil height is 10 The coil turns are set at five. We assume that the magnetizais a z-directed vector of 1 [32]. Fig. 7(c) shows tion vector a simulation result for the distributed magnetic flux density of the fabricated loudspeaker in a -constant plane. The strongest field is located slightly above and below from the plane of the magnet base, and is distributed laterally on the plane of that base. This magnetic-field distribution requires proper coil placement to achieve maximum diaphragm displacement for any given current. Fig. 7(d) shows a comparison of two approaches to simulate the magnetic field: Lorentz force and dipole methods. The

Authorized licensed use limited to: Arizona State University. Downloaded on October 8, 2009 at 19:55 from IEEE Xplore. Restrictions apply.

354

IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 3, NO. 5, OCTOBER 2009

the Lorentz force of up to 43 N, which corresponds to up to 9 Pa of pressure on the diaphragm. The maximum Lorentz force occurs 1.4 mm away from the permanent-magnet plane in the direction. Still, at the permanent-magnet plane , the Lorentz force is 37 N. This is equivalent to generating 110-dB SPL with only 0.13-mW average power consumption. In order to generate 106-dB SPL in a human eardrum, 4 Pa of pressure is required; this is equivalent to 32- N Lorentz force. The simulation results indicate that the MEMS loudspeaker generates sufficient force for achieving the large actuation and SPL generation required for hearing aids. V. DEVICE CHARACTERIZATION The measured coil impedance of the single-turn device is only 1.5 , much smaller than prior MEMS art. The resonant frequency of the diaphragm is simulated and calculated to be 22 kHz as (10) where is a resonant frequency in Hertz, is the mass of the diaphragm in kilograms, and is compliance of the driver’s suspension (the reciprocal of its “stiffness”) in m/N. The power consumption of the MEMS loudspeaker, 0.13 mW, is lower by a factor of at least four when compared to previously reported MEMS loudspeakers, and it is comparable to that of the existing macro-size counterparts [33]. We measured four different types of fabricated MEMS loudspeakers based on the number of layers and turns of coils: 1 1 (one-layer and one-turn), 1 2 (one-layer and two-turn), 2 2 (two-layer and two-turn), and 1 5 (one-layer and five-turn). Estimated SPL is calculated from diaphragm displacements measured by an optical interferometer. A. Displacement

Fig. 7. (a) Schematic of hollow magnetic cylinders (Lorentz force and dipole method models) and a table for coils and a permanent magnet. (b) Side view with geometries. (c) Simulation result of the distributed magnetic flux density for the proposed loudspeaker in a 8-constant plane (units are in Tesla). (d) Simulated results of the Lorentz force and dipole method (units are in Newton).

two approaches demonstrate very similar results, as expected. The magnetic field of 1 T and current of 20 mA rms generate

Mechanical simulation is performed by using ANSYS in order to estimate displacement of the diaphragm. We assume no environmental perturbations, such as air damping, exist and the Lorentz force is uniform across the diaphragm. A 3-D ten-node tetrahedral structural solid 3-D element, Solid92 is used for both diaphragm and coils. Pressure is applied on the top side of the coils to mimic the Lorentz force generated by current and an external magnetic field. Fig. 8(a) shows a schematic 2 device). The diaphragm is view of the microspeaker (1 supported and anchored at the edge. Coils are on the diaphragm, and the Lorentz force is applied from on the coils to generate out-of-plane deflection. Fig. 8(b) shows contour plots of the diaphragm deflection, top and cross-sectional views, upon the Lorentz force of 10 N. The diaphragm deflects to the z-direction by 0.09 m. Fig. 9(a) shows measurements of 1 1 and 1 2 devices and their simulation results. The measurements for the 1 1 devices are very close to the simulation, but the measurements of 1 2 devices have relatively high uncertainty, 18% smaller displacement than the simulation at 88 mA. The discrepancy is likely caused by inaccurate permanent-magnet positioning and sensor focusing, which are addressed in the next section. Fig. 9(b) shows diaphragm displacement of four different device

Authorized licensed use limited to: Arizona State University. Downloaded on October 8, 2009 at 19:55 from IEEE Xplore. Restrictions apply.

JE et al.: COMPACT AND LOW-COST MEMS LOUDSPEAKER

355

2

Fig. 8. (a) Schematic of the microspeaker used in ANSYS simulation (1 2; one layer two turn coils). (b) Contour plots of the diaphragm deflection, top, and cross-sectional views.

types as a function of applied current at 1 Hz. The displacements are linear (0.97 0.99 of ), and the center displacement of the (2 2) at 88 mA. devices ranges from 1.35 m (1 1) to 7 The plot also shows that devices with vertically stacked coils (2 2) show more actuation than those with laterally turned coils (1 2 and 1 5). This is because the Lorentz force is more focused at the center of the diaphragm for vertically stacked coils, and laterally turned coils distribute the force toward the edge of the diaphragm. Fig. 10 shows the displacement comparison over the frequency between simulated at 8-mA (black), 80-mA input currents (red), respectively, and measurement (blue line) of 1 1 devices. The resonant frequency (20 kHz for measurement and 22 kHz for simulation, respectively) is located away from a practical audio range (up to 5 kHz). The discrepancy between the simulation and measurement shows approximately 0.7 (1.6 m for the simulation, 0.9 m for the measurement). We believe this discrepancy may result from device fabrication imperfection, such as stress of the polyimide and electroplated-metal films and the lumped model.

2

Fig. 9. (a) Displacement of various devices at 1 Hz: 1 1 (one layer, one turn coil), 1 2 (one layer, two turn coil), 2 2 (two layers, two turn coils), and 1 5 (one layer, five turn coils). (b) Comparison between measured and simulated results for the 1 1 and 1 2.

2

2

2

2

2

B. Measurement Uncertainty and Reproducibility Measurement uncertainty and reproducibility of the MEMS loudspeaker are characterized in this section. Fig. 11(a) shows

Fig. 10. Displacement comparison over the frequency between simulated (red and black lines) and measured (blue line) of 1 1 (one layer, one turn coil) devices.

Authorized licensed use limited to: Arizona State University. Downloaded on October 8, 2009 at 19:55 from IEEE Xplore. Restrictions apply.

2

356

IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 3, NO. 5, OCTOBER 2009

Fig. 12. Plans for future investigation: Sound pressure level and magnetic flux change by replacing a macro-size permanent magnet with a micromachined permanent magnet, consuming 0.5-mW power with a diaphragm diameter of 3 mm.

Fig. 11. (a) Measurement uncertainty: diaphragm displacement of measured 1 1 (one layer, one turn coil) and 1 5 (one layer, five turn coils) over applied current at 1 Hz, and (b) reproducibility of 1 1 devices at 1 Hz and 1 kHz, respectively.

2

2

2

approximately 27% smaller displacement can be measured from the perfectly located case (2.7 m to 1.95 m, corresponding to approximately 3-dB loss for a 3-mm diameter diaphragm). The second source of measurement uncertainty results from inaccurate focusing of the optical interferometer at the center of the diaphragm. This translates to approximately 23% smaller displacement from actual movement (2.7 m to 2.09 m, corresponding to approximately 2.3-dB loss for a 3-mm diameter diaphragm). Therefore, we expect to measure up to 50% smaller displacement than is possible, which, in a worst-case scenario, corresponds to 5.3-dB loss. This study shows that permanent-magnet placement is a critical issue in the MEMS loudspeaker’s fabrication process; it needs to be addressed in future work by replacing a micromachined permanent magnet [33].

VI. FUTURE WORK diaphragm displacement of 1 1 and 1 5 devices as a function of applied current at 1 Hz. The 1 5 devices show approximately four times larger actuation than the 1 1 devices. Actuation does not increase proportionally to the number of turns; the magnetic flux is distributed laterally, and the effective stiffness of the diaphragm increases from the center to the edge of the diaphragm on multiturn devices due to the different magnetic flux distribution on the diaphragm. In addition, as the number of turns increases, the measurement error also increases. Approximately 10% and 45% uncertainties are shown for the 1 1 and 1 5 devices at 88 mA, respectively. Fig. 11(b) shows the reproducibility of four 1 1 devices measured at 1 Hz and , up to 1 kHz. The displacements are very linear 88 mA in both cases. The reproducibility of the four devices is 36% and 32% at 1 Hz and 1 kHz, respectively. There are two possible sources for the measurement uncertainties. The first stems from positioning the permanent magnet during the last step of fabrication. ANSYS simulation shows that the position of the strongest magnetic field can be shifted by 0.75 mm due to the inaccurate placement of the permanent magnet. The simulation also shows that

In order to reduce measurement uncertainty, increase device reproducibility, and achieve fully micromachined loudspeaker design, we identified future avenues for research: a plan to substitute a micromachined permanent magnet instead of the bulky permanent magnet. The micromachined permanent magnet can be made by various fabrication techniques, such as electrodeposition, spin coating, and sputtering [34]–[40]. Fig. 12 depicts the calculated SPL as a function of the permanent magnetic flux of the techniques. High magnetic flux generates strong Lorentz force and produces large diaphragm displacement, consequently large SPL. The alternative techniques generate 0.2 0.85 T with 0.5-mW input power. Among could generate them, the sputtered permanent magnet of 0.85 T and generate up maximum radial magnetic flux to 117-dB SPL on our MEMS loudspeaker. On the other hand, wax-bonded micromagnets could generate maximum radial of 0.3 T and generate up to 110-dB SPL magnetic flux on our MEMS loudspeaker. The alternatives can be assembled very precisely using a flip-chip bonding technique, eliminating misalignment issues contributing to the measurement uncertainty.

Authorized licensed use limited to: Arizona State University. Downloaded on October 8, 2009 at 19:55 from IEEE Xplore. Restrictions apply.

JE et al.: COMPACT AND LOW-COST MEMS LOUDSPEAKER

VII. CONCLUSION We present an electromagnetically actuated MEMS-based loudspeaker as a highly compact, low-power, low-cost technology suitable for improving the performance of CIC-type hearing aids. Our study shows that a prototype MEMS loudspeaker with 3-mm diameter can achieve large actuation and generate ample SPL in hearing aids. The Lorentz force is 37 N at the permanent-magnet plane, and the MEMS loudspeaker generates more than 0.8- m diaphragm displacement, equivalent to 106-dB SPL, with only 0.13-mW power consumption. The measurement uncertainty is less than 10%. Results of this research suggest that a compact-sized, low-cost, low-power MEMS loudspeaker with sufficient sound generation for hearing-aid applications is within reach. Replacing the macro-sized permanent magnet with a micro-size sol-gel permanent magnet is the next step in our research. The sol gel permanent magnet could be assembled very precisely by using a flip-chip bonding technique, which would eliminate misalignment issues that contribute to measurement uncertainty in this study. ACKNOWLEDGMENT The authors would like to thank Y. Yang and R. Steele of Knowles Electronics, and the staff at the Center for Solid State Electronics Research (CSSER) at Arizona State University for their help. REFERENCES [1] L. Squire, H. Schmolck, and S. M. Stark, “Impaired auditory recognition memory in amnesic patients with medial temporal lobe lesions,” Learning Memory, vol. 8, no. 5, pp. 252–256, 2001. [2] C. Duncan, “Psychophysics and psychology hearing,” Amer. J. Psychol., vol. 92, no. 2, pp. 377–379, 1979. [3] B. Moore, “Psychophysics of normal and impaired hearing,” British Med. Bull., vol. 43, no. 4, pp. 887–908, 1987. [4] J. Dianne, “Hearing loss, speech, and hearing aids,” J. Speech Hearing Res., vol. 36, pp. 228–244, 1993. [5] P. Kricos, S. Lesner, and S. Sandridge, “Expectations of older adults regarding the use of hearing aids,” J. Amer. Academy Audiol., vol. 2, pp. 129–134, 1991. [6] G. Gates and J. Mills, “Presbycusis,” Lancet, vol. 366, no. 9491, pp. 1111–1120, 2005. [7] J. Cohen-Mansfield and J. Taylor, “Hearing aid use in nursing homes. Part 2: Barriers to effective utilization of hearing aids,” J. Amer. Med. Dir. Assoc., vol. 5, no. 5, pp. 289–296, 2004. [8] B. Edwards, “The future of hearing aid technology,” Trends Amplification, vol. 11, no. 1, pp. 31–46, 2007. [9] R. Cox, “Assessment of subjective outcome of hearing aid fitting: Getting the client’s point of view,” Int. J. Audiol., vol. 42, pp. S90–S96, 2003. [10] L. Humes, D. Wilson, N. Barlow, and C. Garner, “Changes in hearing-aid benefit following 1 or 2 years of hearing-aid use by older adults,” J. Speech, Language, Hearing Res., vol. 45, pp. 772–782, 2002. [11] S. Kochkin, “10-year customer satisfaction trends in the U.S. hearing instrument market,” Hearing Rev., vol. 9, no. 10, pp. 14–25, 2002. [12] Hearing Ind. Assoc. Stat. Rep. Hearing Ind. Assoc. (HIA), Washington, DC, 2003. [13] J. H. Won, S. M. Schimmel, W. R. Drennan, P. E. Souza, L. Atlas, and J. T. Rubinstein, “Improving performance in noise for hearing aids and cochlear implants using coherent modulation filtering,” Hear Res., vol. 239, no. 1–2, pp. 1–11, 2008.

357

[14] N. A. Shusina and B. Rafaely, “Feedback cancellation for hearing aids fitted to open ear canals,” IEEE Trans. Audio, Speech, Language Process., vol. 14, no. 2, pp. 658–665, Mar. 2006. [15] M. Ross and D. L. Beck, “Expensive hearing aids: investing in technology and the audiologist’s time.” [Online]. Available: https://www. audiologyonline.com/articles/article_detail.asp?article_id=280 [16] S. Kochkin, “Customer satisfaction with hearing instruments in the digital age,” Hear J., vol. 58, no. 9, pp. 30–43, 2005. [17] S. Nordrum, S. Erler, D. Garstecki, and S. Dhar, “Comparison of performance on the hearing in noise test using directional microphones and digital noise reduction algorithms,” Amer. J. Audiol., vol. 15, pp. 81–91, 2006. [18] M. Valente, Hearing Aids: Standards, Options, and Limitations. : Thieme, 2002. [19] H. Francis, N. Chee, J. Yeagle, A. Cheng, and J. Niparko, “Impact of cochlear implants on the functional health status of older adults,” Laryngoscope, vol. 112, no. 8, pp. 1482–1488, 2002. [20] R. Ramsden, “Cochlear implants and brain stem implants,” British Med. Bull., vol. 63, no. 1, pp. 183–193, 2002. [21] M. Killion, Hearing aid transducers, Encyclopedia of Acoustics. New York: Wiley, 1997, ch. 16. [22] S. Lee, R. Ried, and R. White, “Piezoelectric cantilever microphone and microspeaker,” J. Micro Electromechan. Syst., vol. 4, pp. 238–242, 1996. [23] R. Heydt, R. Pelrine, J. Joseph, J. Eckerle, and R. Kornbluh, “Acoustical performance of an electrostrictive polymer film loudspeaker,” J. Acoust. Soc. Amer., vol. 107, no. 2, pp. 833–839, 2000. [24] M. Cheng, W. Huang, and S. R. Huang, “A silicon microspeaker for hearing instruments,” J. Micromechan. Microeng., vol. 14, pp. 859–866, 2004. [25] S.-S. Je and J. Chae, “An electromagnetically actuated micromachined loudspeaker for hearing aids applications,” in Proc. IEEE Sensors Conf., 2007, pp. 1024–1027. [26] M. Gad-el-Hak, MEMS Applications, 2nd ed. New York: Taylor & Francis Group, 2005. [27] R. Small, “Direct-radiator loudspeaker system analysis,” IEEE Trans. Audio Electroacoust., vol. 19, no. 4, pp. 269–281, Dec. 1971. [28] A. N. Thiele, “Loudspeakers in vented boxes: Part I,” J. Audio Eng. Soc., vol. 19, pp. 382–392, 1971. [29] NanoGage 100: Short range position & displacement sensor. [Online]. Available: http://www.optra.com/images/PS-NanoGage_100.pdf [30] D. G. Gata, W. Sjursen, J. R. Hochschild, J. W. Fattaruso, L. Fang, G. R. Iannelli, Z. Jiang, C. M. Branch, J. A. Holmes, M. L. Skorcz, E. M. Petilli, S. Chen, G. Wakeman, D. A. Preves, and W. A. Severin, “A 1.1-V 270- A mixed-signal hearing aid chip,” IEEE J. Solid State Circuits, vol. 37, no. 12, pp. 1670–1677, Dec. 2002. [31] Signalogic codecs. [Online]. Available: http://www.signalogic.com/ index.pl?page=codec_samples#samples [32] Magnet shop. [Online]. Available: http://www.magnetshop.com/materials.html [33] S.-S. Je and J. Chae, “A compact, low-power, and electromagnetically actuated microspeaker for hearing aids,” Electron. Device Lett., vol. 29, no. 8, 2008. [34] L. Lagorce and M. Allen, “Magnetic and mechanical properties of micromachined strontium ferrite/polyimide composites,” J. Micromechan. Syst., vol. 6, no. 4, pp. 307–312, 1997. [35] J. Lee, J. Park, Y. Oh, and C. Kim, “Magnetic properties of CoFe2O4 thin films prepared by a sol-gel methods,” J. Appl. Phys., vol. 84, no. 5, pp. 2801–2804, 1998. [36] N. Wang, B. J. Bowers, and D. P. Arnold, “Wax-bonded NdFeB micromagnets for microelectromechnical systsms applications,” J. Appl. Phys., vol. 103, p. 07E109, 2008. [37] Y. Zhang, W. Tang, G. C. Hadjipanayis, C. H. Chen, D. Goll, and H. Kronmuller, “Magnetic domain structure in SmCo 2:17 permanent magnets,” IEEE Trans. Magn., vol. 39, no. 5, pt. 2, pp. 2905–2907, 2003. [38] B. Kapitanov, N. Kornilov, Y. Linetsky, and V. Tsvetkov, “Sputtered permanent Nd-Fe-B magnets,” J. Magn. Magn. Mater., vol. 127, pp. 289–297, 1993. [39] T. Speliotis and D. Niarchos, “Deposition of hard magnetic SmCo5 thin films by magnetron sputtering,” J. Phys., Conf. Ser. 10, pp. 175–177, 2005. [40] T. Budde and H. Gatzen, “Thin film SmCo magnets for use in electromagnetic microactuators,” J. Appl. Phys., vol. 99, no. 8, p. 08N304, 2006.

Authorized licensed use limited to: Arizona State University. Downloaded on October 8, 2009 at 19:55 from IEEE Xplore. Restrictions apply.

358

IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS, VOL. 3, NO. 5, OCTOBER 2009

Sang-Soo Je received the M.S. and Ph.D. degrees in electrical engineering from the Georgia Institute of Technology, Atlanta, and Arizona State University, Tempe, in 2000 and 2009, respectively. Currently, he is a Research Engineer with Silicon Audio, LLC, Austin, TX, developing optical microphones. His research interests are microelectromechanical-systems (MEMS) sensors/actuators, nanotechnology, and RF MEMS integrated with barium strontium titanate (BST) material.

Fernando Rivas was born in Santander, Spain. He received the Ph.D. degree in physics from the University of Cantabria (UC), Cantabria, Spain, in 1994. From 1990 to 1997, he was a Researcher with the Electronics Department of the UC. Since 1997 he has been at the University of Jaén, Jaén, first as Assistant Professor and since 2003 as an Associate Professor. In 2007, he was a Researcher with the Electrical and Engineering Department at Arizona State University, Tempe. His research interests include areas in lowfrequency methods in electromagnetic and onboard antennas analysis.

Rodolfo E. Diaz received the B.S. degree in physics from Yale University, New Haven, CT, the M.S. degree in physics from the University of California at Los Angeles (UCLA), and the Ph.D. degree in electrical engineering from UCLA. During 20 years in the aerospace industry, his work spanned many of the disciplines comprising modern electromagnetic engineering, from lightning protection, electromagnetic compatibility on the Space Shuttle, through the design of broadband missile antennas and radomes, to the design, evaluation, and prototyping of electromagnetic composite materials for low observable applications. In 1998, he joined Arizona State University, Tempe, where he is currently Associate Professor in the Department of Electrical Engineering. He directs the Material-Wave Interactions Laboratory, performing research on natural and engineered materials. Prof. Diaz holds 19 patents ranging from the design of broadband radomes to the amplification of magnetic fields.

Jiuk Kwon received the M.Sc. degree from Arizona State University, Tempe, in 2006. His research interests include Class D audio amplifiers, data converters, and digital frequency discriminators.

Jeonghwan Kim is pursuing the M.Sc. degree in electrical engineering, first at Arizona State University and now at Louisiana State University, Baton Rouge. His research interests include microelectromechanical systems (MEMS), acoustic devices simulation, and characterization of the MEMS acoustic devices.

Bertan Bakkaloglu (M’94–SM’08) received the Ph.D. degree from Oregon State University, Corvallis, in 1995. He joined Texas Instruments, Inc. Mixes Signal Wireless Design Group, Dallas, working on analog, RF, and mixed-signal front ends for wireless and wireline communication integrated circuits (ICs). He worked on system-on-chip designs with integrated battery management and analog baseband functionality as a design leader. In 2004, he joined the Electrical Engineering Department, Arizona State University, Tempe, as an Associate Professor. His research interests include radio-frequency (RF) and power-amplifier supply regulators, RF synthesizers, biomedical and instrumentation circuits and systems, high-speed RF data converters, and RF built-in self-test circuits for communication ICs. Dr. Bakkaloglu has been a Technical Program Chair and Steering Committee Member for the IEEE RFIC Conference and an Associate Editor of the IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS.

Sayfe Kiaei (F’02) was a Senior Member of the Technical Staff with the Wireless Technology Center and Broadband Operations at Motorola from 1993 to 2001, where he was responsible for the development of wireless transceiver integrated circuits (ICs) and digital subscriber line transceivers. Before joining Motorola, he was an Associate Professor at Oregon State University, Corvallis, from 1987 to 1993, where he taught courses and performed research in digital communications, very-large scale integrated system design, advanced complementary metal–oxide semiconductor IC design, and wireless systems. Currently, he is a Professor and the Director of the Connection One Center (National Science Foundation I/UCRC Center) at Arizona State University, Tempe. Currently, he Associate Dean of Research at Arizona State University. Dr. Kiaei is a member of IEEE Circuits and Systems Society, IEEE Solid State Circuits Society, and IEEE Communication Society. He has been on the technical program committee and/or Chair of many conferences, including: RFIC, MTT, ISCAS, and other international conferences. He has published many journal and conference papers and holds several patents. His research interests are in wireless transceiver design, and RF and mixed-signal ICs in complementary metal–oxides semiconductor and silicon gernanium SiGe. Prof. Kiaei is the recipient of the Carter Best Teacher Award, Oregon State College of Engineering; IEEE Darlington Award; and Motorola 10X Design award.

Junseok Chae received the B.S. degree in metallurgical engineering from the Korea University, Seoul, Korea, in 1998, and the M.S. and Ph.D. degrees in electrical engineering and computer science from the University of Michigan, Ann Arbor, in 2000 and 2003, respectively. After a couple of years of being a Research Fellow at Michigan, he is now an Assistant Professor of electrical engineering at Arizona State University, Tempe. His areas of interests are MEMS sensors/actuators, MEMS integration with electronics, micropackaging, and bio-MEMS. He received the 1st place prize and the best paper award in DAC (Design Automation Conference) student design contest in 2001. He has published over 55 journal and conference articles, two book chapters, and holds two US patents. He is a NSF CAREER awardee on a MEMS protein sensor array.

Authorized licensed use limited to: Arizona State University. Downloaded on October 8, 2009 at 19:55 from IEEE Xplore. Restrictions apply.