Model-based design of artificial zero power cochlear implant

Mechatronics xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

Mechatronics journal homepage: www.elsevier.com/locate/mechatronics

Model-based design of artificial zero power cochlear implant Jaromír Zˇák a,b, Zdeneˇk Hadaš a,c, Daniel Dušek a,c, Jan Pekárek a,b, Vojteˇch Svatoš a,b, Ludeˇk Janák a,c, Jan Prášek a,b,⇑ a

Central European Institute of Technology, Brno University of Technology, Technická 10, 616 00 Brno, Czech Republic SIX Research Centre, Brno University of Technology, Technická 10, 616 00 Brno, Czech Republic c Faculty of Mechanical Engineering, Brno University of Technology, Technická 2, 616 69 Brno, Czech Republic b

a r t i c l e

i n f o

Article history: Received 30 September 2014 Revised 13 March 2015 Accepted 27 April 2015 Available online xxxx Keywords: Biomechatronics MEMS sensor Energy harvesting Zero power sensor Model-based design Signal processing

a b s t r a c t This paper deals with a model-based design of an autonomous biomechatronic device for sensing and analog signal processing of acoustic signals. The aim is to develop a biomechatronic artificial cochlear implant for people with hearing loss due to damage or disease of their cochlea. The unique artificial electronic cochlear implant is based on an array of microelectromechanical piezoelectric membranes. Oscillations of membranes detect and filter acoustic signals in individual acoustic frequencies. The proposed biomechatronic device of the artificial cochlear implant consists of an active filters array, signal processing electronics, stimulation nerves electrodes and energy harvesting system for autonomous powering of the device. This solution differs from current cochlear implants solutions, which are bulky electronic systems limited by their high power consumption. The multidisciplinary models of the artificial cochlea implant concept are presented. The mechatronic approach based on model seems to be very useful for development of the full implantable cochlear implant which is designed for the sensing and processing of acoustic signals without external energy source. Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction The cochlea is a part of human ear, where acoustic signals from outer air space are transferred to electrical signals and then processed by the brain. The loss of hearing is frequently caused due to damage or disease of cochlea [1–5] and in this cases a cochlear implant is the only possibility to recover hearing, at least partially. The current cochlear implants are complex electronic systems which are composed of outer and inner parts. The outer part is placed outside the head and it consists of the following subsystems:  A microphone to receive the acoustic waves from environment and transfer them into electrical signals.  A speech processor to decompose the signals into simple frequency components.  A transmitter to transmit the electrical signals from the outer part into the inner part of cochlear implant.  A battery for power supplying the cochlear implant.

⇑ Corresponding author at: SIX Research Centre, Brno University of Technology, Technická 10, 616 00 Brno, Czech Republic. E-mail address: [email protected] (J. Prášek).

The inner part is permanently implanted inside of the head and consists of following subsystems:  A receiver to receive the signals from the outer part of the cochlear implant.  Stimulation electrodes to connect to the hearing nerve of brain. Biomechatronic devices such as artificial cochlear implants are developed using modern microelectronic techniques, although there are still some technical issues which need to be solved. For instance, non-full implantability of common cochlear implants is limiting for patients (nowadays, there are only few prototypes of full implantable cochlear implants in the literature [6]) and low number of electrodes are used to reduce energy consumption. Average total power consumption of the implants is around 10 mW and batteries have to be recharged every 12–24 h of operation (depending on stimulation strategies) [7]. Our research presents a unique concept of artificial cochlear implant, which solves the problem of full implantability of this device and power consumption. The proposed artificial cochlear implant is based on a microelectromechanical system (MEMS), which can be fully implantable and has very low power consumption. The main aim of this work is to develop an energy autonomous device, which uses ambient energy in the head area for

http://dx.doi.org/10.1016/j.mechatronics.2015.04.018 0957-4158/Ó 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Zˇák J et al. Model-based design of artificial zero power cochlear implant. Mechatronics (2015), http://dx.doi.org/10.1016/ j.mechatronics.2015.04.018

2

J. Zˇák et al. / Mechatronics xxx (2015) xxx–xxx

self-powering and without the use of external power supply. Several potential sources of ambient energy are analyzed and evaluated in this paper in order to obtain the self-powered cochlear implant. The use of ultra-low power electronics, an active MEMS sensing and a signal processing is required in case that the energy harvesting system have to supply the whole artificial cochlear implant. The cochlear biomechatronic device employing the active MEMS sensing, ultra-low power signal processing and energy harvesting power supply is presented and evaluated in this paper. Fig. 1. Comparison between passive and active basilar membrane.

2. State of art 2.1. Cochlear implant Nowadays cochlear implants use the standard microphones in combination with digital speech processors [8,9]. The disadvantage of standard microphones is their unselectivity for particular frequencies needed for particular nerves stimulation. Therefore the high performance speech processors have to select a desired frequency spectrum. The physical dimensions of the common microphones and their power consumption is not suitable for full implantation. Nowadays, dimensions reduction can be achieved by modern biomedical devices. The miniaturized MEMS microphones have been already developed but the frequency selectivity cannot be achieved yet and for this reason the speech processing is employed. The speech processor decomposes audio signal to the frequency spectrum and each frequencies are processed separately. The total decomposed signal is cumulated until the threshold value of signal power is reached and the nerve is excited by an electric pulse generated by processor. The processor has to prevent interferences between electrodes in the inner ear in case that more signals reaches the threshold value in the same time [10]. Excited signals can be generated synchronously with the defined timing or asynchronously with the priority of the signal with the higher signal power amplitude. A bank of selective filters created by a MEMS acoustic sensor provides solution without the speech processor usage. Some of current MEMS devices are designed as an array of microcantilevers [11,12] which are not suitable for sensing ambient acoustic pressure like the human cochlea. The micro-membrane represents a much more promising shape of resonator for this purpose. The trapezoidal shapes of membranes are developed by other scientific groups involved in artificial cochlear implants development [13– 15]. The trapezoidal shape of the membrane is similar to a basilar membrane which is placed inside the inner ear. The basilar membrane operates as natural mechanical filter where high frequencies excite the membrane on its basal end and low frequencies excite the membrane on its apical end [16]. The real basilar membrane is a highly nonlinear active system [17] which is impossible to be imitated by current MEMS technologies. The comparison between the passive and active basilar membrane vibration is shown in Fig. 1. There is a clearly visible peak in the case of using the active basilar membrane which enables much more accurate resolution in desired frequency range. The aim of our research group is design, develop and fabricate the complex biomechatronic device which respects all patient aspects and allows miniaturization and integration of the cochlear implant into the autonomous MEMS device, which can be fully implantable in the head. In comparison with the trapezoidal membrane an array of micro-membranes as frequency filters with different dimensions provides the appropriate signal decomposition [18]. The basic principle of the signal decomposition by the array of membrane resonators is shown in Fig. 2. The basilar membrane was simulated as the array of connected resonators by springs with

stiffness Ks. A model without connections between resonators achieves better results. The connected springs cause the vibrations of neighbor resonators which are visible in spectrogram as additional false lines. A low mechanical damping of the resonators is inapplicable for the decomposition of non-stationary signals like speech or common ambient sounds because the spectrogram in this case is smudged due to long transient vibrations of resonators.

2.2. Energy harvesting system for biomedical devices Several types of energy harvesting systems are currently developed as energy source for biomedical applications [19–22]. Based on the initial analyses and experience with energy harvesting systems, three fundamental types of energy converters in the head area can be considered: thermal gradient, mechanical movement (shocks) and bending movement of neck muscles or arteries in the head area.

2.2.1. Thermoelectric energy conversion A human body produces waste heat as a result of basal metabolism. This waste heat might be converted into useful electric power by thermoelectric energy conversion based on the Seebeck effect. Temperature difference at the junction of two dissimilar materials produces the electromotive force [23]. Conversion of the small amounts of thermal energy into electricity is called thermoelectric energy harvesting. The common thermoelectric energy harvester consists of thermoelectric module (TEM), power management electronics and thermomechanical integration components. The main thermomechanical integration is usually ensured using the heat sink and heat source with heat spreader. Electric energy generated by the thermoelectric generator is significantly dependent on properties of the materials used in the thermoelectric module. An overview of the physics behind thermoelectric energy harvesting was described in detail previously [23]. Utilization of thermoelectric conversion for the powering of autonomous biomedical implants is a subject of several studies [24–27]. An example of recent effort in the field of bio-implantable thermoelectric generators can be found in the power supply for artificial accommodation system. The proposed device forms a part of the more complex micro-mechatronic system based on adaptive artificial lens. The power consumption of such complex system varies from 5 lW during standby mode up to several mW during the lens actuator operation [26]. A significant development is observed in the field of wearable thermoelectric devices [25]. Measurements performed with 200 mV output using the off-the-shelf TEMs showed the harvested electric power levels of 100–200 lW. Nevertheless, these devices are bulky [27] and generally not applicable for the biomedical implants. Miniaturization of TEMs to the MEMS scale is an issue of contemporary development.

Please cite this article in press as: Zˇák J et al. Model-based design of artificial zero power cochlear implant. Mechatronics (2015), http://dx.doi.org/10.1016/ j.mechatronics.2015.04.018

J. Zˇák et al. / Mechatronics xxx (2015) xxx–xxx

3

Fig. 2. Diagram of the signal decomposition principle based on the array of resonators.

MEMS TEM for bioapplications requires high density array of thermocouples connected electrically in series and thermally in parallel. This configuration allows harvesting of a sufficient voltage even at the low temperature differences. Another issue can be observed in the optimal design of heat flow path – heat sink and heat spreader. Both thermal and electric circuits have to be designed to maintain the matched load conditions. Such conditions ensure the maximum power transfer in both thermal and electric domains [28,29]. TEM for biomedical applications should be easily implantable, flexible, biocompatible, reliable and durable. Biocompatibility can be ensured using the additional biocompatible encapsulation. 2.2.2. Energy harvesting from the mechanical movement The energy generated from the mechanical movement can be harvested by several physical principles of electromechanical conversions: piezoelectric, electromagnetic, electrostatic and magnetostriction physical principle. The mechanical energy harvester is usually based on a resonance mechanism which is excited by ambient energy of mechanical movements [30]. The movement can be in a form of vibrations (usually in engineering applications) or in a form of body shocks (acceleration). The inertia forces provide a relative movement of a seismic mass in the resonance mechanism. This relative movement is converted by any of the physical principle of the electromechanical conversion. The electromagnetic principle represents the promising way of autonomous power source in several medical applications. The energetic analyses of differential equation of the mechanical energy harvester were discussed earlier [31,32]. The model analyzed in Ref. [31] corresponds with lab results and it can be used in technical engineering projects. However, the human body reacts against the relative oscillation inside an energy harvesting resonance mechanism. The human muscles and skeleton provide active damping forces. These forces affect ordinary motion equations and make them inapplicable for power analyses. Unfortunately, several

papers [33,34], do not reflect this fact and presented power analyses are therefore influenced. The damping effect can be neglected for very low weight of the resonance mass as it has been described in a simple experiment reported before [35]. Based on the power analyses [31] seismic mass for electromagnetic conversion has maximal weight around 10 g. 3. Motivation and used approach A presentation of the fully implantable artificial cochlear implant development is the main aim of this paper. Therefore the design of an active acoustic MEMS sensor with very low power consumption and choosing of an energy harvesting system for powering of this ultra-low power sensor without any external energy source is required. This complex artificial cochlear structure has not been studied yet and it represents a unique approach in the field of the cochlear implants development. Fundamental idea of an acoustic signal detection is the full implantable cochlear implant which has to be placed inside of the middle ear space and directly connected to the ossicular chain or to the eardrum by some of the middle ear prosthesis (e.g. from Heinz Kurz GmbH Medizintechnik, Germany). This implant is designed as MEMS sensors array. It consists of several individual active membrane with piezoelectric elements, which are tuned up in the range of hearing frequency. This MEMS sensors array represents a bank of active electromechanical filters as a sensing system for detection of dominant acoustic frequencies. Active MEMS sensors are actuated by acoustics waves, which are propagated through a fluid medium by vibration of the middle ear prosthesis. The wave propagation is similar like in the human cochlea. The full encapsulation of the MEMS sensors array eliminates problems with biocompatibility. A schematic configuration of the proposed fully implantable cochlear implant is shown in Fig. 3. The developed device is implanted inside of the middle ear cavity. The energy harvesting

Please cite this article in press as: Zˇák J et al. Model-based design of artificial zero power cochlear implant. Mechatronics (2015), http://dx.doi.org/10.1016/ j.mechatronics.2015.04.018

4

J. Zˇák et al. / Mechatronics xxx (2015) xxx–xxx

Fig. 4. Basic scheme of full implantable MEMS sensors array.

Fig. 3. The basic scheme of proposed design of biomechatronic artificial cochlear implant.

system can be mounted behind the ear inside of the skull where a receiver of currently used implants is placed. The developed cochlear implant is designed as the array of piezoelectric membranes (individual filters) with different dimensions to achieve different Eigenfrequencies. They correspond with our published mathematical models [31]. The presented results in the developed configuration achieve better results in quality of acoustic signal decomposition than other configurations used by scientific groups [14,15]. Vibrations of membranes are sensed by a piezoelectric element which operates as an active sensor for decomposition of the input excited acoustic signal. It seems to be a great advantage compared with common used speech processors which are based on energetically expensive mathematical decomposition of the acoustic signal. For this reason it is possible to use the ultra-low power energy harvesting system for powering of the artificial cochlear implant. Our design of the cochlear implant uses a fluid medium to transfer the acoustic pressure from the middle ear to the MEMS sensors array similar like the human ear biomechanics (see Fig. 4). The membranes models without the fluid medium were considered to simplify calculations. The influence of the fluid medium on membrane excitation is going to be studied in our future work. Common cochlear implants use range of 12–24 channels. Our design of the MEMS cochlear implant has 24 active MEMS sensors. Each MEMS sensor provides one channel of decomposed signal with own electronics for sensing, processing and electrode nerves excitation. 4. MEMS sensor model 4.1. Concept of MEMS sensor design The developed MEMS sensor uses the active piezoelectric principle for transformation of mechanical vibrations into electrical signals. We suppose to use a thin piezoelectric aluminum-nitride layer on the membrane as a suitable piezoelectric element on the edge of each membrane. The specific dimensions and geometry of the membrane with the piezoelectric element will provide individual channel of the mechanical frequency filter in our design. 4.1.1. MEMS manufacturing The fully implantable cochlear sensors array with 24 individual MEMS sensors is produced by standard micromachining techniques. The process flow of the MEMS sensor fabrication is schematically shown in Fig. 5. The fabrication process starts with the deposition of SiXNY isolation layer on the silicon substrate (step

1). Next step is the deposition and patterning of the AlN piezoelectric layer (step 2). The AlN film with a thickness of 100 nm is deposited by reactive sputtering from an Al target. The CMOS compatible and relatively easy processing of AlN is a major advantage compared with the more commonly used piezoelectric material PZT. The sputter deposition of Au layer for the top electrode finishes the piezoelectric elements fabrication (step 3). On the real samples the Au electrodes will be connected to the bonding pads located on the edge of each MEMS sensor and then wire-bonded to a signal processing electronics. On the backside of the wafer, the SiXNY is patterned as protective mask for KOH wet etching. The membrane is released with KOH etching and the complete release (step 4) is done using top SiXNY etch stop (membrane layer) due to good etch selectivity toward Si and SiXNY. The thickness of the membrane is defined by the thickness of SiXNY layer.

4.1.2. FEM model of piezoelectric MEMS sensor The MEMS sensor represents individual mechanical filter. It consists of a SiXNY square membrane and four piezoelectric AlN elements paced in the edge of the membrane with top Au electrodes for electric signal measurement. The piezoelectric layer is an active part of the sensor which is responsible for transformation of the mechanical strain into an output electric signal. The electric signal is gained by piezoelectric effect of bended AlN layer. This layer has poling axis in orthogonal direction from fabrication process. The MEMS sensor operates in a piezoelectric mode 33. Generally, electrode configuration of a 33 mode is in a form of an interdigital electrode pattern. Due to the piezoelectric element dimensions the sensor model includes only double electrode pattern on the top of piezoelectric layer. This topology is shown in Fig. 6. Vibrations of the membrane and mechanical strain of the active layer generate an electrical potential between the Au electrodes. The model in ANSYS environment was used for analyses of mechanical membrane with piezoelectric layer and electrodes behavior. The membrane dynamics and output electrical signal are calculated. Each of 24 MEMS sensors (see Fig. 6) will be designed and analyzed in ANSYS and finally tested and verified with real samples. The model of the square membrane with dimensions of 0.5 mm  0.5 mm is shown in Fig. 7. Simulation results of this model are discussed in the following subsection.

4.1.3. Results of the piezoelectric MEMS sensor Modal analysis of the MEMS sensor was done for all 24 membranes dimensions as mechanical filter channels. The membrane model with dimensions of 0.5 mm  0.5 mm was chosen for results presentation. The thickness of 100 nm for SiXNY membrane and AlN piezoelectric layer was used in this simulation (see Fig. 8). The AlN piezoelectric layer increases the stiffness of the SiXNY membrane which leads to increasing of membrane Eigenfrequencies. This fact and change of membrane dimensions can be used for tuning up all 24 mechanical filters in audible spectrum.

Please cite this article in press as: Zˇák J et al. Model-based design of artificial zero power cochlear implant. Mechatronics (2015), http://dx.doi.org/10.1016/ j.mechatronics.2015.04.018

J. Zˇák et al. / Mechatronics xxx (2015) xxx–xxx

5

Fig. 5. The process flow of the MEMS cochlea.

Fig. 6. Model of MEMS sensor – membrane with piezoelectric element and gold electrodes operating in mode 33.

Fig. 7. The ANSYS model of MEMS sensor with dimensions of membrane 0.5 mm  0.5 mm.

The harmonic analyses were calculated for all 24 membranes. Membranes were excited by acoustic pressure and the electric signal response was calculated. The harmonic analysis of the presented membrane of 0.5 mm  0.5 mm in operation mode 33 is shown in Fig. 9. The maximum amplitude of the electric potential 0.38 V on the Au electrodes was obtained. The electric potential is affected by the value of the excited acoustic pressure. For this reason several harmonic analyses in frequency domain for several sound pressure levels are calculated and shown in Fig. 9. The results confirmed that the presented membrane with piezoelectric elements can be used for active mechanical filter in the individual sound channel of our developed sensors array for the artificial cochlear implant. 4.2. Design of signal processing electronics The main purpose of the developed electronic circuits is to amplify and transform a low electrical signal, which is generated

by the piezoelectric elements on the MEMS sensor membrane, into an electrical signal, which is distributed to the implanted nerve electrodes. These circuits have to respect several special demands like the ultra-low power consumption, which allows to power the implant by energy harvesting power sources. Then it is possible to develop the fully autonomous and implantable artificial cochlear implant. Another demand is high gain of the circuits due to small voltage signal levels, which are generated by active piezoelectric elements (with respect to common ultra-low acoustic signal pressure levels which are below 50 dB). In accordance with these limitations, the new topology for an ultra-low power MEMS sensor signal processing electronics was designed (see Fig. 10). The MEMS sensor signal processing electronics cannot use standard techniques for the power consumption minimizing such as sleep modes. It can cause a missing of some short time sounds when the sleep mode is active. The main idea of how to decrease processing electronics power consumption is based on useful charge push through electronic circuits directly into implanted cochlear electrodes from power storage element. All charge drawn from energy harvesting generator circuits is delivered into implanted nerve electrodes in the ideal case. The indispensable currents which flow to ground outside of the output nerve electrode is essential but the values of these currents have to be decreased to minimal possible value. The output signal from the processing electronics (see inset in Fig. 10) to the implanted nerve electrodes is structured as bipolar current pulses. The amplitude of these pulses is constant and time delay between them depends on integration of input signal amplitude in the time [36]. The differential input signal is generated directly by MEMS sensor connected to DC offset voltage. This signal is amplified and converted to current value (I1) by an operational transconductance amplifier (OTA) [37]. The integration of the input signal in the time is done by a charge accumulation in CINT. When the accumulated charge reaches the threshold value defined by VREF, output monostable MOSFET buffer connects accumulated internal charge to the output through the separating capacitor CBUF. A portion of the accumulated charge is discharged by I2 through implanted nerve electrodes where positive part of the pulse (I2) is generated. Pulse amplitude is defined by PMOSFET RDSON and total transported charge is defined by controlling comparator hysteresis. The output current I2 also charges the output buffer capacitor CBUF during the positive current pulse generation. Negative I3 current pulse part is generated by the CBUF capacitor discharging when MOSFET output buffer is connected to the

Please cite this article in press as: Zˇák J et al. Model-based design of artificial zero power cochlear implant. Mechatronics (2015), http://dx.doi.org/10.1016/ j.mechatronics.2015.04.018

6

J. Zˇák et al. / Mechatronics xxx (2015) xxx–xxx

Fig. 8. The first Eigenfrequency 9724 Hz and shape of calculated piezoelectric layers on membranes.

Fig. 11. Result of the output signal pulse simulation. Fig. 9. Harmonic analysis of active MEMS sensor.

ground in its stable state. Simulation result of output signal pulse into implanted nerve electrodes is shown in Fig. 11. The bipolar pulse generated using the technique mentioned in previous paragraph is totally charge balanced. The major part of the current from OTA output is used for nerve stimulation by implanted nerve electrode. This method can minimize power

consumption of this device to the lowest value. Only one condition has to be fulfilled. The CBUF capacity has to be greater than CINT to prevent CBUF saturation. The total transconductance of the amplifier was fitted to optimal performance for the maximal simulated voltage change of the sensing element DV = 0.38 V calculated in Section 4.2. The final simulated dependence of the pulse generation frequency on input acoustic pressure level is shown in

Fig. 10. Schematic block of the MEMS sensor signal processing electronics.

Please cite this article in press as: Zˇák J et al. Model-based design of artificial zero power cochlear implant. Mechatronics (2015), http://dx.doi.org/10.1016/ j.mechatronics.2015.04.018

J. Zˇák et al. / Mechatronics xxx (2015) xxx–xxx

Fig. 12. This result was obtained from the final model simulation of the whole MEMS sensor signal processing electronics. Total power consumption of the cochlear implant mainly consists of OTA output (useful) current I1, comparator and OTA supply current, driving current into MOSFET output driver and self-discharge current in capacitors. The maximal expected power consumptions of circuit elements are shown in Table 1. The total power consumption per each channel measurement is about 10 lW at 2 V power supply. The final power consumption optimization for an on-chip realization is necessary in case of more channels will be used. The final power consumption is expected to be lower than the simulation results mentioned in Table 1. The precise power consumption will be specified after future simulations with selected manufacturing microelectronics technology. The accurate simulation of the complex electronic circuit including the energy harvesting generator, power management circuit and MEMS circuits is the aim of our future development. 5. Energy harvesting for artificial cochlea 5.1. Choice of the energy harvesting concept Both thermoelectric and electro-magnetic concepts of energy harvesting can be useful for the developed artificial cochlear implant. The thermal gradient around 5 °C is expected during most of the operation time. Furthermore the cochlear implant can employ the energy harvesting from mechanical energy of a body motion. The mentioned ways of autonomous powering of developed artificial cochlear implant are analyzed on the base model and presented in this paper. The combination of harvesting methods is expected to be used for our application. This solution can provide the zero power cochlear implant, where all consumed energy is harvested from ambient energy of the human head. The proposal of energy harvesting parameters has to be in hand with the design of the signal processing electronics for nerves excitations. The model-based design of energy harvesting systems can provide useful method for the development of this zero power device. The indisputable advantage is the use of individual models of the developed cochlear implant in time domain, where these models can be connected in co-simulation strategy. Whole model can be also used for tuning up and optimization study of this zero power cochlear implant.

7

Fig. 12. Dependence of the output signal pulse frequency on the acoustic pressure level.

Table 1 Expected power consumption of particular electronic subsystems for 2 V power supply. Circuit element

IEXPECTED [lA]

Current type

OTA supply Comparator supply Average useful current MOSFET driving

1–2 <2 0.2 <1

Continuous Continuous Pressure dependent Pulse

Total power consumption

<6

Average

5.2. Thermoelectric analyses A thermal network is appropriate to the above-mentioned configuration which is depicted in Fig. 13. Constant temperature sources of body core temperature (TbodyIN) and ambient environment temperature (Tambient) are connected to the thermal ground. The thermal network consists of the thermal resistance of skull (Rskull), resistance of skin (Rskin) and resistance of thermal convection from skin surface to ambient air (Rconvection). The thermoelectric generator itself is composed of several thermal resistances – thermal resistance of the thermoelectric module (RTEM), thermal resistance of heat sink (Rsink) and heat source (Rsource). The useful temperature difference for the thermoelectric energy conversion is DT. Two basic configurations of thermoelectric module (TEM) placement are considered:  Placement under the skin but outside the skull (variant 1 – dashed arrow number 1 in Fig. 13)  Placement completely outside the head – TEM is attached to the patient’s skin (variant 2 – dashed arrow number 2 in Fig. 13)

Fig. 13. Thermal network of the head thermoelectric generator for the zero power cochlear implant.

The thermal resistance of human tissues between a body core and skin surface changes in time with human physical activity, age, or health conditions [27]. Reference values for thermal conductivity based on the previous research presented in [38] and the approximate thicknesses of particular tissues on the human head are listed in Table 2. These values are used for the simulation based on the model in Ref. [39]. Heat transfer parameters of human tissues are very similar to thermal insulators such as plastics. The implantable thermoelectric generator lies in the optimal design of its heat path. Optimal or more preferably adaptive configuration of the thermoelectric module of the thermal circuit is needed. 5.2.1. Thermoelectric analyses The recoverable amount of waste heat is essential during the initial phase of a thermoelectric module design. The estimated

Please cite this article in press as: Zˇák J et al. Model-based design of artificial zero power cochlear implant. Mechatronics (2015), http://dx.doi.org/10.1016/ j.mechatronics.2015.04.018

J. Zˇák et al. / Mechatronics xxx (2015) xxx–xxx

8 Table 2 Thermal properties of human tissues [38].

Table 3 Available power densities using thermoelectric materials with various ZT’s.

Type of tissue

Thermal conductivity [Wm1 K1]

Thickness [mm]

ZT [–]

Power density [lWmm2]

Skull Skin Fat Muscle Blood Cartilage

0.32 0.37 0.21 0.49 0.52 0.49

7 1.5 4.5 7.5 – –

0.5 1 2

3.28 3.49 3.77

Table 4 Technology boundaries for bio-thermoelectric energy harvesting on the human head.

waste heat produced in the region of human head varies around 15 W excluding the sweat evaporation [38]. This waste heat is normally dissipated into an ambient environment by the means of thermal convection. The assumed head surface area is 0.1414 m2 [38], the average heat flux density through the head skin surface is 106 Wm2 [22]. The previous estimation of the available waste heat can be advantageously used for the computation of harvestable electric power. The basic idea consists in the Carnot engine character of a thermoelectric generator. The absolute boundary of TEM efficiency is given by the Carnot’s theorem for heat engines. This efficiency is further reduced by the material characteristics of the particular thermoelectric module. Material characteristics are covered under the dimensionless figure of merit ZT:

ZT ¼

a2  T ; qk

ð1Þ

where a is the Seebeck coefficient, k is the thermal conductivity and q the electrical resistance of material forming the particular TEM. Higher figure of merit indicates the higher suitability of material for the thermoelectric energy conversion. As could be easily proven from ZT expression, ideal thermoelectric material should have high Seebeck coefficient, low thermal conductivity and low electrical resistance. Increase of the figure of merit is the task for the microstructured materials, superlattices, nanomaterials, low-dimensional materials and thin films. Combination of the above mentioned considerations, the electric power density PdTEM related to the skin surface area A yields in:

PdTEM ðAÞ ¼

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Th  Tc Q 1 þ ZT  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi T c  head ; Th 1 þ ZT þ T Ahead

ð2Þ

h

where Th and Tc are temperatures related to hot and cold sides of the thermoelectric module. Further evaluation of the skin surface power density equation is given in Table 3. The hot side temperature was related to the human body core temperature of 37 °C. The cold side temperature was set to 20 °C. The results are listed for various values of ZT providing the overview of the worst case (ZT = 0.5), advanced bulk material (ZT = 1) and nanostructured thermoelectric material (ZT = 2) scenario. Calculated results for the particularly considered sizes of thermoelectric modules are stated in Table 4. It is evident that the thermoelectric module with a reasonable size of 10 mm  10 mm (100 mm2 footprint) is able to provide hundreds of microwatts of electricity. 5.2.2. Thermoelectric module results The ultimate technology boundaries in the thermoelectric energy harvesting for the zero power cochlear implant were investigated. The absolute maximum of the useful power from the thermoelectric energy harvesting on the human head was determined to be 1500 lW (see Table 4). The proposed TEM would occupy the footprint of 400 mm2. The reasonable dimensions of the thermoelectric generator of 100 mm2 can be considered. Accordingly to

Dimensions [mm  mm]

Footprint [mm2]

55

ZT [–]

Generated power [lW]

25

0.5 1 2

82 87 94

10  10

100

0.5 1 2

328 349 377

20  20

400

0.5 1 2

1312 1396 1508

the realistic idea the TEM with ZT = 1 can produce up to 350 lW power. Then the power budget (see Table 1) of zero power cochlear implant is met. However, serious issues are expected during the research and development of optimal thermoelectric generator and its matched thermal network. This uneasy issue will be solved using the advantages of simulation modeling and model-based design based on our previous work [39].

5.3. Electromagnetic energy harvesting system The principle of developed electro-magnetic conversion is shown in Fig. 14. The second order mechanical differential equations of moving mass (3) is ordinary used for solving displacement x, velocity x_ and acceleration € x of the mass m:

m€xðt Þ þ bm x_ ðtÞ þ kxðt Þ ¼ m€zðtÞ:

ð3Þ

The system is excited by head acceleration and the mechanism operation is effected by mechanical damping bm and stiffness of the mechanism k. The harvesting of electrical energy provides dissipation of energy, which provides feedback by electro-mechanical damper be. This electromagnetic feedback is solved by expression: 2

be ¼

ðBx  N  lÞ ; RC þ RL

ð4Þ

where N is number of coil turns, l is active coil length, Bx is magnetic flux through coil area, RC is resistance of coil and RL is resistance of an electrical load. Due to the assumed printed air coil the impedance of the coil model is neglected in this equation. On the base of Faraday’s law (5) the voltage is induced in the coil and on the base of Kirchhoff’s laws the current flow through electrical load and the harvested power is consumed.

ui ¼

I  ! ~ x_  B  d~ r

ð5Þ

C

This simple theoretical model of the electromagnetic energy harvesting can be solved using the advantages of simulation modeling and model-based design of this mechatronic system as it is presented in [40].

Please cite this article in press as: Zˇák J et al. Model-based design of artificial zero power cochlear implant. Mechatronics (2015), http://dx.doi.org/10.1016/ j.mechatronics.2015.04.018

J. Zˇák et al. / Mechatronics xxx (2015) xxx–xxx

9

Fig. 14. Energy harvesting principle from human body motion.

5.3.1. Model of electromagnetic energy harvester Based on assumed power consumption and volume of mounted place in head area the concept of electromagnetic energy harvester was designed. This concept consists of mechanical flexible element, seismic mass and fixed printed coil in harvester frame. The movable seismic mass consists of magnetic circuit of permanent magnets and ferromagnetic steel poles extensions, which can move against fixed printed coil. This movement induces voltage. The proposed concept is shown in Fig. 15. This model has expected volume around 3 cm3 and total weight of movable mass 10 g. A design of flexible elements provides stiffness of the energy harvester and determines natural frequency. An operation of this harvester is based on excitation by ambient acceleration of mechanical shocks and the relative movement of the magnetic circuit against fixed coil is provided. The flexible element is a metal springy cantilever with suitable shape and material for our application. The natural frequency depends on these parameters. Model shown in Fig. 15 is tuned up to 18 Hz. The lower operating frequency can provide relevant response of the resonance system and consecutively response of the electromagnetic converter [41]. Several different assemblies and materials of springy element are modeled and tuned up for natural frequencies in the range from 10 Hz to 100 Hz. These harvesting frequencies are solved in ANSYS environment with respect to the mechanical Eq. (3). The calculated mechanical parameters are then used in the mechatronic model in Matlab environment. Further the magnetic FEM model (see Fig. 16) of the movable magnetic circuit is analyzed and the value of magnetic flux density is used in model of Faraday’s law (5). The voltage is induced in printed coil during movement of the magnetic circuit. The model of electromagnetic conversion is connected with outputs of mechanical and magnetic models. The utilization of multilayer coil with 25 lm resolution is expected. This technology can provide several hundred coil turns and sufficient induced voltage for used electronics. The mathematical model expect 950 coil turns and inner resistance of coil 1.5 kX. The connected electrical load in electrical circuit causes current flow and the electric current through the coil provides magnetic damping forces. The electromagnetic damping force is determined by overall electrical output power dissipated from this mechatronic system. This damping force affects the mechanical system in feedback and consecutively induces voltage as well. The analysis of an optimal electrical load is essential for the successful energy harvesting from mechanical movements. All individual models are therefore created in Matlab/Simulink environment. This system is excited by the human movement in the head area, which can be measured. The output signal is then used as input for simulation of the energy harvesting system. The simulation model in Matlab/Simulink is shown in Fig. 17.

Fig. 15. Concept of electromagnetic energy harvester design.

Fig. 16. FEM model of magnetic flux density in Tesla units.

The proposal parameters of the presented energy harvesting concept respect free volume in head area and power consumption of the developed cochlear implant. The measured data of head acceleration during usual walking are used as input for the mechanical system and the system response is analyzed. The simulated results of output voltage and output power in time domain are shown in Fig. 18. These results provide response of the energy harvester with optimal electrical load and this model is excited by acceleration during usual walking in time range of 10 s. The advantage of the multidisciplinary model in Matlab/Simulink is connection of individual models like FEM models, which provide the second order differential motion equation. This equation provides the movement of magnetic field in the model. On the base of solving Faraday’s law the model of electronic circuit with electrical load is created. The load represents the

Please cite this article in press as: Zˇák J et al. Model-based design of artificial zero power cochlear implant. Mechatronics (2015), http://dx.doi.org/10.1016/ j.mechatronics.2015.04.018

J. Zˇák et al. / Mechatronics xxx (2015) xxx–xxx

10

Fig. 17. Model of electromagnetic energy harvester in Matlab/Simulink enviroment.

Fig. 18. Simulated outputs of energy harvesting model during 10 s of walking; output voltage (left) and output power (right).

sensing and signal processing electronics of the developed artificial cochlear implant. Another advantage is the fact that this mechatronic model can be linearized. Based on electromechanical analogy the substitute dynamic RCL model will be used for development process of other subsystems. 6. Design of power management electronics Simulink models of the thermoelectric module and the electromagnetic energy harvesting generator provides inputs to the electronic model and it can be designed and optimized for the maximal energy harvesting of ambient energy [42]. From the previous simulation results is evident that output power of energy harvesting system is on a very low level. On the other hand the power consumption of the signal processing and nerve excitation is supposed to be designed as the ultra-low power electronic system integrated on-chip. The mechatronic modeling techniques and the electromechanical analogy can be used for optimal design of the power management electronics and the ultra-low power signal processing electronics. Matlab/Simulink models of thermoelectric and electromagnetic energy harvesting generators can be created as linear RCL circuit.

The RCL circuit represents the model of the autonomous energy harvesting power source for SPICE environment. There is only limitation in linearity of the final model but the model can provide very realistic input for design of the on-chip electronics. The structure of the power management system consists of several sub-blocks (see Fig. 19). Adaptable input circuit, internal storage capacitor, voltage multiplier, main storage supercapacitor and load switching circuit are expected as main blocks of the power management system. Nowadays, electronic circuits with particularly similar principle have been already implemented into commercial designs (e.g. LTC3108 from Linear Technology) and their advantages have been confirmed. The input circuit adjusts power management circuit impedance in accordance to the actual generator state for maximization of harvested efficiency. The circuit also uses the ultra-low power synchronous rectifier. Generated energy is temporarily stored in the internal capacitor. During the startup, energy is delivered directly to the energy storage for input circuits powering. It is done with low efficiency but it is necessary to do that when no energy is in the circuits. Internally stored energy is spasmodically transformed to higher voltage when internal capacitor is charged. This pulse process decreases the power consumption of the DC–DC converting voltage multiplier. After voltage multiplication, energy is stored in the main storage supercapacitor with ultra-low

Please cite this article in press as: Zˇák J et al. Model-based design of artificial zero power cochlear implant. Mechatronics (2015), http://dx.doi.org/10.1016/ j.mechatronics.2015.04.018

J. Zˇák et al. / Mechatronics xxx (2015) xxx–xxx

11

Fig. 19. Structure of the power management system for cochlear implant.

self-discharge properties. Stored charge is redistributed to target system parts by load switching sub-circuit, which enables minimization of charge transfer losses. 7. Conclusions The development of a fully implantable biomechatronic artificial cochlear based on MEMS sensor array is presented. The essential difference between our concept and signal processing electronics in common cochlear implants is the utilization of active sensing elements. The active elements are created as individual MEMS sensors that provide sensing and filtering of acoustic signal in particular acoustic frequency. The whole sensor array affords simple mechanical decomposition and separation of frequency spectrum in 24 channels. Ultra-low power consumption is required for powering the signal processing electronics. This fact can employ energy harvesting methods to supply this biomechatronic device. The present approach based on multidisciplinary models provides a complex view on our development concerning the unique concept of the fully implantable artificial cochlear implant with zero power consumption. The sensing part, signal processing electronic circuits, energy harvesting systems and power management system was designed and simulated. Many simulations were done with respect to multidisciplinary mechanical and electrical models. The verified models will be used for design and fabrication of MEMS sensors array in laboratory conditions. The presented models have a great potential for an optimization study of effective energy harvesting from body environment. The proposed energy harvesting method seems to be able to provide the zero power cochlear implant. Acknowledgements This paper has been supported by the project ‘‘Research of the Micro Electro Mechanical Artificial Cochlea Based on Mechanical ˇ R 13-18219S under the Czech Science Filter Bank’’ GAC Foundation (CSF) and by project FEKT-S-14-2300. "New types of electronic circuits and sensors for specific applications". Research described in this paper was supported by the National Sustainability Program under grant LO1401 for the research infrastructure of the SIX Centre used in this research. References [1] Balkany T, Gantz B, Nadol Jr JB. Multichannel cochlear implants in partially ossified cochleas. Ann Otol Rhinol Laryn Suppl 1988;135:3–7. [2] Donnelly MJ, Pyman BC, Clark GM. Chronic middle ear disease and cochlear implantation. Ann Otol Rhinol Laryn Suppl 1995;166:406–8.

[3] Gantz BJ, McCabe BF, Tyler RS. Use of multichannel cochlear implants in obstructed and obliterated cochleas. Otolaryngol–Head Neck Surg: Official J Am Acad Otolaryngol-Head Neck Surg 1988;98:72–81. [4] Otte J, Schunknecht HF, Kerr AG. Ganglion cell populations in normal and pathological human cochleae. Implications for cochlear implantation. Laryngoscope 1978;88:1231–46. [5] Schmidt JM. Cochlear neuronal populations in developmental defects of the inner ear. Implications for cochlear implantation. Acta Oto-Laryngol 1985;99:14–20. [6] Briggs RJ, Eder HC, Seligman PM, Cowan RS, Plant KL, Dalton J, et al. Initial clinical experience with a totally implantable cochlear implant research device. Otol Neurotol: Official Publ Am Otological Soc Am Neurotol Soc [and] Eur Acad Otol Neurotol 2008;29:114–9. [7] Wei X, Liu J. Power sources and electrical recharging strategies for implantable medical devices. Front Energy Power Eng Chin 2008;2:1–13. [8] Sarpeshkar R, Salthouse C, Sit JJ, Baker MW, Zhak SM, Lu TKT, et al. An ultralow-power programmable analog bionic ear processor. IEEE Trans Biomed Eng 2005;52:711–27. [9] Yodtean A, Thanachayanont A. A micropower CMOS preamplifier for cochlear implant system. In: International technical conference on circuits, systems, computers and communications, Herakleion, Greece; 2008. p. 997–1000. [10] Zajdel T. Asynchronous stimulation for cochlear implants. Electrical and computer engineering. Ohio (USA): The Ohio State University; 2012. p. 66. [11] Knisely K, Grosh K. A MEMS AlN transducer array for use as a cochlear implant. In: 2013 IEEE international symposium on the applications of ferroelectric and workshop on the piezoresponse force microscopy (ISAF/PFM); 2013. p. 240–3. [12] Xu T, Bachman M, Zeng F-G, Li G-P. Polymeric micro-cantilever array for auditory front-end processing. Sens Actuator A: Phys 2004;114:176–82. [13] Chen F, Cohen HI, Bifano TG, Castle J, Fortin J, Kapusta C, et al. A hydromechanical biomimetic cochlea: experiments and models. J Acoust Soc Am 2006;119:394–405. [14] Shintaku H, Nakagawa T, Kitagawa D, Tanujaya H, Kawano S, Ito J. Development of piezoelectric acoustic sensor with frequency selectivity for artificial cochlea. Sens Actuator A: Phys 2010;158:183–92. [15] White RD, Grosh K. Design and characterization of a MEMS piezoresistive cochlear-like acoustic sensor. In: ASME 2002 international mechanical engineering congress and exposition, New Orleans, Louisiana, USA; 2002. p. 201–10. [16] Von Békésy G. Experiments in hearing. McGraw; 1960. [17] Robles L, Ruggero MA. Mechanics of the mammalian cochlea; 2001. [18] Dusek D. Decomposition of non-stationary signals based on the cochlea function principle. Solid State Phenom 2009;147–149:6. [19] Abdi H, Mohajer N, Nahavandi S. Human passive motions and a user-friendly energy harvesting system. J Intell Mater Syst Struct 2013. [20] Delnavaz A, Voix J. Energy harvesting for in-ear devices using ear canal dynamic motion. IEEE Trans Industr Electron 2014;61:583–90. [21] Khaligh A. Kinetic energy harvesting using piezoelectric and electromagnetic technologies—state of the art. IEEE Trans Industr Electron 2010;57:850–60. [22] Lay-Ekuakille A, Vendramin G, Trotta A, Mazzotta G. Thermoelectric generator design based on power from body heat for biomedical autonomous devices. IEEE; 2009. p. 1-4. [23] Rowe DM. Thermoelectrics handbook: macro to nano. Taylor & Francis; 2005. [24] Belleville M, Condemine C. Energy autonomous micro and nano systems. Wiley; 2012. [25] Mitcheson PD. Energy harvesting for human wearable and implantable biosensors. In: 2010 Annual international conference of the IEEE engineering in medicine and biology society (EMBC); 2010. p. 3432–6. [26] Nagel JA, Sieber I, Gengenbach U, Guth H, Bretthauer G, Guthoff RF. Investigation of a thermoelectric power supply for the artificial accommodation system. In: 2010 3rd international symposium on applied sciences in biomedical and communication technologies (ISABEL); 2010. p. 1–5. [27] Leonov V. Simulation of maximum power in the wearable thermoelectric generator with a small thermopile. Microsyst Technol 2011;17:495–504.

Please cite this article in press as: Zˇák J et al. Model-based design of artificial zero power cochlear implant. Mechatronics (2015), http://dx.doi.org/10.1016/ j.mechatronics.2015.04.018

12

J. Zˇák et al. / Mechatronics xxx (2015) xxx–xxx

[28] McCarty R. Thermoelectric power generator design for maximum power: it’s all about ZT. J Electr Mater 2013;42:1504–8. [29] Youn N, Lee H, Wee D, Gomez M, Reid R, Ohara B. Achieving maximum power in thermoelectric generation with simple power electronics. J Electr Mater 2014;43:1597–602. [30] Hadas Z, Singule V. Energy harvesting – opportunity for future remote applications; 2011. [31] Hadas Z, Ondrusek C, Singule V. Power sensitivity of vibration energy harvester. Microsyst Technol 2010;16:691–702. [32] Williams CB, Yates RB. Analysis of a micro-electric generator for microsystems. Sens Actuator A: Phys 1996;52:8–11. [33] Accoto D, Calvano M, Campolo D, Salvinelli F, Guglielmelli E. Energetic analysis for self-powered cochlear implants. In: Conference proceedings: ... annual international conference of the IEEE engineering in medicine and biology society. IEEE engineering in medicine and biology society. Conference; 2009. p. 4860–3. [34] Goll E, Zenner H-P, Dalhoff E. Upper bounds for energy harvesting in the region of the human head. IEEE Trans Bio-Med Eng 2011;58:3097–103. [35] Hadas Z, Vetiska J. Energy harvesting analysis of body motion as energy source for biomedical devices, Svratka, Czech Republic; 2014. [36] Norgia M, Tognola G, Svelto C. Measurement of electrode current pulses from cochlear implants. In: Proceedings of the 21st IEEE instrumentation and

[37]

[38]

[39]

[40]

[41]

[42]

measurement technology conference, 2004. IMTC 04, vol. 1693; 2004. p. 1697–700. Razzaghpour M, Golmakani A. An ultra-low-voltage ultra-low-power OTA with improved gain-bandwidth product. In: International conference on Microelectronics, 2008. ICM 2008; 2008. p. 39–42. ITIS Foundation. Tissue properties: thermal conductivity. [cit. 05.03.2014]. Janak L, Ancik Z, Hadas Z. Simulation modelling of MEMS thermoelectric generators for mechatronic applications. In: Brˇezina T, Jablon´ski R, editors. Mechatronics 2013. Springer International Publishing; 2014. p. 265–71. Hadas Z, Vetiska V, Huzlik R, Singule V. Model-based design and test of vibration energy harvester for aircraft application. Microsyst Technol 2014;20:831–43. Hadas Z, Vetiska V, Singule V, Andrs O, Kovar J, Vetiska J. Energy harvesting from mechanical shocks using a sensitive vibration energy harvester. Int J Adv Rob Syst 2012;9. 1–1. Hadas Z, Kurfurst J, Ondrusek C, Singule V. Artificial intelligence based optimization for vibration energy harvesting applications. Microsyst Technol 2012;18:1003–14.

Please cite this article in press as: Zˇák J et al. Model-based design of artificial zero power cochlear implant. Mechatronics (2015), http://dx.doi.org/10.1016/ j.mechatronics.2015.04.018