This study describes the design of an inductive powering system that combines power transfer with data transmission for implantable biomicrosystem. The implant receives power from an external transmitter through an inductive link between an external power transmission coil and the implanted receiving coil. We designed a transmitter with class E power amplifier to provide maximum power with high-efficiency. It is possible to obtain about 136 mW of power in the secondary system with an efficiency of about 50%. A bit rate of 1 Mbps is achieved for sending command data to the implant. We propose a communication system for the medical implants which uses inductive link. This system is consisted of a class E power amplifier, an ASK modulator, an inductive link and an ASK demodulator.
PDF Abstract XML References Citation
How to cite this article
Implantable stimulators such as pacemakers, defibrillators, cochlear implants, instrumented orthopaedic implants, deep brain and spinal cord stimulators have found a variety of applications in improving the quality of patients lives (Pamela et al., 2006; Burny et al., 2000, Loeb et al., 1998; Wise et al., 2004). Broad research is underway to provide new devices for other applications such as visual prosthesis, recording devices and neuromuscular controllers (Margalit et al., 2002; Stangel et al., 2001; Mohseni and Najafi, 2005; Harrison et al., 2007; Harrison and Charles, 2003; Watkins et al., 2006; Sawan, 2004). Reducing the size and power consumption have been the two major challenges in designing these implantable microelectronics because the implantation procedure should be minimally invasive to the patient. Batteries are not the optimal choice for implantable devices, because their lifetime is limited and they are usually large and leaks can pose a hazard to tissues. For most of these long-term implantable devices, an external transcutaneous wireless link is preferred to an internal battery for the power supply (Galbraith, 1987; Catrysse et al., 2004). Wireless operation is required for most of the biomedical implantable circuits. Many of these devices are powered by inductive coupling in the centimetre range distance over the skin barrier. Some of the biomedical implants, particularly those which interface with the central nervous system, such as cochlear and visual prostheses, need large amounts of data to simultaneously interface with a large number of neurons through multiple channels (Suaning and Lovell, 2001; Wang et al., 2004). Therefore, a high data-rate receiver circuitry that can establish an efficient wireless link between the implant and the external units is highly needed.
Present study field is interested to the cochlear implant (Ghorbel et al., 2006) and implanted micro-sensor system for measurements and recording of neural signal from auditory nerve. The main objective of this study is to supply medical implant with power and data. In this study, we present inductive powering system that has been optimised towards maximal power transfer efficiency and high data rate. The crucial point of this study is to achieve a power efficiency of 50% and 1 Mbit sec-1 data rate.
Figure 1 shows the synoptic diagram of wireless powering and data communication for implantable medical devices.
|Fig. 1:||Block diagram of wireless biomedical implex communication|
The external part is consisting of a modulator and power amplifier. The internal part contains voltage regulator and demodulator. Then, the implanted device is externally controlled and powered by a modulated radio frequency signal. The receiver circuitry of the implant provides the stably regulated voltage and demodulates the data from radio frequency signal (Ben Hmida et al., 2006).
An inductive link, when used in implanted biomedical instruments, is also known as a transcutaneous transformer, since its primary and secondary coils are physically separated and communicate through the electromagnetic field linking the two coils. Extensive research has been performed to analyze and optimize the operation of such inductive links (Gervais et al., 2003) and there exist many applications of inductive links to deliver power and data into implants (Huang and Oberle, 1998).
Equivalent electrical circuit: The equivalent electrical circuit of an inductive link is depicted in Fig. 2. It consists of the self-inductances Ln and resistances RLn of the primary and secondary coil (where n = 1, 2), the mutual inductance M between the two coils, the link load impedance Z2 and a voltage source V1. The equivalent circuit can be described by Eq. 1 through Eq. 3.
An often used measure for the coupling between two coils is the coupling factor or coupling coefficient k, given by:
Physically, the coupling factor equals the fraction of magnetic flux generated by the primary coil L1, which flows through the secondary coil L2 and as mutual inductance is a reciprocal property of two coils, vice versa. The secondary current I2 and voltage V2 can be easily calculated from Eq. 1 through Eq. 3:
|Fig. 2:||Schematic overview of an inductive link|
|Fig. 3:||Equivalent circuit of inductive link transformed into Zequivalent|
The link impedance Zlink as seen from the voltage source V1 in Fig. 2 is also easily calculated from the link equations and is given by:
in which Zequivalent is the equivalent impedance of the secondary circuit as seen by the primary circuit, as shown in Fig. 3:
The expression of the equivalent impedance is given by:
The link efficiency ηlink is defined by the ratio of the power dissipated in the link load impedance Z2 and the power delivered by the voltage source V1. It is given by:
in which In* is the complex conjugate of In and Z2 = R2 + jX2.
Optimization and design choices: It can easily be seen from Eq. 10 that the link efficiency ηlink is maximal when ωL2 = -X2. This can be achieved by tuning the secondary coil with a parallel or series capacitance.
Secondary circuit in series resonance: For a series tuned secondary coil as shown in Fig. 4a, the link load resistance R2 and the link load reactance X2 become
where Rload is the impedance of the load circuit and C2s is the series capacitance. The optimal choice for the series capacitor C2s is then:
The equivalent impedance Zequivalent of the secondary circuit then simplifies to
Secondary circuit in parallel resonance: For a parallel tuned secondary coil, depicted in Fig. 4b, R2 and X2 are given by:
|Fig. 4:||(a) Series tuned secondary coil and (b) Parallel tuned secondary coil|
|Table 1:||Inductive link parameter values|
The optimal choice for the parallel capacitor C2p is:
The equivalent impedance Zequivalent of the secondary circuit then simplifies to
Application: The model of the inductive link, described earlier, is applied by respecting the requirements of our specifications given by Table 1:
Rload = 300 Ω, which checks the condition given by the Eq. 19, i.e., Rload > 170.4 Ω.
One applies the Eq. 6 of the voltage gain for the two cases; series and parallel resonance. By analyzing the curves obtained Fig. 5a, it is noticed that one can achieve a maximum gain of 0.6 for a coupling coefficient about 0.6 for a secondary circuit in parallel resonance. In the case of a series resonance the voltage gain is always lower than that of a parallel tuned secondary circuit.
One applies the Eq. 10 of the link efficiency for the two cases of series and parallel resonance. By analyzing the curves obtained Fig. 5b, it is noticed that one can reach a maximum efficiency about 0.92 for a coupling coefficient higher than 0.2 in the case of a parallel tuned secondary circuit.
The majority of the electronic implants, one uses a coupling coefficient which does not exceed 0.45. In this zone of coupling, the link efficiency given by a secondary circuit tuned in parallel resonance is better than that given by a secondary circuit tuned in series resonance.
Thereafter, we will treat the only case of a parallel resonance since it offers characteristics, such as the voltage gain and the link efficiency, certainly better than that given by a series resonance.
Figures 6a-c give respectively an estimate of the module of voltage gain, link efficiency and the equivalent impedance Zequivalent of the secondary circuit versus the load resistance (275 Ω < Rload < 325 Ω) and the coupling coefficient k.
By analyzing the curve given by Fig. 6a, it is noticed that when the load is increased, the voltage gain also increases, for a coupling coefficient lower than 0.6. It is clear that the load variation, between 275 and 325 Ω, does not have a great effect on the gain voltage module. By analyzing the curve given by Fig. 6b, it is observed that when the load is increased, the link efficiency degrades slightly. It is clear that the load variation between 275 and 325 Ω does not have a great effect on the link efficiency. It is also noticed that the link efficiency reaches its maximum for a coupling coefficient higher than 0.3.
The equivalent impedance Zequivalent of the secondary circuit tuned in parallel resonance, given by equation 18, transferred into primary circuit is plotted Fig. 6c. By analyzing this curve; we notice that when the load is increased, the equivalent impedance of the secondary increases slightly. In conclusion, the load variation between 275 and 325 Ω, does not have a great effect, for a given coupling coefficient, on these three preceding characteristics.
By fixing the coefficient of coupling k = 0.1678 (calculated for an average value of the Rload = 300 Ω and an equivalent impedance of 39.755Ω), we plotted the curve of the equivalent impedance of the secondary circuit transferred to the primary circuit, gain voltage module and link efficiency versus the load resistance variation Fig. 7a-c.
|Fig. 5:||(a) Voltage gain module and (b) Link efficiency vs coupling coefficient|
|Fig. 6:||(a) Voltage gain module, (b) Link efficiency and (c) Zequivalent, vs load resistance and coupling coefficient|
|Fig. 7:||(a) Equivalent impedance, (b) Gain voltage module and (c) Link efficiency vs resistance|
|Table 2:||Variation of Zequivalent, voltage gain and link efficiency for coupling coefficient k = 0.1678|
Table 2 summarizes the results obtained previously, the coupling coefficient is k = 0.1678.
Inductive link simulation: The electronic circuit of inductive link is depicted in Fig. 8. The primary and the secondary circuit are tuned to the carrier frequency of 13.56 MHZ. By using PSpice simulation, we had plot the gain voltage of the inductive link circuit for some coupling coefficient values (k = 0.01; 0.1; 0.13; 0.15; 0.2; 0.3) at a load of 300 Ω, as shown in Fig. 9. We denote that according to this plot, the inductive link behaves as a pass band filter at the central frequency 13.56 MHZ.
The analysis of this simulation shows that:
|•||For a coupling coefficient lower than 0.13, the band width is lower than 1 MHZ.|
|•||For a coupling coefficient upper than 0.15, the band width is upper than 1.5 MHZ.|
Transmission through biological tissues: The transmission through human tissues is governed by standards. The principal standard fixes a limit for Specific Absorption Rate (SAR). The SAR corresponds to the quantity of power absorbed by biological tissues. Several studies are made to determine the effects of the electromagnetic waves on the human body. It is the case of inductive link which transmits energy through biological tissues.
With an aim of simplifying the analysis, it is appropriate to check initially if one can use the quasi-static approximation. It is the case when the operation distance is negligible compared to the wavelength. For medical implants that use an inductive link, the typical operation distance is nearly 1 cm and the operating frequency is equal to 13.56 MHZ.
|Fig. 8:||Electronic circuit of inductive link|
|Fig. 9:||Voltage gain of inductive link|
In this case, the principal losses of energy will not take place due to the reflections, but by the phenomenon of the Foucault currents. To know the losses, it is necessary to know the conductivity of materials crossed by the magnetic field. The complexity of biological tissues led to several studies, including one very complete made by Gabriel and Gabriel (1996), the results of this study give a mathematical model as well as the various constants for each type of tissues.
The complete analysis of the phenomenon is complicated, but a typical case can be illustrated according to results presented by Schuder et al. (1976). The calculation of power, absorbed by biological tissues (skin), according to the Eq. 20, where K[G] and E[G] are complete elliptic integrals of first and second order, respectively. The other parameters used are provided to Table 3. The evaluation of the Eq. 20 is nearly 0.6 mW.
|Table 3:||Parameter values used by equation (20)|
For a case where the total power exceeds 130 mW, the losses are thus lower than 0.5% and can be regarded as being negligible.
CLASS-E POWER AMPLIFIER
The critical element of the external transmitter is the circuit to drive the transmitter coil. There are many types of amplifiers that may be suitable. Among them class D and class-E with both theoretical 100% of efficiency (only ideal switches) are very suitable candidates. The topology chosen is class E. The major advantage of this topology versus other types (like class D or class C) is essentially the easy inclusion of the parasitic drain-source capacitance of the switch transistor into the design optimization.
Circuit analysis: Achieving high power transfer efficiency is one of the challenges that one would face in the design of such systems. For this reason, we use a transmitter with class E power amplifier to provide maximum power transfer with high-efficiency. A structure of class E power amplifier is shown in Fig. 10. A high-speed power switch (MOSFET transistor) drives a resonant network consisting of a parallel capacitor or shunt capacitor (Cd), a transmitter coil (L1), a tuning capacitor (C) and a resistive load (RL) i.e., RL represents the equivalent load of the implanted device with the receiver coil and parasitic resistances of the transmitter coil. When the switch is OFF, the RF choke inductor (RFC) acts as a current source charging the resonant network and creating a transient voltage across the switch. When the switch is ON, its current rises smoothly until the switch is off again (Fig. 11).
|Fig. 10:||Basic circuit of the class E amplifier|
|Fig. 11:||Theoretical wavesforms of the switch voltage (Vds) and current (Id)|
Losses are kept at minimum by having the transistor switch on when both voltage and current are small (Sokal and Sokal, 1975; Kendir et al., 2005). The current and voltage across the transmitter coil are sinusoidal at the resonant frequency of the tank circuit, which has to be the same as the driver switching frequency. The optimum conditions, i.e., Zero-Voltage Switching (ZVS) and Zero-Voltage-derivative Switching (ZVDS) are necessary for the Class E amplifier to achieve high efficiency (Suetsugu and Kazimierczuk, 2004; Raab, 1977; Kazimierczuk and Puczko, 1987). The efficiency of Class E transmitter can reach 100% with theoretical design.
Design of the Analytical Equations for Optimum Operation: The derivations of design equations are carried out under the following assumptions.
|•||The shunt capacitance of the amplifier consists of the MOSFET output capacitance and an external linear capacitance.|
|•||The inductance of the choke coil is large enough to neglect its current ripple.|
|•||The internal resistance of the choke coil is zero; therefore, the dc voltage drop across the choke is zero.|
|•||The loaded quality factor of the output resonant circuit is high enough so that the output current can be considered a sine wave.|
|•||The load resistance includes parasitic resistances of the series resonant circuit; i.e., the resonant circuit is considered to be a pure reactance.|
|•||The MOSFET ON resistance is low enough so that the current and voltage waveforms remain almost unchanged.|
|•||The MOSFET turns on and off very fast so the transition time can be neglected. Since the transistor is turned on at zero voltage in the Class E amplifier, the turn-on transition is negligible.|
|•||The switch duty cycle is 0.5.|
Referring to Fig. 10, the current through the circuit is given by:
Where Idc is the dc input current, Iout is the output current amplitude and θ = ωt. The current through the shunt capacitance is Ip. Id is the current through MOSFET transistor. Figure 11 shows theoretical waveforms of the switch voltage (Vds) and current (Id).
|•||When the switch is ON, i.e., for 0 < θ < π,|
Equation 22 gives :
|•||When the switch is OFF, i.e., for π < θ < 2π,|
The current through the shunt capacitance is given by:
Vds(π) = 0, so we can obtain K1
Imposing the ZVS condition given by Vds(2π) = 0 then
Imposing the ZVDS
condition at the switch turn-on (ωt = 2π)
Since the dc component of the voltage drop across the choke inductor is zero, the average value of the switch voltage is equal to the dc supply voltage. Thus:
Vds is maximal when:
So, calculating drain source voltage:
The output impedance can be decomposed into two constituents: real and imaginary. The real part of the impedance gives a Vdsi voltage. The imaginary part of the impedance gives Vdsq voltage. The Vds(t) function is periodic; so we can calculate the first order coefficients of the Fourier series of this signal.
X > 0, then X = ωLs with Ls is an excess inductance
The output power is done by:
So the class E power amplifier has a theoretical efficiency on 100%.
The quality factor of the resonant circuit is given by:
The condition of resonance of LC circuit is LCω2 = 1, so the expression of the resonator capacitance is given by:
Class-E simulation: To verify the theoretical results, class-E power amplifier was simulated with PSpice. We desire to achieve an output power of 150 mW. So using analytical design equations described earlier, we calculate the different class-E component values. The Table 4 contains the class-E parameter values. The theoretical value of shunt capacitor is 51.46 pF, but this value doesnt give the optimum operation; due to the assumption that Cd consists of the MOSFET output capacitance and an external linear capacitance. Using the method specified by Sokal (2001) to search the optimum operation point, one found Cd = 56.5 pF. The simulated waveforms of the gate-source voltage Vgs and drain-source voltage Vds are shown in Fig. 12.
|Table 4:||Class-E parameter values|
|Fig. 12:||Gate-Source voltage Vgs-Drsin-Source voltage Vds vs time|
|Fig. 13:||Simulation waveforms, (a) Gate-Source voltage, (b) Current through the shunt capacitance, (c) Current through the NMOS transistor and (d) Current through the transmitter coil|
The ZVS operation was achieved at the switching frequency 13.56 MHZ (13.56 MHZ that belongs to the ISM band: Industrial-Scientific -Medical). Using Eq. 39, we calculate the switch peak voltage VDSmax = 11.75 V. The simulation value of VDSmax was 11.67 V. Figure 13 shows simulation waveforms of the gate-source voltage (a), the current through the shunt capacitance (b) and current through the NMOS transistor (c). The current through the transmitter coil, Fig. 13d, is almost sinusoidal due to the high quality factor (QL = 10). The DC power given by the Vdd source (PDC) is 148.2 mW, while the absorbed power by the load RL (Pout) is 134 mW. The drain efficiency is then: DE = Pout/PDC = 90.41%.
The theoretical results were validated by PSpice simulation. In particular, the switch voltage waveform satisfied both conditions for optimum operation. The drain efficiency of class-E transmitter can reach 90.4%, by Pspice simulation, using a model of a real transistor.
POWER AND DATA TRANSFER
Modulator design: In the frequency modulation or phase modulation cases, the modulated signal has a relatively wide band with compared to the amplitude modulation. Furthermore, in the implants case, the power must be transmit on an ISM band which is relatively narrow. Seen that the transmission canal in these implantable systems is the inductive link which is behaves as a selective filter, our choice approved the amplitude modulation.
The amplifier modulator structure presented in the Fig. 17, contains a pair of bipolar transistors Q1 and Q2 (Darlington pair). The voltage source V2 (modulating signal) generates a binary signal with a Tbit = 5 μs. The NMOS transistor M2 works as a switch. The modulation rate can be adjustable with resistance R1. The voltage supply Vdd is fixed at 5V instead of 3.3V in order to recuperate the voltage drop engendered by the modulator ASK. The supply voltage of the class E power amplifier is measured between the transistor Q1 emitter and the ground, as illustrated in the Fig. 14b. The level voltage is 2.92 V when the bit to transmit is 1 and 3.7 V in the case when the message is 0 (Fig. 14a-b). We notice that this voltage variation is around 3.3 V, the voltage for which the amplifier structure was studied. Figure 14c-d the voltage signals at the emitters and receivers coil nodes for a coupling coefficient k = 0.15. The class E power amplifier output voltage is an ASK modulated signal (Fig. 14c). We denote that the received voltage at the receiver coil nodes is also a ASK modulated signal with a modulation rate of 18.25% (Fig. 14d). The supplied power by the Vdd source (PDC) is 274 mW. The received power, by the inductive link, consumed by the resistance R5 is 136.6 mW, where R5 is an equivalent resistance of all the implantable part. That gives a power efficiency of 50%.
ASK demodulator principles: The received modulation signal, at the receiver coil, shown by the bloc diagram of the Fig. 15.
The envelope detector is consisting in a rectifier followed by a filter. The potential comparator for this application is a Schmitts Trigger.
Design of ASK demodulator: The proposed ASK demodulator structure was based on the specific following technology: the AMS 0.35 μm, 3-metal, 2-poly, n-well standard CMOS process. Figure 17 presents all the studied inductive powering system.
The part including transistors M3 and M5 and the capacitor C1 constitute the envelope detector. The transistor M3 is set at a diode stage to realize a simple alternation rectifier. The transistor M5 and the capacitor C1 provide a filter. The circuit formed by transistors M6 to M13 realizes a Schmitts Trigger. Finally, we reverse the signal formed by the two transistors M14 and M15. The demodulated signal is presented in Fig. 16c. The voltage source Vdd2 is in reality a regulated voltage, it is obtained from the received signal in the receiver coil nodes as a Band gap structure.
The simulation results, given by the Fig. 16, have been made at a voltage source V2 (modulating signal) which generates a binary signal with Tbit = 1 μs then a data rate of 1 Mbit sec-1. The output signal of the envelope detector is shown in Fig. 16b.
|Fig. 14:||(a) Binary data signal, (b) Supply voltage of the class E power amplifier, (c) Voltage signal at the transmitter coil and (d) Voltage signal at the receivers coil|
|Fig.15:||Block diagram of ASK demodulator|
|Fig. 16:||(a) Binary data signal, (b) The output signal of the envelope detector and (c) Demodulated Signal|
|Fig. 17:||Complete inductive powering system|
This study presents an inductive powering system that combines power transfer with data transmission. We designed a transmitter with class E power amplifier to provide maximum power with high-efficiency. An analysis and a design of the analytical equations for a high efficiency class E amplifier were presented for optimum operation. The theoretical results were validated by PSpice simulation. The drain efficiency of class-E transmitter can reach 90.4%, by PSpice simulation, using a model of a real transistor. A power supply in the implant is equal 136 mW. Even considering some additional power loss in the DC source converters, this seems to be enough for an integrated low-power system design. This inductive powering system gives a power efficiency of 50%. The data rate transfer achieves 1 Mbit sec-1. Future work will focus on the miniaturisation of the secondary circuit, resulting in an implantable ASIC.
- Ben-Hmida, G., M. Dhieb, H. Ghariani and M. Samet, 2006. Transcutaneous power and high data rate transmission for biomedical implants. Proceedings of the International Conference on Design and Test of Integrated Systems in Nanoscale Technology (DTIS'2006), September 5-7, 2006, Tunis, Tunisia, pp: 374-378.
- Burny, F., M. Donkerwolcke, F. Moulart, R. Bourgois and R. Puers et al., 2000. Concept, design and fabrication of smart orthopedic implants. Med. Eng. Phys., 22: 469-479.
- Gervais, J.F., J. Coulombe, F. Mounnaim and M. Sawan, 2003. Bidirectional high data rate transmission interface for inductively powered devices. Proceedings of the International Canadian Conference on Electrical and Computer Engineering, May 4-7, 2003, IEEE Xplore, pp: 167-170.
- Harrison, R.R. and C. Charles, 2003. A low-power low-noise CMOS amplifier for neural recording applications. IEEE J. Solid State Circuits, 38: 958-965.
- Harrison, R.R., P.T. Watkins, R. J. Kier, R.O. Lovejoy, D.J. Black, B. Greger and F. Solzbacher, 2007. A low-power integrated circuit for a wireless 100-electrode neural recording system. IEEE J. Solid State Circuits, 42: 123-133.
- Huang, Q. and M. Oberle, 1998. A 0.5-mW passive telemetry IC for biomedical applications. IEEE Trans. Solid State Circuits, 33: 937-946.
- Kendir, G.A., W. Liu, G. Wang, M. Sivaprakasam, R. Bashirullah, M.S. Humayun and J.D. Weiland, 2005. An optimal design methodology for inductive power link with class-e amplifier. IEEE Trans. Circuits Sys., I: Regular Papers, 52: 857-866.
- Margalit, E., M. Maia, J.D. Weiland, R.J. Greenberg and G.Y. Fujii et al., 2002. Retinal prosthesis for the blind. Survey Ophthalmol., 47: 335-356.
- Mohseni, P. and K. Najafi, 2005. A 1.48-mW low-phase-noise analog frequency modulator for wireless biotelemetry. IEEE Trans. Biomed. Eng., 52: 938-943.
- Pamela, T.B. and K.D. Wise, 2006. A 32-Site 4-channel high-density electrode array for a cochlear prosthesis. IEEE J. Solid State Circuits, 41: 2965-2973.
- Sawan, M., 2004. Microsystems dedicated to wireless multichannel monitoring and microstimulation: Design, test and packaging. Proceedings of the 7th International Conference on Solid State and Integrated Circuits Technology, October 18-21, 2004, Beijing, China, pp: 1408-1411.
- Stangel, K., S. Kolnsberg, D. Hammerschmidt, B.J. Hosticka, H.K. Trieu and W. Mokwa, 2001. A programmable intraocular CMOS pressure sensor system implant. IEEE Trans. Solid State Circuits, 36: 1094-1100.
- Suaning, G.J. and N.H. Lovell, 2001. CMOS neurostimulation asic with 100 channels, scaleable output and bidirectional radio-frequency telemetry. IEEE Trans. Biomed. Eng., 48: 248-260.
- Suetsugu, T. and M.K. Kazimierczuk, 2004. Analysis and design of class e amplifier with shunt capacitance composed of nonlinear and linear capacitances. IEEE Trans. Circuits Sys., 51: 1261-1268.
- Wang, C.C., Y.H. Hsueh, Y.T. Hsiao, U.F. Ghio, C.C. Huang and P.L. Liu, 2004. An implantable neural interface micro-stimulator chip with external controllability. Proceedings of the Asia-Pacific Conference on Advanced System Integrated Circuits, August 4-5, 2004, IEEE Xplore, pp: 356-359.
- Watkins, P.T., R.J. Kier, R.O. Lovejoy, D.J. Black and R.R. Harrison, 2006. Signal amplification, detection and transmission in a wireless 100-electrode neural recording system. Proceedings of the International Symposium of Circuits Systems (ISCAS). May 21-24, 2006, Island of Kos, pp: 2196-2196.
- Wise, K.D., D.J. Anderson, J.F. Hetke, D.R. Kipke and K. Najafi, 2004. Wireless implantable microsystems: High-density electronic interfaces to the nervous system. Proc. IEEE, 92: 76-97.