Research Article
A 16-electrode Fully Integrated and Versatile CMOS Microstimulator Dedicated to Cochlear Implant
Not Available
Mounir Samet
Not Available
Ahmed Ben Hamida
Not Available
Jean Tomas
Not Available
Profoundly to totally deaf patients with sensorineural hearing loss could benefit from cochlear prosthesis apparatus that could restore partially the hearing. Such patient may have mechanically normal inner ear, but vibrations of the basilar membrane no longer induce a spike pattern in auditory nerve due to damage within sensory hair cells (Tönder et al., 1998; Vanpoucke et al., 2004). This prosthesis can partially restore a sense of hearing through direct electrical stimulation of auditory residual nerve fibbers of the cochlea. Cochlear stimulation induced could initiate a nerve impulse (action potentials) that could be transmitted through synapses into auditory nerve center (Rebscher et al., 1999; Sarpeshkar et al., 2005).
Cochlear implant prosthesis normally includes three main parts: an external digital signal processing DSP unit, a telemetry system for data transmission and an implanted microstimulator coupled to an electrode array system (Germanovix and Toumazou, 2000; Lee and Lee, 2005). The external signal processing unit detects sounds captured from a microphone and converts them to a specific electronic code. This code is transmitted through the skin via a telemetry system to the implanted microstimulator. The stimulator circuit uses the demodulated code to generate electrical stimuli that drive the electrode array which is usually placed close to the residual nerve fibers inside the cochlea (Germanovix and Toumazou, 2000; Georgiou and Toumazou, 2005). The objective of intracochlear stimulation is to replicate the neural activity produced in the normal ear during acoustic stimulation. The conceptual design of the system is illustrated in Fig. 1.
The auditory nerve is organized within the spiral shaped cochlea in a tonotopic manner.
Fig. 1: | Cochlear implant system |
Low frequencies are encoded at the apex of the cochlea while high frequencies are encoded at the base (Tönder et al., 1998; Vanpoucke et al., 2004). Multichannel implants provide electrical stimulation at multiple sites in the cochlea by using an electrodes array (Miyoshi et al., 1999). Thus, different auditory nerve fibers can be stimulated at different places in the cochlea, thereby exploiting the place mechanism for coding frequencies. Globally, quality and intensity of sounds perception using this artificial device could be influenced by variations in the stimulus waveform parameters such as amplitude, pulse-width and frequency (Ghorbel et al., 2002; Germanovix and Toumazou, 2000; Liu et al., 2000).
Most stimulation strategies could be configured using various parameters in order to achieve maximum of flexibility. An implemented algorithm in the external digital speech processing could be accompanied by an aided software tool including the adjustment of these parameters. This clinical aided tool could be beneficial during clinical tests for adapting the apparatus to each pathological case (Tönder et al., 1998; Rebscher et al., 1999; Georgiou and Toumazou, 2005).
In this context of flexibility, adjustability or programmability, our microchip described in this study was designed to produce these independently programmable stimuli. The external part could be easily programmed for shaping the stimuli waveform so that essential parameters could be transmitted to the internal part. Hence, microchip would remotely generate stimuli in real time. Our device was provided by a high degree of flexibility and could be adapted to any other external speech processor. On the other hand, according to clinicians needs and in order to satisfy the various pathological cases, our main objectives of this proposed numeric conception would provide certainly great success in cochlear implant devices (Rebscher et al., 1999; Georgiou and Toumazou, 2005; Suaning and Lovell, 2001; Ghovanloo and Najafi, 2004).
This study proposed a 16-electrode fully integrated and numerical CMOS versatile micro-stimulator which offers a large variety of choices and possibilities in the rehabilitation profoundly to totally deafness. Performances offered by this conception included adjustments at different levels such as stimulus amplitudes which could be ranging up to 1 mA using 10 bits resolution, so at 1 μA steps variation. Stimulus width could also be varied from 0 to 255 μs per stimulation phase with 1 μs steps by using an eight bits control. To ensure a convenient tuning of these waveform parameters, a particular attention was paid to keep the current sources linearity as good as possible. Also, the electrical stimuli were charge balanced biphasic pulses resulting in an almost null residual Direct Current (DC), for protecting biological tissues among charges accumulation (Huang et al., 1999; Sawan et al., 1995).
A specific attention was also given to power consumption which is an important problem to be studied for conceiving an economic device as possible (Sarpeshkar et al., 2005; Georgiou and Toumazou, 2005; Ghorbel et al., 2004a). In fact, as the implanted device receives power and data via inductive link, our device was optimized so that the maximum power consumption was around 1.673 mW and its silicon area was smaller than 0.32 mm2.
COCHLEAR PROSTHESIS SYSTEM
As shown in the Fig. 1, cochlear prosthesis system includes an external unit, the sounds analyser and an internal chirurgically implanted unit, the under-skin implant.
Digitalized sound information, acquired by a miniature microphone for auditory prosthesis, is sent to the digital signal processor DSP. The DSP processes the incoming information in real-time and generates a series of stimulation command-frames that can cause a set of spatiotemporal stimulus pulses based on the adopted stimulation strategy (Germanovix and Toumazou, 2000; Georgiou and Toumazou, 2005).
Fig. 2: | Cochlear prosthesis main functions |
Fig. 3: | Biphasic current stimulus waveform |
The command-frames are arranged in bursts of manchester serial data bit stream, which are modulated into a square-shaped Amplitude Shift Keyed (ASK) signal at 20 MHZ frequency. The modulated signal is then amplified with a high efficiency class-E power amplifier and then inductively transmitted to the internal unit.
A second implanted coil coupled to the primary coil receives the modulated RF power carrier, from witch power and data are recovered using rectifier/regulator and telemetry decoder circuits. The resulting clock and data signals supply the stimulus instructions to the implanted microchip (Liu et al., 2000; Ghovanloo and Najafi, 2004; Ghorbel et al., 2004a; Shabou et al., 2004; Sawan et al., 2005).
The electrical stimulation waveform used for exciting the biological tissue is the charge-balanced biphasic current pulse shown in Fig. 2. Such waveform must be biphasic so to avoid charge accumulation that could cause damage to neural sites (Germanovix and Toumazou, 2000; Liu et al., 2000; Huang et al., 1999; Sawan et al., 1995).
In this study, we were interested primarily on the implanted microstimulator design, which would be connected to a sixteen electrodes array for providing electrical stimulation to the cochleas nervous sites where they are interpreted as sounds.
Implantable microstimulator chip: The block architecture of our proposed programmable microstimulator chip was illustrated in Fig. 4 which was carefully designed to deliver adjustable biphasic current pulses (Fig. 3). This electronic circuit was designed around two main stages: digital controller stage commanding an analogue stimulation stage including programmable CMOS current sources.
Digital processing stage was conceived around one state machine for interpreting all stimuli parameters from the incoming instructions, setting then appropriate commands to the analogue stimulation stage. Sixteen programmable CMOS current drivers operating with high accuracy and substantial flexibility could deliver then stimulation current pulses varying in 10 bit level amplitude, 8 bit range duration as well as programmable stimulation rhythm (Ghorbel et al., 2004a, b; Rekik et al., 2005).
Digital controller stage conception: The digital controller stage was conceived around one finite state machine which processes incoming commands and set the appropriate signals to the programmable current drivers.
Fig. 4: | Microstimulator architecture: Main proposed electronic parts |
Fig. 5: | Packet format definitions for the microstimulator (Ghorbel et al., 2002) |
It includes a 30 bit shifter register with a serial input/parallel output and a command-frame detector, a 26-bit instruction register, one stimulation register, two address decoders, a timing controller and a sequencer.
Prototype architecture of the proposed microstimulator circuit shown in Fig. 4 would be programmed serially using digital clock (CLK), data inputs (Data) and global reset (RAZ) pins. During command-data frame processing, RAZ signal is generated first when power supply (Vcc) is activated which will reset the entire digital controller for initial conditions. Next, incoming digital data would be shifted into a 30-bit serial to parallel shift register at every clock cycle and its content would be latched (Load1 = 1) into a 26-bit instruction register for stimuli generation once the specified header field was detected (Detect=1). Hence, the last 30-bit serial to parallel shift register would be then cleared (Clear0 = 1).
The 26-bit data register content would be forwarded into the stimulation register (Load2 = 1) if this one would be free (No previous data under processing: Buzy = 0) and the 26-bit data register would be then cleared (Clear1 = 1). Appropriate commands would be generated to the analogue stimulation stage in order to deliver programmable biphasic current pulse. After complete stimulus generation, stimulation register would reset (Clear2 = 1) and next instruction could be treated for one another stimulus generation if the 26-bit data register would be full (Full = 1).
Digital controller stage main functions could be summarized for more technical details. This would concern discussion of the communication protocol, the timing generator design and the electrode selection.
Communication protocol: Packet information shown in Fig. 5, which governs the operation of the stimulation, was defined digitally and loaded into the chip serially. It includes a specific code of 30 bits (D0 to D29) witch is distributed in five fields rearranged as follows:
Header: D0D1D2D3 were fixed to (1100) as a chosen header permitting to avoid any confusion during processing and allowing the microstimulator to detect the start of every new frame. It was also necessary to synchronize the transmission command-frames from the outside within the implant. Once the header was detected, the command-frame including all stimulation parameters would be stored in a 26 bit temporary register. Otherwise, the received command-frame is ignored with no further action.
Current amplitude: D4 to D13 are 10 bits field specifying current level to be produced during the cathodic/anodic phase. The maximum stimulation current value is around 1023 μA (210) varying with 1 μA step. To ensure the charge balancing, both cathodic and anodic, amplitude current pulses must use the same (Ia = Ic). Hence, injected charge during anodic phase and the cathodic phase would the same: Qc = Ic. Tc = Ia. Ta = Qa.
Pulse width: D14 to D21 are 8 bits field specifying stimulus pulse width in microseconds allowing then setting of electrical stimulus duration from 0 to 255 μs/phase ranging in 1 μs step. This same field was used for both the anodic and the cathodic phases to avoid charge accumulation into the biological tissue (Ta = Tc).
Active electrode: D22 to D25 are 4 bits field specifying the active electrode address of the stimulation site. The number of bits of this field depends on the number of the electrodes of the stimulator. In our case, one of 16 (24) electrodes could be selected during anodic phase (injected charge in the auditory neurones). Channels addressing was non simultaneous, so that only one site is stimulated at one time in order to avoid interferences. Considering microstimulator design, this would minimize silicon area used.
Return electrode: D26 to D29 are the last 4 bits field specifying the return electrode address. Such electrode assures current return to ensure the injected charge balancing. This field select one of sixteen electrodes (24) and if the addresses of the active and return electrodes are equal, a default return electrode is used.
Timing generator design: Timing generator block shown in Fig. 6 produces two digital timing signals for current drivers: the stimulation enable (Stim) and the direction (Direction) of current stimulation signals.
Fig. 6: | Timing generator |
End of stimulation would be indicated by a reset signal (End Stim.) to the instruction register.
Pulse width field would be shifted from the command-frame and latched into the timing generator register: At this time, stimulation signal (Stim) is asserted high, direction signal (Direction) is set low and the 8-bit counter begins incrementing. The current counter value would be compared with the pulse width parameter value. Once pulse width value reach counters value, the latter would reset and the direction signal would be reversed to high. A second cycle would be restated then and again, once the pulse width value reach counters value, the latter would be reset whereas direction and stimulation enable signals are turned low. At the end of this second cycle, end of stimulation signal (End Stim.) would be generated and the timing generator register would be cleared to zero upon the detection of End Stim. rising edge. The stimulation enable signal remains low, up to a new pulse width field value is latched.
Electrode selection: Our micro-stimulator uses high rate of interleaved pulsatile stimulation when only one stimulation channel could be activated at one time. Such stimulation procedure was the most useful stimulation because it minimizes current spreading and interaction and maximizes patients comfort and hearing ability (Germanovix and Toumazou, 2000).
In the proposed bipolar mode, electrodes would be reassembled in pairs so that one active electrode could be coupled to another return electrode during their functioning. Two 4 bit electrode address decoders are used to specify, respectively the active electrode and the return electrode. If these two decoders inputs were the same, return electrode would be specified automatically (17th electrode used as common reference).
Fig. 7: | Basic architecture of the programmable current source |
Analogue stimulation stage: The second main stage of our microstimulator electronic circuit was the analogue stimulation stage conceived around CMOS programmable current drivers. This device enables stimulation current waveform delivery at each stimulation channel according to specified data frame accurately decoded by the digital controller stage. Generated stimulus waveform was then programmable, charge balanced, biphasic (anodic and cathodic current) proving great flexibility, high degree of accuracy and safe. Such performances, largely verified by different simulation levels, confirm originality in our proposed conception.
Programmable current source was based on a Digital to Analogue Converter (DAC) with 10 bit resolution and piloting sixteen stimulation sites (Fig. 7) (Ghorbel et al., 2004a, b; Ghovanloo and Najafi, 2005). Three major components were then used: 10 bit Digital to Analogue Converter (DAC), one current driver and one output controller.
Output controller used in this architecture consists of an analogue switch array generating fully balanced bipolar stimulus that was necessary to prevent charge accumulation within electrode-tissue interface.
Fig. 8: | 10-bit thermometer-coded DAC architecture |
Since each channel (active electrode and return electrode) would be activated exclusively at one time, only one current DAC is needed. On the other hand, in bipolar mode, only one voltage supply and only one current driver were needed to generate a biphasic stimulus as anodic current and cathodic current.
Fig. 9: | Block diagram of a modified 10-bit thermometer-coded DAC |
Fig. 10: | 5-bit thermometer-coded DAC |
Using a single DAC and a single current driver reduces silicon area in this circuit conception and avoids any possible difference between anodic and cathodic stimulation currents.
DAC sesign: The current steering DAC design was based on an array of matched current sources (10 bit thermometer-coded DAC) (Ghorbel et al., 2004a; Lin and Bult, 1998). Since 10 bits were needed in this design, this would form 210-1 = 1023 current sources to be used, which was relatively complex as design (Fig. 8).
In order to reduce decoding complexity in such thermometer-coded architecture, one 10 bit modified thermometer-coded DAC was proposed as two 5 bit thermometer-coded DAC modules. Hence, 1023 current-sources would be simplified to only two blocs composed each one by 31 identical current sources (62 current sources in total). In this case, it was possible to reduce the number of elements minimizing then silicon area and simplifying the decoding logic. Our proposed 10 bit thermometer-coded DAC block diagram was shown in Fig. 9 including two same 5-bit thermometer-coded DACs.
Digital binary inputs are thermometer-decoded controlling 62 identical current sources. Output current sources, which are switched ON/OFF according to the digital input codes, would be summed and driven into the output driver in order to generate the required analogue stimulating current.
At each 5-bit thermometer-coded DAC, D0D1D2D3D4 used as input of the first thermometer decoder permit the control of 31 current-sources (S1 to S31), using the same current-reference (Iref). Each current source delivers one specified current level referred as ILSB when selected. If D4D3D2D1D0 = 11111, current output would be then Iout = 31 x ILSB. The other bits (D5D6D7D8D9) used as input to the second thermometer-decoder in order to select current sources S32 to S62. These current-sources use the same current reference (32 x Iref, where Iref is the current reference of sources S1 to S31). In this way, each of these current sources (S32 to S62) delivers 32 x ILSBS. If D9D8D7D6D5 = 11111, the output current would be Iout = 31 x 32 x ILSB = 992 ILSB.
In full scale, D9D8D7D6D5D4D3D2D1D0 = 1111111111, maximal output current would be then: Iout(max) = (31+992) x ILSB = 1023 x ILSB.
Figure 10 shows sources dispatching within the proposed 5 bit thermometer-coded converter. Each individual current source was controlled by the thermometer row and column decoder (Ghorbel et al., 2004a; Bastos et al., 1998; OSullivan et al., 2004; Albiol et al., 2004; DeMacro et al., 2003). Any binary combination input equal to P in decimal would involve an output current value equal to Iout = Px ILSB.
Current driver: In our design, the current driver was used mainly to amplify DAC output current (Iout = K x IDAC) and to forward this current to the specified electrode. To have good linearity and better precision, output impedance of the current source should be very high than the considered load representing the auditory nerve impedance (Ghorbel et al., 2004b).
The following electronic conception was proposed for current drive design. The use of wide swing cascode current mirror structure gives satisfactory output impedance to have a good linearity for the stimulation current (Fig. 11). DAC current would be mirrored with a fully cascode current mirror formed with M1-M4 and it would be passed into the reference branch of the wide swing cascode mirror formed with M5 and M6. Output stage M7 and M8 are 10 times wider in order to mirror the current DAC up to full-scale level of 1023 μA into 1 KΩ load (Vanpoucke et al., 2004; Ghorbel et al., 2002; Sawan et al., 1995).
Fig. 11: | Current driver architecture |
Fig. 12: | A H architecture |
For convenient electronic functioning of such device, all conceived transistors should be kept in the saturation region. This means that theses transistors need minimum voltage VDsat across their drain-source terminals, defined by the following equation:
Where ID is the drain current, KP is the intrinsic transconductance and W and L are the MOS transistor width and length, respectively. Obviously, VDsat increases with higher stimulation currents, even though it can be reduced by increasing the width of the MOS transistor at the expense of more area consumption (Ghorbel et al., 2004b; Ghovanloo and Najafi, 2005; Sivaprakasam et al., 2005).
Output controller: The output controller has to produce a bi-directional stimulating current to the targeted electrode in order to avoid charge accumulation in the biologic tissue. This task was done by an H architecture circuit as shown in Fig. 12 and was controlled by simple logic gates (Suaning and Lovell, 2001; Ghorbel et al., 2004a, b; Sivaprakasam et al., 2005).
In the Fig. 12, site i and site j were referred to two different electrode placements. During anodic stimulation of site i, current flow pursue the way M1 Resistance M4. The cathodic stimulation would be then via M3 Resistance M2. Resistance was the equivalent impedance of the considered biological tissue between site i and site j.
Control of current direction in the different stimulation electrodes was assured by four input signals (Fig. 7):
Direction signal was used to determine the current direction through the nerve.
Mode signal was used to choose between two stimulation modes:
• | Mode = 0: Active electrode (anodic current) could be selected (addressed) according to programming, but the return electrode (cathodic current) would be fixed to the common reference electrode (electrode number 0). |
• | Mode = 1: Both active electrode (anodic) as well as return electrode (cathodic) could be selected (addressed) according to programming. |
Select signal was used to select the active electrode.
Select* signal was used to select the return electrode.
Select* would be omitted if Mode=0.
SIMULATION AND RESULTS
Microelectronic conception of the prototype device was designed according to the technology AMI 0.35 μm three-metal two-poly N-well standard CMOS process. The full cochlear micostimulator layout is shown in Fig. 17 and it is 400 μm high and 800 μm wide for a total area of 0.32 mm2. Table 2 summarizes the chip specifications.
Digital controller stage simulation: The aim of the digital controller simulation was merely to identify stimulation commands and to generate appropriate control signals to the analogue stimulation stage.
Table 1: | Summarize for the waveforms shown in Fig. 13 |
The incoming serial data bits stream (Data) are received at 1 MHZ clock rate (CLK), which consists of back to back command-frames in the format as shown in Fig. 5. At start-up, a Power-on-reset (POR) circuit activates a global reset line (RAZ = 0) and initialise the entire digital controller to safely start the digital circuitry from one known state.
The circuit begins then data reception and analyses the incoming data frames. If the stimulation command is detected, a signal detect is set to 1. In this case, the stimulation parameters are memorized and the adequate control signals are generated. Fields t, sp, sp* and i, shown in Fig. 13, specify the stimulation parameters:
• | Field t[7:0] specifies one half of the biphasic pulse width. |
• | Time increment is 1 time CLK period (1 μs). |
• | Field s[3:0] determines the stimulating electrode number |
• | Field sr[3:0] specifies the reference electrode to provide a return path for stimulus. |
• | Field i[9:0] represent the stimulation amplitude to be programmed on the DAC. |
The control signals mode, stim, sens1, ni and pj (where i, j ε{0,16}) drives the analogue stimulation stage to provide charge balanced electrical stimulation pulses:
• | Signal mode determines that the reference electrode is specified by the sp* field (mode = 1) or is fixed as the default return electrode (mode = 0). |
• | Signal stim signal defines the stimulation enabled period; it is equal to the biphasic pulse duration. During this period of time where stim = 1, the microstimulator is in processing. |
• | Signal sens1 determines cathodic and anodic phases of stimulus. The time periods of the cathodic and anodic current pulses must be equal in order to generate a fully balanced bipolar stimulus which is necessary to prevent polarization of the nerve by charge injection and accumulation at the electrode_tissue interface. |
• | Signals ni and pj control the switches dedicated to generate the stimulation current direction via two pairs of transistors (NMOS and PMOS used as switches). In the stimulation period, two electrodes are selected (active and return electrodes) all other electrodes are in high impedance. |
Figure 13 shows the results of simulation of the digital controller. It describes an example of three stimulation commands.
Table 1 summarizes the different signals generated by the digital controller for this example.
Analogue stimulation stage simulation
DAC linearity: Simulation results showed good performances of our designed analogue stimulation stage. Figure 14 shows the DAC output waveforms during digital input increment (D0-D9) at 1MHz frequency rate, beginning from 0 to 1023.
The first DAC uses increments of one LSB = 101 nA whereas the second DAC uses increments of one MSB = 32 x LSB = 3.232 μA.
In the full scale for the first DAC, the output current amplitude is equal to 31 x LSB = 3.128 μA whereas for the second DAC the output current amplitude is equal to 31 x MSB = 100.192 μA. In this case, in the full scale of the whole 10 bit current DAC, the output current amplitude is equal to 3.232+100.192 = 103.424 μA.
Current driver accuracy: Figure 15 shows the input-output characteristics of the driver circuitry with different charges where the input current was varied from 0 to 120 μA with an amplification factor of 10.
Current drive simulation results were satisfactory and showed good performances. Output current has good linearity over the entire range of its ranging when Rload< 3 KΩ (VCC = 3.3V and Imax = 1140 μA).
It could achieve a voltage compliance of 86% (450 mv of 3.3 V) supply voltage, while maintaining high output impedance in the 30 MΩ range to keep the desired stimulation currents constant within 1%.
Fig. 13: | Digital controller Simulation: three stimulation commands analyses waveforms |
Fig. 14: | Output DAC waveforms at 1 MHZ clock rate |
This current driver will have the ability to stimulate higher impedance nerves without saturation problems. This can be explained by the fact the source can support a 2.850 V drop through the nerve without saturation.
Stimulus waveforms: Mixed simulation was used in combining Digital controller stage to the Analogue stimulation stage for the entire functioning of our micro-stimulator. Figure 16 shows stimulation current waveforms passing through 1 KΩ resistor that was chosen for representing biological tissue between two stimulating sites (Vanpoucke et al., 2004; Sawan et al., 1995).
Fig. 15: | Current drive Input-output characteristics waveforms |
Same command-frames shown in Fig. 13 were identically used in this example. First command data frame permitted to select electrode 7 as active and to select electrode 4 as return. As a result, 302.894 μA flows from Site 7 to Site 4 in the first stimulus phase for the 32 μs duration. In the second phase, the same current amplitude value (302.894 μA) flows back from Site 4 to Site 7 during the same time period (32 μs). All of others sites are puts in the high impedance state.
Fig. 16: | Biphasic stimulus currents waveforms |
Fig. 17: | Full cochlear micostimulator layout |
By using only one current driver to produce stimulation current in the biological tissue for the two directions, we ensure then the same amplitude was generated for the cathodic current and for the anodic current. In this way, the charge balanced biphasic pulses are obtained and no charge accumulation exists.
The second command data frame generates a bipolar biphasic current pulse between Site 2 and Site 6 with ±807.693 μA amplitude and 63 μs phase duration. Whereas the third command data frame generates a current pulse between Site 5 and Site 0 (the common return site) with ±504.784 μA amplitude and 45 μs phase duration.
Power consumption: Great attention was given to power consumption since it concerns autonomy of such electronic implanted device. During our microelectronic conception of this proposed microstimulator, we tried to optimise and to minimise several electronic parts so to achieve a high economic degree. Hence, total power consumption of this electronic chip was estimated according to two origins namely:
• | Power consumption for the digital controller stage referred to as Pdigtal. |
• | Power consumption for the analogue stimulation stage referred to as Pdac. |
• | Power dissipated within the biological tissue during stimulation (load), referred to as Pstim. |
We should note the maximum dissipation for one electrode: Full 1034 μA current amplitude injected within an approximate 1 KΩ load as equivalent biological impedance (worst case). So the average power dissipated would around 1.675 mW which the highest dissipated power by the chip.
The full cochlear micostimulator layout is shown in Fig. 17 and it is 400 μm high and 800 μm wide for a total area of 0.32 mm2. Table 2 summarizes the chip specifications.
Table 2: | Microstimulator specifications |
A prototype of fully integrated microstimulator dedicated to flexible and fully programmable stimulation for cochlear implant was conceived. This device including one digital controller stage, one analogue stimulation stage connected to sixteen-stimulation channels was the main electronic part of the cochlear implant.
Our specific integrated circuit design was versatile and totally numerical in order to achieve a high degree of programmability. It could be adapted then to any other external part of one cochlear prosthesis the sound analyser that could be driven by a digital signal processor DSP.
Main functions assured by the inner part during digital processing permitted to determine with great flexibility stimulation current level and width to generate at each specified channel as well as the stimulation rhythm.
Our implantable microstimulator involving electrical stimulation within auditory nerve fibbers included sixteen programmable channels and was provided with several originalities that concerned the numerical stimulus current level ranging around±1 mA with 10 bit resolution, the ~30 MΩ output impedance as well as the numerical current pulse width ranging up to 255 μs with 8 bit resolution.
Microelectronic design of this biomedical device was based on the specific following technology: the AMI 0.35 μm, 3-metal, 2-poly, n-well standard CMOS process. Using 3.3 V supply, the maximum power consumption was estimated to 1.675 mW and the chip size was evaluated to 0.320 mm2.
Finally, it was carefully noted that our design concerning fully programmable neurophysiologic stimulator could be adapted to other medical application involving nerve stimulation.