CHAPTER 2 ANALYSIS OF RETINA INTERFACE
2.3 Measurement Results and Discussion
In the measurement of retina interface impedance, one end of the probe of impedance analyzer was fixed at column 1 of the wire-bonded recorder as the other end moved from column 2 to column 15 to measure interface impedances of different electrode distances. Experiments were done on row A to row H to measure the impedance of multi sizes of electrodes. Frequency bandwidth was between 40Hz and 100kHz with 201 recording points. AC signal was sine wave with 500 mV magnitude, and no DC bias potential was included. The measurement results of frequency to impedance curve are shown in Fig. 2.6. Randomly choose data number 1, 5, 10, 20, 40, 80, 120, 160 and 201 totally 9 points from the recorded points. Frequency and impedance axes are in logarithm scale. Figure 2.6(a) shows the impedance between column 1 and column 6 of the big electrode chip; Fig. 2.6(b) shows the impedance between column 1 and column 7 of the small electrode chip. In each figure, the smaller electrodes have larger impedance at low frequency. But 10-μm and 40-μm side length electrodes did not yield consistent results, so these sizes were not included in Fig. 2.6(b). The impedance of small electrode sets was not always larger than big electrode sets, because these two kinds of sets were separated into two chips.
Therefore, in the following analysis, we studied the results of big electrode chip only.
Besides the impedance between column 1 and column 2 and between column 1 and column 15 of the big electrode chip, other columns yielded consistent results, so these columns were not included in further analysis. These problems may come from that retina did not and very closely attach to the chip thus the measurement results were not very reliable at the two ends of microelectrode chip. In this case, large contact resistance (Rco) was shown in the retina interface that would disturb the measurement results.
Looking into the relationship between retina interface impedance and electrode distance, we compared the relationship of row C at low frequency and at high frequency. Figure 2.7(a) shows the electrode distance to impedance curve at low frequency (40Hz, 0.5kHz, 1kHz and 1.5kHz); Fig. 2.7(b) shows the electrode distance to impedance curve at high frequency (25kHz, 50kHz, 75kHz and 100kHz). No consistent electrode distance to impedance relationship at low frequency, whereas impedance increases with increasing electrode distance at high frequency. These results indicate that the impedance modulus is not highly dependent on Rco at high frequency. Suggesting that Cdl was shunted with Rco, ZCAP was smaller than Rco at
high frequency and thereby Rco didn’t domain the interface impedance.
In other studies, electrochemical impedance spectroscopy (EIS) has been used to electrically characterize the retina interface, but the analyzing process acquired from EIS is complicated. The benefit of using an equivalent model is to develope a better understanding of the physical processes occurring at the biomedically relevant interface. Therefore, a curve fitting tool of the MATLAB was used in this work to analyze the approximate value of Cdl,Rct,Rs and Rre. To simplify the equation, we grouped Rs and Rre into tissue resistance (Ru). From Fig. 2.1, the equivalent equation of retina interface is composed of magnitude portion and phase portion:
2 elements can fit a frequency to impedance curve. Three elements can be found by the tool as three constants. Thus the magnitude of impedance is the function of frequency.
Figure 2.8 shows the fitted Rct and Cdl results of big electrode sets. A large variance (about 10%) was in the fitted results as we analyzed each element by curve fitting tool rather than EIS. For most of the electrode distances, Rct increases with decresing electrode size and Cdl increases with increasing electrode size. Rct and Cdl are not relative to electrode distances. To obtain Rre, we subtracted Rs from Ru. Figure 2.9(a) shows the calculated results of Rs by using the formula in section 2.1. Rs of different electrode distances were viewed as a constant. Figure 2.9(b) shows the fitted Rre
results of big electrode sets. For most of the electrode sizes, Rre increases with increasing electrode distance. But the value of Rre for a smaller size of electrode was less than a bigger one; this fact was opposed to our prediction. We predicted that a bigger size of electrode would contact with bigger retinal area, and the resistance of bigger area is less. However, since large Rco was in the interface between retina and microelectrode chip, the relationship between Rre and electrode size was not correct.
Therefore, contact between retina and microelectrode chip has to be improved by the weight of heavier stuff than filter paper (e.g. cover glass). Therefore, most amount of stimulation current would flow into the tissue without the bad influence of Rco.
Even though the fitted results were not very ideal, defining an approximate range of retina interface impedance is still capable. According to the analyzed results of different electrode distances and multi sizes, three elements were summarized in Table
III. Rre is about 10kΩ to 200kΩ, where as Cdl is about 0.01nF to 1nF and Rct is about 1MΩ to 100MΩ for big electrode sets. Comparing to the work done by Humayun et al.
[22], with similar electrode sizes (50μm side and 50μm diameter), electrode distances (745μm and 800μm), and measurement environment, our results are very similar to theirs. In our work and in [22], Cdl is 0.031nF and 0.033nF; Ru is 32kΩ and 74kΩ, respectively. The comparison between [22] and this work is depicted in Table IV.
Table III Fitting results of retina interface impedance of big electrode sets ι (μm)
Table IV Comparison of Humayun et al. in 2007 [22] and this work Humayun et al., 2007 [22] This Work
Animal tissue Porcine eye Rabbit eye
Electrode type Polyimide thin-film Aluminum
Frequency bandwidth 10Hz to 100kHz 40Hz to 100kHz Measuring instrument Commercial potentiostat
(FAS1, Gamry, Inc)
Impedance analyzer (4294A, Aglient, Inc)
AC signal 10 mV sine wave 500 mV sine wave
DC signal No DC bias Potential No DC bias Potential
Cdl 0.033nF 0.031nF
Ru 74kΩ 32kΩ
(a) (b)
Figure 2.6 Frequency to impedance curve of (a) big electrode sets and (b) small electrode sets. The electrode distance between 80μm x 80μm electrodes is 348μm; the space between 30μm x 30μm electrodes is 387μm.
(a) (b)
Figure 2.7 Distance to impedance curve of 60-μm side length electrodes. Measuring frequency is at (a) 40Hz, 0.5kHz, 1kHz and 1.5kHz (b) 25kHz, 50kHz, 75kHz and 100kHz.
(a) (b)
Figure 2.8(a) Distance to impedance (Rct) curve of big electrode sets (b) distance to capacitance (Cdl) curve of big electrode sets.
(a) (b)
Figure 2.9(a) Distance to resistance (Rs) curve of big electrode sets (b) distance to resistance (Rre) curve of big electrode sets.
CHAPTER 3 Solar Cell Array for Efficient
conduction band, producing a current which travels against diode polarity. This current is called the “photocurrent,” and can forward bias the diode, producing a so-called “photovoltage.” The photovoltage and photocurrent can be utilized to drive a passive load; in this case the cell is operating photovoltaically. It is sometimes desirable to place an external bias voltage across the photodiode; in this case the cell operates photoconductively. In both photovoltaic and photoconductive systems photocurrent is proportional to incoming light intensity. The linear conversion of light into current makes a photodiode an attractive data delivery option. Photovoltaic circuits, which are composed of solar cells, can also be used to deliver electrical power. A photodiode-based (passive) retinal prosthesis may employ either photovoltaic or photoconductive circuits. In the simplest photoconductive design, a bias voltage is provided to a photodiode in series with the retinal tissue load. In the simplest photovoltaic design, a single photodiode, which is so-called “solar cell,”directly drives current into the retina. This approach does not require an external power supply, as each pixel generates current without a bias.
Cross-sections of P+/N-Well solar cells with floating P-substrate in 0.35μm 2P4M technology are proposed in Fig. 3.1. If solar cells are not connected in series, then each solar cell is called single solar cell (SSC). As light irradiates a SSC, if P+ has low voltage (LV) and N-Well is connected to stimulating electrode, then the SSC can directly drive high stimulation current (HI) into a passive load. If N-Well has LV and P+ is connected to a passive load, then anode has high voltage (HV) and the SSC can drive the load with photovoltage. Solar cells can not deliver power to CMOS circuits in this technology, because P-Substrate is usually connected to ground when solar cells and NMOS are in parallel connection. Nonfloating P-Substrate will results in the
leakage current problem of parasitic-BJT as shown in Fig. 3.2.
Photocurrent generated from a passive device could barely stimulate the retina if the output load is too large. Solar cells in series connection have higher photovoltage and thus the stimulation current is larger than those without series connection. Besides, with high photovoltage, solar cells can deliver higher voltage supply to CMOS circuits. To solve the problem depicted in Fig. 3.2, solar cells can be built in the deep N-Well (DNW) of 0.18μm 1P6M technology. The cross-sections of two N+/P-Well solar cells in series connection are proposed in Figure 3.3. P-substrate is kept floating to reduce the parasitic-BJT problem. If two solar cells are connected in series, then each pair of solar cells is called two cascode solar cells (TSC). As light irradiates a TSC, if P-Well has LV and N+ is connected to stimulating electrode, then the TSC can directly drive HI into a passive load. If N+ has LV and P-Well is connected to a passive load, then anode has HV and the TSC can drive the load with photovoltage.
The photovoltage of a TSC is twice of a SSC.
(a) (b)
Figure 3.1 Cross-sections of P+/N-Well solar cells with floating P-Substrate in 0.35μm 2P4M technology. The functions are driving passive loads with (a) photocurrent and (b) photovoltage.
Figure 3.2 Cross-section of P+/N-Well solar cell and NMOS in parallel connection in 0.35μm 2P4M technology. Leakage current is induced by parasitic BJT from nonfloating P-Substrate.
(a)
(b)
Figure 3.3 Cross-sections of two N+/P-Well solar cells in series connection with floating P-Substrate in 0.18μm 1P6M technology. The functions are driving passive loads with (a) photocurrent and (b) photovoltage.
3.1.2 Single and two cascode solar cell array stimulation chips
To increase retinal stimulation current, either photo-sensing region of a solar cell has to be enlarged or many solar cells have to be connected in parallel. According to [25], a solar cell with 4 times photo-sensing regions of a 5μm x 5μm solar cell can only generate about 2 times photocurrent under same light intensity. Based on the cross-sections of Fig. 3.1, the layout view of two P+/N-Well solar cells in parallel connection is shown in Fig. 3.4. The diffusion area and space between each cell is 5μm x 5μm and 1.1μm, respectively. The layout view of single solar cell array stimulation chip (SSCA stimulation chip) is shown in Fig. 3.5. SSCA stimulation chip was composed of 16-pixel SSCA. Each pixel was composed of a solar cell array (SCA) and two ends of SCA were connected to a stimulating electrode and a return electrode.
SSCA was composed of 2540 SSCs in parallel connection. The stimulating electrode was placed in the middle part of the pixel and was surrounded by SSCs. SSCA of each pixel was surrounded by local surrounding return electrode. The dummy metals surrounded whole pixels were cut off four corners for biocompatibility. The summary of the chip is shown in Table V. Implantable SSCA stimulation chip was fabricated in tsmc 0.35μm 2P4M technology. Whole chip size is 2.45mm x 2.45mm.
In order to increase the voltage supply, we connected two solar cells in series to provide twice photovoltage based on the cross-section of Fig. 3.3. Figure 3.6 shows the layout view of two cascode N+/P-Well solar cells connected in parallel. Building solar cells in the DNW takes a lot of area and the space between two N-Wells has to be larger than 5μm, so we enlarged the diffusion area and space between each cell to 10μm x 10μm and 14.1μm, respectively. The layout view of two cascode solar cell array stimulation chip (TSCA stimulation chip) is shown in Fig. 3.7. TSCA stimulation chip is composed of 1-pixel TSCA. Each pixel is also composed of a TSCA. TSCA is composed of 414 TSCs in parallel connection. The stimulating electrode is placed in the middle part of the pixel and is surrounded by TSCs. TSCA of each pixel is surrounded by local surrounding return electrode. The summary of the chip is shown in Table VI. Implantable TSCA stimulation chip was fabricated in tsmc 0.18μm 1P6M technology. Whole chip size is 0.976mm x 0.962mm.
In our past design, the return electrodes of all pixels were connected together and placed far from the stimulating electrodes. Output load of a stimulator increases with decreasing electrode size and increasing return electrode to stimulating electrode distance as mentioned in the previous chapter. Thereby the return electrode of each pixel was designed to be surrounding around whole SCA. Each pixel with its own local return electrode rather than sharing the same return electrode or placing the return electrode in a remote place decouples electric fields of neighboring electrodes and decreases cross-talk and interference [22]. Therefore, in a confined stimulation region, spatial resolution of pixel array should be better. However, we have no ideal whether this return electrode design could enhance subretina stimulation efficiency or not, so simulation on the current distribution in the retina of SCA has to be done.
Table V Summary of the SSCA stimulation chip SUMMARY
Technology Tsmc 0.35μm 2P4M
Power supply voltage No external power supply Photo-sensing area per solar cell 5μm x 5μm
Solar cell number per pixel 2540
Exposured electrode 75μm x 75μm
Photosensing area per pixel 388μm x 388μm
Pixel size 495μm x 495μm
Pixel number 4 x 4 pixels
Chip size 2450μm x 2450μm
Table VI Summary of the TSCA stimulation chip SUMMARY
Technology Tsmc 0.18μm 1P6M
Power supply voltage No external power supply Photo-sensing area per solar cell 10μm x 10μm
Solar cell number per pixel 414 x 2
Exposured electrode size 75μm x 75μm
Photosensing area 754μm x 754μm
Pixel size 930μm x 930μm
Pixel number 1 pixel
Chip size 976μm x 962μm
Figure 3.4 Layout view of two SSCs in parallel connection.
Figure 3.5 Layout view of SSCA stimulation chip. Whole chip size is 0.976mm x 0.962mm.
Figure 3.6 Layout view of two TSCs in parallel connection. Solar cells are built in the DNW.
Figure 3.7 Layout view of TSCA stimulation chip. Whole chip size is 0.976mm x 0.962mm.
3.2 Simulation Results
Regardless of circuit specifics, prosthesis pixels interface with the body through microelectrodes. When SCA is implanted beneath the pigment epithelial layer (PRL) as shown in Fig. 3.8, three main layers are expected to be stimulated: photoreceptor layer, horizontal/bipolar layer and ganglion cell layer. Each layer can be viewed as 2-dimensional spreading resistive network; three layers can be viewed as 3-dimension (3D). In order to simulate current distribution in the retina and to evaluate which kind of return electrodes is the best one for SCA, 3D spreading resistive network model, as shown in Fig. 3.9, has to be involved in the simulation of retinal stimulation current.
SCA directly drives current to the retina from the return electrode to the stimulating electrode.
The surfaces of three main layers of retina are defined as layer 1, layer 2 and layer 3. Layer 1 is closest to the SCA. Impedances of each layer are thought to be equal. where i represents 3 different layers. Assuming that the stimulating electrode of SCA attaches to 3 x 3 sets of resistances and the retinal tissue is composed of 256 x 256 sets of resistances. Stimulation current flows from the stimulating electrode to the return electrode. Potential of the return electrode is represented by ground. Figure 3.10 shows the simplified schematic of the first tissue layer. The resistances are represented by the lines between two intersections to better understand the relationship among different positions. Elements of the matrix are named by x-coordinate and y-coordinate. Origin of the coordinates locates on the middle point of the matrix, and it is also middle point of the stimulation site and probe region.
Stimulation site is where the stimulating electrode attaches to the retina; 3 x 3 nodes surrounding the origin of the resistive network are involved. We probed the current passing through Rti, and probe region involves 5 x 5 nodes surrounding the stimulation site of the resistive network.
In addition to the proposed local surrounding return electrode in this work, there
are two other kinds of return electrodes as shown in Fig. 3.11. One is remote surrounding return electrode which is far away from the stimulation site in ring shape.
The other one is remote single return electrode which is also far away from the stimulation site in square shape. Electrode distance of local return is 2 columns, which is very close to the stimulation site; whereas electrode distance of remote return is 128 columns. Single remote return is at the position (-128, 0).
The simulation model of each solar cell in the circuit is established based on the previous experimental results. As strong light irradiates a 5μm x 5μm solar cell, generated photocurrent is about 10nA [25] as shown in Fig. 3.12(a). In the simulation, we shunted a current source near a solar cell to be the generated photocurrent as shown in Fig. 3.12(b). In comparison of which design of return electrodes can stimulate higher current at retina tissue, we probed the current passing through Rt1.
Simulation results of one pixel of SSCA stimulation chip with three different kings of return electrodes are shown in Fig. 3.13. With 25.4μA photocurrent, current in the stimulation site, i.e. (-1, -1), (-1, 0), (-1, 1), (0, -1), (0, 0), (0, 1), (1, -1), (1, 0), (1, 1), taking the major portion of the generated photocurrent. The minimum current in the stimulating area is at the (0, 0); the maximum current (Imax) is at the four corner, because the equivalent resistance at the latter node is lower than the former node.
Imax of the remote single return structure is 101.1nA; Imax of the remote surrounding return structure is 344.7nA; Imax of the local surrounding return structure is 1085.4nA. The Imax of local surrounding return is 10 times of remote single return structure and 3 times of remote surrounding return structure. Thereby SCA with local surrounding return electrode has higher current efficiency than other return electrode designs.
The simulation method of TSCA is similar with SSCA. Because the photo-sensing region of each solar cell of TSC is four time larger than SSC, we assumed the zero bias photocurrent of each solar cell is 40nA as shown in Fig. 3.14 (a); then 414 TSCs can generate 16.56μA photocurrent. In the simulation, we shunted a current source near the solar cells as the generated current in Fig. 3.14(b). Simulation results of TSCA with three different kings of return electrodes are shown in Fig. 3.15. Imax of the remote single return structure is 217.1nA; Imax of the remote surrounding return structure is 697.5nA. Imax of the local surrounding return structure is 1917.8nA. The Imax of local surrounding return is 10 times of remote single return structure and 3 times of remote surrounding return structure. These ratios are very similar with SSCA.
Even though the current generated from the SSCA is 1.5 times larger than the TSCA, the stimulation current of the latter is twice larger than the former, because
photovoltage of the latter is twice larger than the former. Thereby TSCA with local surrounding return electrode is more current efficient than SSCA. Stimulation current not only changes with different return electrode designs but also decreases with the depth of retinal layer for both SSCA and TSCA. Figure 3.16 shows the SSCA simulation results of stimulation current passing through three different retina layers with different return electrode designs. No matter what kind of return electrode is, stimulation current decreases 1.5 times with increasing retina layers if impedance of each layer is the same.
The quantity of solar cells decides the electrode distance. More solar cells generate more stimulation current but also result in farer electrode distance. To find
The quantity of solar cells decides the electrode distance. More solar cells generate more stimulation current but also result in farer electrode distance. To find