Measurement Device
A wearable measurement device that features solar cells was built. The primary task of this measurement device is to continuously monitor the solar cells’ output power. A key element is that the solar cells are covered by optical filters to simulate subcutaneously implanted solar cells.
Principle
The measurement device includes a circuitry to measure the solar cell’s power output and stores the data on a memory card. Figure 1 shows the block diagram of the measurement device: the ambient light (1) is filtered by dedicated optical filters (2) that exhibit similar optical properties as human skin. The filtered light (3) irradiates the solar cells (4), which are connected to a maximum power point tracker (MPPT, 7). The MPPT maximizes the power output of the solar cells, which is measured by current \(I_{\text{S}}\) [A] (5) and voltage \(U_{\text{S}}\) [V] (6). A microcontroller (8) controls the analog-to-digital converters (ADCs) from the power measurement (5, 6) and stores the data onto a memory card (9).
Optical Filters
To estimate the power output of a subcutaneous solar cell, the solar cells of the measurement device have to be covered by a material that exhibits similar optical properties as human skin. Real non-vital skin would change its optical and biologicals properties rapidly over time and would not allow reliable and reproducible results. Thus, an alternative material with stable properties had to be implemented. Optical properties of human skin are documented well in literature, in vivo
18 and in vitro
1,9,14,16,17 data of human skin are presented, giving the wavelength-dependent absorption coefficient \(\mu_{\alpha }\) and transport scattering coefficient \(\mu '_{\text{s}}\). However, reported data differs, especially in the near infrared range (NIR).1 Moreover, for this application, the major interest is the wavelength-dependent total light transmission ratio (i.e. transmittance = \(T_{\lambda }\)[%]).
To attain more data of the transmittance and to be able to also investigate skin with different layer thicknesses (dermis, subcutis), \(T_{\lambda }\) of several ex vivo domestic pig skin flaps was measured, since anatomical and optical properties of pig skin are similar to human skin.19 In total, 16 skin samples (purchased from the slaughterhouse, size 15 × 15 mm) were cut out at different hairless sites. The total thickness of the samples (epidermis + dermis + subcutis) varied from 1.6 mm up to 6.8 mm. The skin samples were placed on a glass plate and an optically opaque frame with a translucent window of 4.5 × 4.5 mm was placed on the epidermis. Finally, \(T_{\lambda }\) was measured with a spectrophotometer (UV-3600 UV–Vis-NIR Spectrophotometer, Shimadzu, Japan) over the wavelength range of 300 to 1500 nm in 2 nm steps. The measurement was done directly after purchasing the fresh skin and therefore, no significant changes of the optical properties are expected.5
Subsequently, the applicability of the transmittance results to skin-covered solar cells was investigated: The computed short circuit current per solar cell area \(I_{{{\text{SC}}\_{\text{C}}}}\) [A m−2], calculated based on the transmittance results, and the measured short circuit current \(I_{{{\text{SC}}\_{\text{M}}}}\) [A] were compared.
First, \(I_{{{\text{SC}}\_{\text{C}}}}\) for a skin-covered, irradiated reference solar cell with known external quantum efficiency (\({\text{EQE}}\) [%], ratio of incident photons to converted electrons), was calculated. \(I_{{{\text{SC}}\_{\text{C}}}}\) was obtained by integrating the photon flux \(\phi_{{{\text{AM}}1.5{\text{G}}}}\) [# photons m−2 s−1], in our case for the spectrum AM1.5G (100 mW cm−2), the previously measured transmittance spectra \(T_{\lambda } (\lambda )\) and \({\text{EQE}}(\lambda )\) and multiplying the result with the electron charge \(q\) = 1.6 × 10−19 C.
$$I_{{{\text{SC}}\_{\text{C}}}} = q \times \int {\phi_{{{\text{AM}}1.5{\text{G}}}} \left( \lambda \right)T_{\lambda } \left( \lambda \right){\text{EQE}}\left( \lambda \right){\text{d}}\lambda }$$
(1)
Second, the same skin samples were put on the standard reference solar cell (4.5 × 4.5 mm). On top of the skin samples, an optically opaque frame (10 × 10 mm) was stacked to prevent lateral irradiation. The setup was irradiated by a solar simulator (LS0811, LOT-Quantum Design, Germany) simulating 1 sun (100 mW cm−2, AM1.5G). The measured short circuit current \(I_{{{\text{SC}}\_{\text{M}}}}\) was determined with a sourcemeter with four terminal sensing (Keithley 2400, Keithley Instruments, USA) during irradiation. The difference between \(I_{{{\text{SC}}\_{\text{C}}}}\) and \(I_{{{\text{SC}}\_{\text{M}}}}\) will be discussed in the Discussion section.
Based on the obtained transmittance results, a combination of two optical filters that emulate the transmittance profile \(T_{\lambda } (\lambda )\) of 2.3 mm thick skin (Fig. 3) was determined: A 550 nm longpass (FGL550S, Thorlabs, Germany) and an absorptive neutral density filter (NE205B, OD: 0.5, Thorlabs, Germany) were stacked on top of each other. These filters are placed directly on top of the solar cells of the measurement device (Fig. 2). The wavelength-dependent short-circuit current \(I_{\text{SC}} \left( \lambda \right)\) [A m−2] was computed to determine the performance of the optical filters compared to real skin. For full sun irradiation (AM1.5G, 100 mW cm−2), \(I_{\text{SC}} \left( \lambda \right)\) is obtained by the following formula, whereas the transmittance spectra of either the optical filters or the 2.3 mm thick skin is used for \(T_{\lambda } \left( \lambda \right)\).
$$I_{\text{SC}} \left( \lambda \right) = \phi_{{{\text{AM}}1.5{\text{G}}}} \left( \lambda \right)T_{\lambda } \left( \lambda \right){\text{EQE}}\left( \lambda \right)q$$
(2)
Solar Cells
The employed solar cells are monocrystalline silicon cells (KXOB22-12X1L, IXYS Corporation, USA) with an efficiency of 22%. Three cells are connected in series and the resulting total active area is 3.6 cm2. The three solar cells are aligned in parallel which results in a total outer dimension of 21 × 22 mm.
Measurement Circuit
The maximum power point tracker (BQ25570, Texas Instruments, USA) modulates its input impedance to regulate \(U_{\text{S}}\) at 80% of the solar cells’ open-circuit voltage (\(U_{\text{OC}}\)). At \(U_{\text{S}}\) = 80% \(U_{\text{OC}}\), the solar cells exhibit maximum output power. The MPPT obtains the latest value for \(U_{\text{OC}}\) every 16 s, which takes 256 ms and during this time the solar cells' output current \(I_{\text{S}}\) is zero. \(I_{\text{S}}\) is measured by a high-side current-sense circuitry featuring a 1 Ω sense resistor connected in series between the solar cells and the MPPT. The voltage drop over the sense resistor is amplified (ADA4051, Analog Devices, USA) and digitized by a two-channel 16-Bit ADC (ADS1118, Texas Instruments, USA). The system makes use of the ADC’s internal programmable gain amplifier (PGA) to increase the resolution for currents in the µA-range. The circuit can measure currents up to 15 mA with a dynamic resolution between 80 nA and 0.6 µA, theoretically. This range covers full sunlight irradiation (AM1.5G, 100 mW cm−2) considering attenuation of the filters as described previously. The solar cell’s output voltage \(U_{\text{S}}\) is buffered by a unity gain amplifier to minimize measurement induced error and then digitized with a resolution of 0.6 µV by the same ADC, range \(U_{\text{S}}\) = 0–2 V. The microcontroller coordinates the measurement: \(I_{\text{S}}\) and \(U_{\text{S}}\) are measured with a sampling frequency of 1 Hz. Finally, the values (date, time (h:min:s), \(I_{\text{S}}\), \(U_{\text{S}}\), PGA) are stored on a memory card. The electronic components are assembled on a printed circuit board (PCB). The device features a rechargeable battery that enables a continuous measurement for more than two weeks.
To increase the measurement device’s accuracy, individual static errors were determined: The error caused by the input offset voltage of the operational amplifiers as well as the gain error due to resistor tolerances were measured. These data were used to individually correct the data gained from the device measurement results. Finally, the accuracy was specified over the whole measurement range (\(I_{\text{S}}\) = 0–15 mA, \(U_{\text{S}}\) = 0–2 V) using a reference measurement device (Agilent N5705B/N6781A, Agilent Technologies, USA).
Housing
The individual parts of the measurement device are assembled in a 3D-printed housing (outer dimensions: 63 × 46 × 19 mm) that features an elastic cuff for fixation on the upper arm (Fig. 2b). This fixation location results in a similar declination angle of incident light as if the solar cells were implanted in the neck region (as intended to do for a cardiac pacemaker7,15). Figure 2 shows a cross-sectional view of the measurement device: The solar cells (1) are covered by the optical filters (2) that mimic the optical behavior of skin. To avoid shadowing of the housing frame, the solar cells are assembled directly below the filters. The PCB (3), including programming interface, memory card and charging interface as well as the battery (5) are accessible by a removable flap. Clefts are sealed with silicon which makes the device splash-proof.
Study
Design
To validate the feasibility of subcutaneous solar energy harvesting, a long-term validation study with volunteers was performed for the duration of six months. The study participants were recruited via word of mouth. Every study participant wore the measurement device during a period of one week per season in summer (June–August), autumn (September–October) and winter (November–December). The date when the study participant had to wear the device was randomly assigned. The study participants were advised to wear the measurement device on the upper arm during the whole day, from morning to bedtime. In general, the volunteers were instructed to wear the device uncovered (i.e. over the clothes). However, the participants were instructed to cover the measurement device when wearing a neck covering (e.g. scarf, jacket with high collar), in order to simulate a neck implant. Predominant weather and activity was registered using a daily questionnaire (weather: sunny, partly sunny, cloudy, rainy/activity: working indoor, leisure time indoor, working outdoor, leisure time outdoor). Furthermore, discrimination between the age groups of 18–64 years and 65–89 years (retired persons) was made. All data were registered anonymously and no health-related data were recorded. Therefore, this study is not defined as a clinical trial by Swiss law.
Data Analysis
The main result of interest is the output power of the solar cells \(P_{\text{S}}\) [W], which is calculated by multiplying the corresponding values of current \(I_{\text{S}}\) and voltage \(U_{\text{S}}\). This gives a power-profile for the whole day. \(\overline{{P_{\text{D}} }}\) [W] is the arithmetic mean of one day (24 h) and \(\overline{{P_{\text{M}} }}\) [W] is the arithmetic mean of a month (i.e. the mean of all \(\overline{{P_{\text{D}} }}\) during one month). The recordings usually cover 24 h of a day. In rare cases, study participants forgot to wear the logger and reported this in the questionnaire as a lack of data. In this case, data was included in the analysis if at least 12 h were valid (00:00–12:00 or 12:00–24:00), otherwise data was discarded. Calculation and statistical analysis was performed with MATLAB R2015b (Mathworks, USA). The output power for statistical considerations and \(\overline{{P_{\text{M}} }}\) are reported as mean values with standard deviation. A two-sided Wilcoxon signed-rank test was performed for statistical analysis. A p value ≤0.05 was considered significant.