The tracking patch includes two antennas: The first one is used for the communication with the RF beacon and Sigfox network with an operational frequency of 868 MHz and the second antenna that works at a frequency of 1575.42 MHz is used for GPS geolocation. In aim to reduce the cost and bulkiness of our device, we opted for the design of printed antennas [12] . The PIFA (Printed Inverted F Antenna) type was chosen due to its different properties. It is a quarter wave antenna that takes the form of an inverted “F”. The radiating arm is shorted to the ground plane to create an inductive effect and in our case the antenna is fed using a coplanar wave guide (Figure 3). The ground plane dimension has an important role to determine the efficiency of the antenna. For our design, it was chosen to ensure both high integration with acceptable performances.

For a PIFA, The length of the radiating arm must be equal to the quarter of the wave length (λ/4). This criterion is not compatible with the miniaturization constraints of the patch for the ISM and L1 GPS bands: λ/4 = 8.64 cm @ 868 MHz and λ/4 = 4.76 cm @ 1575.42 MHz. To resolve this technological challenge, a miniaturization technique based on the use of meanders was adopted (Figure 4). This method consists of folding the radiating arm many times to minimize the area occupied by the antenna. Modeling works using RF software ADS (Advanced Design System) [13] and HFSS (High Frequency Structural Simulator) [14] were conducted to design the antennas and study the influence of different physical parameters: layers stacking, type of materials, dimensions, meanders form, distance between meanders, coupling phenomena… The ISM and GPS antennas designs using HFSS software are presented in Figure 4.

The proposed antennas were carried out on FR4 substrate and characterized using a vector network analyzer (VNA) E5061B (Agilent Technologies). Figure 5 shows a comparison between different simulated and measured reflection coefficients that pre- sent good agreement. This parameter is expressed in decibels (dB) and describes the reflected power from the antenna. A reflection coefficient that tends to 0 dB means that the energy supplied to the antenna is totally reflected and there is no radiation in this case. This coefficient should be as small as possible and generally we accept a value lower than −10 dB.

The different values of reflection coefficients and bandwidths are presented in the Table 1. The designed antennas provide quite good performances: for the ISM band, the 868 MHz antenna has a reflection coefficient S11 = −16.4 dB with a bandwidth of 16.2 MHz and for the GPS antenna, it has a reflection coefficient S22 = −19 dB with a bandwidth of 39 MHz.

In aim to validate the good functioning of the RF communication with Sigfox network, tests in real cases were carried out. The principle was based on a geographical meshing around our laboratory LAAS-CNRS to select reference points. From each point 10 frames were initially sent and in the end we compared the number of received frames by Sigfox stations implemented in Toulouse (France) to estimate the reception quality.

Two cases were studied, for the first one the tracking patch was not placed on the skin and for the second test the patch was glued on the skin to take into account the influence of the human body.

The results of the first test are resumed in the Figure 6: when the patch is not placed on the skin, for all the tested points the reception quality is greater than 60%. This means that we receive each time more than 6 frames. In addition, for 80% of the reference points we receive all the 10 frames initially sent.

Table 2 presents the number of received frames of all reference points in the first test. As the patch is developed to be glued on the skin, these results will be compared with the second test that takes into consideration the presence of the human body.

For the second test, the same principle of geographical meshing was kept and this time the patch was glued on the human body of a volunteer. The reception qualities of the different tested points are presented in Figure 7 and Table 3.

From Figure 7, we can notice that when the patch is placed on the human body, only two points have a reception quality lower than 60% (points in yellow color). In addition, for 68% of the tested points we are able to receive all the frames initially sent (Ta- ble 3). These tests validate the good functioning of the communication with Sigfox network when the patch is glued on the human body.

In recent years, wireless chargers appeared increasingly on the market. This solution seems the most suitable for our device. It is simple to use, we just need to put the patch on a charging pad without using any cable which allows avoiding the problem of incompatibility between different chargers. Furthermore, by choosing a suitable packaging the waterproof property can be ensured. A comparative study of different standards and protocols known in this field has allowed us to select the Qi protocol using magnetic induction [15] .

As explained in Figure 8, the energy transfer is ensured by magnetic coupling between a receiver (encapsulated patch) and an emitter (charging pad). The Qi protocol adds a unidirectional communication (messages from the receiver to the transmitter) to control the different charging states: full charging of the battery, presence of a Qi compatible receiver, presence of metallic objects…

The wireless charging functionality needs the use of coils to recuperate energy by magnetic induction. Given the patch integration requirements, the commercialized coils cannot be used for our system due to their bulky forms. An example is presented in Figure 9, this coil has a dimension of 3.25 × 4.85 cm² and cannot be integrated in the patch (Figure 9).

To overcome this issue, our efforts were focused on the design of a thin printed coil to meet the integration requirements by placing it under the battery of the patch. A ferrite layer was also added to ameliorate the energy transfer performances. This layer acts

as a magnetic core to channel the magnetic field and avoid the influence of the battery on the coil performances. The final stacking of the different layers constituting the tracking patch is presented in the Figure 10: the PCB of the patch, the battery, the coil and the ferrite layer are encapsulated with a plastic packaging and a medical dressing is also used to avoid any allergic or discomfort problems.

The proposed coil is composed of two layers of copper with a thickness of 17 µm separated by a 50 µm of polyimide substrate. The coil dimensions were chosen to meet the battery area (40 × 18 mm²). The Cu tracks have a width of 520 µm and they are separated by a spacing of 80 µm. Figure 11 presents a 3D view of the proposed printed coil obtained using ADS software. The two layers of copper are connected in series using vias.

Design, simulations and validation of the performances of the printed coil were conducted using ADS and HFSS software. The variations of inductance, resistance, and quality factor of the proposed model are plotted in the Figure 12 as predicted by the both softwares.

From Figure 12, we can notice that simulations predict a self-resonance frequency of 5.3 MHz obtained with ADS and 4.9 MHz computed using HFSS. These values are compatible with the frequency operating range used by the protocol Qi (110 - 205 kHz): they are much greater than the maximum value (205 kHz) and this property ensures constant values of the electrical performances (L, R, and Q) in this range.

For the charging frequency interval, the model simulated with ADS presents an inductance of 10.74 µH, a resistance of 3 Ω, and a quality factor of 3.37 @ 150 kHz. Concerning the results obtained with HFSS software, the coil has an inductance of 10.71 µH, a resistance of 3.25 Ω, and a quality factor of 3.1 @ 150 kHz.

The printed coils were carried out in the cleanroom of our laboratory LAAS-CNRS. A polyimide photoresist HD-4110 [16] was used to form different cover layers and the substrate separating the two layers of Cu tracks. By using this photoresist the different openings (contact pads and vias) were obtained directly by photolithography. In addition, another non photoresist of polyimide PI2611 [17] was also used to form a sacrificial layer between the Si wafer and the different deposited layers by spin coating technique. Unlike our last process described in your article [18] , this time the vias were filled with copper to ensure a better connection between the two copper layers. The copper tracks and vias were made by electroplating via a mold of BPN photoresist to obtain directly the desired forms. This operation was preceded by an evaporation phase of a seed layer Ti/Cu (100 nm/100 nm) needed for the electrolytic deposition of copper. The different technological steps of this process are presented in Figure 13.

A prototype of the printed coil is presented in Figure 14(a). The dimensions of the connecting vias and copper tracks are illustrated in Figures 14(b)-(d). Due to the inclined profile of electroplated copper, two characteristic values are given each time: diameter/distance in the top and the bottom of the vias, tracks, and the inter-tracks spacing.

The printed coil was characterized using an impedance analyzer Agilent 4294A (Figure 15) to study the variations of the electrical performances (inductance, resistance, and quality factor) in function of frequency in aim to validate the simulated values computed using RF softwares.

The measured results are plotted in the Figure 16, the resonance frequency is shifted (3.9 MHz) compared to the simulation results (5.3 MHz obtained with ADS and 4.9 MHz computed with HFSS). This frequency shift does not disturb the coil performances because it is still far from the frequency range used by the Qi wireless charger (110 kHz - 205 kHz). For this range, the measurement results meet the simulated values: L=10.8 μH, R= 3 Ω, and Q = 3.39 @ 150 kHz.

After the validation of the electrical performances of the printed coils, the wireless charging system was integrated in the tracking patch. Several charging pads compatible with the Qi standard were used to validate the interoperability property and the charging current allure. In the Figure 17, the patch is encapsulated using a plastic packaging and is being charged wirelessly using the magnetic induction created by the charging pad. For this test, the charging current value is 10 mA.

A thermal study of the patch is necessary to investigate and quantify temperature changes during the wireless charging (Eddy current created by the magnetic field). This temperature variation must not exceed a maximal value of 45˚C to avoid the deterioration

of the battery performances. To conduct our investigation, a thermal camera “FLIR x6580sc” was used. The patch and the charging pad were placed at a distance of 50 cm from the camera lens and the sampling frequency was adjusted to obtain 5 images/s (Figure 18).

The Analysis of the test results reveals that the maximum temperature noticed during the wireless charging of the patch does not exceed 32˚C that is lower than the maximum value for the using of the battery (45˚C). The variation of the maximal temperature of the patch during 1 hour of charging is plotted in the Figure 19. This figure is obtained directly using the above mentioned camera by selecting the area that witnesses the maximum value of temperature in the thermal image treated by the camera software (Figure 18).

This test confirms that the patch is well protected from the temperature elevation that can be caused by the magnetic field (Foucault currents) during the wireless charging.