13.56 MHz Triple Mode Rectifier Circuit With Extended Coupling Range for Wirelessly Powered Implantable Medical Devices

In this work, a wide input/output range triple mode rectifier circuit operating at 13.56 MHz is implemented to power up medical implants. The proposed novel multi-mode rectifier circuit charges the load for an extended coupling range and eliminates the requirement of alignment magnets. The charging process is achieved in three different modes based on the voltage level of the received signal affected by the distance and the alignment of the inductively coupled coils. Current mode (CM) circuit is activated for loosely coupled coils whereas voltage mode (VM) rectification is proposed for high coupling ratios. Extended coupling range is covered with the activation of half wave rectification mode (HWM) in between CM and VM. The rectifier circuit utilizes these three modes in a single circuit operating at 13.56 MHz according to the receiver signal voltage. The circuit is implemented in TSMC 180 nm BCD technology with 0.9 mm2 active area and tested with printed coils. According to the measurements, the circuit operates in the received power range of 4 to 57.7 mW, which corresponds to 0.10–0.42 coupling range. The maximum power conversion efficiency (PCE) of each operation mode is 51.78%, 82.49%, and 89.34% for CM, HWM, and VM, respectively, while charging a 3.3 V load.


I. INTRODUCTION
W IRELESS power transmission (WPT) is commonly used in biomedical applications such as cochlear implants [1], [2], retinal prosthesis [3], [4], and neural recordings [5]. These implants require high power for neuron stimulation within a limited volume. This power requirement is satisfied by implantable rechargeable batteries. However, the limited energy of batteries restricts their lifetime and they need to be recharged periodically for long lasting operation [6].
Among WPT systems, capacitive coupling is one of the commonly used method in moderate and high power applications (∼ tens of mWs) [7]- [9]. Power and data transmission are achieved by means of the electrical field between the parallel plates. For low and middle power applications, a more suitable method can be preferred for power transfer such as inductive coupling, which offers high speed charging process within a limited space. Fig. 1 shows the general two-coil inductive RF link system. In this system, transmitter coil (T X ) in the transmitter (primary) side transfers the amplified source voltage to the receiver coil (R X ) in the receiver (secondary) side so that the received signal can be rectified and regulated for use in the implanted device underneath the skin. In power transmission, coupling (k) (transmission rate of the power) is affected by T X -R X coil sizes, distance between the link, misalignment, and rotations at the center of the coils. Hence, performance of the WPT system is directly affected from coupling coefficient. To maximize the performance, magnets are used for alignment in conventional devices. Unfortunately, these magnets [10] make implants magnetic resonance imaging (MRI) incompatible so patients with implants need a medical operation before and after the MRI scan to remove and replace the magnets. On the other hand, elimination of magnets makes the coupling coefficient vulnerable to changes in distance between coils and misalignment. Besides, common rectifier circuits cannot operate under variable coupling coefficients. For a magnet free WPT system, multi-mode rectifier circuit is needed. This multi-mode rectifier circuit should rectify the AC signal adaptively for different input levels to compensate coupling change. In addition to coupling This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ limitations, the rectifier circuit should operate either at 6.78 MHz or 13.56 MHz, according to Industrial Scientific and Medical (ISM) band regulations for medical applications [11]. At these frequencies, delay of the circuits become comparable with the switching time and complicate the rectifier design.
In the literature, different rectification approaches for various applications were proposed. These studies differ according to rectification methods, operation voltage, and frequency. The rectification methods are called voltage mode (VM) rectifiers [12], voltage doublers [13], passive rectification, half wave rectifiers and current mode (CM) rectifiers [14]. In VM operation, R X coil and the resonance capacitance C R construct an LC tank in the receiver. In [15]- [17], AC-DC conversion is achieved through cross-coupled PMOS switches and an active diode connected NMOS pair. Similarly, [18] and [19] use the same PMOS and NMOS couple with active feedback structure. The feedback circuit compensates the propagation delay coming from the comparators and adaptively changes the drivers of the NMOS switches. Instead of cross-coupled PMOS pair, latched NMOS pair is used in [20] for rectification. Together with the active rectification methods, passive rectifiers and voltage doubler circuits are used in [21] and [22]. These methods can be used for high coupling ratios. Although large amount of power can be transferred using this technique, misalignment effects distort the operation due to decrease in the coupling between the coils, which decreases input power. Also, the power rating (0.5 W-6 W) and coil sizes make this design unsuitable for use in medical applications [22]. For misaligned and rotated coils with low coupling ratios, CM operation is utilized to rectify AC signal. In this mode, a series LC tank is used in the receiver side to charge the battery/load. CM is reported in different designs [23]- [25]. While [23] aims to improve voltage conversion ratio, [24] offers a very low output power design which cannot be used for high power applications. Furthermore, operation frequencies of 1 MHz [23] and 50 kHz [24] are not suitable for medical applications according to ISM band. In addition to these designs, [25] proposes a voltage-boosted CM operation to charge a 1.1 V battery at 6.78 MHz. However, the single mode studies above, be it VM or CM, cover only a small coupling range resulting from the elimination of alignment magnets. To cover an extended coupling range, a mixed mode rectifier circuit is needed. The mixed mode idea is presented in [26] by switching the operation between VM and CM according to the receiver voltage on R X . Power range of 0.1-20 mW can be covered with this design, but the operation frequency of 1 MHz does not satisfy the frequency requirement determined by ISM band regulations. The dead zone between the VM-CM is eliminated in another design [27]. Yet, these circuits cannot be used in medical applications due to coil sizes and operation frequency limitations.
In this paper, a magnet free extended range triple mode rectifier circuit operating at 13.56 MHz is proposed and implemented to power up implantable medical devices. The main challenges of this design are the operation frequency, and variable input power level. The proposed rectifier circuit includes multi operation modes which are voltage mode (VM), half wave mode (HWM) and current mode (CM) operations and high-speed comparators to overcome these challenges. According to the best of our knowledge, extended coupling range triple mode circuit is the first mixed mode rectification method at 13.56 MHz operation frequency which makes it suitable for medical applications.
The organization of this paper is as follows: The operation principle of each mode and the circuit implementation is discussed in Section II. The measurement results of the triple mode circuit including the in vitro tests and the comparison with the state of the art are presented in Section III. Finally, Section IV concludes the work with discussions.

II. TRIPLE MODE RECTIFIER DESIGN AND CIRCUIT IMPLEMENTATION
The triple mode rectifier schematic for biomedical implants is illustrated in Fig. 2. In the design, cross-coupled PMOS switches, active diode connected NMOS pair and current mode control switch form the main schematic. To get the maximum power conversion efficiency, different modes are activated for different couplings. While CM and HWM use a single ended AC signal, VM rectifies floating AC signal as a full wave rectifier. Therefore, the layout of the overall circuit is drawn symmetrically with respect to the axis of symmetry passing through the switch SW CM . This allows each half of the circuit to match with the other half, improving VM efficiency. For that purpose, resonance capacitor is divided into two equal capacitances, C RP and C RN . Similarly, the dummy grounding switch (SW GD ) is used to preserve symmetry with the grounding switch (SW G ) utilized in CM.
Charging process starts with the generation of MODE SE-LECT 1 (MS 1 ) and MODE SELECT 2 (MS 2 ) signals in the control block ( Fig. 3(a)). These signals, which are determined by the input voltage level, control the operation mode of the circuit. For high voltages (coupling ratios) MS 1 signal become low and the circuit is in VM mode. When the control block sets MS 1 to high, the circuit operates either in CM or HWM. For low voltage levels (coupling ratios), CM is activated. In CM the coils are loosely coupled, and oscillation is required for charging. HWM is used for mid coupling ratios to rectify AC signal with better efficiencies between VM and CM. Utilization of these three modes enables operation for an extended range coupling ratio ( Fig. 3(b)).
In this WPT circuit, there is only one power supply, called LOAD, and it is charged by the rectifier. However, when CM or HWM operation is activated, AC signal takes negative values, and this creates a leakage problem to the substrate. To prevent the leakage, negative dynamic supply generator (NDSG) is implemented to the system. There are two NDSG subblocks in the circuit to maintain the symmetry. One of them takes AC P and ground (GND), while the other one has AC N and GND as inputs. Then, the generated VSS P and VSS N voltages are connected to the body terminals of the main NMOS switches M N1 , M N2 , SW GD ( Fig. 3(f)) and SW G (Fig. 3(i)). In addition to these switches, pass transistor pairs PT 1&2 (Fig. 3(e)) and PT 3&4 (Fig. 3(h)) are used to switch the configuration of the entire system when the operation mode is decided. Optimization of each switch size is performed to maximize the efficiency of the rectifier and minimize switching losses. Moreover, bulk regulation is added to PT 1&2 pass transistors to prevent leakage due to the AC voltage variation. Simulated waveforms of voltage mode, half wave mode and current mode operations are given in Fig. 4(a), (b), and (c), respectively. In each mode, AC signal is rectified to transfer the input power to the LOAD, whose voltage is increased in each charging cycles. Control signals of the M N1-2 and SW CM switches indicate the switching times of the charge conduction to the LOAD. In addition to the simulation results, measurement results of each mode are provided with the detailed circuit description in the following sections.

A. Voltage Mode Rectifier
Fig . 5 shows the simplified version of the triple mode rectifier circuit for voltage mode operation. When the coils are highly coupled (k > 0.26), MS 1 signal becomes zero, and receiver power becomes high enough to charge the load. When VM is activated, PT 1&2 connect AC signals to PMOS gates and form the cross-coupled structure. At the same time, PT 3&4 disconnect the link between LOAD and MP 1&2 transistors' gates. Besides, CM switch turns on and grounding switches (SW G &SW GD ) are deactivated.
In VM operation, M N1 &M P2 and M N2 &M P1 pairs work synchronously in crosslinked states. When AC P is in ground level, AC N is high and M P1 turns off. VM control unit forces AC P to ground during this half period of the oscillation. Hence, M P2 turns on, and power is transferred from AC N to LOAD. In the other half, M N2 and M P1 pair charges the load. During this charging process, NMOS switches are turned on and off by the VM control circuit shown in Fig. 6(a).
The VM subblock is comprised of a common gate comparator (CG CMP), a pulse duration adjustment block, digital control block, and driver circuit. CG CMP, preferred for its high-speed operation property, takes AC signal and ground as inputs and finds the zero crossing points by comparing these signals. While designing the comparator, it was observed that there exists a delay between the switching response and the output of CMP at the operation frequency of 13.56 MHz. Consequently, like [21] an intentional mismatch is inserted to the differential pair transistors M 12&16 and M 18&21 as shown in Fig. 6(b). This mismatch at the input pair corresponds to an offset voltage in the comparison. In other words, AC P is compared with the offset voltage and output is sensed. Switching speed performance of VM comparator in simulations shows that gate control voltage of M N switches has ∼1 ns rising time and ∼0.3 ns falling time. In addition, the gate control signal responds to voltage change in AC signal within 4 ns. When the output of the CMP becomes high, NMOS switch turns on. To complete the switching, an off path should be introduced to the system. For that purpose, an analog pulse width duration adjustment block (Fig. 6(c)) and SR latch is added. When the output of the CG CMP becomes high, the delayed version of the CG CMP signal is generated by the delay cell and fed into the latch structure through a digital control unit, which is utilized to reset the circuit as may be required ( Fig. 6(d)). When pulse duration adjustment signal (T P ) becomes low, M 23 turns on and shorts the CMP output to the ground instantly. Meanwhile, T P cuts off the differential pair current to increase the power efficiency of VM by preventing unnecessary loss of power. As a result, the switching pulse is generated to control the gate voltage of the NMOS switches M N1 and M N2 at the frequency of 13.56 MHz. In the driver circuit ( Fig. 6(e)), three end-to-end connected inverters are utilized, and the size of the transistors are optimized to minimize power loss. Since the coupling ratio determining AC input and LOAD voltage change, charging duration is readjusted by the analog current starved delay block. With this block, conduction time can be adjusted externally to maximize the transferred power to obtain maximum power conversion efficiency (PCE) in the rectification.
As the triple mode circuit includes three different rectification modes, enable signals are added to each subblock to improve efficiency and prevent power loss. MS 1 signal either lets the CG CMP output to go to the driver or enables an external voltage  to control the gates of the NMOS pairs. Also, MS 1 enables and disables the CMP operation, like T P signal. Fig. 7 illustrates the measured waveforms of differential AC signal (AC P -AC N ) and one of the NMOS gate control signal (M N-GATE ) in VM operation. The peaks of the AC signal are clipped due to charge conduction and each NMOS switch turns on and off each cycle accordingly.

B. Current Mode Rectifier
Inductive link may not result in high coupling ratios all the time, due to the misalignment and angle change between the coils. As the coupling decreases, transmitted power to the receiver side becomes low. When the received signal becomes lower than the LOAD voltage, voltage rectification cannot be performed as in VM. Nevertheless, low couplings should be considered for a wide range functional system. Hence, for loosely coupled coils (0.10 < k < 0.20), CM is activated in the triple mode rectifier as shown in Fig. 8. In CM, the cross coupled connection is broken. M N1 and M P2 pair is deactivated and PT 3 converts M P1 switch to diode connection. On the other hand, M N2 switch creates a grounding path for LC resonator. In this way, the resonator oscillates on the ground level. The grounding switch of the mode (SW G ) grounds the right side of the inductor and switch by bypassing the C RN capacitor. In this configuration, the circuit becomes ready to charge the load in CM.
Current mode switch (SW CM ) is the main part of the operation and is controlled by the driving and controlling comparator CMP 2 . When the CM switch is on, AC signal starts to oscillate across the inductor L R and the capacitor C RP with an increased amplitude. During the oscillation, AC signal grows, and power is stored on L R . When the voltage on C RP becomes zero, contact between each side of the control switch (SW CM ) is disconnected. Thus, the maximized inductor current is transferred to the LOAD during the conduction.
As illustrated in Fig. 9(a), AC signal is compared with the ground in a common source comparator ( Fig. 9(b)) and zero crossing points are detected. Zero crossing detection is performed to determine the possible switching times by digitalizing the AC turning points. According to simulation results of CM comparator speed analysis, the gate voltage of SW CM has ∼0.4 ns rising and falling times. Also, AC voltage change can be detected with the output of the comparator in ∼2.5 ns. After the digitalization, the signal is shaped by cycle selection unit to decide the oscillation cycle of the AC signal. Cycle selection unit is controlled externally with 3-bit digital inputs. According to the received voltage level, cycle number can be adjusted from 2 to 7 such that the maximum power transfer can be achieved in different couplings. For the couplings close to 0.20, the signal becomes high enough to charge the load, oscillation cycle can be adjusted as 2-4. However, if the input voltage level is low (∼0.10), more cycles are needed to increase the AC signal level to charge the load. After the predetermined oscillation cycle is reached, SW CM turns off and the connection between the C RP and the virtual ground is broken. The accumulated power on the inductor is transferred to the LOAD through the M P1 diode. To maximize the efficiency and minimize conduction losses, the pulse width of the conduction is controlled by the same current starved delay buffer used in VM for different power levels. Also, using the CM comparator output at the digital control unit, SW CM-GATE signal is generated with the minimized circuit delays. When the conduction is finalized, SW CM turns on and the AC signal starts to oscillate until the next conduction happens. Measured waveforms of oscillating AC signal and the gate voltage of current mode switch (SW CM-GATE ) for 4 cycle oscillation are given in Fig. 9(c). As explained in the figure, once the AC oscillates for 4 cycle, charge conduction occurs, and process is reset, and the circuit begins the next oscillation and  charging process. Also, the measured waveform of AC signal for different oscillation cycles is illustrated in Fig. 10.
Compared to the VM operation, CM works under low couplings; hence, low input powers. However, as the coupling increases, received power level increases and multiple cycle oscillation idea loses its meaning. Instead of waiting for oscillations, load can be charged in each cycle. However, the increase in inductor current with coupling causes conduction losses, and single cycle oscillation means the unnecessary power consumption during the activation and deactivation of each subblock and switch in CM. Hence, instead of CM, different rectification method can be preferred to obtain better performance for higher couplings (k > 0. 20). At this point, voltage mode should be enabled to increase power conversion efficiency by full wave rectification. Yet, as the increase in coupling is not high enough for the circuit to start operating in VM, a third mode becomes necessary as an intermediate step to increase the efficiency between CM and VM to charge the load.

C. Half Wave Mode Rectifier
Increase in coupling ratio causes a decrease in oscillation cycle. With this reduction, the circuit reaches a point where it will start operating with a single cycle oscillation. Hence, the half wave mode, between the CM and VM, is preferred to charge the LOAD as the intermediate step until the VM can start to work.  HWM rectifier is obtained from CM using the MS 2 signal between the coupling ratios 0.20 < k < 0.26 as shown in Fig. 9(a). Once the MS 2 becomes one, CM is switched to HWM (Fig. 11). In this mode, SW CM is always on and connects C RP to the virtual ground. Half wave rectification is performed with the LC tank and current flows through the M P1 diode to power up the LOAD. Compared with CM, complexity of the system is decreased, unnecessary power consumption is eliminated, and HWM gives higher power efficiencies. A drawback of HWM is the diode connected M P1 transistor, which causes a voltage drop and limits the power efficiency. In addition, when the peak-to-peak voltage of the received signal increases, the lifetime of the transistors are limited, and an unsafe condition is created. For this reason, the switching from the intermediate step to VM is done for larger coupling ratios. Fig. 12(a) shows the block diagram of the control unit. At the beginning of the charging process, rectifiers are not active, and mode selection is performed according to the output of the control block. MS 1 and MS 2 signals are generated using the voltage levels of the LOAD and AC signals. The peak detection of the AC is done using the floating AC P and AC N Fig. 11. Schematic of the half wave rectification as an intermediate step in WPT. signals. A negative voltage converter is used to obtain the DC correspondence of the received AC voltage (Fig. 12(b)). During this voltage conversion, a voltage drop occurs in the peak voltage due to the diode connected transistors. To compensate this drop and have a precise comparison with the LOAD voltage, a diode ladder is used (Fig. 12(c)).

D. Control Block
Comparison is performed with a hysteresis comparator and the output is connected to logic block. If the LOAD becomes larger than AC, CM is activated. Contrarily, if LOAD voltage becomes smaller than AC signal, MS 1 and MS 2 signals are adjusted according to the coupling level, i.e., voltage level of the AC signal. The peak detector output is compared with an externally

E. Negative Dynamic Supply Generator
In the proposed work, different modes are combined in a single circuitry. Since the AC oscillates and takes negative values in CM and HWM, possible leakage and reverse current to the bulk of the transistors should be prevented. To this end, bulk regulation is required and a negative dynamic supply generator (NDSG) is added to the overall system ( Fig. 13(a)) based on a standard bulk regulation circuit. The NDSG circuit operation is controlled by MS 1 signal. The circuit operates as a standard bulk regulator in CM and HWM. However, in VM, transition time of the switches are less than the delay of the standard bulk regulation circuit. Therefore, the power efficiency drops significantly with the standard bulk regulation circuit. Furthermore, the AC voltages are clipped by the circuit operation and do not become negative. To prevent power loss, the NDSG circuit shorts the negative supply to ground in VM. By mode dependent dynamic supply, negative supply is generated according to the input voltage level.
In CM and HWM, the NDSG circuit compares AC signal with ground level (GND) and follows the lowest voltage. M 5&6 turn off with MS 1 signal and M 3&4 control the gate voltages of M 1&2 . The measurement result of the NDSG circuit in CM can be seen from Fig. 13(b). The generated negative signal (VSS) follows the most negative signal and prevents the bulk leakage. In the full wave rectification mode, i.e., VM, M 5&6 are activated to turn M 3&4 off. M 2 becomes always on, and M 1 turns off. AC changes between the ground level and the positive voltages. As a result, the NDSG gives the most negative voltage which is ground (GND). Fig. 13(c) shows the experimental results of AC signal and the generated negative signal for VM. With the NDSG circuit, the VM efficiency is not affected from required bulk regulation for the other modes.

III. EXPERIMENTAL RESULTS
The triple mode rectifier circuit is implemented in TSMC 180 nm BCD technology. Fig. 14 shows the chip micrograph of the design with the floor plan. The floor plan consists of voltage mode rectifier, current mode rectifier (including half wave mode), VM-CM switching circuitry and the control block of the system with an active area 0.9 mm 2 . During the measurements, the experimental setup shown in Fig. 15(a) is used. In the test setup, a custom designed coil pair is used for power transfer. A signal generator is used to give input to T X coil. Also, the DC supply is used to bias comparators and delay buffers and to supply voltage to electrostatic discharge (ESD) diodes for the safety of the chip. To avoid grounding problems, AC and DC grounds are isolated from each other. Finally, the oscilloscope monitors the results obtained from the chip test PCB.
The chip is tested with the printed board coil pair in different coupling ratios. The coil pair is designed in ANSYS EM to  maximize the coupling ratio between the coils without using an alignment magnet as explained in the previous work [28]. Coil holder helps to change the position of the coils horizontally. Also, an intentional angle can be introduced to the system to observe misalignment effects. Coil and the holder are given in Fig. 15(b). (1) defines the coupling coefficient which is the ratio of R X and T X voltages. During the measurements, the single coil pair is used with different coupling ratios provided by the position change.
In this design, two C R capacitors are used to preserve the symmetry of the system. As shown in Fig. 16, the normalized resonance voltage at the receiver side makes a peak at 13.56 MHz for the perfectly matched resonance capacitor (156 pF). However, in the proposed system, single C R is used for CM and HWM, and two series connected C R is used for VM operation. Therefore, the resonance capacitor C R is chosen as 220 pF by considering parasitic capacitances. In this way, the peak resonance capacitance is the geometric mean of the two operation capacitances, which are 220 pF and 110 pF. By considering the 12% total efficiency drop, we have minimized the switching losses while combining the three different modes in a single circuit.
In the measurements, 13.56 MHz sinusoidal constant amplitude signal is given to the transmitter coil and the power is transferred to the receiver side. At the receiver side, resonator structure is captured the transmitted AC signal. According to the received signal level, related operation mode is activated externally by the help of the generated MS 1 and MS 2 signals.
Tests are performed from low to high coupling ratios and AC signals are observed. For each coupling coefficient, LOAD voltage is set to 3.3 V by calibrating the load resistance of the system manually, and input and output powers are recorded. While the load resistance is changed, the load capacitance (C L ) is kept constant at 4.7 µF. During the input power measurements, the measurement method proposed in [16] is used (Fig. 17). Two differential probes are used to observe the AC signal and the rectifier current. A 10 Ω sampling resistor and a 30 pF compensation capacitor are used to observe the input current (I REC ), and the differential AC (AC P -AC N ) input voltage. Using the measured current and the voltage values the input power is calculated. In these calculations, rectifier current is taken into consideration instead of the LC current. Hence, using a sampling resistor affects the power on LC tank, but it does not cause change in the rectifier power. As a result, the power conversion efficiency is obtained without measurement setup distortion. Fig. 18 shows the measured output power and power conversion efficiency of the triple mode rectifier circuit with respect to different received signal powers. The PCE tests are performed with a constant T X power by changing the coupling coefficient. The system can operate with different coil distances (2 mm to 10 mm) and misalignments (0°to 20°angle variations).
In CM, rectified input power starts from 4 mW with the corresponding output power of 233 µW and increases up to 17 mW with an output power of 8.8 mW. Maximum power efficiency for CM is 51.78%. Compared with other rectification methods, CM provides lower power efficiencies. This is due to the power consumption of the circuit, which becomes comparable to the output power. Also, in CM, LOAD can only be charged in once cycle for each different number of oscillations whereas charge conduction is achieved in each cycle of VM. This causes power efficiency drops. However, medical implants consume low power and 8.8 mW output power with 51.78% efficiency becomes important when there is no other rectification method. When the boundary between the CM-HWM is reached, AC signal grows enough to charge the LOAD each oscillation cycle. Accordingly, half wave mode is activated to obtain higher power efficiencies. This mode constructs a bridge between the CM and VM, and HWM enables the overall system to cover the power and the coupling range. For HWM, input power changes from 18 mW to 29 mW with 10.58-24.11 mW output power. Also, the efficiency of HWM circuit changes from 58.81% to 82.49% for the measured output power range. The increase in coupling range and the received power cause the circuit to go into VM. The full wave rectification mode gives the best efficiency values compared to the other two modes. The VM efficiency changes from 82.30% to 89.34%. At the same time output power Analysis of coil pair and the rectifier circuit is performed in air. However, receiver side of the system should be in the human body for medical purposes. Thus, in vitro tests are needed to observe the effects of environment change on the performance of inductive link and the rectifier circuit. The test setup shown in Fig. 19 is used for in vitro experiments. R X and T X coils are separated by the coil holder with 6 mm distance (d) . The volume between the coils is filled by air, saline and beef environments  as in [29]. The saline is the solution of distilled water with 9 mg/L NaCl, and the beef is sirloin steak with 6 mm thickness. As Fig. 19 shows the setup for beef, saline solution is also tested with the same way. Another package of saline and beef is also used behind the R X coil to emulate implant environment. According to experiments, it is observed that coupling coefficient for the normalized received voltage drops slightly when the environment is changed from air to saline and beef ( Fig. 20(a)). Triple mode rectifier still can operate in an extended input power range.
Medical devices are coated with biocompatible materials before the end-use. Hence, in addition to environment change, effects of the surface coating of the coils to their performances are analyzed in experiments like [29]. Surface of the PCB coils is coated with ∼0.4 mm silicone. Due to the parasitic effects from the coating material, the value of resonance capacitance of the coils is distorted, and the coupling coefficient decreased compared to uncoated case in air with 6 mm separation distance ( Fig. 20(b)). The triple mode rectifier circuit can still operate in each mode with the coated coils. Since the rectifier performance is specified with the receiver side power level, by increasing the voltage level of the transmitter side, this drop can be compensated. As stated in [30], in the RF link performance, surface coating becomes more important than the environment change, and it should be considered to prevent performance drops.
The comparison with the state of the art WPT systems are given in Table I. While all references in the comparison table is about inductive coupling, [9] is based on capacitive coupling WPT method and can be preferred for high power applications. On the contrary, [24] is suitable for very low power applications with 50 kHz frequency in single mode operation. Even if [25] operates at 6.78 MHz frequency, it can charge low voltage batteries. Both [16] and [20] offer rectifier circuits in VM and are limited by high couplings. Although [27] proposes a mixed mode rectifier circuit, coil sizes make the design unsuitable for medical devices. None of these studies combine more than one operation at 13.56 MHz for medical applications. In addition, the performance of most of these studies has been maximized for a single coil distance. However, it cannot be guaranteed to keep the position of the coils fixed during power transmission in the patient body. Therefore, rather than a single point, a distance range should be proposed to compensate possible voltage changes. Operation under different distances corresponds to the extension in the input and output range. For instance, [28] presents a mixed mode rectifier at the operation frequency of 13.56 MHz for medical implants in a distance range. Yet, the efficiency of the circuit is low, and it occupies a larger area. The proposed work presents a triple mode rectifier circuit which operates with 4 mW -57.7 mW input range to charge 3.3 V implantable batteries at 13.56 MHz, which is not proposed before. With this novel design, medical implants can be charged without alignment concerns in an extended coupling range.

IV. CONCLUSION
The implemented triple mode rectifier circuit takes the input through the RF link at the operating frequency 13.56 MHz and converts AC signal to DC voltage in different modes for different coupling levels. For low coupling ratios, caused by misalignment and rotation between the coils, oscillating AC signal is used. In this mode, the charge is accumulated on the receiver coil and transferred to the load in the charging cycle. Full wave rectification is used to perform AC-DC conversion at high couplings. An extended coupling range is covered with the addition of a third mode, called half wave rectifier mode. The symmetry of the design is preserved both in circuit implementation and layout drawings. According to the measurement results of the proposed extended range triple mode rectifier circuit, 0.10-0.42 coupling range is covered for 4 mW -57.7 mW input power to charge 3.3 V LOAD with 0.233-51.54 mW output power range. The maximum efficiency achieved for CM, HWM and VM is 51.78%, 82.49% and 89.34%, respectively. The implemented rectifier circuit can charge the load from 2 mm to 10 mm distance and 0°to 20°angle variations with the maximized efficiency.
In vitro experiments show that the RF link and the rectifier circuit can still be utilized for charging the load of the system with the slight drop on the coupling coefficient, which can be compensated by increasing the voltage level of transmitter side within the safe limits. As a future work control block of the system can be supported by another decision unit such as ADC to automize the cycle selection operation in CM. Besides, charge conduction duration of VM and CM can be controlled internally with a feedback circuitry.