Far-field RF power extraction circuits and systems

[0001] Not Applicable.

[0002] Not Applicable.

[0003] The present invention relates generally to power extraction and more particularly to circuits and systems for extracting power from radio frequency (RF) signals.

[0004] Devising efficient methods for extracting DC power from electromagnetic radiation has become an important necessity for a number of applications involving self-powered devices, such as Radio Frequency Identification (RFID) tags and bionic implants. The operating range of such self-powered devices has been severely limited by the failure of existing power extraction techniques to successfully extract power from radio frequency (RF) signals having relatively low power levels. The problem of extracting DC power from electromagnetic radiation has two basic parts: collecting the incident radiated power, and then converting the collected power to DC signals which are usable by the self-powered devices.

[0005] Converting RF energy from RF signals at different frequencies is a relatively difficult problem particularly when the RF signals have relatively low power levels. Fundamentally, this problem arises because frequency conversion is generally a nonlinear operation, but all practical systems are essentially linear for small signals. In addition, nonlinear devices normally used for rectification have exponential nonlinearities with relatively large “dead zones” near the origin, i.e., nonlinear devices can be non-responsive in response to signals having voltage and current levels which are close to zero. Severe constraints can also be imposed when it is desirable to provide a self-powered device which is relatively inexpensive and environmentally robust. This limitation precludes the use of exotic devices and structures.

[0006] A method and apparatus for performing far-field power extraction are disclosed. The method includes receiving an electromagnetic radiation signal, rectifying the signal to produce a direct current (D.C.) voltage and providing the D.C. voltage to a circuit. A far-field power extraction circuit includes an antenna for receiving an electromagnetic radiation signal, a rectifier for rectifying the electromagnetic radiation signal. The circuit may further include a charge pump for amplifying the rectified voltage, an impedance matching network for coupling the antenna to the rectifier and a feedback tuning circuit for optimizing performance of the extraction circuit.

[0020] The present invention comprises various combinations of different circuits and techniques to efficiently extract power from electromagnetic signals having relatively low electromagnetic field strengths in the far field of the transmitter, thereby substantially reducing the power threshold required for operation of self-powered devices.

[0021] Referring now to FIG. 1, a block diagram of a system 10 for extracting power from electromagnetic radiation includes an antenna 20 for receiving a radio frequency (RF) signal from which direct current (DC) power will be harvested. A matching network 30 is coupled to the antenna 20 to impedance match the antenna 20 to the remainder of the system, thereby obtaining efficient power transfer. A switched rectifier 40 operates on the impedance matched differential RF signal and converts the signal to one or more DC levels. The output of the switched rectifier 40 is coupled to a group of charge pumps 50, which increase the DC levels of the voltage. The output of the charge pumps is coupled across a load capacitor 60 to provide the DC output voltage. A feedback tuning circuit 70 is coupled to the output voltage and feeds into the impedance matching network 30 to provide continuous tuning of the impedance matching network to get the maximum possible DC output.

[0022] Referring now to FIG. 2, the antenna 20 is shown. The antenna bandwidth is defined as the frequency range over which the reflection coefficient of the antenna into a specified load impedance is less than some specified value, typically −10 dB (VSWR=2). This antenna has improved bandwidth over a simple linear dipole, while occupying approximately the same area. In a preferred embodiment the antenna 20 thus comprises a wideband dipole antenna.

[0023] It should be appreciated that although a particular dipole antenna design is shown in FIG. 2, it may be desirable to provide the antenna as a planar loop, dipole or fractal antenna. In one embodiment, the antenna 20 is provided on a flexible substrate. One objective for power collection applications is to provide a planar antenna on a flexible substrate that produces the maximum possible open circuit voltage V_OC across the antenna terminals 21 and 22 for a given incident field strength. In addition, the antenna design has to have sufficient bandwidth to withstand bending and proximity effects, such as attachment to a dielectric surface, without moving too far off resonance. Since passive RFID systems typically use backscatter modulation to communicate with the tag reader, another factor of interest is optimization of the Radar Cross Section (RCS) modulation capabilities of the antenna.

[0024] To address these issues, the use of one or more of planar loop, dipole, bow-tie and fractal antennas is presented. Loop antennas are advantageous because most proximity (near field) effects in practice are caused by dielectric materials. Since the near field energy of loop antennas is primarily stored in the magnetic field, they are typically less susceptible to these effects than other antenna types. Bow-tie antennas are desirable when a large impedance matching bandwidth is desired, but typically require large amounts of area in order to achieve this bandwidth. Fractal antenna structures are of interest in this application since they allow the bandwidth to be increased without consuming more area, or by reducing the area required to achieve a given bandwidth. Photonic Band Gap (PBG) substrates which reduce losses due to surface wave propagation in the flexible substrate may also be utilized. PBG substrates have electrical properties (like dielectric constants) which are periodic functions of space. Solutions of Maxwell's equations in such a medium have a ‘stop band’, or forbidden frequency range, where no surface wave propagation is possible. This may be utilized in our application by making the stop band lie in the frequency range where the main surface wave modes propagate, thereby preventing energy loss due to such (undesirable) modes and improving the efficiency of power extraction.

[0025] Referring now to FIG. 3, a plot 25 of a simulated reflection characteristic for the antenna 20 is shown. In graph 25, S₁₁ is the reflection coefficient of the antenna and the substrate material used for simulations was FR-4. When simulated with a commercially available Method of Moments (MoM) based field solver, and using a criteria of a voltage standing wave ratio (VSWR) of two (VSWR≦2) this antenna 20 exhibited a fractional bandwidth of approximately 40% at a center frequency of about 900 MHz. In this implementation, the antenna occupied an area of approximately 5″×2.5″.

[0026] Referring again to FIG. 2, another technique used in performing far-field RF power extraction comprises utilizing package parasitics to increase the input voltage levels to the rectifier. The far field case, when the input amplitude of the RF signal is not large enough to efficiently operate typical rectifying devices such as Schottky diodes are of particular interest. To overcome this problem, the high-Q input matching network 30 is used to passively amplify the input RF voltage. The inductance L and capacitance characteristics C(V) of network 30 are adjusted to include the effects of parasitic inductances and capacitances introduced by the chip packaging. In this way, a use for these normally unwanted parasitic characteristics has been found, as they now function as part of the matching network 30. Of course, increasing the Q also decreases the frequency range over which the system can operate and increases its sensitivity to environmental conditions, which cause the resonant frequency to drift with time. This necessitates the use of active resonant frequency control. This is implemented by the feedback tuning network shown in FIG. 1.

[0027] Another technique used in performing far-field RE power extraction comprises using traveling wave architectures for distributed voltage amplification and rectification. The matching network 30 is provided having both high voltage gain and high bandwidth at the input of the rectifier by using a cascade of exponentially tapered inductor-resistor-capacitor (L-R-C) transmission line segments. Each segment acts like a low pass filter with a certain Q and cutoff frequency. All segments have essentially the same Q, but have exponentially tapering cutoff frequencies. The cutoff frequency of the n-th section is given by:
f_n=exp(−n/N_nat)

[0028] where f_n and N_nat are constants. Such a technique is useful for attaining high gain from many low-gain stages.

[0029] Referring now to FIG. 4, a graph 35 which shows the AC transfer characteristic of the input matching network 30 for a loaded Q of 2 and a resonant frequency of 900 MHz is shown. The curve shows that a maximum output is achieved at approximately 900 MHz. The input to the matching network is assumed to be the antenna, and its output is fed to the switched rectifier.

[0030] Referring now to FIG. 5, a rectifier circuit 40 is shown. One aspect of the present invention for performing efficient power extraction comprises using parametric-amplifier-like topologies and capacitive nonlinearities for rectification instead of exponential resistive nonlinearities. Exponential resistive nonlinearities (e.g., diodes) have traditionally been used for rectification but are unsuitable at low power levels. To solve this problem, parametric-amplifier-like topologies and capacitive nonlinearities are used. FIG. 5 shows how a nonlinear capacitor (e.g. a varactor) 41 is used for rectification. The varactor 41 (which can be a reverse-biased PN junction or a MOS capacitor) has a capacitance characteristic C(V), where V is the voltage of the control terminal. V is varied at the same frequency as the input RF signal. The RF signal is applied differentially, and, as an example, V may be tied to the upper plate of the varactor. In that case, C(V) will be different on the positive and negative halves of the RF signal cycle.

[0031] Inductors 42 and 43 and varactor 41 form a high-Q circuit and are chosen to resonate at the input frequency for some value of V. Since V varies as the RF, the resonant frequency and gain of the resonator will also be different on the positive and negative halves of the RF cycle. This asymmetric signal gain leads to the development of a DC component V_DC of voltage across the capacitor 46, i.e., rectification.

[0032] The power extraction system can be adaptively adjusted for optimal performance by using floating gate transistors as adaptive elements. The threshold voltage of floating gate transistors can be changed by adding or subtracting charge from the floating gate. A lower threshold voltage improves the performance of the switching rectifier and charge pumps described in the following sections by increasing the rectified current for a given input RF amplitude.

[0033] In addition, the highest Q that can be used for the input LC tank shown in FIG. 2 is limited by resonant frequency variations caused by environmental factors and manufacturing tolerances. This limits the passive voltage gain obtainable from the tank. To overcome this problem, a high-Q system which maintains a constant resonant frequency by adapting the C(V) characteristic appropriately is utilized. This can be done by using a floating gate MOS capacitor for C(V).

[0034] Still another technique for performing far-field RF power extraction requires using switching rectifiers to avoid voltage drops associated with diode rectifiers. By using the differential RF inputs to operate transistors as switches and not as diodes, the threshold voltage drop associated with diode rectifiers is reduced considerably. FIG. 5 shows a circuit that achieves this. When the phase of the RF input is such that gate of the upper PMOS transistor 44 is low, it turns on, drawing current from the high side of the RF input, thereby charging the load capacitor 46. The other transistor 45 is off during this phase. During the opposite RF phase, the roles of the two transistors are reversed, but the load capacitor is still charged upwards. The circuit thus acts as a full wave rectifier and charges the load capacitor C_L 46 towards the positive envelope of the input RF voltage V_RF sin(ωt). The final rectified DC voltage is determined by the resistance of the transistors and the load resistance connected to the output. It should be evident that by replacing the PMOS transistors with NMOS ones, the charging direction of the load capacitor can be reversed, i.e., charge can be made to flow out of the load capacitor, thereby decreasing the output voltage. Thus by adding a pair of NMOS devices 144, 145 in parallel with the PMOS pair 44 and 45, a negative voltage (referenced to the common mode voltage of the RF input) is generated, giving a total output DC voltage of approximately 2V_RF Note that MOSFET devices, being truly bidirectional devices, are ideally suitable for this circuit, where they are operated as switches This makes the integrated implementation of the circuit using standard low cost IC fabrication processes, such as CMOS, feasible.

[0035] FIG. 6 comprises a graph 48 which shows the simulated rectification curve 49 for the nonlinear capacitance when it was implemented as a MOS capacitor in a 0.5 μm CMOS process. The RF input amplitude was V_RF at 900 MHz. It can be shown that if the MOS devices operate in the sub-threshold region, this technique, which works by parametrically modulating the capacitance of a nonlinear capacitor 41, allows the DC current charging the load capacitor C_L 46 to increase by a factor of cosh

[0036] $\left(\frac{\kappa V_{\mathrm{DC}}}{V_T}\right),$
where K is the sub-threshold body bias coefficient and the thermal voltage

[0037] $\left(\frac{\kappa V_{\mathrm{DC}}}{V_T}\right),$

[0038] FIG. 7 shows the results of a SPICE simulation 140 of the switching rectifier implemented in a 0.5 μm CMOS process with W/L=300 for both PMOS and NMOS devices. The input RF voltage amplitude V_RF was 0.5 V at 900 MHz. The charging time for a load capacitance C_L =50 pF was 1 μs.

[0039] Another technique for performing far-field RF power extraction uses charge pumps to increase the rectified output voltage. Since the input RF amplitude is extremely low (insufficient to operate the circuitry needed by the tag), charge pumps are used to increase the output DC voltage. FIG. 8 shows a single charge pump cell 50 using cross-coupled charge transfer switches. The differential RF input voltages φ and φ are used to pump charge unilaterally through the pump capacitors C_P 55, 56, thus making V_OUT>V_IN. The circuit works as follows. When φ is high (=V_RF in this case), the bottom PMOS transistor 54 and the upper NMOS transistor 51 are turned on, while the other two transistors 52 and 53 are turned off. The other end of the pump capacitor 56 connected to φ is charged to its previous value (=V_IN)+V_RF Current flows through the PMOS transistor 54 from this node, charging up the output towards (approximately) V_IN+V_RF. At the same time, the upper NMOS transistor 51 charges the other end of the pump capacitor 55 connected to φ to V_IN The whole procedure is repeated during the opposite phase, when the RF input polarities are reversed. During this phase, the upper PMOS transistor 52 and the lower NMOS transistor 53 turn on, the other two transistors 51 and 54 turn off and the output is again charged towards V_IN+V_RF.

[0040] Ideally, V_OUT=V_IN+V_RF, i.e., a single cell acts as a voltage adder—it adds the RF amplitude to the input voltage. By cascading N of these cells in series, the output voltage is increased under no load conditions to (N+1)V_RF, where the input voltage to the first cell is assumed to be V_RF. Practically, parasitic capacitances to ground at the charge pumping nodes and increasing body bias effects on the NMOS devices limit the available voltage gain. This assumes that a typical n-well CMOS process is being used. This limitation can be removed if a more expensive dual-well process is used instead. In addition, by reversing the input and output terminals, the same circuit can be used to pump charge in the reverse direction and thereby generate large negative voltages. By combining charge pumps pumping in opposite directions with the switching rectifier described previously (which is assumed to generate approximately 2 V_RF across its outputs), an output DC voltage of 2(N+1)V_RF can be generated.

[0041] FIG. 9 shows the result of a SPICE simulation 57 of the switching rectifier 40 of FIG. 5 combined with the charge pump 50 shown in FIG. 8 implemented in a 0.5 μm CMOS process. Width (W)/Length(L)=60 for the devices in the switching rectifier, the pump capacitance C_P=1 pF and 2 cascaded cells were used to pump charge in each direction (N=2). The overall system was similar to that shown in FIG. 1, with C_L=10 pF. The input RF amplitude V_RF was 0.55 V at 900 MHz.

[0042] The apparatus may utilize solar and other sources of ambient power for starting up the power collection module. For best performance, the power collection module can adapt to changing environmental conditions, for example by automatically adjusting the resonant frequency of the antenna using a feedback tuning network (shown in FIG. 1). However, such adaptation loops consume power, which may not be available from the transmitted RF. To ensure startup power to run the adaptation loops, alternative sources of ambient power (apart from the transmitted RF) may be used. The amount of power required from these sources is small (typically in the nano watt range) since the system only needs to adapt slowly (time scale of milliseconds to seconds) and can thus be designed to have very low power consumption. Promising power sources include solar radiation (utilizing light energy using on-chip solar cells), ambient mechanical vibrations (using MEMS transducers), thermal gradients (using thermoelectric materials) and the like. FIG. 10 shows a solar cell structure 150 which can be used to provide startup power for the power collection module. The solar cell structure uses 3 solar diodes 152 in series to charge load capacitor 154, thereby providing a D.C. voltage across the load capacitor 154.

[0043] A feedback tuning network 70 is shown in FIG. 11. The objective is to maximize the output DC voltage of the system, subject to a poorly known and possibly time varying open circuit RF amplitude across the antenna terminals. The primary cause of this is antenna resonant frequency variations. The resonant frequency (and other properties, such as the radiation pattern) of the antenna on the power collection module can vary by significant amounts because of variations in nearby environmental conditions. One way to minimize the effects of this variation is to use a broadband antenna, but this lowers the input Q and decreases the power up range. To prevent this, a feedback controller 70 is used to regulate the antenna resonant frequency (FIG. 11). The controller uses derivative-based control to regulate the antenna resonant frequency. The controller comprises a state machine which operates on the temporal derivative of the rectified RF voltage and tries to keep it positive, i.e., increasing in time. The controller outputs a voltage which is used to control the capacitance of a MOS varactor, thereby controlling the resonant frequency of the antenna.

[0044] The controller shown in FIG. 11 can be implemented using subthreshold CMOS logic and consumes very little power. The controller can be powered off the rectified RF supply or from an alternative source of ambient energy, such as a solar cell. Two controllers running in different frequency ranges can be used to adapt the resonant frequency to variations that occur on different time scales. For example, a slow control loop could adjust to mechanical movement around the power collection module, while a faster loop could adjust to more rapidly varying quantities, such as multipath fading effects on the received RF signal.

[0045] The output of the switching rectifier (V_ENV, which is the DC voltage to be maximized) is provided to slope detector 210. Slope detector 210 comprises a PMOS device 202 having a source coupled to the output of the rectifier and charge pumps (V_ENV), a gate coupled to the clock generator 260 and a drain coupled to storage capacitor 200. The storage capacitor 200 is coupled between the drain of PMOS device 202 and a reference ground. A buffer 212 has a first input coupled to the output of rectifier and charge pumps and a second input coupled to the storage capacitor 200, Buffer 212 provides a first buffer output and a second buffer output. A comparator 214 receives the buffer outputs and provides a comparator output. A latch 216 receives the output of comparator 214 and a clock input, and provides the slope detector output. In use, the output of the rectifier and charge pumps is sampled and held on the capacitor C_S 200 of slope detector 210. This value is compared with the actual value of V_ENV. This operation is a discrete time approximation to the time derivative, and the output C of the slope detector 210 is a 1-bit estimate of the slope of V_ENV.

[0046] A predictor circuit 220 comprises a latch 222 receiving a clock input and a data input and providing an output to exclusive-or gate 224. The exclusive-or gate 224 also receives the slope detector output and provides a predictor output which is also coupled to data input of latch 222. In use, the predictor takes the current value of C, combines it with information about the previous correction made to the antenna resonant frequency and generates a control signal. This control signal is fed into an integrator 230.

[0047] Integrator 230 comprises a PMOS device 232 receiving a positive bias voltage at a gate, having a source coupled to a reference voltage VP and providing an output at a drain. A second PMOS device 234 has a source coupled to drain of PMOS device 232, a gate coupled to the output of predictor 220 and a drain providing an output of the integrator 230. Integrator 230 further includes an NMOS device 238 receiving a negative bias voltage VN at a gate, having a source coupled to a reference ground and provides an output at a drain. The second NMOS device 236 has a source coupled to NMOS device 238, a gate coupled to output of predictor 220 and a drain couple dot the drain of PMOS device 234 and also providing an output of integrator 230.

[0048] The integrator 230 output voltage V_C controls the antenna resonant frequency by changing the capacitance of the MOS varactors 240, 250 connected across the antenna output terminals. In the preferred implementation that is shown in FIG. 11, the predictor control law is defined as:
ΔV_C,n+1=C⊕ΔV_C,n
where ΔV_C,n+1 is the new correction to be made to V_C, ΔV_C,n was the previous correction, and ⊕ denotes the logical XOR operation. This control law is that of a simple ‘bang-bang’ controller. Every time the controller makes a right decision (V_ENV increases and its slope C is positive), it repeats it on the next time step. Every time the controller makes a wrong decision (V_ENV decreases and its slope C is negative), it reverses its previous decision on the next time step. An oscillator (clock generator) 260 generates the sampling and timing signals for the rest of the system. Necessary current and voltage biases are generated by a bias generator 270. Typically this takes the form of a supply-independent current reference circuit.

[0049] The physical structure of an RF power extraction system is shown in FIG. 12. The input RF signal is captured by the antenna 20. The antenna impedance is matched to the rest of the system using an appropriate matching networks 30. The signal is then passively amplified using a high-Q resonator and fed to CMOS circuitry 80 which rectifies it and generates one or more DC voltages. These DC voltages are then used for powering other circuits on the chip 80, such as an RFID transceiver. The entire system is disposed on a flexible substrate 90.

[0050] The present application has focused on RFID tags operating in the UHF frequency band since it is of commercial importance, but the applicability of the present inventions is not confined to this application or frequency band. Since they are of a fairly general nature, the innovations described in thus application can be applied over a broad range of frequencies and power levels for various self-powered applications.

[0051] Currently, the minimum RF power threshold for self powered devices is in the 50–60 μW range. By way of implementing one or more of the techniques and circuits described above, this threshold is reduced to 3 μW or below. This results in a concomitant increase in the maximum read range by a factor of 4 over current designs.

[0052] Having described preferred embodiments of the invention it will now become apparent to those of ordinary skill in the art that other embodiments incorporating these concepts may be used. Additionally, the software included as part of the invention may be embodied in a computer program product that includes a computer useable medium. For example, such a computer usable medium can include a readable memory device, such as a hard drive device, a CD-ROM, a DVD-ROM, or a computer diskette, having computer readable program code segments stored thereon. The computer readable medium can also include a communications link, either optical, wired, or wireless, having program code segments carried thereon as digital or analog signals. Accordingly, it is submitted that that the invention should not be limited to the described embodiments but rather should be limited only by the spirit and scope of the appended claims. All publications and references cited herein are expressly incorporated herein by reference in their entirety.