The rest of the thesis is organized as follows: an overview of solar energy harvesting applications technique is introduced in the Chapter 2. The characteristics of photovoltaic (PV) cell and a maximum peak power tracking technique are introduced.
The detail circuit of previous power management system for solar energy harvesting applications is also introduced.
The techniques of improving the stability and transient response of voltage regulator are introduced in the Chapter 3. The new voltage regulators with high current efficiency and fast load regulation are proposed in the Chapter 3.
The voltage doubler and Dickson charge pump are introduced in the Chapter 4. The techniques of improving the pumping efficiency of charge pump are also introduced.
For improving the power efficiency of charge pump which generates ultra high voltage, a novel connect scheme is proposed in the Chapter 4.
An integrated power management system for solar energy harvesting applications is proposed in the Chapter 5. The power management system contains the new voltage regulator in Chapter 3 and the new charge pump in Chapter 4. The power management system has two power sources. Thus, a control unit is proposed to decide who will supplies energy to the power management system. The control unit also helps the power management system efficiently using the energy from battery.
Chapter 2
An Overview of Solar Energy Harvesting Applications
In this chapter, it introduces the overview of solar energy harvesting applications.
The photovoltaic (PV) cell would be described in Section 2.1. A maximum peak power tracking would be introduced in Section 2.2. Section 2.3 and Section 2.4 introduce two power management systems for solar energy harvesting applications.
Finally, Section 2.5 introduces the battery management system for solar energy harvesting applications.
2.1 Photovoltaic (PV) Cell
The photovoltaic cell is used to convert the light energy into electrical energy.
The photovoltaic cell is a nonlinear device and can be represented as a current source model, as shown in Fig. 2.1. The traditional I-V characteristic of a photovoltaic cell, when neglecting the internal shunt resistance, is given by the following equation [6]:
Where Ioand Voare the output current and output voltage of the photovoltaic cell, respectively, Igis the generated current under a given insolation, Isatis the reverse saturation current, q is the charge of an electron, K is the Boltzmann’s constant, A is the ideality factor for a p-n junction, T is the temperature (K), and Rsis the intrinsic series resistance of the solar array.
Fig. 2.1 Equivalent circuit of a photovoltaic cell.
The saturation current of the photovoltaic cell varies with temperature according to the following equation [6]:
3
Where Ior is the saturation current at Tr, T is the temperature of the photovoltaic cell (K), Tr is the reference temperature, EGOis the band-gap energy of the semiconductor used in the solar array, KI is the short-circuit current temperature coefficient and λ is the insolation in mW/cm2. In literature, instead of the I-V characteristic given by (2.1) the following I-V characteristic is used to compute the output voltage of the PV cell:
ln
q o s a tFrom above equations, the electric power generated by a photovoltaic cell fluctuates depending on the solar radiation value and temperature, as shown in Fig. 2.2.
Fig. 2.2 P-V curve of photovoltaic cell under different temperature.
2.2 Maximum Power Point Tracking and Control Algorithm
2.2.1 MPPT Process
The electrical characteristic of the PV under a given insolation is illustrated in Fig. 2.3. For different loadings, the PV cell will operate at different point, either in current source region or in voltage source region, where the output current or voltage almost keeps a constant. In other words, the internal impedance of the solar array is low on the right side of the curve and high on the left side. The maximum output power point of PV cell exists at the crossing point of the two regions. According to the maximum power transfer theory, the power delivered to the load is maximum when the source internal impedance matches the load impedance. Thus, the impedance seen from the converter side (which can be adjusted by controlling the duty cycle) needs to match the internal impedance of the solar array if the system is required to operate close to the maximum power point of the photovoltaic cell.
Most traditional dc/dc converters have a negative impedance characteristic inherently, due to the fact that their current increases when voltage decreases. This behavior is due to the constant input power and the adjustable output voltage of the power supply. If the system operates on the high-impedance (low-voltage) side of the PV cell characteristic curve, the output voltage of PV cell will collapse. Therefore, the PV cell is required to operate on the right side of the curve to perform the tracking process. Otherwise, the converter will operate with the maximum duty cycle, and the output voltage of PV cell will only change with the insolation. Thus, the system cannot achieve maximum power tracking and might even mistake the present operating point for the maximum power point.
0 200 400 600 800 1000
Fig. 2.3 The characteristic curves of PV cell.
Fig. 2.4 The flowchart of MPPT control.
The control flowchart of the maximum power tracking system is shown in Fig. 2.4[1]. If a given perturbation leads to an increase in output power of PV cell, the next perturbation is made in the same direction. In this way, the maximum power tracker continuously seeks the maximum power point.
2.2.2 Control Algorithms for MPPT
Two control algorithms often used to achieve the maximum power point tracking are the perturbation and observation method and the incremental conductance method.
Although the incremental conductance method offers good performance under rapidly changing atmospheric conditions, four sensors are required to perform the measurements for computations and decision making [7]. If the sensors or the system require more conversion time, a large amount of power loss will result. On the contrary, if the sampling and execution speed of the perturbation and observation method is increased, then the system loss will be reduced. Moreover, this method only needs two sensors, which results in the reduction of hardware requirement and cost.
Two different control variables are often chosen to achieve the maximum power control [8].
1) Voltage Feedback Control: This method assumes that any variations in the insolation and temperature of the PV cell are insignificant and that the constant
reference voltage is and adequate approximation of the true maximum power point.
The output voltage of PV cell is used as the control variable for the system. The system keeps the array operating near its maximum power points by regulating the output voltage of PV cell and matches to a fixed reference value.
The control method is simple, but it has the drawback of neglecting the effect of the insolation and temperature of the PV cell. This method cannot be widely applied in the battery energy storage systems. Therefore, the control method is only suitable for use under the constant insolation condition, such as a satellite system, because it cannot track the maximum power points of the PV cell when variations in the insolation and temperature occur.
2) Power Feedback Control: The actual output power of PV cell, instead of its estimate from measurements of other quantities, is used as the control variable.
Maximum power control can be achieved by forcing the derivative (dP/dV) equal to zero under the power feedback control. A general approach to the power feedback control is to measure and maximize the power at the load terminal. However, it maximizes the power to the load, not the power from the PV cell. A converter with MPPT offers high efficiency over a wide range of operating points. The full power may not be delivered to the load completely, due to the power loss for a converter without MPPT. Therefore, the design of a high-performance converter with MPPT is a very important issue.
2.3 Ultra Low Voltage Power Management and Computation Methodology for Energy Harvesting Applications
An ultra low voltage power management system for energy harvesting applications is illustrated in Fig. 2.5[2].
Fig. 2.5 Ultra low voltage power management system for energy harvesting applications.
The ultra low voltage power management system consists of four blocks, the energy harvesting mechanism, the power management system, the computation module and the charge-based control unit. A generic energy harvesting mechanism scavenges the energy from the environment, converts the energy into an electrical energy source and gives out an unregulated source voltage. Solar energy system is one example.
In this application, the output voltage of the energy harvesting mechanism is in the range of 100mV. The unregulated voltage is then input to the power management system. The power management system consists of a high conversion ratio integrated charge pump which steps up the voltage to >1V, and this unregulated voltage (Vs) then drives the controller such that the charge pump is self-powered once it is jump-started. The jump-start circuit can be a small primary battery that only needs to supply energy during start-up, and will be cut off from the system when self-powered operation is in place. To reduce the cost, the unregulated voltage (Vs) will be used directly by the computation unit. Therefore the computation unit has to cater for the fluctuation of the supply voltage. Because of the unstable source of the input energy, the energy may not be enough to sustain continuous operation of the computation.
The charge-based control unit will make sure the energy available is enough for an atomic operation of the computation. Also it will trigger the calculation when enough energy is available.
2.3.1 Power Circuit
The power circuit of this power management system is shown in Fig. 2.6[2]. The power circuit mainly has two parts, a four stage 16x exponential charge pump circuit, and a clock generator. Since the voltage source Vin is around 80mv~200mv, the circuit needs a start up circuit, which only functions at the beginning of the circuit running. Once the circuit is started, the generated high voltage source Vout will provide the energy to the clock generator and cut the switch between the start-up voltage and the circuit. The circuit will then form an energy loop and be self powered, and at the same time provide energy to the outside circuit.
Fig. 2.6 Block diagram of the power circuit.
In the charge pump, two sets of clock driving signals are used: (φ1, φ2) and (φ1h,φ2h). The generators of these signals are shown in Fig. 2.7[2]. To save energy, in the generator for (φ1, φ2), all the inverters, except the last one of the buffer stage, has a swing between Vdd and Vin. For (φ1h,φ2h), the conventional level shifter is eliminated and a CMOS inverter is used to generate a driving signal swing between 2Vout and 4Vin. By doing so, the clock swing is reduced and the energy can flow back to some nodes in the charge pump circuit, which leads to more energy saving.
2.3.2 Computation Module
The power supply becomes unstable without the regulator. Variation in supply voltage affects the delay of the circuit and may cause timing problem for the computation. Therefore it is better if the computation unit can track the change of the supply voltage and automatically adjust the performance of the digital circuit to cater for the change. A self-time asynchronous pipeline design [9] is used to implement the computation module. In this case the operation of a pipeline stage is not dependent on a global clock but only on the completion of the previous stages. It is more robust over various operating conditions for its locally-generated timing signals and it is more suitable for the design to track with an unstable supply voltage. To simplify the asynchronous hand-shaking protocol and to cater for the static CMOS library, bundle delay method is used to implement the asynchronous pipeline. The bundle delay is designed to be a little bit larger than the computation delay with a safety margin for the process variation. When the supply voltage fluctuates, the bundle delay will automatically track the change and synchronize the computation in the pipeline. This can work over a large fluctuation in the supply voltage.
Fig. 2.7 Clock generator.
2.3.3 Charge-Based Control Unit
Due to the unpredictable nature of the energy source, sometimes the energy available at a particular time interval may not be enough for a certain computation and if system carries out the computation, the computation may not be able to finish and even worse the voltage may drop to a level that some of the stored data will be lost.
To cater for this situation, a charge-based computation methodology is used. The charge required at different voltage levels for a certain atomic computation is estimated and stored in the device. The atomic computation refers to a computation that will results in data stored in the memory, or data output to external world and does not need to be used again. The atomic computation will only be triggered when the harvested energy can provide the charge for it with some margin. In addition, a supply-side paradigm of computation is used. For some very low energy applications, such as environmental data collection using wireless sensor device, the computation does not have a hard deadline requirement. Based on the energy available, the system can decide whether an operation should be triggered and executed. Moreover, the computation can be prioritized, both in task level or bit level (an example of bit level priority is multiple bit resolution). Depending on the energy available and the priority, different computations will be executed based on the decision of the charge-based control unit.
2.4 A Micro Power Management System with Maximum Output Power Control for Solar Energy Harvesting Applications
The micro power management system with maximum output power control for solar energy harvesting applications is shown in Fig. 2.8[3].
Fig. 2.8 Block diagram of micro power management system with maximum output power control for solar energy harvesting applications.
To cater for different light intensities, the micro power management system tracks the voltage of the PV cells and the battery to determine whether the charge pump is used or bypassed. Under low light intensity, the PV voltage is low and the charge pump steps up the voltage either for charging the rechargeable battery or powering the computation circuit. At the same time, the optimal power tracking unit monitors the charge pump output power and determines the adjustment of the system operating parameter. Based on the adjustment decision, the control unit tunes the operating frequency of the charge pump in order to maximize the system power output.
Rechargeable battery is used to support the system continuous operation even when the light source is out.
2.4.1 Optimal Power Tracking Unit
In order to supply maximum power to charge the battery or to directly drive the computation circuit, the optimal power tracking unit is used to monitor the amount of power flowing out of the charge pump. The detail circuit is shown in Fig. 2.9[3].
Fig. 2.9 Circuit of the optimal power tracking unit. (①normal working phase, ② sensing phase, ③evaluation phase)
The optimal power tracking unit monitors the charge pump output power and generates the adjustment decision for the switching frequency of the charge pump so that the system is working around the optimal output power point. The optimal power tracking unit contains two parts. Since the system output voltage is regulated by the battery and the voltage of the battery changes very slowly, thus, maximizing the system output power is equivalent to maximizing the system output current. The circuit uses current sensor to measure the charge pump output current. Based on the measured current value, the tracking unit checks whether the system is at the optimal point and generates the corresponding decision signal to tune the system parameters.
Basically a generic hill climbing algorithm is used in the circuit for the optimal point tracking. The charge pump switching frequency affects the system output current.
Hence the tracking unit generates the corresponding voltage value to vary the switching frequency through a VCO in the control unit. The tracking unit consists of a current sensor. With the biasing current through MN1and MN2, the source voltages of MP3 and MP4 are clamped the same. When the current sensor power supply VO is connected to the charge pump output, due to the size ratio of MP1and MP2, about 1/N of the total output current flows to the resistor RSthough the transistor MS. Hence the output current level from the charge pump is reflected by the voltage drop across RS
(VS), which is then sent to the decision generation circuit. The rest of the output current would flow to the battery or the computation circuit through MP2 at node VOUT. Since the output current from the charge pump is in pulse shape, a smoothing capacitor is connected to the node VO to obtain a smooth current profile for the accurate measurement of the average value of the output current through the current sensor. The sensed current which is represented by the value of VS, is sent to the decision generation circuit to implement the generic hill climbing algorithm. By comparing the current sensed current value with the previous one, the circuit can determine the direction of the change in output current and make the decision on whether to increase or decrease the switching frequency. This action continues and the output current will oscillate around the maximum current point finally. In the decision generation circuit, the current sensed VS value and the previous value are stored in the sample capacitors, CPP1 and CPP2, alternatively. If CPP2 holds the previous value, the transmission gates T3 and T4are off while T1and T2are on. The current sensed VSwill be stored in CPP1, and Vcheckwill output the comparison results from VO+of the comparator. Vcheckis equal to logic ‘1’ if the current sample value is larger than the previous one. In the next sample period, T1and T2 will be turned off while T3and T4 are on, and CPP1 holds the old sample while CPP2stores the current VS value. Vcheck will output the comparing results from VO- for this sample period.
This value is XNORed with the previous adjustment decision which is stored in a
Fig. 2.10 Control unit for the charge pump.
D-flipflop and the new decision value Vactionwill be updated at the rising edge of the control pulse SE. The logic value of the decision signal Vaction indicates whether the charge pump switching frequency should be increased or decreased. Depending on the Vaction logic value, the capacitor CVCO is either charged by the current ib2 through transistor ME1, or discharged by the current ib2 through transistor ME4, during the control pulse interval of SE. In this way, the voltage of VVCO is either increased or decreased and it is then sent to the control unit to adjust the switching frequency through the VCO.
In order to reduce the power overhead, the optimal power tracking unit operation is divided into three phases during each sample period, which is shown in Fig. 2.9. Here, the sample period means the time interval between 2 consecutive frequency adjustments. During each period, the transmission gate control signal SPP is either high or low, which determines whether the previous sampled VS or the current one is
In order to reduce the power overhead, the optimal power tracking unit operation is divided into three phases during each sample period, which is shown in Fig. 2.9. Here, the sample period means the time interval between 2 consecutive frequency adjustments. During each period, the transmission gate control signal SPP is either high or low, which determines whether the previous sampled VS or the current one is