You are on page 1of 13

254

IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 1, NO. 3, SEPTEMBER 2011

Efcient Design of Micro-Scale Energy Harvesting Systems


Chao Lu, Student Member, IEEE, Vijay Raghunathan, Member, IEEE, and Kaushik Roy, Fellow, IEEE
(Invited Paper)

AbstractMicro-scale energy harvesting has emerged as an attractive and increasingly feasible option to alleviate the power supply challenge in a variety of low power applications, such as wireless sensor networks, implantable biomedical devices, etc. While the basic idea and system composition of micro-scale energy harvesting systems have been explored and applied in a number of prototypes in recent years, designing micro-scale efcient energy harvesting systems require an in-depth understanding of various design factors and tradeoffs. This paper provides an overview of the area of micro-scale energy harvesting and addresses various challenges and considerations involved from circuit, architecture and system perspectives. This paper explicitly highlights several subtle but essential design differences that distinguish energy harvesting systems from battery-powered embedded systems. Moreover, we envision the necessity and importance of developing a simulation tool that enables design space exploration and quick performance evaluation at the early design phase. The practical issues, challenges and considerations for implementation of this envisaged simulation tool is discussed in this paper. Index TermsDesign and optimization consideration, energy harvesting, micro-scale, simulation and evaluation tool.

I. INTRODUCTION

APID advances in nanoscale integration have resulted in a new class of highly miniaturized electronic systems (e.g., smart dust sensors [1], biomedical implants [2], [3], etc.) that enable new application domains. Energy storage element (e.g., battery) is conventionally used for powering electronic systems. However, since the space permitted for battery integration in these miniaturized systems is quite tiny (and hence limited energy capacity), the energy storage element will be quickly depleted after a short time and then these electronic systems will stop operation. Frequent battery replacement is impractical in these micro-scale systems, since it is prohibitively expensive (e.g., a large sensor network consisting of thousands of distributed sensor nodes) or infeasible (e.g., invasive surgery is required for battery replacement of implanted pacemakers every six to seven years [4]). Loss of power in a biomedical implant due to a drained energy supply can have serious and potentially
Manuscript received January 09, 2011; accepted April 22, 2011. Date of publication September 01, 2011; date of current version November 09, 2011. This work was supported in part by National Science Foundation and in part by Intel Corporation. This paper was recommended for publication by Guest Editor E. Macii. The authors are with the Department of Electrical and Computer Engineering, Purdue University, West Lafayette, IN 46323 USA (e-mail: lu43@purdue.edu; vr@purdue.edu; kaushik@purdue.edu). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/JETCAS.2011.2162161

life threatening consequences. As a result, one key design challenge in these miniature systems is to conveniently provide the required power for long-lived, maintenance-free operation. A promising approach to overcoming the problem of limited energy availability in these micro-systems is to scavenge energy from ambient sources, such as solar radiation, thermal gradients, mechanical motions or vibrations, etc. In the past few years, a number of micro-scale energy transducers have been fabricated to convert energy from other modalities into electrical energy. The use of energy harvesting techniques signicantly prolongs overall system lifetime and in some cases, it is possible to eliminate the dependence on battery and directly contributes towards mitigating the energy limitation problem [5], [6]. While the basic idea of environmental energy harvesting has been widely explored and applied in some prototypes, either in chip level [7][10] or board level [11][15], the design of optimized energy harvesting systems require an in-depth understanding of various design factors and tradeoffs, which has not yet been fully investigated and explored in literature. For instance, a power converter is a key component in micro-scale energy harvesting systems. It plays a vital role in determining the amount of electrical power that can be delivered to an energy storage element. Even though power converters have been extensively used in battery-powered embedded systems, the optimization objective in those systems is very different from microscale energy harvesting systems. Existing optimization techniques for power converters in battery-powered systems, when used for energy harvesting systems, will result in substantial performance degradation and represent nonoptimal solutions. In this paper, therefore, we will review and discuss various design factors that help to optimize circuit-level design of power converters for micro-scale energy harvesting systems. In addition to addressing circuit-level optimization for efcient power conversion, we also discuss numerous design tradeoffs and considerations at the architecture level. Although researchers keep on ameliorating the performance or efciency of individual system building component, however, as revealed in [11] and [66], it is absolutely necessary to coordinate the entire system behavior from architecture perspective. This is because these building components operate interactively. Improving one component often results in severe performance degradation of other components. As a result, the energy harvesting efciency of entire system is often worse than before. In this paper, we will address several design tradeoffs and discuss the impact of these building blocks on architecture level.

2156-3357/$26.00 2011 IEEE

LU et al.: EFFICIENT DESIGN OF MICRO-SCALE ENERGY HARVESTING SYSTEMS

255

Fig. 1. Block diagram of a micro-energy harvesting system.

Nowadays, energy harvesting system design usually follows a developing manner of trial-and-error approach. Hence, system performance heavily depends on the designers experience. Faced with a great number of design parameters or choices, even an experienced designer needs to spend much time and effort for guring out a better design solution. In addition, it is very difcult to quickly assess and evaluate the entire system performance at the early design phase. Therefore, design solutions can only be evaluated at a later phase of the design. If the evaluation shows a functional failure or poor efciency, a new design cycle can be required. Such design methodology involves high complexity and suffers from long time-to-market. To deal with this design challenge, we envision the necessity and the importance of developing a system-level design simulation tool, which can allow designers to quickly evaluate the impact of various design choices and parameters at the early design phase, while architecting micro-scale energy harvesting systems. As the simulation tool enables systematic design space exploration, it can signicantly reduce the inherent design complexity and greatly shorten the development time. The rest of this paper is organized as follows. In Section II, we review the state-of-the-art research on micro-scale energy harvesting systems. Then, we discuss system components and optimization techniques. In Section III, we highlight several essential design differences that distinguish micro-scale energy harvesting systems from battery-powered embedded systems. From Section IV to VI, we address and elaborate various crucial design considerations and tradeoffs from circuit, architecture, and system perspectives, respectively. We also illustrate these design considerations using a case study in Section VII. Finally, the conclusions are drawn in Section VIII. II. MICRO-SCALE ENERGY HARVESTING SYSTEM A. System Overview Fig. 1 shows the generic block diagram of a micro-scale energy harvesting system. It comprises ve main blocks: the micro-scale energy transducer, the power converter, the control unit, the energy buffer, and the application unit. This system involves the basic processes of energy conversion, transfer, storage, and consumption, as denoted in Fig. 1. The energy transducer converts ambient power from another modality into electrical power, which will be used for powering the application unit or recharging the energy storage buffer. The energy transducer may be based on a single energy conversion mechanism [16][20] or a hybrid mechanism [21], [22]. As the ambient energy sources are often unstable and varying with time,

the power output from an energy transducer is also changing. Due to this unstable power supply characteristic, an energy transducer is impractical to directly power the application unit. The power converter is used to condition the power output from the energy transducer. The goal of the power converter is to extract power from the energy transducer and transfer as much as power as possible to the energy buffer or application unit. The control unit plays a crucial role in reducing the system power loss and maximizing the overall system efciency. It ensures maximum power point (MPP) operation at all times by running a tracking scheme. The energy buffer usually has a large envaries with ergy storage capacity and its terminal voltage internal energy or charge status. The energy buffer that is re, supplies the required plenished by the harvested current to the application unit. The application loading current unit (e.g., a sensor node) dissipates the harvested power. Note a control unit could also be embedded in the application unit to manage the energy consumption from energy buffer. Next, these building components will be addressed and discussed in detail. B. Energy Transducer Environmental energy sources are ubiquitous in our surroundings. Examples of these energy sources include solar radiation, air ow, mechanical motion or vibration, thermal gradients, radio-frequency (RF) transmission, etc. A variety of micro-energy transducers have been developed to convert energy from other modalities into electrical energy [16][22]. While there has been (continues to be) extensive research from the device perspective to improve the fabrication cost, energy conversion efciency, and power density of these micro-energy transducers, it is crucial for system designers to be aware of their electrical characteristics in-depth in order to understand their impact on the system being powered. Although some physical and mathematical models have been presented in literature to characterize micro-scale energy transducers, these models are cumbersome, computationally intensive, and incompatible with circuit simulation software (e.g., SPICE). Hence, in the reminder of this subsection, we provide an overview of various energy harvesting modalities and their electrical characteristics. Then, we focus on describing how these transducers can be modeled from electrical perspective. 1) Micro-Photovoltaic Cell: A photovoltaic (PV) cell is a device that converts the light energy directly into electricity by the photovoltaic effect. Fig. 2 shows the equivalent electrical circuit model of a PV cell [23], which is composed mainly of a is the genercurrent source and a forward biased diode. is the parasitic series resistance, and ated photocurrent, is the equivalent shunt resistance. and are the output current and terminal voltage of the PV cell, respectively. Based on the circuit shown in Fig. 2, the output current of a PV cell can be expressed as

(1) is the reverse saturation current, is the electron Here, charge, is a dimensional factor, is the Boltzmann constant,

256

IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 1, NO. 3, SEPTEMBER 2011

Fig. 2. Electrical model of a PV module.

Fig. 4. Illustration of operation mechanism of a micro-TEG.

Fig. 3. Measured I -V and P -V characteristics of a PV cell.

and is the operating temperature. Experiments have been conducted for a commercial PV cell (#1-100 [19]) to validate this model. The PV cell was characterized under weak light (indoor) conditions. It was illuminated using a 40-W light bulb and the distance between them was adjusted to emulate varying light intensities. Various resistive loads were connected to the PV cell. The output voltage and current were measured. Fig. 3 plots the - curve of the PV cell obtained using (1) and measured exvalues t well perimentally. We can see that the measured with the values predicted by the electrical model. of the PV cell as a Fig. 3 also plots the output power function of its terminal voltage. As is evident from the gure, for a given light irradiance, there exists an optimal output voltage for the PV cell at which is maximized (e.g., 0.29 V for 784LUX). This point on the - curve is the PV cells MPP. Note that the MPP changes signicantly as the light intensity

changes. The goal of MPP tracking schemes for the PV cell is to ensure that the PV cell operates at its MPP at any given time. It can also be seen that the harvested power is very limited (in to 1.1 mW). Obviously, we the range of several hundred would like as much of this power as possible to be available to the load. Therefore, the power budget for an MPP tracking scheme in such a system is severely constrained (e.g., at most ), which requires the MPP tracking subsystem to be a few very carefully designed. 2) Micro-Thermoelectric Generator: Micro-thermoelectric generator (TEG) is scalable, reliable and does not require any moving part like vibration energy transducers. As a consequence, it is very appealing in micro-scale energy harvesting systems, such as human body powered biomedical devices. Micro-TEGs typically consist of multiple couples of p-type and n-type thermoelectric legs, which can output electrical energy by employing the temperature gradient between the hot surface (e.g., human body) and the cold surface (e.g., ambient air). These thermocouples are usually connected electrically in series and thermally in parallel to effectively make use of the limited surface area. When there is a temperature difference across it, the seebeck effect causes the moving of charged carriers and generates a terminal voltage. Fig. 4 illustrates the operation mechanism of a TEG. Micropelt MPG-D751 is a good example of small scale TEG devices [20]. The current and voltage values for different across it were obtained by the simulation tool supported by the manufacturer, as plotted in Fig. 5. The output power varies as . The maximum a function of output voltage for different output power is maintained when its output voltage is around half of the open circuit voltages. The open circuit voltage of a TEG is proportional to the number of leg pairs, the actual temperature difference and the seebeck coefcient, as shown in the equation below (2) It is evident that a TEG can be modeled as a voltage source in series with an internal resistor with the voltage source being proportional to . Such a model can be expressed using (3) and (4), where is a constant internal resistor (3) (4) 3) Micro-Fuel Cell: Micro-fuel cell is a viable alternative power source that converts fuel energy into electrical energy by

LU et al.: EFFICIENT DESIGN OF MICRO-SCALE ENERGY HARVESTING SYSTEMS

257

Fig. 6. V-I and P-I characteristics of a micro-fuel cell.

Fig. 7. Piezoelectric lm model and verication.

Fig. 5. Simulation results of TEG MPG-D751 (1T = 1

 5 K).

chemical reaction. A fuel cell is considered as a green power source, because the outputs of chemical reaction are environmentally clean. Fuel has a much higher energy density. For example, the energy density of a methanol is ve times higher than that of a lithium ion battery. Thus, it can achieve longer lifetime for the same weight or volume. With the increasing advances of MEMS technology, the size of fuel cells has been shrinking to chip scale [18] and integrating with circuit to form a system-in-package platform [24]. Fig. 6 shows a typical V-I characteristic (solid line) of a micro-fuel cell. When the current density increases from zero, the micro-fuel cell passes through three distinct operation regions: activation, ohmic and concentration polarization. Fig. 6 also shows the estimated output power (dashed line). It is obvious that the maximum power point (MPP) is located in the region of ohmic polarization. Most existing fuel cell models assume constant fuel ow and unchanging concentration [25]. As a result, these models are only applicable to predict steady-state, time-independent behaviors. An electrical model for a micro-scale direct methanol fuel cell was developed in [26]. This proposed model is capable of prediction of runtime, large/small signal steady state or transient responses. 4) Micro-Vibration Energy Transducer: Low level mechanical vibrations commonly occur in various household or industrial places, such as small microwave oven, oors, etc. Many of them have been measured [27] and their fundamental frequencies are less than 200 Hz. Electromagnetic, electrostatic and piezoelectric conversions are three kinds of possible methods for mechanical energy conversion. The rst approach gener-

ates an ac current in a coil, when there is a relative motion between the coil and magnetic eld. Due to the existence of bulky moving coil and magnet, this method is hard to be implemented in small volume systems [28]. Electrostatic conversion is performed through a vibration-driven variable capacitor. Vibrations cause the capacitance variation and lead to an electrical energy generation [29]. This mechanism is easy for micro-scale integration, but a separate voltage source is needed for system start-up. Piezoelectric energy transducers utilize the strain and deformation of a piezoelectric material. As it does not require a separate voltage source and is easily used in compact integration, this method has been gaining much attention. Fig. 7(a) shows the equivalent electrical model of a piezoelectric lm [7] and the generated current source is shown as (5) where depends on the vibration magnitude, size, and material of the piezoelectric lm, is the vibration frequency, and are the internal capacitance and resistance of the lm, respectively. is almost constant under a wide range of vibrais usually very large in practice. The output tion frequencies. voltage of a piezoelectric lm under vibration thus depends on the lms geometry, piezoelectric properties, the mechanical vibration level and the output load. This electrical model has been widely used in many research cm cm mm piezoelectric works [7], [30], [31]. A was used to conduct experiments for lm with verifying this model in [7]. The lm was mounted on a vibration platform and the vibration frequency was set at 60 Hz. Various resistive loads were connected to the lm and the peak-to-peak output voltages were recorded. Based on the piezoelectric model, the peak-to-peak output voltage is given by (6) The simulation data and measured data are plotted in Fig. 7(b). It can be observed that the electrical model ts

258

IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 1, NO. 3, SEPTEMBER 2011

Fig. 8. Equivalent circuit for a RF antenna.

Fig. 9. Diode bridge based ac/dc rectier.

well with the measurement data. Therefore, this model can be used in design and simulation of vibration energy harvesting systems. 5) Micro-Electromagnetic Energy Transducer: Antenna is the commonly used RF energy receiver, which converts surrounding electromagnetic waves into electrical energy. Generally, an antenna is designed for a specic frequency or a narrow frequency band. Fig. 8 shows its equivalent circuit, which consists of a RF-induced voltage source and internal impedance . The magnitude of the RF-induced voltage source depends on the ambient electromagnetic strength. The internal impedance is sensitive and varying with the received frequency. Impedance matching between the antenna and rectier is very important, therefore, an antenna is followed by a matching network to provide the maximum power transmission to the rectier. Then, the rectier performs the ac-dc conversion for the input RF power. C. Power Converter We review the research progress involved in power converter design for low power energy harvesting systems. Various optimization techniques is addressed and discussed in-depth. 1) AC/DC Rectier: The authors of [32] investigated to power an application unit by using the ac power harvested from vibrations. However, since dc supply is dominant for powering application units, ac-dc rectication is essential in piezoelectric or RF energy harvesting systems. A full-wave rectier is better than a half-wave one due to better utilization of the converter power. The basic circuit conguration of a full-wave rectier is a diode bridge, as shown in Fig. 9. The input sinusoidal is the output of a piezoelectric material or a voltage RF impedance matching network. The existence of two serial diodes in the conduction path signicantly decreases the rectier efciency. As a result, several circuit design techniques have been presented to improve the power efciency. Researchers rst presented to use Shockley diodes in the bridge structure [33]. As a Shockley diode has a lower threshold voltage, the power conversion efciency can be enhanced. To further reduce the voltage drop, active diodes were proposed to replace passive ones. In [34], the authors designed an active diode that is composed by a large PMOS transistor and a comparator. An external voltage supply was required as the power supply of the comparator. Thus, this design cannot be deployed

Fig. 10. Multiple-stage ac/dc rectier.

in passive energy harvesting systems where no backup energy source is available and active diodes are unable to start up initially. The researchers presented a PMOS cross-coupled, NMOS active diode driven ac/dc rectier [35]. This design is limited by the minimum amplitude of input voltage, which has to be higher than the sum of threshold voltages of a PMOS transistor and a NMOS transistor. Later, the authors of [7] presented a hybrid ac/dc rectier, which integrates passive and active rectier together. The passive rectier is used for self-starting up, while the active diode kicks in to replace the passive diode when the rectier has started up. Next, an efciency-enhanced ac/dc rectier was presented for biomedical implants in [36], where an unbalanced-biasing comparator was used to minimize the reverse leakage current. In [37], a bias-ip ac/dc rectier that uses a shared inductor was presented for piezoelectric energy harvesting. When the output magnitude of a piezoelectric material or an antenna is extremely low (e.g., 100 mV), multiple-stage ac/dc rectier is a good option to boost the output voltage, as shown in Fig. 10. First, a design model for improving the efciency of RF to dc power conversion was presented in [38]. Then, a novel multiple-stage ac/dc rectier for RFID applications was presented in [39]. In particular, the effects of conduction angle, leakage current and body effect were elaborated. Appropriate approximations were made to derive an analytical expression for output voltage and power efciency. Later, the researchers of [40] presented a oating gate technique for RF-dc rectier, whose power conversion efciency was signicantly improved due to a lower threshold voltage that can be programmed. 2) DC/DC Power Converter: There are several variants of dc/dc voltage converters (charge pumps, buck/boost converters, or low dropout regulators), each with its own distinct advantages and disadvantages. Charge pumps or buck/boost converters, although efcient, introduce power supply noise due to their inherent switching nature, making them more suitable for noise tolerant digital systems. The presence of bulky off-chip inductors in buck/boost converters causes large EMI noise and increases system volume and cost. Consequently, buck/boost converters have become less attractive in micro-scale energy harvesting systems that have

LU et al.: EFFICIENT DESIGN OF MICRO-SCALE ENERGY HARVESTING SYSTEMS

259

stringent constraints on system size. Low dropout regulator produce less power-supply noise, which is attractive for analog or RF systems that are noise sensitive. However, its dropout voltage has to be very low to maintain high power efciency. In contrast, charge pumps only consist of low-cost on-chip capacitors and transistors. This advantage makes it much desirable for low-cost full integration. Therefore, in this paper, we focus on discussing design issues and considerations for charge pump power converters. Charge pump power converters have been widely used in many battery-powered embedded systems to generate a high level voltage. The design considerations for a charge pump in these applications are output voltage ramp-up time, output ripple voltage, line or load transient response, power conversion efciency, etc. [41], [43]. However, as addressed in [53], in micro-scale energy harvesting systems, the most substantial design objective is the charge transfer capability instead of power conversion efciency. This results in new design optimization challenges. The charge transfer capability of a charge pump power converter depends on its implementation technology, input and output voltages, circuit topology, transistor sizing, and the number of stages it has. The rst proposed charge pump circuit is a Dickson charge pump that has a linear topology [42]. After that, a variety of optimized linear charge pumps were presented for low voltage applications [43], [44]. As studied in literature, the performance of linear topology charge pumps degrades substantially when used with an ultra-low voltage energy transducer. In order to further improve the charge transfer capability, cross-coupled charge pumps were proposed [45], [46]. However, when the input voltage is lower than the threshold of transistors, the crosscoupled conguration fails to function normally. Later, an exponential topology was presented to extract energy from ultra low voltage energy transducers [47][49]. However, because the charge sharing paths from energy transducers is quite limited (only exist in the rst stage), the charge transfer capability is severely impeded. In [50], the authors presented a new tree topology that has a reduced charge sharing time, leading to an improved charge transfer capability. Besides, simulation results shows this proposed design has an increase of up to 30% in the harvested output power, compared to linear topologies. In [51], various power losses inside a power converter were modeled, and an analytical design methodology was presented for optimal transistor sizing. A charge pump power converter with variable number of stages was presented for higher harvested output power [52]. The authors proposed a runtime digital control algorithm to tune the optimal number of stages. The optimal number of stages of a charge pump was theoretically analyzed and derived to maximize the charge transfer capability in [53]. 3) MIMO Power Converter: Since the energy harvested from a single energy source is usually varying with time and environment, a heterogeneous energy transducer has becoming increasingly attractive, where multiple energy transducers are combined to operate together, as shown in Fig. 11. This combination can potentially decrease the temporal variability in generated power and increase the total amount of harvested energy in a given time duration, especially if the modalities involved are carefully selected.

Fig. 11. Multiple-input multiple-output energy harvesting.

The simple design choice of implementing independent power converters for each energy transducer is neither economical nor form-factor efcient. In [54], the researchers proposed a load-matching approach to maximize the power utilization for multiple supply systems. Some research about single-inductor multiple-input multiple-output (SIMIMO) power converters have been reported recently. The adoption of a single inductor reduces the system implementation cost. In [55], a single inductor MIMO power mixer-charger-supply system was proposed for a hybrid fuel cell-lithium ion source. The authors of [56] presented a dual-input dual-output boost converter that consists of two sub-converters to provide two regulated outputs for the application and energy buffer. Further, a novel MIMO power converter [57] was proposed to achieve a maximum power extraction from the energy transducer when it performs the charge transfer function. D. Control Unit The control unit produces the required control signals for the entire system and ensures maximum power point (MPP) operation at all times by running a MPP tracking scheme. The MPP for an energy transducer varies with the strength of the environmental energy sources (e.g., light intensity for PV cells). The most substantial consideration for MPP tracking in micro-scale energy harvesting systems is to ensure that MPP tracking introduces minimal power overhead. This is because the output power from a micro-scale energy transducer is very limited to begin with (in the range of tens of to a few mW). As much of this power as possible should be delivered for use by the system being powered. 1) Design Time Component Matching: DTCM is a very simple approach and is adopted in the Heliomote solar harvesting wireless sensor node [59]. In this approach, the output of the PV cell is directly connected to a rechargeable battery with appropriate reverse current protection. Therefore, the battery terminal voltage dictates the operating point of the PV cell. Near-MPP operation is achieved through careful selection of the specic PV cell and battery used. Although the approximate nature of this method results in a sub-optimal operating point, there is zero tracking overhead. This enables this scheme to perform almost as well as a more precise operating voltage. 2) Reference Voltage Tracking: This MPP tracking method originated from empirical data analysis. Experimental measurements on PV cells have found an approximately linear relationand its open ship between the PV cells MPP voltage circuit voltage as (7)

260

IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 1, NO. 3, SEPTEMBER 2011

where is a parameter that depends on the specic PV cell. This method is suitable for micro-scale energy harvesting systems because it involves simple open-loop control and does not require any intensive computations. The drawback is that the PV cell is periodically disconnected from the load, which causes temporary power loss to the load. Further, there is a hardware cost involved in time multiplexing between normal operation and the open circuit mode of operation. To address these disadvantages, an improved design was presented in [60], where an additional tiny PV cell is used in the system as a pilot cell. The open-circuit voltage of the pilot cell is used in place of the open-circuit voltage of the main PV cell. This eliminates the necessity for doing any open circuit voltage sensing on the main PV cell. However, with this approach, the pilot cell should be carefully chosen to ensure that is close to that of the main PV cell. 3) Hill-Climbing/Perturb and Observe (P&O): Hillclimbing and Perturb & Observe methods essentially have the same operating principle. An MPP tracking procedure is periodically initiated. This procedure involves applying a small perturbation to the power converter; either by varying the duty cycle of a boost/buck converter [12] or the switching frequency of a charge pump [9]. This perturbation results in a small change in the operating point of the PV cell as well as its output power. Assume, for illustration, that the perturbation results in an increase in the terminal voltage of the PV cell. The output power is recorded and compared to the power output before the perturbation. If the perturbation results in a power increase, another perturbation in the same direction is performed. If the perturbation results in a power decrease, a perturbation in the opposite direction is performed that results in a decrease in terminal voltage. The process is continuously repeated until the MPP is reached. In steady state, the system oscillates around the MPP. 4) MPP Tracking for Hybrid Transducers: For hybrid energy transducers, MPP tracking becomes more complicated since each transducer may have its own distinct MPP point. If each energy transducer is connected to its own power converter and energy buffer, the MPP tracking process for each transducer can be completely independent because achieving system MPP is equivalent to achieving MPP for each energy source simultaneously. However, when a MIMO power converter and less energy buffers are used, MPP operation for one transducer may lead to less efcient operation of another. Achieving a system-wide optimal operating point is more challenging in this case and involves a careful analysis of the relative power outputs of the individual energy transducers and their inter-dependence. Very few MPP tracking schemes for hybrid energy transducers have been introduced in literature. An energy harvesting platform that involves two distinct energy transducers was presented in [12], where each energy transducer corresponds to a particular energy buffer, thus, MPP tracking is carried out independently. A RFID system was presented to scavenge ambient harvested energy by using a solar cell and a RF antenna [61]. Both energy transducers share one energy storage element and no MPP tracking is involved in this design.

E. Energy Storage Buffer In micro-scale energy harvesting systems, either rechargeable batteries or super-capacitors are viable for temporally buffering the harvested power. Super-capacitor is an electrochemical capacitor with higher energy density, and typically has capacitance in the range of tens of millifarads to hundreds of farads. They do not suffer as much from rate-capacity and aging problems that are severe challenges for rechargeable batteries. However, they usually have a signicantly higher leakage power loss than batteries. Experimental measurement in [65] demonstrates the leakage power of a super-capacitor is exponentially related with the terminal voltage. For example, the GreenCap [82] with capacitance of 200 F has a leakage current of 0.25 mA when the terminal voltage is 2.2 V. However, the leakage current rises to 2.25 mA, when the terminal voltage becomes 2.5 V. This fact indicates a voltage shift of 0.3 V gives rise to eight times increase of the leakage current. F. Application Unit While the previous discussion mainly focuses on energy extraction, transfer and storage, this subsection explores the complementary problem of ensuring that the harvested energy is used in an efcient manner by the electronic system being powered. The ideal case is when the system is able to run completely off the energy-harvesting source itself. We will review some performance/power scaling methods that enables the design of harvesting-aware application unit. These proposed techniques adapt the performance and power consumption of the application unit at runtime in response to the spatial and temporal variations in harvested energy. In [70], the authors presented an adaptive algorithm for harvesting-aware duty cycling of wireless sensor nodes. The authors choose to use duty-cycling between active and low power modes for the purpose of performance/power scaling. In [71], the authors proposed to operate the application unit in an energy neutral mode. A high-level model for characterizing time-varying energy transducers is developed. Based on this model, harvesting-aware power management techniques for a single sensor node or a sensor network was described. The research of [72][74] primarily focused on designing adaptive scheduling and voltage/frequency selection algorithms for energy harvesting systems. More research works on adaptive power management for energy harvesting embedded systems can be found [75][77]. At the circuit level, more sophisticated power scaling methods, such as ultra dynamic voltage scaling (UDVS), have been investigated. Previous research has demonstrated sub-threshold operation in various digital applications [78]. The basic idea behind UDVS is to design the application unit that may operate in the super-threshold or sub-threshold region based on the energy harvesting status. In [79], the authors investigated a robust UDVS system by using an adaptive body-biasing technique that dynamically adjusts the -ratio between the pull-up and pull-down transistors. In [80], the authors investigated a voltage scalable and process tolerant arithmetic unit. Adaptive clock stretching is used to achieve aggressive scaling of supply

LU et al.: EFFICIENT DESIGN OF MICRO-SCALE ENERGY HARVESTING SYSTEMS

261

Fig. 12. Block diagram of a battery-powered embedded system. Fig. 13. Block diagram of a micro-energy harvesting system.

voltage, thus, facilitate harvesting-aware applications. In [81], a discrete cosine transform (DCT) architecture that allows aggressive voltage scaling for low-power dissipation was presented. III. COMPARISON WITH BATTERY-POWERED SYSTEMS As we will discuss later, micro-scale energy harvesting systems comprise several building blocks that are commonly adopted in battery-powered embedded systems. Without careful comparison, people tend to think both kinds of systems are similar and overlook their subtle but essential differences. Therefore, in this section, we rst review the basic design issues of battery-powered embedded systems and then make in-depth comparison to distinguish them. A. Battery-Powered Embedded System Fig. 12 shows the block diagram of a battery-powered embedded system. The battery is a storehouse of energy with xed total capacity. The power converter is used to convert the battery voltage to the required output voltage , which is supposed to be regulated around the reference voltage by the control unit. The power demand of the application unit is translated into a power demand on the battery. The control unit generates all kinds of necessary required control signals for efcient power conversion and voltage regulation. Note that there does not involve any energy harvesting process in these systems. B. Design Consideration Comparison 1) Design Merits: For battery-powered systems, for simplicity, we ignore the leakage power loss of battery. Thus, the duration of system operation can be modeled as (8) (9) Here is the output power dissipated by the application unit, is the power conversion efciency of the power converter, and is the energy stored in the battery. As is determined by the application unit and is xed for a given battery, (8) indicates higher power conversion efciency results in more energy available for use by the application unit and longer system operation. Therefore, the design objective for power converters in battery-powered systems is to waste as little power as possible in this voltage conversion process, which is equivalent with a design goal of maximizing the power conver). It is noted that according to sion efciency (i.e.,

(9), the power conversion efciency varies with the loading cur. A given loading current corresponds to a particular rent input current and hence a value of power conversion efciency. In the context of micro-scale energy harvesting systems, energy transducers cannot be viewed as capacity-limited energy sources, because they will never run out of energy as long as the environmental energy source is present. Further, unlike energy storage buffers, energy transducers are not capable of storing converted energy. At any given point of time, a transducer is capable of producing a maximum amount of electrical power, dein Fig. 13. If all of this power cannot be extracted noted as from the transducer immediately, it will be lost forever (unlike in batteries where any unused energy remains in the battery and is available for later use). Therefore, in order to reap the maximum benet from energy harvesting, it is necessary to extract as much power as possible from the energy transducer at any point in time. Consequently, one key design goal for a power converter in micro-scale energy harvesting systems is to max. Only a part of the eximize the extracted input power tracted input power can be transferred to the energy buffer. This in Fig. 13) amount of delivered output power (marked as can be modeled as (10) (11) Here is the ratio of actual input power over the maximum available power converted by the energy transducer, and is the power conversion efciency of the power converter. The entire system energy harvesting efciency is dened as the ratio of divided by the maximum the net harvested output power [7]. In input power extracted from the energy harvester order to obtain as much as usable harvested power for the application unit, the system energy harvesting efciency (i.e., also the product of and ) is a crucial design merit that needs to be maximized. From the above expressions, it can be observed that a power converter that operates at lower power conversion but extracts higher input power from the energy efciency transducer is preferable to a power converter that operates at higher conversion efciency, but is only able to extract little input power from the transducer. From Fig. 13, it is apis virtually isolated from parent that the loading current the power converter by the energy buffer. The system energy harvesting efciency is not signicantly affected by the loading current of application unit. 2) Optimal Number of Stages: According to [62] and [63], a charge pump power converter, when used for battery-powered

262

IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 1, NO. 3, SEPTEMBER 2011

systems, has an optimal number of stages for maximum power conversion efciency as (12) (13) Here is a technology-dependent parameter and is calculated as the ratio of the parasitic capacitance of a stage capacitor to and are the input and the stage capacitance itself. output voltages of the power converter, respectively. Equations (12) and (13) were derived for generic output loads and only capacitive loads, respectively. Since the value of is usually very small, both of them can be approximated as (14) Taking into account the actual hardware implementation, the optimal number of stages should be an integer close to the value computed using (14). However, as shown in [53], this estimated number of stages for charge pump power converters, when used for energy harvesting systems, result in a substantial degradation . It is because in the amount of harvested output power this optimal number of stages corresponds to a peak value of but a very low value of . Therefore, the product of and , which determines the entire energy harvesting efciency, is not optimized. The authors of [53] addressed this problem and proposed a new expression that represents the optimal number of stages (15) Circuit simulation demonstrates this proposed expression results in a substantial increase in the net harvested output power . IV. CIRCUIT LEVEL DESIGN CONSIDERATIONS The circuit-level design considerations are mainly targeted at optimizing the power converter parameters to enhance its output power capability, which depends on the input and output voltages, implementation technology, circuit topology, transistor sizing, and the number of stages it has. By choosing a set of proper parameters, the output power capability can be enhanced. In prior work, a number of attempts have been made to improve the circuit topology [43][50], estimate the optimal transistor sizing [51], and derive the best number of stages [53]. However, up to date there is still lacking of a systematic design methodology to analyze and jointly optimize these parameters. Developing such a design methodology requires in-depth understanding of the interaction and impact of these design parameters. In addition, it is of great importance to theoretically model the impact of these design parameters on the energy harvesting efciency. Since the output voltage of an energy transducer usually changes with the variation of environmental energy sources, according to (15), therefore, the corresponding optimal number of stages is not a constant. As a result, adaptively adjusting the number of stages is an attractive and viable option to maintain

high efciency. Up to date, very few researchers ever discuss how to design a recongurable charge pump with low design overhead. For instance, for passive RFID applications where there is no energy storage buffer, the harvested RF energy is dissipated by the load immediately. The ac/dc rectier is usually designed for the worst case, where minimum amount of power is received by the antenna. In these systems, the number of stages is often optimized and xed for this worst case. However, in practice, the received power by the antenna signicantly varies with the distance between the RFID tag and transponder. Thus, in most cases, the power efciency of an ac/dc rectier is far away from the best value, due to the use of nonoptimal number of stages. Therefore, designing new circuit structures that can dynamically adapt the number of stages with low overhead is a potential research area. V. ARCHITECTURE LEVEL DESIGN CONSIDERATIONS One key architecture-level design consideration is to model the behavior of each building block and to explore their impact on the system energy harvesting performance. As we have discussed before, these building blocks are closely related and interacted. Improving one component does not imply the entire system performance is also enhanced. For example, the net in Fig. 13) depends on the status harvested output power ( of ambient energy source, the design parameters of power converter, the switching frequency generated by the control unit, the energy buffer voltage and so on. A stronger ambient energy source indicates an increase of the maximum possible converted in Fig. 13). However, even if other design parameters power ( or choices do not change, it cannot guarantee that the net harvested output power is larger than before. This could be either due to a worse impedance matching between the energy transducer and power converter, or variation of the optimal number of stages. Therefore, it is imperative to quantify the impact of each building block on system performance. It is also noted that the size of an energy storage buffer is crucial to the overall system performance. In some cases, the self-discharge leakage current is comparable with the harvested charging current. Thus, this non-negligible leakage current may signicantly degrade the energy storage efciency and reduce the amount of energy usable for powering the application unit. As we will demonstrate in Section VII, the optimal size of an energy storage buffer is affected by a number of factors (e.g., the leakage model of energy buffer, the prole of application unit, the energy harvesting behavior). It is often required to perform careful analysis to determine the optimal size of energy buffer for micro-scale energy harvesting systems. Cold-booting involves the issue of self-starting up from zeroenergy energy storage at the architecture level and several existing solutions for cold-boosting have been discussed in [68]. Taking into account of the non-idealities of super-capacitor and battery, it is possible to develop simple models to analyze the relative energy efciency of both kinds of storage devices. Essentially, if both devices start with the same amount of energy in them, the energy stored in the capacitor degrades at a faster rate than the energy stored in the battery. After some threshold amount of time, the available energy in the capacitor degrades to a point where the battery, even with its low round-trip efciency

LU et al.: EFFICIENT DESIGN OF MICRO-SCALE ENERGY HARVESTING SYSTEMS

263

would result in a higher deliverable energy capacity. Therefore, the choice of which energy storage device to use is also dependent on the temporal prole of power consumption by the load. Finally, note that it is possible to integrate both a super-capacitor and a rechargeable battery to form a hybrid energy buffer [69]. VI. SYSTEM LEVEL DESIGN CONSIDERATIONS To ensure efcient micro-scale energy harvesting and utilization, the entire system must be optimized in a holistic way from the design of constituent circuits, selection of energy buffer type and size, to power management implementation at the application level. From system perspective, it is absolutely necessary to develop simulation models of various system components to enable systematic design space exploration. As very low level of abstraction (e.g. SPICE models) tends to be complex and extremely time-consuming for simulation, we envision the necessity and importance of developing a high-level CAD simulation tool to capture the inherent design complexity and deal with numerous involved design parameters and choices. Such a simulation tool is supposed to consider all possible underlying factorsenergy transducer characteristics, power converter design parameters, energy storage buffer type and size, the MPP tracking algorithm and implementation, and current prole of application unit. This simulation tool enables designers to quickly evaluate the impact of various design choices and parameters at early design phase. The challenges at different levels of system hierarchy, from circuit design to simulation methodology, need to be carefully investigated. This simulation tool will also benet system designers who decide to prototype an energy harvesting system by using off-the-shelf products available in the market, instead of designing custom ICs by themselves. However, the truth is product manufacturers usually do not provide enough data in the datasheet. Only a few curves for limited number of cases are given. Thus, one key challenge is to investigate the behaviors and interfaces of these off-the-shelf products for a series of operation conditions, and develop high-level equivalent simulation models that contain essential information for simulation. When these simulation models are successfully generated, they will be imported into the high-level CAD simulation tool for quick system-level evaluation. The establishment of such a simulation tool saves much time and effort to select appropriate building blocks among numerous off-the-shelf products available in the market. The authors of [67] attempted to design a simulator for system-level performance simulation. This simulator models the generic behavior of each block and supports entire system simulation. However, this proposed tool does not involve the design and optimization of specic circuit-level parameters, such as transistor sizing or number of stages, etc. Hence, it does not give much insight to help design optimized components. VII. ILLUSTRATIVE CASE STUDY To better illustrate and understand the proposed design optimization, we have designed and simulated a micro-scale energy harvesting system using IBM 65-nm CMOS technology. HSPICE simulations were carried out using the BSIM models.

Fig. 14. Super-capacitor powered sensing application.

TABLE I SIMULATION RESULTS OF DURATION OF SYSTEM OPERATION

Fig. 15. Micro-scale solar energy harvesting system.

The rst simulation was designed for a super-capacitor powered sensing application, as shown in Fig. 14. A 100 F GreenCap [82] was used as the energy buffer and its leakage current model was experimentally extracted and used in the following simulations. The initial terminal voltage was 2.5 V. Assume the sensor dropped below 2.3 V. The current procannot operate when le of the sensor node was assumed to be of burst style with and negligible sleep current. active current and Simulation was carried out for two duty cycles ( 10%) and the results are recorded in Table I. The second simulation was carried out for solar energy harvesting assisted sensor system, as shown in Fig. 15. A 9 cm PV cell (#3-1.5-100 [19]) was used as the energy transducer. We chose the light irradiance based on the averaged measurement data recorded in Las Vegas, Nevada on November 1, 2010 [83]. Three charge pump power converters with number of stages varying from two to four were implemented. All of them adopted the same circuit topology and transistor size as in [53]. Optimal clock frequencies were generated by the control unit and applied to these power converters for maximizing net . The setups for the super-capacitor and output power senor node were identical to the last simulation. The simulation results are recorded in Table I. From Table I, we can see that when the duty cycle is low , the use of energy harvesting techniques results in an increase of 135%206% in the duration of system operation, compared to results of simulation 1. It is noted that there for simulation 2. When is an optimal number of stages , the benet of energy harthe duty cycle is higher vesting is marginal (the duration of operation extended by 8%). This phenomenon can be explained as follows. The total current drawn from the super-capacitor can be expressed as (Fig. 15) (16)

264

IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 1, NO. 3, SEPTEMBER 2011

TABLE II SIMULATION RESULTS OF DURATION OF SYSTEM OPERATION

to consider design challenges at all levels of design abstractioncircuit, architecture, and system. VIII. CONCLUSION Environmental energy harvesting represents a promising approach to powering a variety of micro-scale low power applications. These applications are feasible to operate autonomously for several months or years, when efcient energy harvesting techniques are applied. This paper gives an overview study of the state-of-the-art of micro-scale energy harvesting systems. Various design challenges, tradeoffs and considerations from circuit, architecture and system perspectives have been introduced and discussed. We explicitly provide a comprehensive comparison between energy harvesting systems and battery-powered embedded systems. Moreover, we envision and address the design considerations for developing a CAD simulation tool, which enables the quick evaluation and design space exploration of energy harvesting systems at the early design phase. We believe that, to ensure micro-scale energy harvesting systems operate efciently, the entire system must be optimized in a holistic way from the design of the constituent circuits to system architecture. REFERENCES
[1] B. A. Warneke and K. S. J. Pister, An ultra-low energy micro-controller for smart dust wireless sensor network, in Proc. ISSCC, 2004, pp. 316317. [2] S. Y. Lee and S. C. Lee, An implantable wireless bidirectional communication microstimulator for neuromuscular stimulation, IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 52, no. 12, pp. 25262538, Dec. 2005. [3] R. F. Yazicioglu et al., Ultra-low-power biopotential interfaces and their applications in wearable and implantable systems, Microelectron. J., vol. 40, no. 9, pp. 13131321, 2009. [4] U.S. Nat. Inst. Health, How Will a Pacemaker Affect My Lifestyle? [Online]. Available: http://www.nhlbi.nih.gov/health/dci/Diseases/pace/pace_lifestyle.html [5] L. Mateu and F. Moll, Review of energy harvesting techniques and applications for microelectronics, in Proc. SPIE Microtechnol. New Millennium, 2005, pp. 359373. [6] V. Raghunathan and P. H. Chou, Design and power management of energy harvesting embedded systems, in Proc. ISLPED, 2006, pp. 369374. [7] C. Lu, C. Y. Tsui, and W. H. Ki, Vibration energy scavenging and management for ultra low power applications, in Proc. ISLPED, 2007, pp. 316321. [8] H. Shao, C. Y. Tsui, and W. H. Ki, A micro power management system and maximum output power control for solar energy harvesting applications, in Proc. ISLPED, 2007, pp. 298303. [9] H. Shao, C. Y. Tsui, and W. H. Ki, The design of a micro power management system for applications using photovoltaic cells with maximum output power control, IEEE Trans. Very Large Scale Integr. (VLSI) Syst., vol. 17, no. 8, pp. 11381142, Aug. 2009. [10] E. Carlson, K. Strunz, and B. Otis, A 20 mV input boost converter with efcient digital control for thermoelectric energy harvesting, IEEE J. Solid-State Circuits, vol. 45, no. 4, pp. 741750, Apr. 2010. [11] X. Jiang, J. Polastre, and D. Culler, Perpetual environmentally powered sensor networks, in Proc. IPSN, 2005, pp. 463468. [12] C. Park and P. H. Chou, AmbiMax: Efcient, autonomous energy harvesting system for multiple-supply wireless sensor nodes, in Proc. SECON, 2006, pp. 168177. [13] F. Simjee and P. H. Chou, Everlast: Long-life, supercapacitor-operated wireless sensor node, in Proc. ISLPED, 2006, pp. 197202. [14] C. Park and P. H. Chou, TurboCap: Batteryless, supercapacitor-based power supply for mini-FDPM, in Proc. ESSCAP, Rome, Italy, 2008. [15] C. Y. Chen and P. H. Chou, DuraCap: A supercapacitor-based, power-bootstrapping, maximum power point tracking energy-harvesting system, in Proc. ISLPED, 2010, pp. 313318.

The leakage current of a super-capacitor varies with is deits size and terminal voltage, the harvested current termined by the PV cell and power converter, and the load curis largely affected by the active current and duty rent cycle. When the duty cycle is very low, the leakage current and harvested current are comparable to the load current. Hence, there is large benet from energy harvesting. When the duty cycle is larger (e.g., 10% in simulation 2), the harvested current is much smaller than the load current and hence, the load current becomes dominant. Thus, energy harvesting has minimal impact on improving the lifetime extension. These simulations illustrate the importance of exploring circuit and architecture level design optimizationsthe number of stages for power converters, leakage dissipation of energy buffers, and power dissipation of application units. The third simulation was to evaluate the impact of energy buffer size. Here the number of stages of the power converter was xed at three. Various super-capacitor sizes (50 F, 100 F, and 150 F) were used for simulations. The resultant simulation results are summarized in Table II. ), It is noted that when the duty cycle is high (i.e., the duration of system operation is almost proportional to the energy buffer size. However, when the duty cycle is low (i.e., ), a super-capacitor with smaller size results in a longer operation time. According to (17), a smaller super-capacitor implies less stored energy, and hence, is easy to drain off (17) However, if we take into account of the non-ideality of a super-capacitor, the system duration can be approximated as

(18) It is noted that the load current and the harvested curare not inuenced by the super-capacitor size. Recall rent is proportional to the super-cathat the leakage current pacitor size. When the duty cycle is high (e.g., 10%), is , therefore, [expression (18)] monotonically larger than is much increases with the super-capacitor size. Further, if and , (18) implies that the is proporlarger than tional to the super-capacitor size, as shown in Table II for . On the other hand, when the duty cycle is low (e.g., 1%), is less than . According to (18), the reduction of super-capacitor size corresponds to an increase of , which was also veried by the simulation results in Table II for . This simulation illustrates that the energy buffer has a crucial impact on determining the duration of system operation. From the above case study, it is clear that to ensure efcient design of micro-scale energy harvesting system, it is important

LU et al.: EFFICIENT DESIGN OF MICRO-SCALE ENERGY HARVESTING SYSTEMS

265

[16] W. J. Choi, Y. Jeon, J. H. Jeong, R. Sood, and S. G. Kim, Energy harvesting MEMS device based on thin lm piezoelectric cantilevers, J. Electroceramics, vol. 17, no. 24, pp. 543548, 2006. [17] R. G. Egbert, M. R. Harvey, and B. P. Otis, Microscale silicon thermoelectric generator with low impedance for energy harvesting, in Proc. 5th Eur. Conf. Thermoelectrics, 2007, pp. 219221. [18] K. L. Chu, S. Gold, V. Subramanian, C. Lu, M. A. Shannon, and R. I. Masel, A nanoporous silicon membrane electrode assembly for on-chip micro fuel cell applications, J. Microelectromech. Syst., pp. 671677, 2006. [19] Solar Made Inc. [Online]. Available: http://solarts.biz/ [20] Micropelt, Inc. [Online]. Available: http://www.micropelt.com/ [21] D. Choi et al., Piezoelectric touch-sensitive exible hybrid energy harvesting nanoarchitectures, Nanotechnology vol. 21, no. 40, 2010 [Online]. Available: http://iopscience.iop.org/0957-4484/21/40/ 405503 [22] T. Suzuki, K. Yoshikawa, and S. Momose, Integration of organic photovoltaic and thermoelectric hybrid module for energy harvesting applications, in Proc. IEDM, 2010, pp. 720723. [23] C. Lu, S. P. Park, V. Raghunathan, and K. Roy, Efcient power conversion for micro scale energy harvesting, in Proc. DATE, 2010, pp. 16021607. [24] E. O. Torres and G. A. Rincon-Mora, Energy harvesting system-in package microsystem, ASCE J. Energy Eng., pp. 121129, Dec. 2008. [25] D. Yu and S. Yuvarajan, A novel circuit model for PEM fuel cells, in Proc. APEC, 2004, pp. 362366. [26] M. Chen and G. A. Rincon-Mora, A compact electrical model for microscale fuel cells capable of predicting runtime and I-V polarization, IEEE Trans. Energy Convers., vol. 23, no. 3, pp. 842850, Sep. 2008. [27] S. Roundy, P. K. Wright, and J. Rabaey, A case study of low level vibrations as a power source for wireless sensor nodes, Comput. Commun., pp. 11311144, 2003. [28] R. Amirtharajah and A. P. Chandrakasan, Self-powered signal processing using vibration-based power generation, IEEE J. Solid-State Circuits, vol. 33, no. 5, pp. 687695, May 1998. [29] S. Meninger, J. O. Mur-Miranda, R. Amirtharajah, A. Chandrakasan, and J. Lang, Vibrationto-electric energy conversion, IEEE Trans. Very Large Scale Integr. (VLSI) Syst., vol. 9, no. 1, pp. 6476, Jan. 2001. [30] G. K. Ottman, H. F. Hofmann, A. C. Bhatt, and G. A. Lesieutre, Adaptive piezoelectric energy harvesting circuit for wireless remote power supply, IEEE Trans. Power Electron., vol. 17, no. 5, pp. 669676, Sep. 2002. [31] G. K. Ottman, H. F. Hofmann, and G. A. Lesieutre, Optimized piezoelectric energy harvesting circuit using step-down converter in discontinuous conduction mode, IEEE Trans. Power Electron., vol. 18, no. 2, pp. 696703, Mar. 2003. [32] J. Wenck, R. Amirtharajah, J. Collier, and J. Siebert, AC power supply circuits for energy harvesting, in Proc. VLSI Symp., 2007, pp. 9293. [33] U. Karthaus and M. Fischer, Fully integrated passive UHF RFID transponder IC with 16.7-W minimum RF input power, IEEE J. Solid-State Circuits, vol. 38, no. 10, pp. 16021608, Oct. 2003. [34] T. T. Le, J. Han, A. V. Jouanne, K. Mayaram, and T. S. Fiez, Piezoelectric micro-power generation interface circuits, IEEE J. Solid-State Circuits, vol. 41, no. 6, pp. 14111420, Jun. 2006. [35] Y. H. Lam, W. H. Ki, and C. Y. Tsui, Integrated low-loss CMOS active rectier for wirelessly power devices, IEEE Trans. Circuits Syst. II, Express Briefs, vol. 53, no. 12, pp. 13781382, Dec. 2006. [36] S. Guo and H. Lee, An efciency-enhanced CMOS rectier with unbalanced-biased comparators for transcutaneous-powered high-current implants, IEEE J. Solid-State Circuits, vol. 44, no. 6, pp. 17961804, Jun. 2009. [37] Y. K. Ramadass and A. P. Chandrakasan, An efcient piezoelectric energy harvesting interface circuit using a bias-ip rectier and shared inductor, in Proc. ISSCC, 2009, pp. 296298. [38] R. E. Barnett, J. Liu, and S. Lazar, A RF to DC voltage conversion model for multi-Stage rectiers in UHF RFID transponders, IEEE J. Solid-State Circuits, vol. 44, no. 2, pp. 354370, Feb. 2009. [39] J. Yi, W. H. Ki, and C. Y. Tsui, Analysis and design strategy of UHF micro-power CMOS rectiers for micro-sensor and RFID applications, IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 54, no. 1, pp. 153166, Jan. 2007. [40] T. Le, K. Mayaram, and T. Fiez, Efcient far-eld radio frequency energy harvesting for passively powered sensor networks, IEEE J. Solid-State Circuits, vol. 43, no. 5, pp. 12871302, May 2008.

[41] D. Ma, L. Su, and M. Somasundaram, Integrated interleaving SC power converters with analog and digital control schemes for energy-efcient microsystems, J. Anal. Integr. Circuits Signal Process., vol. 62, pp. 361372, 2010. [42] J. F. Dickson, On-chip high voltage generation in MNOS integrated circuits using an improved voltage multiplier technique, IEEE J. SolidState Circuits, vol. 11, no. 3, pp. 374378, Jun. 1976. [43] F. Su, W. H. Ki, and C. Y. Tsui, Gate control strategies for high efciency charge pumps, in Proc. ISCAS, 2005, pp. 19071910. [44] F. Pan and T. Samaddar, Charge Pump Circuit Design. New York: McGraw-Hill, 2006. [45] P. Favrat, P. Deval, and M. J. Declercq, A high-efciency CMOS voltage doubler, IEEE J. Solid-State Circuits, vol. 33, no. 3, pp. 410416, Mar. 1998. [46] T. Ying, W. H. Ki, and M. Chan, Area-efcient CMOS charge pumps for LCD drivers, IEEE J. Solid-State Circuits, vol. 38, no. 10, pp. 17211725, Oct. 2003. [47] L. K. Chang and C. H. Hu, An exponential-folds design of a charge pump styled DC/DC converter, in Proc. PESC, 2004, pp. 516520. [48] L. K. Chang and C. H. Hu, High efciency MOS charge pumps based on exponential-gain structure with pumping gain increase circuits, IEEE Trans. Power Electron., vol. 21, no. 3, pp. 826831, May 2006. [49] C. Y. Tsui, H. Shao, W. H. Ki, and F. Su, Ultra-low voltage power management circuit and computation methodology for energy harvesting applications, in Proc. ASPDAC, 2006, pp. 9697. [50] C. Lu, S. P. Park, V. Raghunathan, and K. Roy, Efcient power conversion for ultra low voltage micro scale energy transducers, in Proc. DATE, 2010, pp. 16021607. [51] C. Lu, S. P. Park, V. Raghunathan, and K. Roy, Analysis and design of ultra-low power thermoelectric energy harvesting systems, in Proc. ISLPED, 2010, pp. 183188. [52] I. Doms, P. Merken, R. P. Mertens, and C. V. Hoof, Capacitive powermanagement circuit for micropower thermoelectric generators with a 2.1 W controller, in Proc. ISSCC, 2008, pp. 300302. [53] C. Lu, S. P. Park, V. Raghunathan, and K. Roy, Stage number optimization for switched capacitor power converters in micro-scale energy harvesting, in Proc. DATE, 2011. [54] C. Park and P. H. Chou, Power utility maximization for multiplesupply systems by a load-matching switch, in Proc. ISLPED, 2004, pp. 168173. [55] M. Chen and G. A. Rincon-Mora, Single inductor multiple input multiple output (SIMIMO) power mixer-charger-supply system, in Proc. ISLPED, 2007, pp. 310315. [56] N. M. Sze, F. Su, Y. H. Lam, W. H. Ki, and C. Y. Tsui, Integrated single-inductor dual-input dual-output boost converter for energy harvesting applications, in Proc. ISCAS, 2008, pp. 22182221. [57] H. Shao, C. Y. Tsui, and W. H. Ki, A single inductor dual output DC-DC converter with hybrid supplies for solar energy harvesting applications, in Proc. ISLPED, 2009, pp. 6974. [58] C. Lu, V. Raghunathan, and K. Roy, Maximum power point considerations for micro-scale solar energy harvesting systems, in Proc. ISCAS, 2010, pp. 273276. [59] V. Raghunathan, A. Kansal, J. Hsu, J. Friedman, and M. Srivastava, Design considerations for solar energy harvesting wireless embedded system, in Proc. IPSN, 2005, pp. 457462. [60] D. Brunelli, L. Benini, C. Moser, and L. Thiele, An efcient solar energy harvester for wireless sensor nodes, in Proc. DATE, 2008, pp. 104109. [61] S. S. Clark, J. Gummeson, K. Fu, and D. Ganesan, Towards autonomously-powered CRFIDs, in ACM Workshop Power Aware Comput. Syst., Big Sky, MT, 2009. [62] G. Palumobo and D. Pappalardo, Charge pump circuits: Power consumption optimization, IEEE Trans. Circuits Syst. I, Fundam. Theory Appl., vol. 49, no. 11, pp. 15351542, Nov. 2002. [63] G. Palumbo and D. Pappalardo, Charge pump circuits with only capacitive loads: Optimized design, IEEE Trans. Circuits Syst. II, vol. 53, no. 2, pp. 128132, Feb. 2006. [64] B. Srensen et al., Renewable Energy Focus Handbook. New York: Academic, 2008. [65] C. Renner, J. Jessen, and V. Turau, Lifetime prediction for supercapacitor-powered wireless sensor nodes, in Proc. FGSN, 2009, pp. 5558. [66] Y. Kim, N. Chang, Y. Wang, and M. Pedram, Maximum power transfer tracking for a photovoltaic-supercapacitor energy system, in Proc. ISLPED, 2010, pp. 307312.

266

IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS, VOL. 1, NO. 3, SEPTEMBER 2011

[67] J. Jeong, A practical theory of micro-solar powersensor networks, Ph.D. dissertation, Dept. Comput. Sci., Berkeley Univ., Berkeley, CA, 2009. [68] P. H. Chou and S. Kim, Techniques for maximizing efciency of solar energy harvesting systems, in ICMU Conf., Seattle, WA, 2010. [69] M. Pedram, N. Chang, Y. Kim, and Y. Wang, Hybrid electrical energy storage systems, in Proc. ISLPED, 2010, pp. 363368. [70] J. Hsu, S. Zahedi, A. Kansal, M. Srivastava, and V. Raghunathan, Adaptive duty cycling for energy harvesting systems, in Proc. ISLPED, 2006, pp. 180185. [71] A. Kansal, J. Hsu, M. Srivastava, and V. Raghunathan, Harvesting aware power management for sensor network, in Proc. DAC, 2006, pp. 651656. [72] J. Lu, S. Liu, Q. Wu, and Q. Qiu, Accurate modeling and prediction of energy availability in energy harvesting real-time embedded systems, in Proc. Int. Conf. Green Comput., 2010, pp. 469476. [73] S. Liu, Q. Wu, and Q. Qiu, An adaptive scheduling and voltage/frequency selection algorithm for real-time energy harvesting systems, in Proc. DAC, 2009, pp. 782787. [74] S. Liu, J. Lu, Q. Wu, and Q. Qiu, Load-matching adaptive task scheduling for energy efciency in energy harvesting real-time embedded systems, in Proc. ISLPED, 2010, pp. 325330. [75] C. Vigorito, D. Ganesan, and A. Barto, Adaptive control for dutycycling in energy harvesting-based wireless sensor networks, in Proc. SECON, 2007, pp. 2130. [76] J. P. Recas, C. Bergonzini, D. Atienza, and T. Rosing, Prediction and management in energy harvested wireless sensor nodes, in Proc. 1st Int. Conf. Wireless Vitae, 2009, pp. 210. [77] C. Moser, J. Chen, and L. Thiele, Power management in energy harvesting embedded systems with discrete service levels, in Proc. ISLPED, 2009, pp. 413418. [78] H. Soeleman, K. Roy, and B. C. Paul, Robust subthreshold logic for ultra-low power operation, IEEE Trans. Very Large Scale Integr. (VLSI) Syst., vol. 9, no. 1, pp. 9099, Feb. 2001. [79] M. Hwang and K. Roy, ABRM: Adaptive -ratio modulation for process-tolerant ultradynamic voltage scaling, IEEE Trans. Very Large Scale (VLSI) Syst., vol. 18, no. 2, pp. 281290, Feb. 2010. [80] S. Ghosh, D. Mohapatra, G. Karakonstantis, and K. Roy, Voltage scalable high-speed robust hybrid arithmetic units using adaptive clocking, IEEE Trans. Very Large Scale Integr. (VLSI) Syst., vol. 18, no. 9, pp. 13011309, Sep. 2010. [81] G. Karakonstantis, N. Banerjee, and K. Roy, Process-variation resilient and voltage-scalable DCT architecture for robust low-power computing, IEEE Trans. Very Large Scale Integr. (VLSI) Syst., vol. 18, no. 10, pp. 14611470, Oct. 2010. [82] SAMWHA, Gold Capacitors Technical Guide [Online]. Available: http://www.samwha.com/electric/templatedirs/guest/list_pdf1/DP.pdf [83] Measurement and Instruction Data Center [Online]. Available: http:// www.nrel.gov/midc

Vijay Raghunathan (M08) received the B.Tech. degree in electrical engineering from the Indian Institute of Technology, Madras, India, in 2000, and the M.S. and Ph.D. degrees in electrical engineering from the University of California, Los Angeles, in 2002 and 2006, respectively. He is currently an Assistant Professor in the School of Electrical and Computer Engineering at Purdue University, where he leads the Embedded Systems Lab. Prior to joining Purdue University, he was a visiting researcher at NEC Laboratories America, Princeton, NJ, from August 2005 to August 2006. His research interests include the design of embedded computing systems, system-on-chip architectures, and wireless sensor networks with an emphasis on low power design and reliable system design. He has co-authored a book chapter, several journal and conference papers, and has presented full-day and embedded tutorials on the above topics. He serves on the organizing and technical program committees of several leading ACM and IEEE conferences in the areas of embedded systems, VLSI design, and wireless sensor networks. Dr. Raghunathan is a recipient of an NSF CAREER award, the Edward K. Rice Outstanding Doctoral Student Award from the UCLA School of Engineering and Applied Sciences, and the Outstanding Masters Student Award from the UCLA Electrical Engineering Department. He also received the design contest award at the ACM/IEEE International Symposium on Low Power Electronics and Design in 2005, the best student paper award at the IEEE International Conference on VLSI Design in 2000, and a best paper award nomination at the ACM/IEEE International Symposium on Low Power Electronics and Design in 2006.

Chao Lu (S10) received the B.S. degree in electrical engineering from the Nankai University, Tianjin, China, in 2004, and the M.S. degree from the Department of Electronic and Computer Engineering, the Hong Kong University of Science and Technology, Hong Kong, in 2007. Since 2008, he has been pursuing the Ph.D. degree at Purdue University, West Lafayette, IN. His research interests include design of micro-scale energy harvesting systems, and power management IC design for ultra low power applications. Mr. Lu was the recipient of the Best Paper Award of the International Symposium on Low Power Electronics and Design in 2007.

Kaushik Roy (F11) received B.Tech. degree in electronics and electrical communications engineering from the Indian Institute of Technology, Kharagpur, India, and the Ph.D. degree from the electrical and computer engineering department of the University of Illinois, Urbana-Champaign, in 1990. He was with the Semiconductor Process and Design Center of Texas Instruments, Dallas, where he worked on FPGA architecture development and low-power circuit design. He joined the electrical and computer engineering faculty at Purdue University, West Lafayette, IN, in 1993, where he is currently a Professor and holds the Roscoe H. George Chair of Electrical and Computer Engineering. His research interests include Spintronics, VLSI design/CAD for nano-scale silicon and non-silicon technologies, low-power electronics for portable computing and wireless communications, VLSI testing and verication, and recongurable computing. He has published more than 500 papers in refereed journals and conferences, holds 15 patents, graduated 50 Ph.D. students, and is co-author of two books on low-power CMOS VLSI design (Wiley and McGraw Hill). He is Purdue University Faculty Scholar. He was a Research Visionary Board Member of Motorola Labs (2002) and held the M. K. Gandhi Distinguished Visiting faculty at Indian Institute of Technology (Bombay). Dr. Roy received the National Science Foundation Career Development Award in 1995, IBM faculty partnership award, ATT/Lucent Foundation award, 2005 SRC Technical Excellence Award, SRC Inventors Award, Purdue College of Engineering Research Excellence Award, Humboldt Research Award in 2010, and best paper awards at 1997 International Test Conference, IEEE 2000 International Symposium on Quality of IC Design, 2003 IEEE Latin American Test Workshop, 2003 IEEE Nano, 2004 IEEE International Conference on Computer Design, 2006 IEEE/ACM International Symposium on Low Power Electronics and Design, and 2005 IEEE Circuits and system society Outstanding Young Author Award (Chris Kim), 2006 IEEE TRANSACTIONS ON VERY LARGE SCALE INTEGRATION (VLSI) SYSTEMS best paper award. He has been in the editorial board of IEEE Design and Test, IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS, and IEEE TRANSACTIONS ON VERY LARGE SCALE INTEGRATION (VLSI) SYSTEMS. He was Guest Editor for Special Issue on Low-Power VLSI in the IEEE Design and Test (1994) and IEEE TRANSACTIONS ON VERY LARGE SCALE INTEGRATION (VLSI) SYSTEMS (June 2000), IEEE Proceedings Computers and Digital Techniques (July 2002).