Simulation of adaptive power management circuit for hybrid energy harvester and real-time sensing application

Received Nov 18, 2019 Revised Dec 22, 2019 Accepted Jan 28, 2020 Many wireless sensor network (WSN) applications, nowadays, require realtime communication, which demands cautious design consideration to resolve inherent conflicts between energy efficiency and the need to meet Quality of Services (QoS), such as end-to-end delay communications. Numerous innovative solutions are proposed such as Real-time Power-Aware Routing (RPAR) protocol, which dynamically adapts transmission power to meet specified communication delays at low energy cost. Hence, to enable real-time communication with RPAR protocol, an adaptive Power Management Circuit (PMC) using hybrid energy harvester to support WSN real-time communication is proposed. In this paper, a high-level architecture of the proposed PMC is discussed, which consists of Thermal Energy Generator (TEG), and Piezoelectric Energy Harvester (PEG) as energy providers, with low-power Maximum Power Point Tracking (MPPT) feature enabled. Preliminary simulations which analyze and characterize TEG and PEG system are conducted separately to determine the optimal design parameters to support the conventional WSN QoS requirement. Next, both systems will be integrated into a single PMC implementation prior to fabrication and lab characterization.


INTRODUCTION
With the recent advancement in wireless sensor applications, such as Internet of Things (IoT), ambience intelligence, and wireless sensor networks (WSN), it is expected to have 50 billion of sensor nodes by the year 2025 [1,2]. These large number of sensor nodes are normally required to operate for a long period of time, ranging from months to years. In addition to that, many WSN applications require real-time communication. For example, a surveillance monitoring system needs to alert authorities instantaneously (or within few seconds) of suspicious intruders. However, supporting real-time communications in WSNs is very challenging due to inherent conflict between stringent requirement of end-to-end delay and energy efficiency [3,4]. All these challenges require a very careful design consideration to manage energy resources.
Batteries are commonly used to power up WSN sensors. Although they are designed to be lowpower, eventually their batteries will be drained out, and are required to be replaced, hence incurring high maintenance cost. Therefore, energy harvesting devices have now become a preferred solution to provide cheap, clean, and unlimited alternative energy. Conventionally, energy harvesting system is mainly built upon three major components. First component is energy generator, which is used to harvest energy from Maximum Power Point Tracking, when discussing efficiency for energy harvesting devices, one should always consider the followings: 1) conversion efficiency; 2) transfer efficiency; 3) buffering efficiency; and 4) consumption efficiency. In generic term, transfer efficiency considers amount of loss due to power transfer activity from energy source to load. Hence, there is a technique, called Maximum Power Point Tracking (MPPT), which is typically used to minimize power mismatch between energy source and load [13]. Currently, there are several MPPT techniques developed. One of them is leveraging digital signal processing [12]. Another MPPT technique is based on impedance matching [14,15]. This approach is not only simple to implement, but also consume very low power. The basic idea is to match between input impedance, which varies from one energy generator device to another, and inductor value of boost converter. However, due to huge tolerance of inductor itself, this technique is not be very accurate. Another MPPT technique used is called fractional open-circuit voltage (FOC) [16][17][18]. The open-circuit voltage (Voc) of a specific energy generator is sampled periodically to determine its maximum power point voltage (VMPP), leaving this energy generator electrically open. In this paper, we will focus on FOC technique as it is proven to achieve ~99.5% of efficiency.
Common PMC Architectures for HEH System, There are few prior works done to design PMC for HEH system. The simplest method is called "complementary," as defined by [19], which mainly collects energy from a primary source, while secondary transducers is merely used to power up auxiliary circuitry, such as biasing and boot-up circuitry. As shown in Fig. 5, the Piezoelectric Generator is used during cold start-up, in which at that time, the power generated by the Thermal Generator is not sufficient to sustain the whole operation. However, this simple method does not implement maximum power point tracking (MPPT), which is crucial to ensure maximum harvesting efficiency. Besides, another common technique to harvest multiple source of energy is called "Power ORing" architecture [20][21][22], which offers modular approach, connecting multiple sources in parallel through diodes. Each of energy sources is independent from each other, hence could perform its own MPPT technique. The diodes is used to ensure self-synchronized operation. However, there are several disadvantages linked to this method. First, there is additional power loss due to forward voltage drop on the diodes, as well as having independent MPPT could increase cost and size of the system. The "Power ORing," architecture.
Last but not least, another more sophisticated technique commonly used in managing multisource energy harvesting is called voltage level detection method. For example, in [23] the charging of a micro battery is taken either from the voltage generated by a thermal or RF harvesting subsystem, depending on which exhibits higher voltage value. However, this cannot be applied for loads that require constant voltage supply [24]. In [25], an improved technique is used: each input is connected to the output at pre-defined period of time, provided its respective voltage surpasses the determined voltage threshold.

RESEARCH METHODOLOGY 2.1. Specification definition
Since each and every WSN applications has its own requirement, determining specific Quality of Services (QoS) is a crucial step. Table 1 shows a QoS for specific Structural Montoring Application.

Architecture Definition
In this paper, we propose an adaptive PMC, which is capable to integrate inputs from two energy harvesters, which are TEG & PEG. This proposed PMC integrates some of the elementary techniques from the previous works into one single architecture. For example, the proposed PMC applies basic "complementary" technique to meet specified workload requirement. For structural monitoring application, which stays in sleep mode for most of the time, TEG is used to provide energy during sleep mode, while PEG is during active mode. Besides, the proposed PMC also implements MPPT technique for TEG sub-system. Unlike "Power ORing," the TEG and PEG is not independent from each other -they are controlled by Pulse Generator, which already have pre-determined timing information between sleep and active mode. This Pulse Generator is used to control the "ON" and "OFF" state for each sub-system. The high-level architecture of proposed PMC design is shown in Figure 1. The proposed PMC architecture consists of several main blocks, and components. For example, Pulse Generator is used to control switching between TEG and PEG supply. The pulse is generated based on the loading conditions, i.e. sleep mode = TEG is activated; active mode = PEG is activated. Enable Generator block, which consists of 2-stage comparator and latch circuitry is critical to ensure TEG system operates within tolerated supply, hence maximizing power transfer from TEG harvester to the load. PWM Generator, and the Error Amplifier are used as a control loop to ensure PEG system is regulated into desired voltage. Full-wave rectifier, together with ripple-cancellation capacitor, C1, are the major component of the AC-to-DC voltage converter. Comparator and Phase Delay blocks are added to ensure seamless transition between TEG and PEG subsystem   The main challenges to design PMC for HEH system is to integrate sub-systems seamlessly. Therefore, to ensure smooth transition, the PMC required definite overlapped time between two-subsystems, as shown in the

RESULTS AND ANALYSIS
A feasibility study of the proposed PMC architecture is carried out using HSPICE software 1) To characterize design parameters for PEG system in order to meet QoS during active mode; 2) To analyze feasibility of TEG to meet efficiency target.

Characterization of PEG
PEG is basically a generator that converts vibration generated from steps into electrical energy. In this [26]. PEG system typically consists of two parts: AC-to-DC converter, and DC-to-DC boost converter. For AC-to-DC converter, we are evaluating for optimal capacitor to minimize voltage ripple. Meanwhile, for boost converter, we are tuning duty cycle, switching frequency, and inductor to find the optimal solutions. For this simulation, we assumed 5-layer piezoelectric thick films as our PEG device, which has 5.3x higher output current vs. single layer PEG device, 34x smaller matching impedance [27].
First, we characterized the rectifying circuit. Figure 5 shows the converted DC voltage, together with its corresponding voltage ripple for varying capacitance value. Based on this results, minimum capacitance of 10uF is required to meet sub-mV voltage ripple. We, then characterized DC-DC boost converter by sweeping crucial design parameters. Figure 6 to Figure 7 show sensitivitities of different design parameters such duty cycle, switching frequency, and inductor values. Based on the preliminary simulation results, a recommendation for optimal design of the PEG system to meet specified QoS is shown in Table 2.

Characterization of TEG
Unlike the PEG System, TEG efficiency is very crucial to ensure we can meet >80% efficiency requirement. Based on Yoon, et al. [7], due to the linear I-V characteristics of TEG, Maximum Power Point (MPP), for each TEG device can easily be determined by (1) and (2).
Hence, by knowing the RT & Voc of any TEG device, we can determine its corresponding Vmpp. Figure 8 shows MPP Voltage (Vmpp) for TEG device with Voc = 6 V and RT = 300 KOhm.  Figure 9 shows that maximum power can be harvested from this speficic TEG is ~3.5 V, meaning that if we can regulate our TEG system to ~3.5 V, our power efficiency will be almost 100%. However, achieveing 100% efficiency is nevertheless unrealistic. Hence, for this study, we are only targeting efficiency of >80%, which is typical for buck-boost platform regulator. Therefore, we can allow Vsto1 to be varied slightly from its optimal Vmpp. Other than that, for this architecture to work decently, we need to guarantee PEG can sustain sleep current for long period of time. This is actually related to capacitor selection of Csto1. The bigger the capacitor, the longer it can sustain. However, this could impact the charging time of capacitor as well. Figure  10 shows linear correlation between sustain time vs. capacitance value, assuming Vsto1 varies between 4.25 V and 2 V (i.e. >80% efficiency) for 175 uA loading current.

CONCLUSION
In summary, based on preliminary simulation results, we can see line of sight to meet conventional WSN QoS requirement for structural monitoring application. Optimal design parameters for PEG has also been identified to support output current of 3.5 mA during active state. Besides, TEG is projected to sustain output current of 175 uA, which is required during sleep state, for >600 ms, with efficiency of >80% using 47 uF capacitor. In addition to that, relationship of output voltage of TEG to power efficiency has also been characterized in this paper. Next step is to integrate both systems and to implement our power-state transition policy, which is key component of our proposed PMC architecture.