An Autonomous Wireless Sensor Platform based on Microgenerators 1 D. Maurath1, T. Hehn1, D. Spreemann2, Y. Manoli1,2 Chair of Microelectronics, Department of Microsystems Engineering (IMTEK), University Freiburg, Germany 2 Institute for Micromachining and Information Technology, HSG-IMIT, Villingen-Schwenningen, Germany Abstract: This paper presents a novel wireless autonomous system platform for sensor applications, which is only based on energy scavenging power supply. This autonomous system platform needs neither any external electrical energy supply nor any added battery or external pre-charging. The functionality comprises high efficient voltage conversion, power management, sensor data acquisition and radio communication in order to be able to participate in a network of distributed autonomous sensor nodes. It is presented how the system was designed and implemented. Key Words: micro-generator, power management, autonomous system, wireless, sensor platform 1. INTRODUCTION By now, a vast diversity of smart autonomous sensor systems (ASS) are developed and applied in various scenarios and applications. Despite all the desirable advantages, which those systems possess, the long-term and permanent power supply relies mainly on batteries only. An eligible and helpful improvement of such ASS is energy harvesting with micro-generators. However, it is quite challenging to design an ASS, which does not have any auxiliary power source, but only a microgenerator. Most systems either focus especially on adjustment of the converter duty cycle, in order to achieve maximum conversion efficiency, or on optimized application and radio implementation [1-3]. This novel system presented here (Fig. 1) performs comparable features with less energy consumption (Table 1) but only supplied by a micro generator. The generator for the presented system has a tuneable frequency characteristic for excitation frequencies in the range of 30 Hz – 60 Hz [4]. Within this frequency range the output power is around 230 µW - 825 µW. However, the generator’s open-circuit output voltage is not higher than 2.1 V̂ and the internal generator resistance is around 1-3 k: . Thus, the generator voltage varies dramatically when a load current is applied. Despite this drawback, a reliable power system is realized with conversion efficiency from 65 % - 82 %. The paper is sectioned in two main parts. Firstly, it shows efficient generator interfacing, voltage conversion and power management circuitry. Additionally, basic design rules are introduced, which help to optimize the power system. The second part concerns the application electronic. rectifier RCB generator antenna MSP430 C boost converter CC2420 voltage supervisor Figure 1: This picture illustrates electronics for voltage conversion, system control and radio communication. Subsequently, measurement results are presented and discussed. Finally, a short outlook towards further intended future research topics is added. 2. POWER SUPPLY CIRCUITRY The power supply circuitry is composed of four sections. As shown on Fig. 2, between the rectifier and the voltage converter a reverse current barrier and a pre-charge stage is included. The main energy storage buffer Cb is located in-between these two internal stages. At active periods, the boost stage is enabled and supplies the application with VDC. During sleep periods, the application is supplied directly by the buffer Cb, thereby the current is bypassed via supervisor device (TPS3619), which is not shown in Fig.3. 281 rectifier D1 reverse current barrier SP1 D2 D3 Vgen MN1 MN2 VR + CRCB pre-charge stage SP2 U1 VZ Cb + D4 LBoost VL Vin EN U2 CFF RFB1 VFB TPS61120 CC1 VDC Vout RFB3 RFB2 Cout Px.y SN1 Figure 1: Schematic view of the rectifier, pre-charge and boost stage. The rectifier will cause only one forward diode voltage drop and the pre-charge stage enhances start-up reliability. 2.1 Efficient Passive Rectification Due to the relatively low generator output voltages the diode forward-bias voltage losses of common diode-only rectifiers are not acceptable when micro-generator based supplies are used. Therefore, a combination of Schottky diodes and NMOS transistors is applied (D1, D2, MN1, MN2). On the on hand, this gives the opportunity of only one diode forward-bias voltage loss. On the other hand, the reverse current is much higher compared to diode-only rectifiers. This is caused by the transistors, which do not completely close and by the use of Schottky diodes, which tend intrinsically to higher reverse currents compared to pn-diodes. 2.2 Reverse Current Barrier These adverse reverse currents are nearly completely blocked by the implementation of the reverse current barrier (RCB). In order to achieve a net power loss reduction, not only a very low power comparator U1 is used, but also a high resistive comparator input network. Thereby, diode D3 charges the small capacitor CRCB, which gains a voltage higher than the resistor divider node, shortly after the generator starts working. As soon as the generator stops or has only an output voltage lower then the buffer Cb, the voltage at CRCB drops, whereby the comparator turns off switch SP1 [5]. 2.3 Voltage Conversion, Start-up and Reliability Inductive switch-mode converters exhibit high efficiency but less reliability at start-up. In order to ensure a reliable usage of the inductive converter even though critical start-up performance, a precharge stage is used. In addition to this, stable loadtransition regulation has to be performed as well. Therefore, a feed-forward capacitor CFF is added. 282 Due to correspondence of power efficiency and application supply voltage, the converter output voltage is re-programmed within the active time. Thereby, during data acquisition the output voltage is around Vout = 3.3 V and during radio transmission the voltage is decreased to Vout = 2.2 V. This is realized by a switchable third feedback resistor, as shown in Fig. 2. 2.4 Pre-Charging Circuitry The inclusion of the pre-charging stage (PCS) has two main intentions. Firstly, the boost stage needs especially at start-up relatively large currents, which exceeds the ability of the generator. Secondly, due to large internal generator resistance, an increased current load lowers Vgen. This may result in start-up failure of the boost stage, as well as reduced rectification efficiency is more probable. Thus, a buffer has to be sufficiently charged before the boost stage starts conversion. As the buffer voltage initially rises from the uncharged state, voltages VZ and VR rise as well. However, once the Zener voltage is crossed, diode D4 keeps the voltage VZ constant, while VR rises further. So, the comparator switches and turns on SP2. 2.5 Design Methodology In order to implement the system concept according to a given generator and an intended application, the buffer capacitance, the trigger levels of the PCS and the cycle times have to be dimensioned reasonably. The minimum size for the buffer is calculated by the equation Cb t Qactive 'Vb Qactive l V gen VµC ,min (1) where Qactive is the average application load current times the active time Tactive, and VµC,min is the minimum required operation voltage. This voltage is equal to the turn-off voltage of the PCS. Whereas, the PCS turn-on voltage is set to Von t VµC ,min Qstart Cb (2) where Qstart is the average load current times the initialization time. A further issue is the fact, that the converter needs a certain input-power during startup time. In order to achieve a large current conduction angle and rather better impedance matching a small buffer capacitance is preferable. However, if the generator size and thus, its output energy, is reduced, also the PCS turn-on level as well as the buffer capacitance has to be increased. This is because Cb has to provide the energy difference between the generator output and the converter input energy. Otherwise, the buffer voltage will drop below the turn-off voltage more likely in case of smaller generators. Additionally, for high efficient conversion an inductance of only 15 µH is applied even though small load currents. By reducing inductance one has to take care about the ccm/ dcm mode crossing as the inductor current drops below I L, rms d 2'I L , where 'I L is the inductor ripple voltage. However, a smaller inductor causes less resistive losses. Hence, due to a reduced inductor resistance a notably increased efficiency is achieved, as Fig. 3 illustrates. 3. POWER AND SYSTEM MANAGEMENT The efficient and reliable operation of measurements and communication can be notably improved by reasonable system management and parameter selection. Therefore, the management concepts are explained in a quite general manner, whereas the realized design values are mentioned in Tab. 1. 3.1 Dynamic Processing Adaptation In practical usage generator excitation may vary and so the average available power is non-constant. This requires the ability of dynamic processing adaptation. Therefore, certain system parameters, like sleep time Tsleep, communication activity as well as sensor resolution are dynamically adapted with respect to the available power. 3.2 Sensor Data Acquisition and Computing Since the efficiency of boost converters are load current dependent, during the active time Tactive, preferably all task are executed simultaneously. Consequently, the sensors are triggered simultaneously. The controller is programmed in an event-driven style. Thus, good modularity and efficient operation is possible. All events are triggered by an interrupt, which allows setting the controller in low power mode after an accomplished operation. The next event will re-wake the controller by an interrupt. Moreover, modularity is very helpful in combination with the usage of different sensors or RF modules, which are connected with SPI or I²C interfaces. So, different sensors can be plugged easily to the platform. 3.3 Radio Interface and Communication Radio communication is realized by a CC2420 RF transceiver chip. This chip has very short startup times and offers efficient operation with little protocol overhead. For low power consumption requirements packet payload length and power amplifier gain is adapted with respect to current operating conditions. Hence, the transceiver is set in power down mode if no radio communication is required. Thus, data reception is only possible at short timeslots after a data packet was transmitted. 4. SYSTEM SET-UP AND RESULTS The specific parameters of the actually built system are shown in Tables 1 and 2. Especially the duty cycle D and the averaged consumption results in Tab. 2 and Fig. 3 are maximum values, measured with an ideally excited generator. Thereby, the transceiver is only used after the power supply system has steady-state condition. For less generator power, the duty cycle is decreased, the transceiver power amplifier gain is reduced and the sensor performance is minimized by reducing measurement time and current. As described above, the conversion efficiency not only depends on the load current, but also depends strongly on the inductance value and inductor resistance. Nevertheless, the presented efficiencies are strongly related to Vgen Ri Cb Lboost 4.1 Vpp 2.98 kȍ 470 µF 15 µH Tactive Tsleep D rms 6I sleep 190 ms 810 ms < 0.2 < 5.5 µA CFF 20 nF rms 6I active 1.6 mA Von 1.9 V 6Prms 497 µW Table 1: Power supply parameters of the built energy harvesting based sensor platform. 283 state PV [µW] Transceiver sleep 0.04 CC2440 active 16800 µC sleep 0.2 MSP430 active 1980 Sensors + sleep 4,1 Peripheries active 2700 EV,total time [ms] 995.5 4.5 876 124 830 170 EV [µJ] 0.0398 75.6 0.18 245.5 3.32 175 499.6 Table 2: Average power consumption for one measurement cycle and transmission of 1000 bits every one second is about 500 µW. efficiency 67.1k, 15H 0.6 0.4 3k, 2.2mH 0.2 67.1k, 2.2mH 0.0 1.8 2.2 2.6 3.0 Vbuffer 3.4 [V] Figure 3: Measured conversion efficiency over the entire input voltage range as well as with different resistive loads and inductances. appropriate usage of the converter. That means the converter is only used in active periods and never during sleep periods. The mentioned resistor values correspond to representative currents, at active period only. Fig.4 illustrates the general transient power consumption profile. The transmission power and sensor power, PTx and Psens respectively, are highly related to the used parameter settings. 5. CONCLUSION Altogether, it is shown that the presented system works with reasonable parameter settings. Hence, efficient and reliable operation of a microgeneratoronly driven autonomous system can be obtained with reasonable effort. The possibility to drive the system with smaller generators by adapting operating parameters and circuit dimensions is also given. The most valuable results of the work are the 284 Tx PTx sensors Psens Pcalc P sleep µC sleep t sens t sleep t calc tTx sleep t t cycle Figure 4: Averaged power profile of the wireless sensor during one measurement/ transmit cycle. portability to other designs and the developed and proved building blocks for the power management and conversion circuitry. Also additional implementation techniques and flexibility is shown, with regards to the sensor attachment and high performance data acquisition possibilities as well as for the controller programming and self-controlled operation. 3k, 15H 0.8 P ACKNOWLEDGEMENT This work is supported by the German Research Foundation (Deutsche Forschungsgesellschaft – DFG) under Grant Number GR 1322. REFERENCES [1] A. Chandrakasan, R. Min, M. Bhardwaj, S. Cho, A. Wang, Power aware wireless microsensor systems, European Solid State Conference 2002, 24-26 Sept. 2002, pp 47 - 54 [2] H. Dubois-Ferrière, L. Fabre, R. Meier, P. Metrailler, TinyNode: a comprehensive platform for wireless sensor network applications, 5th international conference on information processing in sensor networks, 2006, pp. 358 – 365 [3] J. Polastre, R. Szewczyk, D. Culler Article, Telos: enabling ultra-low power wireless research, IPSN 2005, No. 48, pp. 364-369 [4] D. Spreemann, B. Folkmer, D. Maurath and Y. Manoli, Tunable transducer for low frequency vibrational energy harvesting, 20. EuroSensors 2006, 17-20. Sep. [5] T. Hehn, Energy Harvesting Based Sensor System, diploma thesis, University of Freiburg, IMTEK, 2006