178 IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 45, NO. 1, JANUARY /$ IEEE

Transcription

1 178 IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 45, NO. 1, JANUARY 2010 A 5.2 mw Self-Configured Wearable Body Sensor Network Controller and a 12 W Wirelessly Powered Sensor for a Continuous Health Monitoring System Jerald Yoo, Student Member, IEEE, Long Yan, Student Member, IEEE, Seulki Lee, Student Member, IEEE, Yongsang Kim, Student Member, IEEE, and Hoi-Jun Yoo, Fellow, IEEE Abstract A self-configured body sensor network controller and a high efficiency wirelessly powered sensor are presented for a wearable, continuous health monitoring system. The sensor chip harvests its power from the surrounding health monitoring band using an Adaptive Threshold Rectifier (ATR) with 54.9% efficiency, and it consumes 12 W to implement an electrocardiogram (ECG) analog front-end and an ADC. The ATR is implemented with a standard CMOS process for low cost. The adhesive bandage type sensor patch is composed of the sensor chip, a Planar-Fashionable Circuit Board (P-FCB) inductor, and a pair of dry P-FCB electrodes. The dry P-FCB electrodes enable long term monitoring without skin irritation. The network controller automatically locates the sensor position, configures the sensor type (self-configuration), wirelessly provides power to the configured sensors, and transacts data with only the selected sensors while dissipating 5.2 mw at a single 1.8 V supply. Both the sensor and the health monitoring band are implemented using P-FCB for enhanced wearability and for lower production cost. The sensor chip and the network controller chip occupy 4.8 mm 2 and 15.0 mm 2, respectively, including pads, in standard 0.18 m 1P6M CMOS technology. Index Terms Body sensor network, continuous health monitoring, dry electrode, electrocardiogram (ECG), planar-fashionable circuit board (P-FCB), rectifier, self-configuration, wearable network, wireless power transmission. I. INTRODUCTION WITH the aging of societies around the world, chronic diseases are becoming the major cause of death. An important perspective on the chronic diseases is that the sooner they are detected, the better quality of life the patients will have. Since the diseases have asymptomatic or intermittent properties, long-term continuous health monitoring is essential in detecting and treating the diseases [1], [2]. Moreover, continuous health monitoring during normal life will satisfy the National Institute of Health (NIH) s 4Ps of future medicine: Predictive, Personalized, Preemptive, and Participatory [3], [4]. There are several methods to continuously monitor health in everyday life. The Holter Monitor system records electrocardiogram (ECG) on 24-hour basis, and, currently, this is the Manuscript received April 27, 2009; revised July 10, Current version published December 23, This paper was approved by Guest Editor Kevin Zhang. The authors are with the Department of Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon , Korea ( jerald@kaist.ac.kr). Digital Object Identifier /JSSC most powerful method to detect irregular heart activities during everyday life. However, typical Holter Monitors have 6 to 10 wires that connect the electrodes to the system, which annoys patients. Moreover, for chronic use, conventional Ag/AgCl wet electrodes may irritate the skin, and the signal quality degrades as the electrolyte gel dehydrates [5]. A novel wearable healthcare system based on knitted integrated sensors [6] is able to capture physiological signals, such as ECG, respiration, and activity. Nevertheless, sophisticated knitting and interconnection of conductive yarn result in relatively high production and maintenance costs. Another useful approach is the digital plaster for Body Area Networks (BAN) [7]. This smart approach meets the convenience requirement of continuous health monitoring: the patient can monitor the ECG data and throw the plaster-type sensor away afterwards. However, requirements for security and resilience to interferers are stringent for health monitoring, and therefore, using open-air wireless industrial, scientific and medical (ISM) bands is not suitable for a health monitoring body sensor network (BSN) [8]. Furthermore, and most importantly, some people are reluctant to patch batteries directly on their skin all day because of potential hazard from explosion, chemical leakage, chemical response with sweat, and heating. To meet the convenience and reliability requirements of continuous health monitoring at once, this paper presents a self-configured wearable health monitoring system based on efficient wirelessly-powered adhesive bandage patch sensors that continuously monitor the ECG and other vital signals at selected locations on the body with low power consumption [9]. The disposable adhesive bandage patch sensor removes the need for a battery, as it is wirelessly powered by a surrounding health monitoring chest band. The sensor IC consumes 12 W, and it employs an Adaptive Threshold Rectifier (ATR) with 54.9% efficiency. The nested chopper readout circuit shows input referred noise performance of To ensure continuous monitoring without skin irritation, dry fabric electrodes are used. The self-configuration enables the network controller to automatically detect, provide power, and record vital signals from only the selected sensors around the body. The paper is organized as follows. Section II describes the concept and the system architecture of the proposed continuous health monitoring system. In Section III, the Planar-Fashionable Circuit Board (P-FCB) process and the dry P-FCB electrode characteristics are discussed in detail. Section IV shows the design of an adhesive bandage type sensor IC, and Section V /$ IEEE

2 YOO et al.: A 5.2 mw SELF-CONFIGURED WEARABLE BODY SENSOR NETWORK CONTROLLER 179 Fig. 1. Proposed continuous health monitoring system with a wearable body sensor network (BSN). presents the network controller for self-configuration. After the implementation results in Section VI, Section VII concludes the paper. II. SYSTEM ARCHITECTURE Fig. 1 shows the proposed continuous health monitoring system with the wearable BSN. It is composed of two parts: the adhesive bandage type P-FCB sensor and the health monitoring chest band. The health monitoring chest band, with a 12 4 inductor array and the network controller, is worn over the chest, where the sensors are attached at arbitrary locations. The network controller on the chest band automatically finds the locations and types of sensors (self-configuration) and provides power only to selected sensors; it can continuously operate and collect data from sensors for up to eight days without replacing the battery (1/2 AA type, 1000 mah capacity). For convenience and safety, the battery is not integrated with the sensor; instead, power is wirelessly provided by the surrounding health monitoring chest band. The basic concept is to take the power overhead from the sensors, moving it to the relatively power-sufficient health monitoring chest band. To ensure continuous monitoring without irritating the skin, dry P-FCB fabric electrodes are used at the sensors. Motion artifacts are minimized by an adhesive bandage patch that tightly sticks to the skin. The sensor patch is disposable, so it is convenient to use. Fig. 2 illustrates the components of an adhesive bandage patch sensor. A 4-turn, octagonal P-FCB inductor ( and ) is screen printed on a cloth with electrodes. Then, the sensor IC is wire bonded on to a cloth, and the molding is applied. When the molding hardens, the cloth is folded in half and attached to an adhesive bandage base. Details on P-FCB technology will be described in Section III, and the architecture of the sensor IC will be discussed in Section IV. The sensor chip adopts a CMOS-only Adaptive-Threshold Rectifier (ATR) with the dual-mode power transmission scheme to harvest power from electromagnetic waves at MHz and 400 MHz. It has a sensor readout circuit capable of handing various types of vital signals. The specific version of the chip Fig. 2. Adhesive bandage sensor. described in this paper is optimized for ECG signals. The controller chip activates only the selected inductors and transacts data with the selected sensors automatically for health monitoring, so careful alignment between the reader and sensor is not necessary. The health monitoring chest band is composed of an array of 12 4 P-FCB inductors and a network controller chip. P-FCB allows the chest band to closely adhere to the body, so a subject will feel comfortable during everyday monitoring. The pitch between the inductors is determined to maximize the coverage. With the zigzag inductor array configuration shown in Fig. 1, all of the chest area around the body is covered. Power is transmitted to the patch sensors through these inductors, and the monitored health data is captured by the inductors at the same time. The detailed architecture of the network controller chip will be described in Section V. III. DRY ELECTRODES BY PLANAR-FASHIONABLE CIRCUIT BOARD Planar-Fashionable Circuit Board (P-FCB) is a planar printed circuit technology on fabric. Silver paste is directly screen printed on fabric to form patterned electrodes, passive elements, or circuit boards [10], [11]. Fig. 3(a) shows the scanning electron microscope (SEM) photograph of the P-FCB cross section (1500 magnified), and Fig. 3(b) shows the surface of the P-FCB (100 magnified). Silver paste is applied with a thickness of around m, with resolution down to 200 m. Chips can be directly wire bonded and molded on fabric as well. Therefore, patch sensors can be easily made at low cost, while obtaining good wearability and flexibility. Ag/AgCl electrodes with electrolyte gel (Fig. 4(a)) are commonly used as biopotential electrodes in ambulatory vital signal monitoring, such as ECG, electroencephalogram (EEG), electrooculogram (EOG), and electromyogram (EMG) [12]. The electrolyte gel forms a conductive path between the skin and an electrode, ensuring a good transmission of the electric signal

3 180 IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 45, NO. 1, JANUARY 2010 Fig. 3. (a) Cross section and (b) surface of Planar-Fashionable Circuit Board (P-FCB). Fig. 5. Measured skin-electrode impedance. Fig. 4. Biopotential electrodes: (a) Ag/AgCl (wet), (b) PCB (dry), and (c) P-FCB (dry). to the front-end electronics. However, for chronic use, the wet type Ag/AgCl electrode suffers from signal quality degradation as the electrolyte gel dehydrates. Moreover, there are toxicological concerns regarding electrolyte gel, such as skin irritation [5], [13]. As an alternative to wet type electrodes, various dry electrodes were proposed for long term monitoring [14] [16]. The metal electrode formed on PCB [14] [Fig. 4(b)] is free from skin irritation, but since PCB is stiff, its wearability is low. The Biopotential Fiber Sensor (BFS) successfully solves stiffness by adopting metal-coated fabric strands as an electrode [15]. However, because of the complicated thin-film process (electrochemical deposition with the rapid dipping technique), the production cost is high. Another smart approach is the nonwoven fabric active electrode [16], but it is connected by cumbersome wires, and, hence, is inconvenient to use. The P-FCB electrode solves the problems mentioned above: since the electrode is dry, it is safe and free from skin irritation. Fig. 4(c) shows the proposed P-FCB electrodes. Due to the flexibility of the fabric, it follows well the contours of the skin, and the effective contact area increases. As a result, the signal quality is better than that of [14]. In addition, the thick-film printing does not involve any complicated process, so the production cost can be lowered when compared to [15]. There is also a drawback in using dry electrodes. The skin-electrode contact impedance is an important factor in electrodes, and if this impedance is too large, the loading due to read-out electronics degrades the signal strength, and more noise will be induced. Unfortunately, dry electrodes have higher contact impedance than wet type electrodes. Fig. 5 shows the measured skin-electrode contact impedance of P-FCB and wet electrodes. Most of the vital signals lie between Hz, where the impedance of wet type electrodes is around k. On the other hand, dry P-FCB electrode impedance ranges between 60 k 800 k, where the smaller the electrode size, the bigger the impedance. Therefore, the Fig. 6. Sensor IC architecture. proposed system is designed to deal with high skin-electrode impedance, and this will be discussed in the following section. IV. ADHESIVE BANDAGE TYPE SENSOR IC As described in Section II, the adhesive bandage sensor patch integrates a sensor IC. Fig. 6 shows the architecture of the sensor IC. It is composed of an Adaptive Threshold Rectifier (ATR), a nested chopper ECG Instrumentation Amplifier (IA), an ADC, an amplitude shift keying (ASK) modulator, an oscillator, and a power level detector. The power level detector stores information on the received ASK signal strength and recovered VDD level. This information is transmitted back to the network controller by the ASK modulator. A pair of P-FCB electrodes is capacitively coupled to the IA input. The ATR generates power from the incoming carrier signal, either in the High Frequency (HF, MHz) or Medical Implant Communications Service (MICS, 400 MHz) band; the HF band is for patch sensors, whereas the MICS band is supported for implantable sensors. The generated voltage is regulated and dispatched to the clock generator, the IA, and the ADC. Due to the low efficiency of P-FCB inductors, the amount of generated power is very limited, so the overall power budget of the sensor is designed to stay below 20 W. As a result, the ADC consumes 2 W, clock generator 2 W, regulator 3 W, and Analog Front-End (AFE) consumes only 5 W. Overall, the

4 YOO et al.: A 5.2 mw SELF-CONFIGURED WEARABLE BODY SENSOR NETWORK CONTROLLER 181 average power consumption of the sensor IC is 12 W when active. A nested chopper scheme is adopted in IA to improve SNR [17]. Although the sensor itself can process various vital signals, the IA in this version of the sensor is optimized for ECG to verify its operation. The captured ECG signal is amplified at IA and fed into ADC; the converted data is packetized and transmitted to the network controller by backscattering using ASK. For a sensor power source, we use wireless power transmission using inductive coupling and the ATR. There are several ways to provide power to a sensor patch: 1) Battery: Small batteries such as a flexible, paper-like battery [18] can be used for a sensor. However, chronic health monitoring over the long term forces the employment of a large capacity battery, making it inconvenient and expensive. Moreover, and perhaps most importantly, when it comes to sensors that will be attached directly to skin, some people are reluctant to use batteries because of the safety issues. 2) Thermal Energy Harvester: Body heat is a good candidate to provide power since human body temperature is stable at C. Several studies have demonstrated the feasibility of scavenging thermal energy from the body [19], [20]. However, the generated power level is around a few W cm with the load attached [19], which is too small to operate a sensor IC that will consume at least tens of Ws. [20] shows sufficient power levels of up to 30 W cm, but thermopiles are too expensive to be used in disposable sensors. Therefore, a thermal energy harvester alone would not be sufficient for the sensors. 3) Vibration Energy Harvester: Vibration is another candidate for the sensor power source. The electrostatic transducers or piezoelectric harvesters [21] can generate as much as W at 1.8 khz. Although the power level is large enough to operate a sensor IC, continuous high-frequency vibration is required to generate power, and, therefore, a vibration energy harvester is not appropriate for body-worn sensors. 4) Solar Cell: Solar energy cannot be used in this case since the patch sensor will be surrounded by clothes almost all the time. 5) Wireless Power Transmission Based on Inductive Coupling: Inductive coupling and RF can provide power to the sensor. RF link can transmit power over longer distance than the inductive coupling does, but it is more sensitive to interferences [8]. For example, RF in ISM band such as ZigBee and Bluetooth may interfere with Wireless LAN (WLAN) because they share the same frequency range [8], [22]. On the other hand, inductive coupling link is less sensitive to RF interference and can withstand more under the same level of unintended signal because it requires proximity between the transmitter and receiver. Since the healthcare system requires high security and interference-resilience, we choose to adopt inductive coupling to provide power to patch sensors. Even if unintended signal activates the patch sensor, the sensor response is automatically blocked by the network controller since it is physically not connected to the network controller. A. Adaptive Threshold Rectifier (ATR) Maximizing the rectification efficiency and the sensitivity are important in implementing wirelessly powered rectifiers. However, in this study, sensitivity is not a major problem due to Fig. 7. (a) Conventional CMOS rectifier and (b) battery-assisted CMOS rectifier [23]. Fig. 8. CMOS rectifier with ferroelectric capacitors [24]. the short distance cm between the sensor and reader coil. Therefore, we focus on increasing the efficiency. Typically, CMOS rectifiers are preferred because Schottky diodes or ferroelectric capacitors are expensive and difficult to integrate with the standard CMOS process. The proposed Adaptive Threshold Rectifier (ATR) boosts up the efficiency while using the CMOS process with Metal-Insulator-Metal (MIM) capacitors only. Hence, the cost can be lowered. The efficiency losses of a rectifier mainly originate from the threshold drop of the diodes within the rectifier [Fig. 7(a)]. As a result, dead zones occur, and the ideal maximum output voltage of the conventional CMOS rectifier is limited to, where is the maximum input signal amplitude [Fig. 7(a)]. To increase the rectification efficiency, various methods were proposed. A battery-assisted rectification scheme [23] [Fig. 7(b)] compensates for the drop at diode-connected M3 and M4 by adding a small battery with enough voltage to override the drop. This simple but powerful method increases the maximum output voltage to, and, as a result, the efficiency is increased as well. However, this method has a fundamental limitation: it requires batteries in the sensor, which is inappropriate in our case. In [24], a rectifier with ferroelectric capacitors was introduced (Fig. 8). It cancels out the threshold drop of the diode-connected transistors MP and MN by adding pre-applied voltage between gate and source of the diode-connected transistors, using ferroelectric capacitors and biased transistors, and. The ferroelectric capacitor has 10-fold better permittivity than the oxide cap: consequently, area reduction is achieved and parasitic capacitance is reduced. However, as in

5 182 IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 45, NO. 1, JANUARY 2010 Fig. 9. Adaptive Threshold Rectifier (ATR): (a) OFF and (b) ON. Fig. 10. Cascaded Adaptive Threshold Rectifier (ATR). the previous case, this smart idea requires additional ferroelectric capacitors and cannot be integrated into the standard CMOS process; thus it is too expensive to adopt in patch sensors. Fig. 9 shows the schematic diagram of the proposed ATR. The drop of the diode-connected transistor is minimized to improve the rectification efficiency. Transistors M1 M4 form a CMOS bridge rectifier. At startup, SW1 SW4 are connected to the drain of each transistor (ATR off, Fig. 9(a)). As the incoming signal is applied to the ANT and ANT nodes, the VP node voltage gradually increases. When the generated VOUT goes above, the power-on-reset (POR) signal triggers SW1 SW4. Then, the diode-connected devices M5 M8, which have voltage difference of between the drain and source, are attached to gate of M1 M4 [ATR on, Fig. 9(b)]. Since M5 M8 are matched to M1 M4, the drop of M1 M4 are cancelled out, making the dead zone virtually zero. As shown in Fig. 10, four ATR blocks are cascaded to generate 1.8 V at HF band or 1.6 V at MICS band. B. Nested Chopper Instrumentation Amplifier Fig. 11 shows the architecture of the sensor readout front-end with nested chopper IA. The architecture is composed of an input chopper stage, a preamplifier, a demodulator, and a Programmable Gain Amplifier (PGA). Due to the large dry P-FCB electrode impedance, large bias resistors ( each) are used at the input to minimize the loading effect. Thermal noise induced by the large bias resistors is low pass filtered by two off-chip DC blocking capacitors (100 nf each). In the IA, the nested chopper scheme [17] is adopted to achieve low noise and high SNR operation. At the preamplifier, the M3, M4 pair configures a positive feedback loop to enhance the gain of the preamplifier. The transconductance of this pair is designed to be 0.9 of the M4, M5 pair and 0.16 of the M1, M2 pair so that the M3, M4 pair contributes only 5% of the total noise, which is equal to a 16% noise contribution of the input pair. Consequently, the positive feedback adds a negligible amount of noise,

6 YOO et al.: A 5.2 mw SELF-CONFIGURED WEARABLE BODY SENSOR NETWORK CONTROLLER 183 Fig. 11. Sensor readout front-end with Nested Chopper Instrumentation Amplifier (IA). Fig. 12. Noise removal with the Nested Chopper IA. and, in addition, it does not degrade the power noise efficiency. The measured input impedance of the open-loop topology IA is 9.8 M, which is sufficiently high to avoid loading effects for the dry electrodes. As the vital signal level at the P-FCB electrode is 10 to 500, the readout front-end is designed to have a preamplifier gain of 60 db. Since the reference voltage of the 10-bit ADC is 0.8 V, the resolution is equal to 800, and this translates into a maximum tolerable 0.8 input referred noise level. To suppress the 1/f noise, the chopping frequency should be sufficiently higher than the 1/f corner frequency, so a 10 khz clock is used at the inner chopper. However, chopping spikes caused by the charge injection at the inner chopper (operating at 10 khz) leave a residual offset after the preamplifier and the inner chopper. Therefore, another chopper operating at lower frequency (625 Hz) is added to form a nested chopper topology, and suppress the residual offset (Fig. 12). The preamplifier with the diode-connected load pair M5, M6, together with the demodulator and LPF, have a gain of 40 db, while the PGA has 6 to 20 db of gain. V. NETWORK CONTROLLER The health monitoring chest band has a network controller on it. The main role of the network controller is to automatically configure the sensor location around the body, provide power to the configured sensors, and finally, collect the sensor data to storage. Fig. 13 shows the architecture of the network controller with a 12 4 inductor array. It is composed of an inductor array controller, a dual-band power transmitter with an 8-step adaptive power controller, an ASK demodulator, an MCU, a compression block, an AES-128 accelerator, SRAM, and an I C interface. The network controller can communicate with up to 48 sensors. In every programmed period (0.2 to 24 h), the array controller scans the inductor array for self-configuration. According to the signals from the sensors, the network controller automatically selects the HF (for patch-type sensors such as ECG, temperature, skin conductivity, etc) or MICS (for implantable sensors) band. The data rate can vary from 10 kb/s to 120 kb/s, depending on the sensor types. An important feature of the network controller is the self-configuration. The health monitoring band is worn over a subject s chest, and the inductor array with the inductors automatically finds sensors attached at arbitrary positions around the chest. Fig. 14 describes the self-configuration algorithm based on sequential scanning algorithm. At system reset, the inductor array controller sequentially connects the power transmitter to each inductor, from (X0, Y0) to (X11, Y3). Then, the controller provides power to the selected position. An inductor is selected when the X address and the Y address are connected to the power transmitter. If a sensor exists beneath the current position, it will generate supply voltage and respond with an acknowledgment packet (ACK) including the sensor information; in this case, the network controller immediately configures the sensor type. If there is no response, the network controller waits for 168 ms for time expiry before moving on to next position. If more than one inductor detects a sensor, a first-come, first-served method is applied: the first inductor that detects the sensor takes control of the sensor. Fig. 15 shows an example of the self-configuration process at (X3, Y1) position. At Step 1, the inductor at (X3, Y1) position in the health monitoring band is selected, and at Step 2, the carrier signal is provided through the inductor. For Step 3, the adhesive bandage sensor placed underneath the inductor is activated and responds with the ACK packet via ASK load modulation. In Step 4, the backscattered information is demodulated, and the network controller configures the position. Lastly, in Step 5, the network controller moves on to the next position (X3, Y2). The 8-step adaptive power controller controls the transmitting power signal strength from 1 to 8 to compensate for the fluctuation in received power at sensors [Fig. 16(a)]. The received power level detector in the sensor encodes the power level received into the data packet that will be backscattered to the network controller. As the network controller decodes the packet,

7 184 IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 45, NO. 1, JANUARY 2010 Fig. 13. Network controller architecture. Fig. 15. An example of self-configuration process. Fig. 14. Self-configuration algorithm. it tracks the current power level (recovered VDD) at the sensor and adapts the strength of the transmitting power. At the beginning, the adaptive transmitter drives the inductor at maximum strength, and as the data packet from the sensor is received, the adaptive transmitter starts to lower the transmitted power level to minimize power consumption. If the sensor and the chest band inductors are misaligned, a higher power level will be necessary to provide sufficient power to the sensors; simulation shows that power increases 4-fold when two inductors are misaligned by 15 mm. When the inductors are aligned, the network controller consumes 5.2 mw while in operation. However, if there is too much misalignment, the network controller alerts an error signal, and, in this case, the chest band needs to be reinstalled. Fig. 16(b) shows the adaptive power transmitter. CTRL [7:0] controls the amount of current that will drive the attached inductor. The sensor signal and the configuration information (time, position, type of sensors) are compressed by a quadratic level compression algorithm with an average compression ratio of 8.4:1 in order to reduce the data memory size [25]. To protect the privacy of the personal health information, the sensor data are encrypted by an AES-128 accelerator [10] before being stored in data memory. Through a high-speed mode I C diagnosis interface, a healthcare expert can access the collected data for healthcare assistance. The network controller is powered by a single 3.6 V, 1000 mah, 1/2-AA size battery; it is then regulated down to 1.8 V as a supply. With the 5.2 mw measured minimum power consumption of the network controller, this translates into eight days of continuous ECG monitoring without replacing the battery. This is sufficient for 7-day ambulatory monitoring that is essential for tracking heart diseases [2]. VI. IMPLEMENTATION AND MEASUREMENT The health monitoring chest band (Fig. 1) and the adhesive bandage type sensor patch (Fig. 2) are implemented using P-FCB technology. The health monitoring band is

8 YOO et al.: A 5.2 mw SELF-CONFIGURED WEARABLE BODY SENSOR NETWORK CONTROLLER 185 Fig. 16. (a) Adaptive power control. (b) Power transmitter. Fig. 17. Conversion efficiency of Adaptive Threshold Rectifier (ATR). Fig. 19. Input referred noise PSD of IA. Fig. 18. Rectified voltage (VDD) versus frequency. mm large, and it has an array of 12 4 inductors on it. There are two types of sensors: one is mm, and the other is mm large. Fig. 17 shows the measurement results of the ATR within the sensor chip. The ATR peak efficiencies are measured as 54.9% and 45.2% for HF (13.56 MHz) and MICS (402 to 405 MHz) bands, respectively. The ATR improves the rectification efficiency by 18.1% at 6 dbm input power compared to that using expensive ferroelectric capacitors [24] or that using battery at sensor node [23]. Fig. 18 plots the measured rectified voltage (VDD) versus the input signal frequency when the input power is fixed to 0 dbm. The generated VDD is 1.8 V up to 200 MHz before dropping sharply around 500 MHz to 1.0 V at 1 GHz. This drop is due to a leakage current caused by parasitic capacitance at the rectifier input. However, since our band of interest is below 400 MHz, this is not a major concern. At 0 dbm input, VDD increases Fig. 20. Measured ECG signal and VDD ripple at sensor. by 0.75 V when ATR is switched on. Fig. 19 shows the input referred noise power spectral density (PSD) of the implemented IA. The noise floor is 47 nv Hz, and this is equivalent to 0.51 over vital signal bandwidth ( Hz). As shown in Fig. 20, the VDD ripple after regulation is suppressed down to 30 mv (1.6% of generated VDD) while the chip is in operation. The measured ECG waveforms by the P-FCB sensor are also shown in Fig. 20. The measurement results in Fig. 21 show that the network controller scans (X2, Y3), (X3, Y0), (X3, Y1), and (X3, Y2) positions sequentially. In this specific time window during the self-configuration phase, (X3, Y0) and (X3, Y2) do not reply until a timeout (168 ms), whereas (X2, Y3) and (X3, Y1) respond and are successfully configured. Table I summarizes the performance and key features of the proposed system. The average power consumption is 5.2 mw for the network controller chip with a 12 4 inductor array, and

9 186 IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 45, NO. 1, JANUARY 2010 Fig. 21. Self-configuration results for (X2, Y3) to (X3, Y2). TABLE I PERFORMANCE SUMMARY TABLE II COMPARISON WITH PREVIOUS WORKS 12 W for the sensor chip when the sensor interfaces, the ECG IA, and the ADC, are in operation. The transceiver uses the HF band for patch sensors, and MICS band for implantable sensors. Table II compares previous works and this work. The rectifier efficiency is the highest (54.9%) at the HF band. Furthermore, the proposed system is implemented using a standard CMOS process without any expensive additional process. Fig. 22 shows the chip micrographs of the network controller and the sensor IC. The network controller chip and the sensor chip are fabricated in 0.18 m 1P6M standard CMOS technology and occupy 15.0 mm and 4.8 mm area respectively, including pads. Fig. 22. Chip micrographs.

10 YOO et al.: A 5.2 mw SELF-CONFIGURED WEARABLE BODY SENSOR NETWORK CONTROLLER 187 VII. CONCLUSION A wearable healthcare system for continuous health monitoring over long period is presented and implemented using Planar-Fashionable Circuit Board (P-FCB) technology. The proposed system exploits the adhesive bandage type patch sensor that is wirelessly powered by a health monitoring chest band. A dry P-FCB electrode is used at the sensor, and, unlike conventional wet electrodes, it is free from skin irritation problems, and thus, suitable for long-term monitoring. The Adaptive Threshold Rectifier (ATR) within the sensor IC improves the rectification efficiency up to 54.9% at the HF band, and, with the help of the nested chopper instrumentation amplifier, the integrated noise at the input of the sensor is 0.51 for vital signal bandwidth. The network controller integrated within the health monitoring band adopts a self-configuration scheme to automatically locate the position of sensors worn on arbitrary locations around the body, and it provides power to the selected sensors only. The sensor IC (4.8 mm area) and the network controller (15 mm area) are both fabricated in 1P6M 0.18 m CMOS technology and consume only 12 W and 5.2 mw average power, respectively. ACKNOWLEDGMENT The authors would like to thank M.D. Yeonchang Ryu, Prof. Shiho Kim, and Mr. Soonbae Kim for their helpful comments and technical supports. REFERENCES [1] I. Korhonen, J. Parkka, and M. Van Gils, Health monitoring in the home of the future, IEEE Engineering in Medicine and Biology Magazine, vol. 22, pp , May-Jun [2] D. Jabaudon, J. Sztajzel, K. Sievert, T. Landis, and R. Sztajzel, Usefulness of ambulatory 7-Day ECG monitoring for the detection of atrial fibrillation and flutter after acute stroke and transient ischemic attack, Stroke, Journal of American Heart Association, vol. 35, pp , May [3] E. A. Zerhouni, NIH in the post-doubling era: Realities and strategies, Science, vol. 314, pp , Nov [4] E. A. Zerhouni, NIH aims to transform findings into clinical changes Future medicine: Predictive, personalized, preemptive, and participatory, in A Column in U. S. Medicine. :, [5] A. Searle and L. Kirkup, A direct comparison of wet, dry and insulating bioelectric recording electrodes, Physiological Measurement, vol. 21, no. 2, pp , May [6] R. Paradiso, G. Loriga, and N. Taccini, A wearable health care system based on knitted integrated sensors, IEEE Trans. Information Technology in Biomedicine, vol. 9, no. 3, pp , Sep [7] A. C.-W. Wong, D. McDonagh, G. Kathiresan, O. C. Omeni, O. El-Jamaly, T. C.-K. Chan, P. Paddan, and A. J. Burdett, A 1 V, micropower system-on-chip for vital-sign monitoring in wireless body sensor networks, in IEEE Int. Solid-State Circuits Conf. (ISSCC) Dig. Tech. Papers, Feb. 2008, pp [8] J. Yoo, S. Lee, and H.-J. Yoo, A 1.12 pj/b inductive transceiver with a fault-tolerant network switch for multi-layer wearable body area network applications, IEEE J. Solid-State Circuits, Nov [9] J. Yoo, L. Yan, S. Lee, Y. Kim, H. Kim, B. Kim, and H.-J. Yoo, A 5.2 mw self-configured wearable body sensor network controller and a 12 W 54.9% efficiency wirelessly powered sensor for continuous health monitoring system, in IEEE Int. Solid-State Circuits Conf. (ISSCC) Dig. Tech. Papers, Feb. 2009, pp [10] H. Kim, Y. Kim, and H.-J. Yoo, A 1.12 mw continuous healthcare monitor chip integrated on a planar-fashionable circuit board, in IEEE Int. Solid-State Circuits Conf. (ISSCC) Dig. Tech. Papers, Feb. 2008, pp [11] Y. Kim, H. Kim, and H.-J. Yoo, Electrical characterization of printed circuits on the fabric, IEEE Trans. Advanced Packaging. [12] J. G. Webster, Medical Instrumentation: Application and Design, 3rd ed. : John Wiley & Sons, Inc., 1998, ch. 5. [13] M. Connolly and D. A. Buckley, Contact dermatitis from propylene glycol in ECG electrodes, complicated by medicament allergy, Contact Dermatitis, vol. 50, no. 1, p. 42, Apr [14] L. Yan, N. Cho, J. Yoo, B. Kim, and H.-J. Yoo, A two-electrode 2.88 nj/conversion biopotential acquisition system for portable healthcare device, in IEEE Asian Solid-State Circuits Conference (A-SSCC) Proc. Technical Papers, Nov. 2008, pp [15] S. S. Lobodzinski and M. M. Laks, Biopotentional fiber sensor, Elsevier Journal of Electrocardiology, vol. 39, pp. S41 S46, [16] T.-H. Kang, C. R. Merritt, E. Grant, B. Pourdeyhimi, and H. T. Nagle, Nonwoven fabric active electrodes for biopotential measurement during normal daily activity, IEEE Trans. Biomedical Engineering, vol. 55, no. 1, Jan [17] A. Bakker, K. Thiele, and J. H. Huijsing, A CMOS nested-chopper instrumentation amplifier with 100-nV offset, IEEE J. Solid-State Circuits, vol. 35, pp , Dec [18] PowerPaper Apr. 20, 2009 [Online]. Available: [19] C. Lauterbach, M. Strasser, S. Jung, and W. Weber, Smart clothes self-powered by body heat, in Avantex Symposium, [20] V. Leonov, P. Fiorini, S. Sedky, T. Torfs, and C. Van Hoof, Thermoelectric MEMS genrerators as a power supply for a body area network, in Proc. 13th Int. Conf. Solid-State Sensors, Actuators and Microsystems (Transducers 2005), Jun. 2005, pp [21] M. Renaud, K. Karakaya, T. Sterken, P. Fiorini, C. Van Hoof, and R. Puers, Fabrication, modelling and characterization of MEMS piezoelectric vibration harvesters, Sensors and Actuators A: Physical, vol , pp , Jul. Aug [22] Z. Alliance, ZigBee and wireless radio frequency coexistance, in ZigBee White Paper. :, [23] T. Umeda, H. Yoshida, S. Sekine, Y. Fujita, T. Suzuki, and S. Otaka, A 950-MHz rectifier circuit for sensor network tags with 10-m distance, IEEE J. Solid-State Circuits, vol. 41, pp , Jan [24] H. Nakamoto, D. Yamazaki, T. Yamamoto, H. Kurata, S. Yamada, K. Mukaida, T. Ninomiya, T. Ohkawa, S. Masui, and K. Gotoh, A passive UHF RF identification CMOS tag IC using ferroelectric RAM in 0.35-m technology, IEEE J. Solid-State Circuits, vol. 42, pp , Jan [25] H. Kim, Y. Kim, and H. Yoo, A low cost quadratic level ECG compression algorithm and its hardware optimization for body sensor network system, in Proc. IEEE Engineering in Medicine and Biology Conference (EMBC), Aug. 2008, pp Jerald Yoo (S 05) received the B.S. and M.S. degrees in Department of Electrical Engineering from the Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea, in 2002 and 2007, respectively. He is currently working toward the Ph.D. degree at KAIST. As a chief researcher at the Semiconductor System Laboratory in KAIST, he developed low-energy transceivers and wirelessly powered sensors for body sensor network, and an embedded processor for PRAM. His current research interests include low energy wearable body area network transceiver, wireless power transmission and low-power biomedical microsystem. Long Yan (S 07) received the B.S. and M.S. degrees in electrical engineering from Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea, in 2007 and 2009, respectively. He is currently pursuing the Ph.D. degree in electrical engineering at KAIST. He has worked on developing the low power FSK transmitter for body channel communication, and low noise, wirelessly powered sensor for wearable body sensor network. His current research interests include design of energy-efficient biomedical micro-system for body area network.

11 188 IEEE JOURNAL OF SOLID-STATE CIRCUITS, VOL. 45, NO. 1, JANUARY 2010 Seulki Lee (S 07) received the B.S. and M.S. degrees in Department of Electrical Engineering from Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea, in 2007 and 2009, respectively. She is currently working toward the Ph.D. degree in Electronic Engineering at KAIST. Her current research interests include the inductive coupling transceiver design and near-field communication. Yongsang Kim (S 07) received the B.S. and M.S. degrees in Department of Electrical Engineering from Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea, in 2004 and 2009, respectively. His research includes printing circuit board techniques on the fabric and low power sensor unit design for wearable computer system. Hoi-Jun Yoo (M 95-SM 04-F 08) graduated from the Electronic Department of Seoul National University, Seoul, Korea, in 1983 and received the M.S. and Ph.D. degrees in Department of Electrical Engineering from the Korea Advanced Institute of Science and Technology (KAIST), Daejeon, in 1985 and 1988, respectively. His Ph.D. work concerned the fabrication process for GaAs vertical optoelectronic integrated circuits. From 1988 to 1990, he was with Bell Communications Research, Red Bank, NJ, where he invented the two-dimensional phase-locked VCSEL array, the front-surface-emitting laser, and the high-speed lateral HBT. In 1991, he became a manager of the DRAM design group at Hyundai Electronics and designed a family of fast-1 M DRAMs to 256 M synchronous DRAMs. In 1998, he joined the faculty of the Department of Electrical Engineering at KAIST and now is a full professor. From 2001 to 2005, he was the director of System Integration and IP Authoring Research Center (SIPAC), funded by Korean Government to promote worldwide IP authoring and its SOC application. From 2003 to 2005, he was the full time Advisor to Minister of Korea Ministry of Information and Communication and National Project Manager for SoC and Computer. In 2007, he founded System Design Innovation & Application Research Center (SDIA) at KAIST to research and to develop SoCs for intelligent robots, wearable computers and bio systems. His current interests are high-speed and low-power Network on Chips, 3-D graphics, Body Area Networks, biomedical devices and circuits, and memory circuits and systems. He is the author of the books DRAM Design (Seoul, Korea: Hongleung, 1996; in Korean), High Performance DRAM (Seoul, Korea: Sigma, 1999; in Korean), Low-Power NoC for High-Performance SoC Design (CRC Press, 2008), and chapters of Networks on Chips (New York, Morgan Kaufmann, 2006). Dr. Yoo received the Electronic Industrial Association of Korea Award for his contribution to DRAM technology in 1994, the Hynix Development Award in 1995, the Design Award of ASP-DAC in 2001, the Korea Semiconductor Industry Association Award in 2002, the KAIST Best Research Award in 2007, the Asian Solid-State Circuits Conference (A-SSCC) Outstanding Design Awards in 2005, 2006 and 2007, and the DAC/ISSCC Student Design Contests Award in 2007 and He is an IEEE fellow and serving as an Executive Committee Member and the Far East Secretary for IEEE ISSCC, and a Steering Committee Member of IEEE A-SSCC. He was the Technical Program Committee Chair of A-SSCC 2008.