Outline for Proc. of the IEEE paper on the µCluster

Manuscript Received: 11/24/97

Revised: 1/14/98

A GENERIC MULTI-ELEMENT MICROSYSTEM FOR PORTABLE WIRELESS APPLICATIONS

Andrew Mason, Navid Yazdi, Abhijeet V. Chavan,

Khalil Najafi, and Kensall D. Wise

ABSTRACT

An open-architecture microsystem that can be populated with a variety of sensors and actuators is described. The microsystem is designed for low-power wireless applications where small size and high sensor accuracy are important. It consists of an in-module microcontroller connected to multiple front-end transducers through an intramodule sensor bus. An external interface allows internally processed data to be output through either a hardwired I/O port or an RF transmitter. The present microsystem is configured for environmental monitoring and measures temperature, barometric pressure, relative humidity, and acceleration/vibration. It occupies less than 10cc, consumes an average of 530µW from 6V, and transmits data up to 50m. System features such as active power management, the intramodule sensor bus, generic bus interface circuitry, and in-module sensor compensation based on bivariate polynomials are discussed.

The authors are with the Center for Integrated Sensors and Circuits, Department of Electrical Engineering and Computer Science, The University of Michigan, Ann Arbor, MI 48109-2122. This work was sponsored by the Defense Advanced Research Projects Agency under Contract No. J-FBI-92-149.

Introduction

Driven by rapid advances in microcomputers and global connectivity, many of the most important emerging markets for microelectronics require the ability to gather information from the non-electronic world [1,2]. Examples include health care (diagnostic and therapeutic devices, prosthetics), automotive systems (smart vehicles and smart highways), automated manufacturing (including smart VLSI process tools), environmental monitoring and control devices, defense systems, and many consumer products. Using integrated circuit technology and extensions of it, integrated sensors and microactuators are being developed to provide the necessary input/output devices. These are increasingly being realized combined with hybrid or monolithic circuitry on a common substrate [3-6] and have come to be known as MicroElectroMechanical Systems (MEMS). Merging these devices with increasingly-powerful digital signal processing electronics now makes it possible to go beyond simple analog readout circuitry and form complete closed-loop microsystems in very small highly-integrated modules. These autonomous microsystems are capable of gathering data from the physical world, converting it to electronic form, compensating it for interfering variables and nonlinearities, and either acting on the information directly or transferring it to other systems [7-9]. Such microsystems can be expected to have a significant and pervasive impact on a large number of applications during the coming decade.

As used in this paper, a 'microsystem' is defined as a collection of highly-integrated devices that contains transducers along with appropriate interface circuitry and is capable of performing multiple tasks autonomously as well as responding intelligently to various commands from a host system. This definition helps to identify the class of devices being considered and illustrates the differences between a microsystem and a less complex “smart” sensor. Reflecting general trends in microelectronics, many of the emerging applications for microsystems demand small portable wireless devices [10-12] having high accuracy and the ability to perform in distributed instrumentation systems collecting data over a broad physical area [13]. It is significant that in applications as diverse as personal health monitors, implantable neural prostheses, sensors for VLSI process control, sentries for battlefield awareness, and sensors for unmanned air reconnaissance, a single generic open system architecture applies. Such systems are customized by the sensors and actuators populating them and by the software used in their resident microcontrollers. While not every application shares the same power or telemetry range requirements, a common system framework can be shared and provides a basis for defining the standards that will be required in order for these systems to fully evolve [14,15].

The eventual realization of many such microsystems will be as small (1cc), portable, low-power (<1mW), highly-accurate sensing modules capable of bidirectional communication over considerable distances (1-2km). While not yet meeting all of these objectives, the microsystem reported in this paper represents a significant step in this evolution. It illustrates a generic architecture for such devices along with operational features such as in-module data compensation and the use of self-testing at the transducer level. The resulting microinstrumentation system ("µCluster") is configured for environmental monitoring [9], including use as a microweatherstation for local, regional, and global weather forecasting.

Generic Microsystem Architecture

The open architecture shown in Fig. 1 partitions each individual microsystem so it can be populated to fit the requirements of different applications without limiting the transducer technologies that can be accommodated. Each microsystem can be used as part of a distributed network of such devices connected through a system bus to a remote host. This external bus may be either hardwired or wireless. Within each microsystem node is signal processing electronics that operates under stored program control to perform preset tasks and respond to commands received from a host system. In-module memory stores the control programs as well as sensor-specific code that describes the operation of each front-end device. The control electronics connects to the transducer front-end through an intramodule sensor bus which allows read and write instructions to be issued to the transducers. Each of the front-end sensor/actuator chips contains an interface to this bus along with the transducers and their associated readout and/or control circuitry. A microcontroller provides in-module signal processing and memory as well as power management circuitry and the I/O hardware for both the external system bus and the intramodule sensor bus.

Several bus-compatible “smart” sensors (similar to the front-end sensors here) have been reported for use as microsystems [5,16]. However, many emerging sensors measure multiple parameters and are programmable, electronically trimmed, and self-testing. For these devices, as well as for systems containing multiple sensing chips, it is very attractive to utilize an intramodule sensor bus along with an embedded microsystem controller [17]. The microcontroller performs in-module signal processing and permits the microsystem to respond to commands from a host. It also compensates the sensor data and can make in-module decisions based on this data to off-load the host system [12]. The microsystem can be programmed with more or less knowledge of the overall system task being performed, depending on the application.

For an open module architecture with a “plug and play” approach to front-end transducer population, a standard interface between the microcontroller and front-end devices is necessary to allow different sensor manufacturers to design to a common system specification. Whether the sensor is implemented as a single chip or as a two-chip hybrid [1], it is necessary to have a standard sensor bus connecting the controller to the sensing nodes. Figure 2 illustrates the signal path for the sensor data and shows the possible partitions to accommodate a sensor bus that divides the front-end from the controller. With each option the signal path consists of the transducer itself, analog signal conditioning (amplification) circuitry, analog-to-digital conversion (ADC), digital signal processing, and finally the microsystem controller. In (a) the bus receives the direct transducer output while all signal processing circuitry is in the control block. This option is not appropriate because of the difficulties in busing unbuffered sensor signals (representing aF or fF of capacitance change), and because, in order to maintain an open architecture, it would require many different analog readout circuits to be built into the control circuitry. In options (b), (c), and (d), the difference is the amount of signal processing done at the front-end. Many microcontrollers available today have built-in ADCs and digital signal processing capabilities so that option (b) is a good choice that shares available resources. However, a truly open architecture should also accommodate devices matching options (c) and (d) since these do not force additional requirements on the microcontroller. Thus, using a microcontroller with a built-in ADC, the architecture shown in Fig. 1 represents a very generic structure accommodating a variety of sensor nodes. Any front-end device with the appropriate bus interface circuitry can be added to such a microsystem, providing the desired “plug and play” capabilities.

Microsystem Components

The generic architecture introduced above defines a basic set of control, communication, and front-end components that will now be discussed in more detail to illustrate issues important to microsystem development. These issues will be presented as they relate to the prototype µCluster [9]. This microinstrumentation system was designed to demonstrate how the generic architecture could be used to build a microsystem that is small in size, dissipates very low power, utilizes wireless communication, and provides highly-accurate fully-compensated sensor data. As shown in Fig. 3, the µCluster consists of external interfaces for both hardwired and wireless communication, a control block containing a microcontroller and power management circuitry, and several front-end sensing nodes. The control block is connected to the sensors through an intramodule sensor bus and sensor interface chips. The sensors, measuring temperature, barometric pressure, relative humidity, and acceleration/vibration, are all capacitive so that the associated power dissipation is limited to the analog readout circuits used; all of these transducers are self-testing using embedded thermal or electrostatic actuation. Operating under stored program control, the µCluster performs periodic scans of these embedded transducers. The scan rate is programmable and can be adaptive, based on the rate of change in the sensor readings. The sensor data is compensated by electronically programming the front-end readout circuitry (for coarse trims of offset and slope) and in software using prestored compensation algorithms. The resulting measurements are stored internally until output to a remote host system. Between scans, the µCluster utilizes a low-power “sleep mode” to maximize battery life, but the sensor bus is still monitored for event-triggered interrupts that can wake the system.

The Microcontroller

The microcontroller unit (MCU) used in such microsystems should provide a microprocessor for signal processing and stored-program control, I/O interfaces to both the external system bus and the intramodule sensor bus, and an on-chip ADC and timing hardware to accommodate different sensor data formats. Adequate on-chip ROM, RAM, and EEPROM are important as well as overall power dissipation, which is important in spite of the relatively low duty cycles in many applications. For the present µCluster, the Motorola 68HC11 is used. It has an 8b processor, an RS-232 compatible UART type interface for external I/O, a synchronous serial peripheral interface (SPI) adapted for sensor bus communications, an 8b ADC for converting analog sensor data, signal timing hardware for frequency-encoded sensor data, a variety of on-chip memory, and a low-power mode that maintains RAM data while drawing less than 50µA [18]. The active power dissipation is about 90mW with a 2MHz bus clock. EPROM (96kb) is used to store control and sensor-specific program code; EEPROM (4kb) is used to store operation and compensation data for individual sensors; and RAM (4kb) is used for the temporary storage of sensor data and program variables.

The 68HC11 was not specifically designed for these microsystem applications; indeed, few manufacturers have so far addressed this area in spite of the fact that it promises to be a very high-volume market. A microcontroller designed specifically for such microsystems should have all of the basic subsystems noted above but should minimize hardware, including I/O ports, to keep the power consumption low (<10mW active; <20µW standby). Since a direct trade-off between clock speed and power dissipation exists in the digital electronics, and because the readout time for many transducers is on the order of 100µsec or more (during which the processor will generally be on), a processor speed of 5-15MHz should be adequate. The controller should also have an ADC with =12b accuracy; for some applications, 14-16b accuracy is needed. Hence, a 16b processor is also desirable. Low-power low-voltage circuit techniques [19] should be used along with the smallest possible device dimensions to help reduce load capacitances. The ability to selectively shut down MCU subsystems when not in use can further reduce power dissipation. Finally, the ability to programmably control clock frequencies would allow transducer front-end operations to occur more slowly while processing data at the fastest possible rate. Although some of these specifications are perhaps aggressive, the technology available today, if applied to a microsystem-specific controller, could achieve them.

Power Management

In battery-powered microsystems, it is important to minimize power consumption wherever possible. In the transducer front-end, capacitive sensors should be used since these devices offer high sensitivity yet consume no power and can be read out rapidly using low-power circuit techniques. For wireless operation, it is important to use low-power telemetry hardware, understanding the direct trade-off between power and communication range. While many efforts are currently underway to reduce the power consumption of MCUs [20] and wireless communication devices [21], system-level approaches to power management are also important. A common method for conserving power that was adopted in the µCluster is to power down unused subsystems. Since the MCU here is periodically shut down, a separate power management chip (PMC) was necessary. This all-digital CMOS circuit can implement all of the necessary functions with less than 10µA of supply current. As shown in Fig. 4, the PMC contains an on-chip clock generator and timing hardware that can measure eight discrete time periods between 15 seconds and 5 minutes based on a 4b code input from the MCU. This allows the MCU to enter a low-power “sleep” mode for a preset duration, after which the PMC wakes the MCU. Because many of the parameters measured by the µCluster change slowly, most front-end nodes can also be powered down while not in use. These devices are powered through a 5V reference available on the sensor bus. This voltage reference is controlled by the PMC and can be turned on and off by input codes sent from the MCU. Likewise, the RF transmitter can be shut-off by commands from the MCU. Thus, the MCU, through the PMC, controls when and where power is available on the µCluster. For some front-end devices, however, it is important to monitor the environment continuously. Such sensor nodes must be very low in power and are driven by the main system supply. If an event is detected that requires attention while the MCU is asleep, an interrupt line on the sensor bus is used to wake the MCU through the PMC. A front-end device that uses this interrupt feature is discussed in a later section of this paper.

Power management features will certainly be necessary in most low-power microsystems [22] and will normally be integrated directly within the microcontroller along with selectively-activated voltage references, low-power timing circuitry, and interrupt recognition. To further reduce power consumption, a 3V supply is desirable but will put demanding constraints on the analog circuits in the transducer front-end. Due to the limited current available from long-life battery supplies, it is necessary to reduce the maximum active (peak) load by individually enabling different circuit blocks to periodically perform desired tasks. Advances in battery materials and packaging continue to provide cells with improved energy density, weight, and volume [23], yet further improvements will be necessary as portable systems increase in performance while shrinking in size. Efforts using MEMS-based micro power supplies [24] illustrate new approaches to meeting the challenges of these small low-power microsystems.

Wireless Communications

Wireless communications eliminates the need for costly wired networks and allows for rapid and simple microsystem deployment. Bidirectional communication with a range of a mile or more is needed for many applications, yet with current technology, these features would require more space and power than are available in the µCluster. However, the portable revolution now taking place is driving wireless technology with similar demands so that devices compatible with these needs should soon be available [21]. RF technology has moved from purely analog to a mix of analog and digital components, providing a more efficient use of bandwidth, lower power dissipation, and improved noise immunity [25]. Several technologies are currently being investigated for use in RF transceivers including CMOS, Si bipolar, Si/SiGe heterojunction bipolar (HBT), and GaAs MESFET and HBT devices. The passive components (inductors, capacitors, filters, and resonators) used often determine the size and performance of such systems, and current research efforts in MEMS aimed at performing these functions mechanically [26] should make it possible to realize the entire transceiver monolithically, reducing size, weight, cost, and power. Low-power CMOS wireless systems have been introduced [27], and commercial products such as the pager wristwatch [28] are driving the technology in the same direction as the microsystems considered here.

For the µCluster, the twin goals of small size and low power had a direct impact on the achievable communication range. To minimize power consumption, wireless operation was limited to data output only, and the HX1005 transmitter from RFM Inc., a commercially available component in a very small package, was selected. This device operates with a 315MHz carrier frequency using amplitude-shift-keyed modulation with a 1kHz bit rate. The device has an average active power dissipation of 10.5mW, with a range of about 50m and a volume of 0.27cc.

A Standard Sensor Bus

The sensor bus provides access to multiple front-end sensing nodes while using a limited number of interconnects. Each of these nodes can contain many addressable elements, including sensors, actuators, and programmable interface circuits. For the µCluster, a nine-line intramodule sensor bus has been adopted, although many other options are available [29]. This bus is based on the Michigan Serial Standard [30] and is very similar to the transducer independent interface (TII) of the recently developed IEEE-P1451 standard for sensor systems [31,32]. This bus includes three power leads, four lines for synchronous serial communication, a data output line, and a data valid/interrupt signal. The three power leads are the system supply, ground reference, and a switched reference voltage for devices that can be powered-down during a low-power system sleep mode. For synchronous communication with the front-end devices, a chip enable and data strobe provide handshaking for use with a programmable-frequency clock and a serial data line. The data out line is used for sensor data output, which can be either an analog voltage, a frequency encoded pulse train, or a serial bit stream. The last line of the sensor bus is data valid, which signals the MCU when valid data is present on the data out line. Data valid is also used as an interrupt from the front-end devices that can initiate an event-triggered sensor readout when the bus is otherwise idle. A timing diagram illustrating the relationships among several of these signals is shown in Fig. 5.

As implemented in the µCluster, the data format for commands sent to the front-end nodes consists of a 4b chip address followed by an 8b instruction byte that contains a 3b command and a 5b element address. The 3b command determines if the instruction is a sensor read request or one of several write commands. The 5b element address is used to access multiple devices within each front-end node. In the case of a write command, data in 4b or 8b packets will follow the 8b instruction byte. Write commands can be used to write data to front-end actuators or to set control bits on programmable interface chips as discussed below. In the case of a read command, the sensor bus clock can be used by front-end readout circuitry; the clock frequency is set by MCU software. When valid data is ready, the interface circuitry signals the controller through the DV bus line, placing the data on the shared DO line. Output to the DO line is disabled once the controller has received the sensor data.

The Transducer Front-End

The transducer front-end of a microsystem consists of one or more sensing nodes. Each node contains sensors and/or actuators with appropriate readout circuitry to interface with the intramodule sensor bus. For open architectures, the front-end can be populated based on the requirements of the selected application, allowing the same basic system to be used for a variety of applications. In the µCluster, each front-end sensing node has an assigned 4b address code, allowing up to 16 such nodes to be accessed within the same microsystem. Each node can contain multiple sensors and/or actuators, limited only by the sensor bus instruction format. As described above, the data format used on the µCluster contains a 5b element address that allows 32 readable elements and 32 writeable elements within each sensing node. Because the front-end transducers in the present µCluster are all capacitive, a generic interface chip can be used to connect all of these devices to the sensor bus. This generic interface chip and the individual sensors used for environmental monitoring in the µCluster are discussed below.

A Generic Capacitive Sensor Interface Chip

For environmental monitoring applications, the µCluster utilizes several capacitive sensors, which are attractive due to their low power, high sensitivity, and self-test capabilities. While numerous readout circuits for capacitive microsensors have been introduced [33-38], for multi-element microsystems such as the µCluster, a low-power generic interface that can accommodate the broad range of base capacitances and sensitivities associated with different transducers is needed. Such an interface has been developed [39]. This sensor interface chip reads out capacitive devices very quickly, communicates with the microcontroller through the sensor bus, supports self-test and self-calibration, supports multi-ranging for a single sensor, and dissipates low power. The chip also includes a temperature sensor as discussed below.

Figure 6 shows a block diagram of the capacitive sensor interface chip. The serial data instructions that are transmitted from the MCU over the sensor bus are received, decoded, and stored by the bus interface unit and are applied to control the other circuit blocks illustrated in Fig. 6. Data written to the interface chip is stored in registers within the bus interface unit. To interface with capacitive sensors, this chip utilizes a low-noise front-end charge integrator to read out the difference between the sensor capacitance and a reference capacitor [34]. An input multiplexer allows the chip to interface with up to six sensor elements. Furthermore, the chip can be digitally programmed to operate with one of three reference capacitors, either external or internal. The on-chip reference capacitors are laser trimmable from 0.15pF to 8pF. Programmability of the reference capacitors allows the chip to interface with capacitive sensors having a wide range of base capacitance and also provides offset control.

The analog signal path consists of the input multiplexer, an input charge integrator, a gain stage, an output sample/hold circuit, and an output multiplexer. The gain stage can be programmed on-line to one of three gain settings. These gains can be used to accommodate sensors with different sensitivities or can be used for multi-ranging a single sensor and can be fine tuned by laser trimming the on-chip capacitors. The overall readout sensitivity can be varied from 0.23mV/fF to 73.5mV/fF using both digital programming and laser trimming, providing an effective gain variation from 1 to 312. The output multiplexer controls access to the sensor bus for both the capacitive readout circuitry and the on-chip temperature sensor. Additionally, the interface chip supports self-test and self-calibration through a 3b on-chip DAC that generates a variable-amplitude two-phase clock for driving the sense and reference capacitors in each input charge integration cycle. The variable amplitude clock can also be used to apply a varying electrostatic force to the sensor for self-test and self-calibration.

The interface chip has been fabricated using a standard 3µm single-metal double-poly p-well CMOS process. Figure 6 shows a photograph of the chip, which measures 3.2mm x 3.2mm. It dissipates less than 2.2mW from a single 5V supply and can resolve input capacitance variations of less than 1fF with a readout time of 60µsec [39].

Temperature Sensor

To measure temperature in close proximity to the other sensors in the microsystem, a temperature sensor has been integrated on the interface chip. This sensor provides data that can be used to digitally compensate for the temperature sensitivity of any other sensors connected through this chip. The temperature sensor utilizes the temperature dependence of the drain current of an MOS transistor in weak inversion [40]. The charging current for the capacitively-loaded Schmitt input stage of a ring oscillator is set by a PMOS transistor biased for subthreshold operation. Since this charging current is temperature dependent, the frequency of the oscillator provides a measure of the local temperature. A typical device displays a sensitivity of 4msec/°C at 60°C and 33msec/°C at -20°C with a resolution better than 0.5°C across the tested range. Although the sensor is highly nonlinear, it is easily calibrated using the digital techniques discussed later in this paper. While many integrated circuit temperature sensors exist [41], this technique provides a direct digital output, very low power dissipation, and can easily be implemented in a standard CMOS process.

Barometric Pressure Sensor

The major challenge in designing the barometric sensor is to achieve very high resolution over a wide dynamic range in both temperature (-20 to +60°C) and pressure (600 to 800Torr). The targeted resolution of 25mTorr here is equivalent to about one foot of altitude shift at sea level and represents 15b. In order to achieve wide dynamic range and high resolution simultaneously, a multi-transducer vacuum-sealed capacitive pressure sensor has been developed [42,43]. This device uses multiple diaphragms to segment the overall pressure range as shown in Fig. 7. The bossed diaphragm diameters vary from 920 to 1100µm at a thickness of 2.4µm and a gap separation of 9.8µm. The sensor is fabricated using bulk micromachining and a silicon-glass dissolved-wafer process [43]. In order to measure barometric pressure, the microcontroller first reads out the smallest of the diaphragms, which serves as a global sensor spanning the entire measurement range at relatively low resolution. After determining the approximate pressure, the controller then selects the appropriate segment transducer, each of which covers a range of about 50Torr. As pressure increases, the diaphragm deflection varies according to the fourth power of its diameter, causing the largest of the diaphragms to approach the glass first. As the gap approaches zero, a very high pressure sensitivity is achieved. When one diaphragm touches the glass, it is strain relieved, and the next diaphragm in sequence becomes the measurement device of interest. The transducer capacitance is read out using the switched-capacitor charge integrator on the interface chip. The typical pressure sensitivity of this device is 27fF/Torr (3000ppm/Torr) for a resolution of 25mTorr with accuracy limited by nonlinearity and temperature sensitivity. Before compensation, the temperature coefficient of offset (TCO) of the transducer alone at 750Torr is 3900ppm/°C and the temperature coefficient of the pressure sensitivity (TCS) is about 1000ppm/°C. After compensation, the worst case TCO for the entire system is 40ppm/°C.

Relative Humidity Sensor

Another environmental parameter of interest is relative humidity (RH). A humidity sensor with small size (<0.5cm²), broad dynamic range (25-90% RH), high resolution (±2% RH), a low temperature coefficient, and low power consumption is necessary to accurately sense relative humidity changes in many µCluster applications. By applying high-aspect-ratio micromolding [44] and electroplating technologies, a capacitive hygrometer with the structure shown in Fig. 8 has been fabricated. The spaces between a series of interdigitated electrodes are filled with a thin polymer film material. The dielectric constant of the polymer varies as a function of moisture, causing the capacitance between the two electrodes to change with ambient relative humidity. This capacitance change can be read out by the same interface circuit used with the other transducers on the µCluster. From a variety of humidity sensitive polymer materials, the polyimide DuPont PI 2723 has been selected because its photosensitivity simplifies the fabrication process while it provides good thermal stability, a linear response to humidity, and high-speed moisture absorption. Compared to other humidity sensors [45], this device utilizes high-aspect-ratio electrodes in order to obtain high sensitivity in a small area.

The hygrometer has been tested over a wide range of ambient relative humidity, temperature, and measurement frequency, and the results are shown in Fig. 9. This sensor, with a base capacitance around 6pF, exhibits a sensitivity of 4.5fF/%RH with good linearity over the desired humidity range. The response time is about 30sec, which is a limiting aspect of the current device. When used with the sensor interface circuit described above, the output voltage shows a linear response, a sensitivity of 9.6mV/%RH, and a resolution of better than 2%RH. Although the polyimide alone has only a slight temperature dependence (20-50ppm/°C), the overall sensor output shows a temperature sensitivity of approximately 30mV/°C. This temperature sensitivity is due largely to the fact that at higher temperatures there is more moisture in the air at the point of saturation than at lower temperatures with the same relative humidity. Thus, the capacitance of the hygrometer increases with temperature even when the humidity level is held constant. Although this effect limits the usable gain of the readout circuit, the shifts are predictable and can be digitally compensated. Current research efforts to improve this humidity sensor are focused on reducing the response time of the device and reducing moisture-induced charge leakage at the sensor-readout interface.

Integrated Accelerometers

The µCluster incorporates sensors for the measurement of acceleration and/or vibration as well as a low-power programmable threshold accelerometer array that provides an event-triggered wake-up call when the system is in sleep-mode or an interrupt when it is in the normal mode of operation.

A linear torsional accelerometer is used for z-axis measurements [46]. This device, shown in Fig. 10, uses interdigitated electrode fingers and a capacitive overlap area that varies with acceleration to achieve high linearity, a high pull-in voltage, and low damping. The device is made very sensitive by using a torsional beam suspension and a very narrow air gap without limiting the measurement range. Figure 10 also shows SEM views of this device. The sensor is fabricated using a three-mask dissolved-wafer process [47]. It consists of a 12µm thick boron-doped silicon inertial mass suspended 7.5µm above the glass substrate by two narrow, high-aspect-ratio 12µm x 3µm torsion beams anchored to the glass substrate. A large number of 300µm long capacitive sense fingers are attached to the end of the proof mass, opposite the support beams. These fingers are separated by a 2µm air gap from fixed fingers anchored to the substrate. The accelerometer provides a sensitivity of 20fF/g over a range of ±4g, and a bandwidth of 30Hz when packaged at atmosphere. The sensor is read out using the sensor interface chip discussed above to provide a sensitivity of 300mV/g. The fabrication process is very simple, while the device structure provides a large bandwidth without special packaging or damping control, does not rely on a large-area parallel-plate gap (which helps circumvent problems with stiction), and has a high sensitivity. In addition, its process is compatible with that of the barometric pressure sensor, allowing its integration with other sensors into a single chip.

Although continuous measurement of acceleration can be performed using the above accelerometer (e.g., to track position changes), it is also desirable to measure single acceleration events (impacts) over a much larger range using an array of acceleration-activated switches. Each of the devices in the threshold accelerometer array functions as a switch that is triggered by an acceleration greater than the threshold level of the sensor [48]. Each sensor consists of an inertial mass, which is suspended by a compliant spring and is separated from the support frame by an air gap as shown in Fig. 11. A shock or acceleration greater than the threshold of the device causes the mass to deflect, making contact with the frame. A metal strip on the bottom surface of the mass shorts metal contacts on the frame and closes a sense circuit, acting as an event-triggered switch. The threshold level is a function of the mass of the inertial element, the spring constant of the suspension, and the air-gap dimension separating the contacts on the mass from the frame. By altering one or more of these design parameters, different thresholds can be achieved. The accelerometer array is also fabricated using the dissolved wafer process [47] and uses a p⁺⁺ inertial mass suspended by an oxide suspension beam anchored to the glass substrate. The device employs gold contacts on both the p⁺⁺ mass and the glass substrate, as shown in Fig. 11. The array used in the µCluster contains devices with seven different threshold levels covering a range of 1.5g to 1000g [48].

The array is interfaced to a micropower circuit chip capable of sensing switch closure and alerting the microcontroller of the occurrence of an event. Because of extremely low standby power, the array and its interface chip can be used for continuous monitoring of environmental shock or vibration while the remainder of the microsystem is in sleep mode. Figure 11 also shows a block diagram of the interface circuit [48]. Activation of an element of the sensor array causes an instantaneous high to low transition at the input of the corresponding interface channel, which is amplified and latched. Higher reliability and fault-tolerance is achieved by using channels with triple redundancy. Each of these redundant channels consists of three individual sense channels and a majority voter. The interface chip can also be programmed to set the overall threshold level of the array by enabling the desired elements. All the control and address signals of the chip are issued by the system microcontroller through the sensor bus and the bus interface chip. This chip has been fabricated in a standard 3µm single-metal double-poly p-well process. It measures 2.2mm x 2.2mm, dissipates less than 300µW from a 5V supply, and can operate up to 4MHz [48]. Figure 11 shows the hybrid assembly of the sensor and circuit chips.

Software Features

Having a microprocessor embedded within the microsystem provides several advantages, including the ability to alter system operation quickly and easily through changes in software and the ability to perform digital signal processing on the sensor data. Additionally, sensor-specific software routines and/or measurement parameters can be downloaded to the microsystem, tailoring operation to the set of transducers being utilized. An extension of this concept is to provide a set of test routines that allow each device to be tested within the system to calibrate the entire data path, from the sensor to the digital data stored in the controller. This feature is especially useful for obtaining high accuracy in conjunction with the highly-programmable bus interface chip presented above. The ability to test and calibrate the entire microsystem in closed-loop fashion not only improves performance but also simplifies final testing.

In the µCluster, the main control routines are stored in EPROM with several variables allocated to EEPROM so that various operation sequences can be selected for different applications. EEPROM also contains a block of code for each sensor which determines how it is read out and how often, how it can be self-tested, how its data should be processed (including parameters for digital compensation), and what to do if the sensor does not appear to be functioning correctly. RAM is used for program variables and the temporary storage of sensor data. The basic operating sequence for the µCluster is to scan each front-end device, calibrate (and compensate) the incoming data using one of several algorithms, output the data through the wireless transmitter, program the PMC for the desired sleep interval, shut down most front-end devices and the RF transmitter, and enter sleep mode. During sleep mode the PMC will monitor any front-end devices that remain active and wake the system when the sleep interval has expired to initiate another scan. To obtain the data needed while minimizing the amount of energy taken from the battery supply, the sensor scan rate (or sleep mode duration) is adaptively set based on the measurements taken. That is, when the parameters being measured are changing quickly, the scan rate is increased, but when the measured parameters appear to be steady, a very slow scan rate is used. The ability to interrupt sleep mode for an event-triggered response, as in the threshold accelerometer, is an important feature for low-power systems, providing a means for monitoring desired parameters while much of the system is shut down.

Another important software feature that affects overall performance is the digital compensation of the sensor data. For microsystems, the built-in digital signal processing capabilities can be used to eliminate complex calibration circuitry and achieve as much as two orders of magnitude of error reduction over circuit techniques and laser trims, which have difficulty following nonlinearities [49-51]. Look-up tables and polynomial evaluation are the two most common methods [52] for digital compensation, although a combined piecewise polynomial approach has also been demonstrated [53]. Look-up table algorithms [54] offer good accuracy but are very demanding on system memory. An attractive alternative is polynomial evaluation [54-55], which uses significantly less memory than look-up table methods but is generally slower, with speed set by the operating frequency of the MCU. Polynomial evaluation also makes it possible to perform both calibration (offset/slope) and compensation (e.g., for temperature) simultaneously using bivariate equations and can produce highly accurate results even for sensors with a very nonlinear response.

On the µCluster, a variety of standard calibration/compensation algorithms were established, including look-up tables, bivariate polynomials, and piecewise polynomials. The two most interesting examples are calibration of the highly nonlinear temperature sensor and compensation of the pressure sensor. The temperature sensor on the µCluster has a response that is an inverse log function, which is very difficult to fit accurately even with high-order polynomials. However, if the response is broken into segments, the device can be accurately described by four second-order equations each spanning 20°C. With similar accuracy, the device can be fit by eight first-order equations covering 10°C as shown in Fig. 12. Here, the response of the sensor is shown along with each of the linear approximations used to calibrate it. For the same error, this piecewise polynomial approach allows the device to be calibrated about 100 times faster than a single fifth-order polynomial and requires only two-and-one-half times the memory.

The barometric pressure sensor uses a bivariate polynomial for calibration and for compensation of temperature sensitivity. Considering the trade-offs among accuracy, speed, and required memory, the fourth-order bivariate polynomial

Pressure = [b₀ ± b₁(V) ± b₂(V²) ± b₃(V³) ± b₄(V⁴)

± b₅(T) ± b₆(T²) ±b₇(T³) ± b₈(T⁴)

±b₉(TV) ± b₁₀(T²V) ±b₁₁(TV²) ±b₁₂(TV³) ± b₁₃(T³V) ± b₁₄(T²V²)]

was implemented, where V is the sensor output voltage after analog-to-digital conversion and T is the measured temperature. Using floating-point software routines, this equation can be evaluated in 8.3msec and the fifteen coefficients, b_i, can be stored in 45 bytes of MCU memory. For similar accuracy with a look-up table, approximately 9kB of the 13kB available on the MCU would be required, although results could be obtained at least ten times faster. Figure 13 shows the output voltage measured from one of the barometric segment transducers at two temperatures (right axis) and the results after compensation (left axis), which match very closely with the actual pressure. By compensating for nonlinearity and temperature sensitivity, each 50Torr sensing element is accurate to within ±100mTorr in the range of 600-800Torr. Accuracy approaching the 25mTorr device resolution can be obtained using higher order polynomials, but at substantial cost in memory and evaluation time.

Results

As an example of a portable wireless microsystem, the µCluster has provided a great deal of information that helps identify the strengths and weaknesses of present technologies and this approach in general. Table 1 lists a number of important parameters along with the results obtained from the µCluster and some longer-term goals set at the beginning of this paper. Although some of the longer-term goals have not yet been reached, the µCluster was successful in demonstrating the feasibility of a generic microsystem architecture and features such as sensor bus communication, use of a programmable generic interface circuit, and in-module digital calibration and compensation for several sensors. A variety of power management techniques have also been implemented. While the power dissipation goal of less than 1mW was achieved, this was only possible at slow sensor scan rates using an RF transmitter with a limited range. Figure 14 plots the power consumption of the µCluster versus scan interval and shows that the total average power of this configuration is dominated by the MCU. A sharp drop in the overall power consumption can be seen as the scan interval approaches one minute. For longer intervals, a limit of about 400µW is reached where power consumption is limited by the sleep-mode current of the MCU and, to a lesser degree, the continuous supply to the power management components. In comparison, the power consumption of the transmitter and the front-end devices is negligible, although at very fast scan rates the transmitter begins to play a larger role. Analysis of the same system with a 3.3V supply shows a factor of two improvement, but any further reduction in power consumption can only be achieved by integrating power management functions directly on the MCU to minimize all standby currents. As the performance demands for microsystems increase, micropower operation will be an important continuing challenge.

For microsystems to approach the 1cc wristwatch goal listed in Table 1, packaging technology appears to be the greatest issue. Unpackaged, the µCluster components and mounting substrate occupy 10cc while a packaged module with batteries fills 50cc. Considering the components themselves, only 50% combined volume is occupied by control, communication, and front-end devices while the remaining 50% is used for components such as the hardwired I/O connector and the on/off switch, which could be made smaller with improvements in packaging. Battery size is also a significant concern, since a single 3V lithium coin cell with sufficient energy for microsystem applications occupies about 1.6cc by itself.

In constructing the present µCluster, the components were mounted on a custom three-layer printed circuit board as shown in Fig. 15. One version of the µCluster has been used in a number of artillery exercises with the U.S. Marine Corps, where it has been benchmarked against existing meteorological equipment. The microsystem has flown on an unmanned aerial vehicle (UAV) at the Naval Research Laboratory, where µCluster data was transmitted through a satellite link while the UAV was in flight. An ocean buoy has also been modified for µCluster measurements at the Naval Command, Control, and Ocean Surveillance Center. These experiments have demonstrated the effectiveness of the µCluster and shown the importance of similar microsystems to a variety of emerging applications.

Conclusion

This paper has reported a microinstrumentation system for the measurement of environmental temperature, barometric pressure, humidity, and acceleration. It uses a generic open architecture that permits it to be customized for a given application through the choice of front-end sensors and through the control software resident in the embedded microcontroller. The MCU periodically scans the sensors, calibrates and compensates their data, and communicates the resulting information to the outside world using either a hardwired system bus or a wireless link. The scan rate is programmable and adaptive. An associated power management chip switches the peripheral sensors off when not in use and allows the MCU to operate in a low-power sleep mode between scans. The MCU communicates with the front-end sensors and actuators using a nine-line sensor bus which permits up to 16 different front-end sensing nodes, each with 64 directly addressable elements, to be used in the microsystem. The system thus supports electronic trims, variable readout gain, subranging, and sensor self-test. A generic CMOS readout chip interfaces between the capacitive sensors and the sensor bus. Sensor outputs in the form of frequency-encoded digital waveforms and analog-amplitude formats are accommodated.

The microsystem module operates over an ambient temperature range from -20°C to +60°C with a power dissipation of 530µW at a typical sensor scan interval of 1min. Digital compensation within the µCluster improves sensor performance significantly compared to analog trims. Temperature is measured with an accuracy of 1°C, pressure to ±100mTorr over the range from 600 to 800Torr, and humidity to ±2%RH over the 25-90%RH range. Acceleration is measured ±0.1g over the 0 to ±4g range while an array of threshold accelerometers continuously monitors for impact events with threshold levels from 1.5g to 1000g. The microsystem occupies an unpackaged volume of only 10cc. It operates from two 3V lithium coin cells (275mA-hr) with a life of up to 1 month and with a communication range of 50m. Such microsystems should continue to improve significantly in size and performance during the coming decade and promise to see wide application in a broad array of important products.

Acknowledgments: The authors would like to thank Wayne G. Baer, Arjun Selvakumar, Uksong Kang, Richard De Souza, Chingwen Yeh, and Farrokh Ayazi for their assistance in various aspects of this project. The efforts of Charles M. Kiers (Navy Science Assistance Program) and Robert R. Whitlock (Naval Research Laboratory) in testing and demonstrating the µCluster in the field are also gratefully acknowledged.

References

[1] J. H. Huijsing, “Integrated Smart Sensors,” Sensors and Actuators A, 30(1-2), pp. 167-174, 1992.

[2] K. D. Wise, “Microelectromechanical Systems: Interfacing Electronics to a Non-Electronic World,” Digest, IEEE Int. Electron Device Meeting, pp. 11-18, December 1996.

[3] J. Bryzek, “MEMS: A Closer Look,” Sensors Magazine, pp. 4-9, July 1996.

[4] L. Spangler and C. J. Kemp, “A Smart Automotive Accelerometer with On-Chip Airbag Deployment Circuits,” Digest, Solid-State Sensor and Actuator Workshop, Hilton Head Island SC, pp. 211-214, June 1996.

[5] H. Baltes, H. Haberli, P. Malcovati and F. Maloberti, “Smart Sensor Interfaces,” Digest, IEEE Int. Symposium on Circ. and Systems, Atlanta GA, vol. 4, pp. 380-383, May 1996.

[6] E. Yoon and K. D. Wise, “An Integrated Mass Flow Sensor with On-Chip CMOS Interface Circuitry,” IEEE Trans. Electron Devices, vol. 39 (6), pp. 1376-1386, June 1992.

[7] S. Middlehoek and S. A. Audet, Silicon Sensors, Academic Press, London, 1989.

[8] K. D. Wise, “Integrated Microsystems: Device and Technology Challenges,” (Invited Plenary), Proc., European Solid-State Device Research Conf. (ESSDERC), The Hague, pp. 15-24, September 1995.

[9] A. Mason, N. Yazdi, K. Najafi, K. Wise, “A Low-Power Wireless Microinstrumentation System for Environmental Monitoring”, Digest, Int. Conf. on Sensors and Actuators (Transducers’ 95), Stockholm Sweden, pp. 107-110, June 1995.

[10] B. J. Hosticka, “CMOS Sensor Systems,” Digest, Int. Conf. on Solid-State Sensors and Actuators (Transducers ‘97), Chicago, pp. 991-993, June 1997.

[11] K. D. Wise, “Integrated Microinstrumentation Systems: Smart Peripherals for Distributed Sensing and Control,” Digest, IEEE Int. Solid-State Circ. Conf., San Francisco, pp. 126-127, February 1993.

[12] C. J. Koomen, “Technologies for the Multimedia City,” (Invited Plenary), Proc., European Solid-State Device Research Conf. (ESSDERC), The Hague, pp. 25-36, September 1995.

[13] J. E. Brignell, “Sensors in Distributed Instrumentation Systems,” Sensors and Actuators, 10(3-4), pp. 249-261, 1986.

[14] K. D. Wise, “Device and Technology Challenges for Integrated Sensors, (Invited Plenary), IEEE Device Research Conf., Charlottesville VA, June 1995.

[15] K. D. Wise and N. Najafi, “The Coming Opportunities in Microsensor Systems,” (Invited) Digest, IEEE Int. Conf. on Solid-State Sensors and Actuators (Transducers ‘91), San Francisco, pp. 2-7, June 1991.

[16] K. M. Mahmoud, R. P. van Kampen, M. J. Rutka, and R. F. Wolffenbuttel, “A Silicon Integrated Smart Pressure Sensor,” Digest, Int. Conf. on Solid-State Sensors and Actuators (Transducers ‘93), Yokohama, Japan, pp. 217-220, June 1993.

[17] E. J. Hogenbirk, H. J. Verhoeven, and J. H. Huijsing, “An Integrated Smart Sensor for Flow and Temperature with I²C Bus Interface Based on Thermal Sigma-Delta Modulation,” Digest, Int. Conf. on Solid-State Sensors and Actuators (Transducers ‘93), Yokohama, Japan, pp. 792-795, June 1993.

[18] MC68HC11E9 (HCMOS Microcontroller Unit) Technical Data Manual, Motorola Inc., 1991.

[19] A. Chandrakasan, S. Sheng, and R. W. Brodersen, “Low-Power CMOS Digital Design,” IEEE J. Solid State Circ., vol. 27, pp. 473-484, April 1992.

[20] J. D. Meindl, “Low Power Microelectronics: Retrospect and Prospect,” Proc. IEEE, vol. 83(4), pp. 619-635, April 1995.

[21] A. A. Abidi, “Low-Power Radio-Frequency IC’s for Portable Communications,” (Invited), Proc. IEEE, vol. 83(4), pp. 544-569, April 1995.

[22] B. Davari, R. H. Dennard and G. G. Shahidi, “CMOS Scaling for High Performance and Low Power - The Next Ten Years,” (Invited), Proc. IEEE, vol. 83(4), pp. 595-606, April 1995.

[23] R. A. Powers, “Batteries for Low Power Electronics,” (Invited), Proc. IEEE, vol. 83(4), pp. 687-693, April 1995.

[24] P. B. Koeneman, I. J. Busch-Vishniac, and K. L. Wood, “Feasibility of Micro Power Supplies for MEMS,” J. Microelectromechanical Systems, vol 6(4), pp. 355-362, December 1997.

[25] L. E. Larson, RF and Microwave Circuit Design for Wireless Communications, Norwood MA: Artech House, Inc., 1996.

[26] C. T-C. Nguyen, L. P. B. Katehi, and G. M. Rebeiz, “Micromachined Devices for Wireless Communications,”Proc. IEEE, (this issue), June 1998.

[27] K. Bult, A. Burstein, D. Chang, M. Dong, M. Fielding, E. Kruglick, J. Ho, F. Lin, T. H. Lin, W. J. Kaiser, R. Murkai, P. Nelson, F. L. Newburg, K. S. J. Pister, G. Pottie, H. Sanchez, O. M. Stafsudd, K. B. Tan, C. M. Ward, G. Yung, S. Xue, H. Marcy, and J. Yao, “Wireless Integrated Microsensors,” Digest, Solid-State Sensor and Actuator Workshop, Hilton Head, SC, pp. 205-210, June 1996.

[28] A. Kompolt and J. Brewer, “Designing a Wristwatch Pager Creates Its Own Special Challenges,” Wireless Systems Design, pp. 37-41, March 1997.

[29] J. R. Moyne, N. Najafi, D. Judd and A. Stock, “Analysis of Sensor/Actuator Bus Interoperability Standard Alternatives for Semiconductor Manufacturing,” Proc., Sensors Expo, Cleveland OH, pp. 375-386, September 1994.

[30] N. Najafi and K. D. Wise, “An Organization and Interface for Sensor-Driven Semiconductor Process Control Systems,” IEEE Trans. Semiconductor Manuf., vol. 3 (4), pp. 230-238, November 1990.

[31] S. P. Woods, “The IEEE-P1451 Transducer to Microprocessor Interface,” Sensors Magazine, pp. 43-48, June 1996.

[32] L. H. Eccles, “A Brief Description of IEEE P1451.2,” Proc., Sensors Expo, Detriot MI, pp. 81-90, October 1997.

[33] K. Watanabe and W.S. Chung, “A Switched Capacitor Interface for Intelligent Capacitive Transducers,” IEEE Trans. on Instrumentation and Measurement, Vol. IM35(4), pp. 472-476, December 1986.

[34] Y. E. Park and K. D. Wise, “An MOS Switched-Capacitor Readout Amplifier for Capacitive Pressure Sensors,” Proc., IEEE Custom Integrated Circuits Conf., pp. 380-384, May 1983.

[35] Y. Cao and G. C. Temes, “High-Accuracy Circuits for On-Chip Capacitance Ratio Testing or Sensor Readout,” IEEE Trans. on Circuits and Systems II: Analog and Digital Signal Processing, vol. 41(9), pp. 637-639, September 1994.

[36] S. L. Garverick, M.L. Nagy, N.K. Rao, D.K. Hartsfield, and A. Purushotham, “A Capacitive Sensing Integrated Circuit for Detection of Micromotor Critical Angles,” IEEE J. Solid-State Circ., vol. 32(1), pp. 23-30, January 1997.

[37] CSEM2003 (Capacitive Sensor Interface with Gain and Offset Adjustment) data sheet, CSEM (Centre Suisse d’Electronique et de Microtechnique SA), 1995.

[38] ISC721 (Capacitive Sensor Interface Circuit) preliminary data sheet, Irvine Sensors Corporation, 1997.

[39] N. Yazdi, A. Mason, K. Najafi, K. Wise, “A Low-Power Generic Interface Circuit for Capacitive Sensors,” Digest, Solid-State Sensor and Actuator Workshop, Hilton Head Island SC, pp. 215-218, June 1996.

[40] E. Vittoz and O. Neyroud, “A Low-Voltage CMOS Bandgap Reference,” IEEE J. Solid-State Circ., vol. SC-14(3), pp. 573-577, June 1979.

[41] G. C. M. Meijer, “Thermal Sensors Based on Transistors,” Sensors and Actuators, 10, pp. 103-125, 1986.

[42] Y. Zhang and K. D. Wise, “A High-Accuracy Multi-Element Silicon Barometric Pressure Sensor,” Digest, Int. Conf. on Sensors and Actuators (Transducer ‘95), Stockholm Sweden, pp. 608-611, June 1995.

[43] A. V. Chavan, K. D. Wise, “A Batch-Processed Vacuum-Sealed Capacitive Pressure Sensor,” Digest, Int. Conf. on Solid-State Sensors and Actuators (Transducers ‘97), Chicago, pp. 1449-1452, June 1997.

[44] A. B. Frazier and M. G. Allen, “Metallic Microstructures Fabricated Using Photosensitive Polyimide Electroplating Molds,” IEEE J. Microelectromechanical Systems, vol. 2, pp. 87-94, 1993.

[45] N.Yamazoe, "Humidity Sensors: Principles and Applications," Sensors and Actuators, vol.10, pp.379-398, 1986.

[46] A. Selvakumar, F. Ayazi and K. Najafi, “A High Sensitivity Z-Axis Torsional Silicon Accelerometer,” Digest, IEEE Int. Electron Device Meeting, San Francisco CA, pp. 765-768, December 1996.

[47] Y. Gianchandani and K. Najafi, "A Bulk Silicon Dissolved Wafer Process for Microelectromechanical Devices," J. of Microelectromechanical Systems, vol. 1(3), pp. 77-85, June 1992.

[48] A. Selvakumar, N. Yazdi and K. Najafi, “A Low Power, Wide Range Threshold Acceleration Sensing System,” Proc., IEEE Microelectromechanical Systems Workshop (MEMS 96), San Diego, CA, pp. 186-191, February 1996.

[49] O. Machul, D. Hammerschmidt, W. Brockherde, B.J. Hosticka, E. Obermeier, and P. Krause, “A Smart Pressure Transducer with On-Chip Readout, Calibration and Nonlinear Temperature Compensation Based on Spline-Functions,” Digest, IEEE Int. Solid-State Circ. Conf., pp. 198-199, February 1997.

[50] J. A. Lloyd, H-S. Lee, L. Parameswaran, M. A. Schmidt, and C. G. Sodini, “An Adaptive Calibration Technique for Micromachined Pressure Sensors,” Digest, Int. Conf. on Solid-State Sensors and Actuators (Transducers ‘97), Chicago IL, pp. 295-298, June 1997.

[51] J. Bryzek, “Modeling Sensor Performance for Smart Transducers,” Proc., Sensors Expo, Cleveland OH, pp. 501-507, September 1994.

[52] X. Li, G. C. M. Meijer, and G. W. de Jong, “A Self-Calibration Technique for a Smart Capacitive Angular-Position Sensor,” IEEE Instr. and Meas. Tech. Conf., Brussels Belgium, pp. 774-777, June 1996.

[53] P. T. P. Tang, “Table-Lookup Algorithms for Elementary Functions and Their Error Analysis,” Proc., IEEE Symp. on Computer Arithmetic, pp. 232-236, June 1991.

[54] S. B. Crary, W. G. Baer, J. C. Cowles, and K. D. Wise, “Digital Compensation of High-performance Silicon -Pressure Transducers,” Sensors and Actuators, A21-A23, pp. 70-72, 1990.

[55] Y. Yoshii, A. Nakajo, H. Abe, K. Ninomiya, H. Miyashita, N. Sakuri, M. Kosuge, S. Hao, “1 Chip Integrated Software Calibrated CMOS Pressure Sensor with MCU, A/D Convertor, D/A Convertor, Digital Communication Port, Signal Conditioning Circuit and Temperature Sensor,” Digest, Int. Conf. on Solid-State Sensors and Actuators (Transducers ‘97), Chicago IL, pp. 1485-1488, June 1997.

Table 1: µCluster Characteristics and Eventual Goals.

	Parameter	Result		Eventual Goal
Microsystem Volume (with PCB)			10 cc		1 cc
Packaged Volume (with batteries)			50 cc		10cc
Power Supply			External (hardwired)		3V coin cell
			(or) 2, 3V/275mA-hr coin cell batteries
			(or) Standard 9V battery
Telemetry Range			50m		1 mile
Telemetry Frequency / Modulation			315 MHz / ASK		GHz
Average Power Dissipation*			530µWatt		<1mWatt
Portable Operating Life*			30 days		30 days
Measurement Aperture Time			70µsec		50µsec
Maximum Scan Rate (with output)			10 Hz		1kHz
Sensor Scan Interval			1 minute (typical)
			Adaptive and Event Triggered
Temperature			-20 to +60ºC ±0.5ºC		-20 to +60ºC ±0.2ºC
Barometric Pressure			600 to 800 Torr ±100mTorr		500 to 800Torr ±25mTorr
Relative Humidity			25 to 90%RH ±2%RH		15 to 95%RH ±2%RH
Vibration/Acceleration			-4 to +4g ±0.1g		-4 to +4g ±1mg
*Operating Mode Dependent