Most sensor-circuit integration in Japan is probably still hybrid, as it is in the United States. Researchers in Japan have been in the forefront of monolithic sensor-circuit integration for many years, however, and the push toward monolithic system implementations is continuing, especially in areas such as automotive, where volumes are high enough to recover the costs of advanced process development and where very high reliability is required. There is also considerable interest in monolithic integration for medical devices, where small overall size is a principal development goal. The situation is virtually the same in the United States. This review of efforts in Japan will revisit some of the examples used in Chapter 3, but with greater emphasis on the levels of integration involved and the circuit functions performed.
One of the first sensors with on-chip electronics was produced by Toyota in 1984 (Sugiyama et al. 1983) and is shown in Figure 5.6. It combines a piezoresistive pressure sensor with on-chip bipolar circuitry to produce outputs encoded as voltage amplitude (1 to 4 V) or frequency (210 kHz to 240 kHz) for a pressure range of 0 to 750 mmHg. The readout circuitry is used to temperature compensate the output signal to a level of less than 0.06%/°C. The chip size is 3 mm x 3.8 mm. That same year, a piezoresistive pressure sensor with on-chip electronics was also reported by Hitachi (Yamada et al. 1983).
Figure 5.6. Piezoresistive pressure sensor with on-chip bipolar readout electronics reported by Toyota in 1983, one of the earliest examples of an integrated MEMS sensor with integrated electronics.
The use of on-chip electronics with pressure sensors is well illustrated in a series of papers from Toyota that were reported between 1986 and 1993. Figure 5.7 shows the diagram of a 32 x 32-element array of pressure sensors used as a tactile imager. The array is organized in x-y fashion as in a memory, each cell containing a full piezoresistive pressure sensor along with decoding electronics to allow the cell to be read out to a differential pair of analog signal leads. As first reported (Sugiyama et al. 1986), the device used a bulk micromachined cavity formed by undercutting from the front as shown in Figure 5.8. The cavity was vacuum sealed by depositing a CVD dielectric over the etch access holes; however, it was found to be difficult to get high yield from this structure due to the relatively large cavity produced and the associated production of hydrogen bubbles by the etch, which tended to become trapped due to the small access holes. More recent devices have used a surface micromachined structure, also shown in Figure 5.8, in which a thin film of sacrificial polysilicon is laterally etched away to produce the cavity. This produces a much shallower cavity, but also one that is higher in yield. The diaphragms are 100 mm x 100 mm in size, with a cell pitch of 250 mm and an overall chip size of 10 mm x 10 mm. The force sensitivity is about 100 mV/g-mm(2). The circuitry here is primarily utilized to provide on-chip addressing capability, and consists of about 16,000 MOS transistors, although an on-chip amplifier is also included. Similar pressure sensors (Sugiyama et al. 1986) with diaphragms 100 mm in diameter have produced a sensitivity of 10 mV/V/kPa.
Figure 5.7. Organization of the Toyota tactile imager readout electronics. The device implements a 32 x 32-element array of pressure sensors with on-chip selection electronics. The overall die size is 10 mm x 10 mm.
Figure 5.8. Pressure sensor evolution in the Toyota tactile imager from a bulk micromachined device (left) to a surface-micromachined structure (right). Both devices are vacuum sealed at wafer level.
Digital compensation has begun to be reported in a few efforts worldwide in which either computed polynomials (Najafi et al. 1988) are used or shift registers are loaded and used to produce analog trim voltages on-chip, using digital-to-analog converters (Hammerschmidt et al. 1993). In the former case, the compensation is done in software in the digital domain, whereas in the latter the signal path remains analog. In a recent paper, researchers at Toyoda Machine Works and Toyota have reported a digitally-compensated capacitive pressure sensor with on-chip CMOS circuitry (Nagata et al. 1992). The sensor works as a two-chip hybrid with a limited amount of electronics on the sensor and more extensive electronics on the processing chip. The sensor chip encodes the capacitance as an output frequency that is temperature compensated by adjusting the charging current of the oscillator. The span and offset are compensated using the external digital interface chip to adjust the timing window width and preset count. This approach implements compensation in the digital domain, but keeps it in hardware. The circuit organization for this device is shown in Figure 5.9.
Figure 5.9. Circuit organization in the digitally-compensated pressure sensor from Toyoda Machine Works and Toyota.
While the Toyota activities probably represent the most aggressive efforts in Japan at monolithic sensor-circuit integration, there are several other efforts that should be mentioned. Yokogawa Electric Corporation is a leading manufacturer of instrumentation, measurement equipment, and industrial process controls. Yokogawa manufactures flowmeters, pressure sensors, power meters, valves, and many other items of interest. In the area of solid-state devices (MEMS), the company produces a piezoresistive pressure sensor and a high-performance resonant pressure sensor, the latter being of considerable interest. The DPharp pressure sensor (Ikeda et al. 1990) consists of a silicon diaphragm in which an H-shaped silicon resonator is embedded. The resonator is formed within its own vacuum-sealed cavity by a series of four selective silicon epitaxial growth steps. Figure 5.10 shows a view of an exposed beam and the process sequence used to produce it. The resonator is excited by the application of an alternating current in a DC magnetic biasing field, with a variable-gain AGC amplifier used to maintain a constant vibration amplitude, as noted in Figure 5.10. Two resonators are used, positioned on the diaphragm so that they respond differentially, increasing and decreasing their resonant frequencies as the diaphragm stress changes in response to externally applied pressure. The resonator beam is 5 mm thick and 500 mm long. Figure 5.11 shows the circuit configuration used with this device (Saigusa et al. 1992). There is no on-chip circuitry; however, a hybrid external drive chip provides the necessary drive signals. Temperature is also measured within the DPharp module, and an EPROM is used to store sensor parameters for external compensation in an accompanying converter section that contains a microprocessor. This organization, while hybrid and not yet highly integrated, is very similar to the configuration of Part c of Figure 5.1. The DPharp is arranged to cover ranges of 0 to 400, 0 to 1,000, and 0 to 13,000 mmH(¯2)O, meeting an accuracy specification of ±0.1% for each range, and with typical published errors well below this level. This device represents a state-of-the-art high-end pressure sensor, and appears to be evolving along the track shown in Figure 5.11.
Hitachi has recently (Suzuki et al. 1991) developed a bulk-micromachined force-balanced accelerometer for automotive applications that uses the feedback arrangement shown in Figure 5.12. The cantilevered accelerometer is held electrostatically in its neutral position, and the output of the device is taken as the pulse-width modulated output signal. Offset and sensitivity calibrations are implemented using a digital trimming circuit in which a ROM is programmed using the Zener-Zap technique. The accuracy is ±3% over a 0 to ±1 g or 0 to ±2 g measurement range. Again, with this device, the readout circuitry is hybrid.
Seiko Instruments produces a variety of integrated electronic devices for use in equipment ranging from consumer products (watches, portable multilingual dictionaries), information systems (thermal printers, pagers, telephone sets), and various types of production equipment. They include several small micromotors (£1 cmOD, not lithography based). Seiko is clearly interested in the emerging area of MEMS and possible medical products, including portable monitoring devices to allow remote patient monitoring, possibly as part of a worldwide health network. The feasibility of such systems is also being studied in the United States.
Figure 5.10. Structure and excitation scheme used for the Yokogawa DPharp resonant pressure sensor.
Figure 5.11. Module electronics used for signal readout and compensation in DPharp.
Figure 5.12. Circuit organization in the Hitachi force-balanced accelerometer.
The technology levels needed to focus on sensor-circuit integration and system partitioning issues are clearly very high, making it difficult for most universities to contribute in this area. However, it is clear that significant contributions are being made at Tohoku University, which has excellent facilities for both sensor and circuit fabrication. Both capacitive pressure sensors (Matsumoto, Shoji, and Esashi 1990) and capacitive accelerometers (Matsumoto and Esashi 1992) have been reported. The pressure sensor contains a CMOS capacitance to frequency converter and is hermetically sealed using an electrostatic glass to silicon bond. The device has a nominal output frequency of 65 kHz at 760 mmHg, and a sensitivity of about -30 Hz/mmHg. The thermal offset and sensitivity drifts are about 0.05%FS/°C and 0.09%FS/°C, respectively, with a long-term drift of ±2 mmHg/week. The capacitive accelerometer is based on a silicon proof mass suspended by bulk undercut beams. An on-chip phase-locked loop circuit is used for force-balancing. The force-sensitivity is 200 Hz/g, with a nominal output frequency of 70 kHz. The integration levels on both of these chips is modest, probably consisting of less than 100 transistors; however, with the process established, the integration level could be easily pushed higher if desired.