SENSOR-CIRCUIT INTEGRATION AND SYSTEM PARTITIONING

Kensall D. Wise

INTRODUCTION

The development of MEMS in the United States has been primarily lithography-based and has focused on extensions of silicon integrated circuit technology. The efforts have been rooted in electrical engineering, slowly involving more and more individuals from the mechanical engineering community. In Japan, the overall situation has been similar, although several important programs, especially those funded by MITI, have been based in mechanical engineering. As a result, these programs have emphasized nonlithographic technologies and have been somewhat less tied to silicon as a core material. The emphasis on silicon in the United States has made the integration of electronics with sensors a natural progression, and the monolithic merging of sensors, actuators, and interface circuitry has been an important focus for activities for some time. This has also been true in Japan in many companies and in a few universities, but has not been true for the remainder of the Japanese community, which has focused on other areas. This may result in more short-term payoffs from sensor-circuit integration in the United States, but also reflects less work on longer-term areas such as micromachines, for which there are really no comparable U.S. efforts yet.

The integration of interface electronics with transducers, either in monolithic or in hybrid form, is very important, as shown in Table 5.1 (Wise 1993). Integrated sensors typically produce outputs extending down to the microvolt range (limited by noise), and seek to resolve changes in resistance or capacitance of milliohms or femtofarads, or less. The output signals are continuously changing (analog) and are typically sensitive to secondary variables such as temperature, in addition to the primary parameter for which the device was intended. The zero-point outputs (offsets) and the sensitivities (slopes) are typically temperature dependent and may also be nonlinear. Worse still, the outputs at some level will vary over time. From a systems point of view, it would be preferable to deal with signals in the range of tenths of volts to volts, which are robust in the face of environmental factors such as humidity and electromagnetic noise. Since virtually all such parameters will eventually be processed by computer, the final signal should be in a digital format and any secondary-variable sensitivity (e.g., to temperature) should be eliminated along with any nonlinearity problems. Finally, there should be some way of checking the sensor in its operating system to see if the device is still giving correct data (the device should be self-testing over at least a portion of its operating range). Controlling critical industrial process equipment or invasive biomedical systems on the basis of devices that are not testable is clearly undesirable. Thus, the interface circuitry performs an essential role in any complete system.

Table 5.1
The Roles of Integrated Signal Processing Circuitry
in Interfacing With Integrated Transducers and MEMS

In the past most sensors and actuators have been regarded as stand-alone components to be plugged into larger systems. However, the advantages of integrating the sensors and actuators as part of smart modules containing additional electronics are largely lost with this approach. On the other hand, adding electronics to the sensing module makes sense only if it buys increased system performance. Thus, the sensor or actuator must be considered as a system element in order to justify making it smart. This can be difficult to do in large companies and still harder when the sensors are produced as components in a different company altogether.

The need for amplification and multiplexing with sensors has been well accepted for many years, and such circuitry has been added either in monolithic or hybrid form to a great many devices. Figure 5.1 shows a progression of sensor-circuit integration beginning with a transducer having simple in-module signal conditioning (Part a of Figure 5.1). Until recently, this signal conditioning was limited to amplification along with some temperature compensation. The output was one-way analog, with a typical range of 0 to 5 V. This signal fed into a remote analog-to-digital converter (ADC) and microcomputer and then, perhaps, to a hierarchical control system. Some sensors have evolved to the situation in Part b of Figure 5.1, where more than one transducer is located in-module and can be addressed digitally from the processor. This has been true of some array devices and of devices with independent temperature readout. Many arrays are of the type in Part a of Figure 5.1, however, where an on-chip or external clock serially multiplexes the various element signals to the remote processing electronics.

Part c of Figure 5.1 shows a more integrated implementation of the electronics, where the first stage of computer processing and control has been moved to the sensing/MEMS module itself (Najafi and Wise 1990). The module is now truly smart and is able to respond to various commands received externally over a digital sensor bus. It responds digitally with sensing signals that meet all of the desired system characteristics shown in Table 5.1. Using onboard memory, correction for temperature sensitivity and device nonlinearities can be done in software, allowing at least an order of magnitude improvement when compared to hardware trim techniques because of the ability to more precisely fit nonlinear response characteristics using computed polynomials or lookup tables. Such systems are not yet fully realized in the United States, although a number of companies are experimenting with them in various forms. Most involve a number of chips, assembled in hybrid form using surface mount or multichip module (MCM) technology. Within five years, however, many such modules will probably evolve to a two-chip hybrid, with a front-end sensor/actuator/MEMS chip coupled to an embedded microcontroller/microprocessor chip. With the sensing/MEMS node smart, such nodes can also operate autonomously in data gathering and self-testing, storing data in memory, perhaps doing some data interpretation, and responding rapidly to commands received over the external bus.


Figure 5.1. Evolution of sensor-circuit integration and system partitioning. More electronics are being included in the sensor/MEMS module, allowing local signal processing and a digital bus output.

As sensor-circuit integration evolves, the track depicted in Figure 5.1 appears to be evolving in several different industries in the United States. This is true in the automotive industry and in several companies involved with industrial control systems. In both cases, distributed sensing/control is required. SEMATECH is currently working on sensor bus standards for use in the semiconductor process equipment industry, and several bus standards have been proposed for automotive systems. Indeed, bus standards are an important prerequisite to this evolution since the hierarchical control system must be able to work with the sensors as integrated system elements. Avoiding the necessity of point-to-point wiring, realizing a digital output format, and obtaining greater precision are three important goals for such systems. In Japan, the situation seems to be quite similar. Of the companies visited, those involved with automotive systems (Toyota, Nippondenso) or distributed control systems (Yokogawa) were most involved with issues of sensor-circuit integration and system partitioning, whereas the more mechanically-centered organizations were not, either because they have simply chosen to emphasize a different area or because of the relatively high investments required to provide the microelectronics. Table 5.2 lists industrial sites the JTEC panel visited and identifies those thought to be particularly active in this area.

Table 5.2
Industrial Sites Visited by JTEC MEMS Panel


* Asterisk indicates sites where sensor-circuit integration is an important focus.


Published: September 1994; WTEC Hyper-Librarian