As noted above, it is evident that there is considerable effort directed at sensor-circuit integration in Japan, especially in connection with automotive and medical applications. Automotive is the current principal driver, with medical a possible emerging force during the coming decade. The evolution to "instruments on a chip" is viewed in Japan as likely in the longer term, with nearer-term efforts probably limited to sensing systems and actuators used primarily in connection with the sensors (as in force-balanced accelerometer structures and thermally-based flowmeters). The combination of sensors, actuators, and electronics to do control is viewed as most likely to proceed first in optical devices, due to the compatibility of such systems with hermetic packaging and low drive forces. Both of these viewpoints seem much in line with thinking in the United States.
The JTEC panel's Japanese hosts expressed the view that systems will be primarily hybrid at first, evolving slowly to monolithic implementations as power levels and device process sequences permit. The elimination of parasitics was viewed as a significant motivation for monolithic devices, with integration levels expected to remain modest (less than a few thousand devices) during the remainder of this decade. Embedded microcontrollers were viewed as very likely in high-end devices (and indeed are already present in devices such as the DPharp), migrating into low-end devices as cost allows. It was noted that such devices may be required in order to achieve compatibility with distributed bus-organized systems. Bus standards were viewed as important and likely for automotive applications, but do not seem to be a significant focus yet for other areas.
The question of sensor calibration is clearly one being pursued through a number of approaches, both in hardware and software, and no clear preferences were evident in the JTEC site visits. Some researchers felt that calibration/compensation was better done in hardware, and that the issue was either not very important or not a significant effort in MEMS. Others felt that it was an important issue, particularly with regard to system partitioning, and that it was likely to be done in software in the future. It is expected that industry will converge around a few standard MEMS processes (or at least this seems to be a desirable goal), and that this will be determined in part by the way systems are partitioned in the future.