A 1.7mm3 MEMS-on-CMOS tactile sensor using human-inspired autonomous common bus communication

M. Makihata, M. Muroyama, Y. Nakano, S. Tanaka, T. Nakayama, U. Yamaguchi, H. Yamada, Y. Nonomura, H. Funabashi, Y. Hata, M. Esashi
2013 2013 Transducers & Eurosensors XXVII: The 17th International Conference on Solid-State Sensors, Actuators and Microsystems (TRANSDUCERS & EUROSENSORS XXVII)  
A bus-connected tactile sensor system composed of MEMS-CMOS integrated force sensors was developed. A capacitor-to-digital convertor for force sensing, a data reduction processor and a serial bus communication controller are implemented by a laboratory-designed integrated ASIC. These functions enable the tactile sensor to be connected serially, and to autonomously transmit the sensing data using CSMA (Carrier Sense Multiple Access) protocol (Fig. 1) . A novel MEMS-CMOS integration technology[1]
more » ... ation technology[1] was applied, so that the integrated tactile sensor can be directly mounted on a flexible printed circuit board. The chip size is 2.54mm × 2.54mm × 0.27mm, i.e. 1.7mm 3 , and there are 16 through silicon interconnection using the lateral tapered groove (Fig. 2) . The digital data from the completed tactile sensor contains 32-bits force sensing data, which corresponds to an external force linearly with 8bit/N resolution at 6.25Hz sampling (Fig. 3 ). Since the drastic decrease of data collision rate was expected, human-inspired threshold-based operation and adaptation were carried out to reduce the non-critical tactile data (Fig. 4) . Finally, the serial bus network was demonstrated to estimate the network performance ( Fig. 5 ) Figure 1 shows the overview of the tactile sensor system. The flexible cable is a common bus line composed of 2 power lines (V18, V33), 1 signal line (DATA) and 1 ground line (GND). Reducing wires and data quantity are critical problems to achieve the whole body tactile sensation for robot [2]. To address the problems, we adopted a human-inspired autonomous common bus communication. Figure 2 shows the fabricated integrated device. The surface mountable tactile sensor were prototyped by a fabrication process using "through silicon groove" interconnection technology and wafer bonding. Figure 3 shows the measured output signals from the completed integrated device. The data field includes converted force data. Since force-to-capacitance conversion by the MEMS and capacitance-to-frequency conversion by the circuit canceled their parabolic characteristic of transduction each other, the linear response of the digital output with external force was obtained. This result well agrees with mechanical and electrical simulations. The deformation of the diaphragm was 300nm and capacitance change was 10fF, when applying 1N normal force. The demonstration of data reduction was examined using a chip as shown in Fig. 4 . After the tactile sensor was initialized, data was transmitted continuously at a frequency of 5.5kHz ( Fig. 4 (a) ). By the threshold operation, the number of the tactile data was reduced by 59% ( Fig. 4 (b) ). In the threshold-based operation, weak force data below the pre-defined threshold was filtered and not transmitted on the bus. A further data reduction of 37.4% (i.e. 96.7% in total) was achieved in conjunction with adaptation ( Fig. 4 (c) ). By the adaptation, transmission interval time was lineally increasing during applying constant strong forces over the threshold value. Although autonomous bus communication using CSMA protocol improves response time to external forces compared with the polling protocol, the packet collision will become a critical problem as the packet quantity increases. Therefore, the collision rate was experimentally measured in a small system composed of the three sensor nodes. Figure 5 shows the collision rate and throughput as a function of the packet generation rate, which are calculated from the acquired bus signals. In this experiment, every sensor node transmits the signals without the data reduction. The highest data throughput (73%, 1Mbps) was obtained when all of the sensors transmit each data at 4.6kHz. We obtained the fact that the collision rate is negligible when the throughput is less than about 50%. Since 90% reduction of tactile data can be achieved by the adaption like Fig.4 , both of low-collision and fast response data transmission can be realized in this system. REFERENCES: [1] M.
doi:10.1109/transducers.2013.6627370 fatcat:nd53y2slt5ffxnahjvmdkkjwfe