This paper reports a 3-axis fully integrated differential capacitive tactile sensor surface-mountable on a bus line. The sensor integrates a flip-bonded complementary metal-oxide semiconductor (CMOS) with capacitive sensing circuits on a low temperature cofired ceramic (LTCC) interposer with Au through vias by Au-Au thermo-compression bonding. The CMOS circuit and bonding pads on the sensor backside were electrically connected through Au bumps and the LTCC interposer, and the differential capacitive gap was formed by an Au sealing frame. A diaphragm for sensing 3-axis force was formed in the CMOS substrate. The dimensions of the completed sensor are 2.5 mm in width, 2.5 mm in length, and 0.66 mm in thickness. The fabricated sensor output coded 3-axis capacitive sensing data according to applied 3-axis force by three-dimensional (3D)-printed pins. The measured sensitivity was as high as over 34 Count/mN for normal force and 14 to 15 Count/mN for shear force with small noise, which corresponds to less than 1 mN. The hysteresis and the average cross-sensitivity were also found to be less than 2% full scale and 11%, respectively.

Covering a whole surface of a robot with tiny sensors which can measure local pressure and transmit the data through a network is an ideal solution to give an artificial skin to robots to improve a capability of action and safety. The crucial technological barrier is to package force sensor and communication function in a small volume. In this paper, we propose the novel device structure based on a wafer bonding technology to integrate and package capacitive force sensor using silicon diaphragm and an integrated circuit separately manufactured. Unique fabrication processes are developed, such as the feed-through forming using a dicing process, a planarization of the Benzocyclobutene (BCB) polymer filled in the feed-through and a wafer bonding to stack silicon diaphragm onto ASIC (application specific integrated circuit) wafer. The ASIC used in this paper has a capacitance measurement circuit and a digital communication interface mimicking a tactile receptor of a human. We successfully integrated the force sensor and the ASIC into a $2.5\times 2.5\times 0.3$ mm die and confirmed autonomously transmitted packets which contain digital sensing data with the linear force sensitivity of 57,640 Hz/N and 10 mN of data fluctuation. A small stray capacitance of 1.33 pF is achieved by use of 10 $\mathsf{\mu }$m thick BCB isolation layer and this minimum package structure.

Robot tactile sensation can enhance human–robot communication in terms of safety, reliability and accuracy. The final goal of our project is to widely cover a robot body with a large number of tactile sensors, which has significant advantages such as accurate object recognition, high sensitivity and high redundancy. In this study, we developed a multi-sensor system with dedicated Complementary Metal-Oxide-Semiconductor (CMOS) Large-Scale Integration (LSI) circuit chips (referred to as “sensor platform LSI”) as a framework of a serial bus-based tactile sensor network system. The sensor platform LSI supports three types of sensors: an on-chip temperature sensor, off-chip capacitive and resistive tactile sensors, and communicates with a relay node via a bus line. The multi-sensor system was first constructed on a printed circuit board to evaluate basic functions of the sensor platform LSI, such as capacitance-to-digital and resistance-to-digital conversion. Then, two kinds of external sensors, nine sensors in total, were connected to two sensor platform LSIs, and temperature, capacitive and resistive sensing data were acquired simultaneously. Moreover, we fabricated flexible printed circuit cables to demonstrate the multi-sensor system with 15 sensor platform LSIs operating simultaneously, which showed a more realistic implementation in robots. In conclusion, the multi-sensor system with up to 15 sensor platform LSIs on a bus line supporting temperature, capacitive and resistive sensing was successfully demonstrated.

This paper reports a 3-axis MEMS-CMOS integrated tactile sensor for surface-mounting on a flexible bus line. This 3-axis sensor uses a bran-new CMOS LSI with capacitive sensing circuit and other extended functionalities (e.g. configurability and a robust clock data recovery algorithm). The sensor is composed of a flip-bonded CMOS substrate with a sensing diaphragm and a special low temperature co-fired ceramics (LTCC) substrate with vias. These substrates are electrically and mechanically connected by Au-Au bonding, forming sealed differential capacitive gaps. The completed sensor outputs coded 3-axis digital signals according to applied 3-axis force with small cross-sensitivity and hysteresis.

Accurate multi-axis inertial sensing is required for self-positioning estimation in automatic driving and robot posture control. In particular, a small bias drift is an important requirement for these applications. The bias drift is primarily caused by the external stress from the package. To reduce the bias drift in micro electro-mechanical systems (MEMS) sensors, a stress isolation suspension (SIS) structure inside the sensor chip is considered promising for miniaturization and cost reduction. To easily design an effective SIS, we propose a new design procedure based on the topology optimization method (TOM). The number of elements for the TOM in MEMS devices is often large owing to narrow lines and gaps such as release holes and comb-shaped electrodes, which increasing the computational cost of a TOM using an iterative algorithm. Before the TOM iterative calculation, we first identify the dominant external force mode affecting the bias drift using a single finite element analysis (FEA) with precise detailed geometry. Based on this, the TOM model is simplified. This combination of FEA and TOM allows a biasreducing topology-optimized SIS (T-SIS) structure to be designed with reduced computational cost. For validation, we applied the T-SIS design procedure to a capacitive accelerometer for z-axis bias drift reduction and compared its performance to accelerometers without T-SIS. The experimental results indicate that the T-SIS reduces the z-axis bias drift from ±811 mG to ±9.9 mG (temperature range: -30 °C to +70 °C), which is 1/80 of the bias drift of the sensor without T-SIS.