Since the earliest days of microelectromechanical systems (MEMS) sensor products, the goal of easily combining the sensing structures made using MEMS processing with the required signal conditioning has driven many sensor R&D labs. If successful, the solution would make the resulting devices more useful in a broader range of applications. In the 1980s, the circuitry was analog and the primary MEMS technology was bulk micromachining for pressure sensors. When accelerometers were required by the automobile industry as crash sensors to trigger airbag deployment (circa 1990s), the key technologies changed. Surface micromachining was used to create the capacitive accelerometers and the signal conditioning circuitry was CMOS.
Today, both bulk and surface micromachining are used for creating sensing structures, but when the lowest power consumption and smallest structures are required, the choice is always surface micromachining. Surface micromachining is also used to create the broadest range of sensing devices including accelerometers, gyroscopes, magnetometers, pressure sensors and more. Since these devices ultimately interface to microcontrollers (MCUs), CMOS circuitry is used for signal conditioning and, if necessary, signal conversion. While CMOS processing has industry standards and is available from several third-party wafer fabs, MEMS processes are not standardized and remain highly unique for many sensor companies. At the same time, many MEMS process steps are not compatible with CMOS processing and require dedicated manufacturing facilities. Consequently, the goal of a wafer fab process capable of making the MEMS structures at the same time as CMOS circuitry without complicating the high-volume CMOS process has continued to be elusive in spite of many different approaches to combine the two. Either additional pre- or post-processing steps are involved in existing processing approaches.
Recently, one company claims to have solved the problem. Nanusens, a sensor company headquartered in England, successfully produced both the MEMS sensor structure and its detection circuitry at the same time within a chip using standard CMOS processes.
Figure 1. (a) Top view of an integrated sensor with circuitry design and (b) 3D cross-sectional of the wafer. Image Source: Nanusens.
Employing patent pending techniques, the process solution uses single vapor hydrogen fluoride (vHF) to etch away part of the silicon oxide in the back end of line (BEOL) of a CMOS process common in today’s wafer-level chip scale packaging (WLCSP), to create the MEMS device. A bottom metal plane and a top metal plane with an array of small holes allows the vHF to get inside the MEMS cavity. Combined with fully digital detection circuitry, the solution can be scaled down to implement the desired process node.
The resulting devices have a very high yield, reliability and performance and can be very cost effective since they take advantage of high-volume CMOS processing with no or minimal added processing steps. For example, from a performance perspective, the all-digital detection circuit provides very fast on/off switching of 3 microseconds compared to 300 microseconds or even several milliseconds in conventional analog transconductance/charge amplifier or similar circuits. The ultrafast on/off detection circuitry results in sub micro-amperes of current consumption on the 180 nm test chip. This level is much lower than the state-of-the-art current consumption in the market and could significantly increase the battery life in end applications.
The process technique can be applied to many different devices including accelerometers, gyroscopes, pressure sensors, ultra-sound sensors, loudspeakers, magnetometers, microphones, RF switches, tuneable capacitors and many more.
Understanding Smart Sensors, 3rd Edition