Periodic Reporting for period 1 - 0-drift (Towards no-drift sensors with on-chip self-calibration)
Okres sprawozdawczy: 2023-10-01 do 2026-03-31
Autonomous navigation will be an indispensable part of our daily lives in the near future. Data from multiple sensors, i.e. GPS, camera, LIDAR, RADAR, and inertial sensors, are fused in navigation algorithms. Unlike any other sensor, navigation with inertial sensors is error-proof if a no-drift inertial sensor could be invented. The gyroscope and accelerometer outputs are integrated for navigation, and drift leads to error accumulation over time. The drift is the main barrier for utilizing MicroElectroMechanical System (MEMS) inertial sensors in demanding autonomous navigation applications. This project (0-drift) aims to address the drift problem in MEMS gyroscopes. 0-drift targets to develop an ideal sensor with just white noise through innovative on-chip self-calibration methods. Currently, inertial sensors are considered unreliable and deployed only to stop the vehicle when all other sensors fail due to weather conditions. If 0-drift succeeds, the navigation algorithms could rely on the error-proof, only gravity-referenced inertial sensors. The navigation would be more robust and could resume under all conditions. For example, cameras might fail in the dark, GPS signal can be jammed or lost, but the inertial sensor operation is independent of light, weather, and location conditions. A no-drift, low-cost MEMS sensor would enable indoor navigation where GPS cannot be used and also disrupt the precision inertial sensor market by replacing much more expensive and bulky counterparts. There has been ongoing research, and the MEMS inertial sensors have become more stable, mainly due to noise improvements. However, limited research focuses on the basic drift problem. 0-drift will approach the problem in an unconventional way by introducing a self-calibrating sensor. The only way a MEMS sensor can interact with its environment is through the device anchors. Environmental stress and temperature variations cause the device anchors to shift, affecting device dynamics and resulting in drift. 0-drift will develop a sensor that applies a stress stimulus to the device and extracts its own drift transfer function by measuring the on-chip stress and the MEMS sensor output. A novel MEMS fabrication flow, a new integrated electronic readout circuit, new control and calibration methods, along with an innovative sensor, will be developed in the context of 0-drift.
We integrated a ring gyroscope and on-chip capacitive stress sensors on the same chip. Stress is a tensor with multiple shear and normal components in all directions. An efficient calibration requires an accurate assessment of the MEMS device stress state. Therefore, we employed a ring gyroscope design that enables tight integration of stress sensors due to its sparse structure. MEMS gyroscopes are the most complex MEMS devices and measure the rotation through energy coupling between the two eigenmodes of a resonant body. Drift calibration is not a straightforward task, as drift is a small signal that can be easily overshadowed by unintentional errors. We have developed analytical models to gain a detailed understanding of the drift dynamics. Our analytical model interpolates the anchor displacements from the measured stress and predicts the key gyroscope parameters, taking into account device imperfections and temperature variations. Our model accounts for the nonlinear strain and extensibility of the ring, as well as thermoelasticity to capture temperature effects. Our experimental and analytical modeling results showed good agreement, providing us deeper understanding of drift dynamics and setting the path for physics-aware drift calibration. The scale factor stability is equally important as the offset stability for inertial navigation. Commonly utilized temperature calibration results in ~1000ppm scale factor accuracy. The proposed on-chip stress calibration method significantly outperformed the conventional methods by achieving <100ppm scale factor temperature accuracy. In addition, implementing a mode-matching loop for sensitivity enhancement and combining it with on-chip stress calibration resulted in a no-drift offset stability of 0.6°/h for a wide 60°C temperature range. These results have become a solid proof for the efficiency of on-chip stress calibration.
MEMS gyroscope short-term noise performance has almost reached its theoretical limits; however, long-term drift still remains as a bottleneck for inertial navigation. Suppressing long-term drift is a challenging task, and 0-drift aims to eliminate drift using a novel on-chip stress calibration method. We acknowledge that drift will occur, and our goal is to calibrate for it. We see the scale factor temperature calibration with on-chip stress sensors work as the most important breakthrough result of the 0-drift. The gyroscope scale factor depends on multiple electrostatic and mechanical parameters. Conventional calibration methods employ an average temperature measurement, which cannot predict these device parameter variations with enough resolution. The distributed on-chip measurements accurately captured the key device parameters and resulted in orders of magnitude better temperature stability than the commercial counterparts. The reported temperature stability of 30ppm with no indication of drift sets the current MEMS state of the art stability. Along with experimental results, our state of the art analytical model unveils the physics behind the environmental stress and temperature effects on the device. We were able to distinguish ppm level signals without finite element analysis (FEA), which allowed us to observe all the electrostatic and mechanical variations within a micro-scale sensor. 0-drift targets to achieve self stress calibration at the end of the project, and the analytical model will be a major building model for the calibration.