Search

This paper reports on real-time offset calibration of a three-axis gyroscope whereby the angular rate sensor is part of an IMU in the field. If the IMU is in motion, the orientation changes of the gyroscope can be compared with those from other calibrated attitude sensors to identify the offset of the gyroscope. However, the acceleration sensor of an IMU in the field is typically uncalibrated and thus prevents an exact offset calibration during motion. In this paper, a new multi-model error state Kalman filter is proposed in order to determine the gyroscope's offset. This filter exclusively uses an uncalibrated accelerometer for determining the gyroscope's offset. Consequently, the resulting calibration system is only partially observable which in turn influences the observability of the requested offset parameters. However, the latter conditions are the typical situation encountered in an IMU in the field. Therefore, the observability conditions as well as the stability of the multi model error state Kalman filter will be analyzed in the second part of this paper.

This paper reports on realtime offset calibration of a three-axis gyroscope, especially, offset calibration of a gyroscope exclusively based on the redundant information of one three-axis accelerometer. The self calibration is based on the knowledge that the output of any gyroscope equals zero if the inertial measurement unit is in a static position. For this purpose, a sophisticated movement recognition has been developed using unaligned acceleration signals containing sensitivity and offset errors. The proposed movement recognition identifies not exclusively human caused activities, which is state of the art, but rather extends the sensibility to small, constant yaw rates which are close to the ground noise floor of the static gyroscope. These very small, constant yaw rates are in the range of possible offsets and not identified through previously presented movement recognitions. The effectiveness of the proposed movement recognition overcomes this drawback and enables the complete rejection of the environmental dependent offset of any gyroscope. The performance of the auto calibration algorithm has been evaluated by means of simulations, human movement patterns and movement patterns generated through a turn table.

In the EU FP7 project "ADOSE" a new approach has been used to develop a new cost efficient technology by adapting a volume proven integrated MEMS process for the production of a suspended thermodiode array. As low-cost was to be the key feature of the project the focus was set on fully semiconductor compatible production without the need for dedicated equipment and on the use of cost efficient MEMS technologies like wafer-level vacuum encapsulation. A first proof-of-concept integrated FIR array with 42 x 28 pixels was already published [1], now the final ADOSE design with 100 x 50 pixel resolution is produced and under evaluation. While the ADOSE chip targets the requirements of a low-end “Hot-Spot-Detector”, many applications require higher resolutions. As the technology used in ADOSE inherits some limitations from its pressure sensor ancestor, we advance the technology to get rid of these and achieve a low-cost and high-res imager. Here we present the results from ADOSE and give an outlook on the new BMBF-funded Spitzencluster MicroTEC Sudwest project RTFIR.

A new low-cost process for mid resolution FIR-Arrays for thermal imaging in the LWIR range (8- 14 μm) has been developed. Integrated array devices with 1k and 5k pixel array size have been designed and manufactured using this process which is fully compatible with semiconductor production. The devices are suitable to support detection of humans e.g. in security and automotive safety applications.