US-20260126290-A1 - DYNAMIC BIAS COMPENSATION FOR MEMS MOTION SENSORS

US20260126290A1US 20260126290 A1US20260126290 A1US 20260126290A1US-20260126290-A1

Abstract

MEMS sensors output data that may be susceptible to systematic bias that diminishes the accuracy of the data. This disclosure is directed to a method that compensates for this bias so that the output is more accurate and representative. Once a movement of the proof mass of the MEMS sensor outputs a specific characteristic signal, and that signal is identified, the signal is dynamically processed to eliminate and/or reduce the offset. Some applications may require fast settling times, high accuracy of offset removal, and robustness to vibrations and shocks. These requirements typically conflict with each other, meaning that the faster the offset removal, the lower its accuracy, and the sensor is more susceptible to errors due to vibration and shock. In order to tradeoff these requirements, the output of the MEMS sensor is filtered with some non-linear and/or time variant processing, followed by averaging the signal during a startup phase over a steadily increasing number of sampling periods. Once the startup phase is complete, the offset compensation signal is determined and removed from the initial output signal from the MEMS sensor. By utilizing the disclosed method, it is possible to maintain a high accuracy of offset estimation, a fast settling time, while also mitigating the effects of external perturbations (e.g., shocks, vibrations) on the system during the offset estimation process.

Inventors

Vito Avantaggiati

Assignees

INVENSENSE, INC.

Dates

Publication Date: 20260507
Application Date: 20251030

Claims (20)

1 . A method for dynamically compensating for offset in a microelectromechanical system (MEMS) sensor, comprising: receiving an output signal based on a movement of a proof mass of the MEMS sensor in response to a force; monitoring the output signal for at least one signal characteristic, wherein the output signal is discarded until the at least one signal characteristic is identified by the monitoring; processing, once the at least one signal characteristic has been identified, the output signal to determine an offset compensation signal that compensates an offset in the output signal, wherein the processing comprises: performing a first filtering of the output signal; averaging a first set of samples of the first filtered output signal over an initial sampling period of a plurality of sampling periods to generate the offset compensation signal for the initial sampling period; increasing from the initial sampling period to a first sampling period; averaging a second set of samples of the first filtered output signal over the first sampling period to generate the offset compensation signal for the first sampling period; repeating the increasing of the sampling period during a startup phase; averaging an additional set of samples of the first filtered output signal over each increased sampling period to generate the offset compensation signal for each increased sampling period; determining that the startup phase is complete; sampling the first filtered output signal over a fixed sampling period once the startup phase is complete to generate a plurality of fixed sample sets; and averaging each of the fixed sample sets over a fixed averaging period once the startup phase is complete to generate the offset compensation signal for each fixed averaging period, wherein the fixed averaging period is greater than or equal to a longest sampling period of the increased sampling periods, and wherein the fixed sampling period is lower than or equal to the fixed averaging period.
2 . The method of claim 1 , wherein the identification of the at least one signal characteristic comprises an amplitude of the output signal not exceeding a predefined limit.
3 . The method of claim 1 , wherein the identification of the at least one signal characteristic comprises the output signal derivative not exceeding a predefined limit.
4 . The method of claim 1 , wherein the processing once the at least one signal characteristic has been identified comprises an initiation of the startup phase.
5 . The method of claim 1 , further comprising performing a second filtering, wherein the offset compensation signal is further based on the second filtering applying a low pass filter to the averaged sample sets.
6 . The method of claim 5 , further comprising dynamically modifying the second filtering, wherein the dynamically modifying comprises adjusting a bandwidth of the low pass filter for different sampling periods of the plurality of sampling periods.
7 . The method of claim 1 , wherein each increase of the sampling period comprises doubling of an immediately prior sampling period.
8 . The method of claim 7 , wherein an increased period for the averaging associated with each increase of the sampling period is double of an immediately prior period for averaging an immediately prior sample set.
9 . The method of claim 7 , wherein the fixed sampling period is a multiple of a longest sampling period of the increased sampling periods, wherein the multiple of the longest sampling period is two, wherein the fixed averaging period is a multiple of a longest averaging period associated with the increased sampling periods, and wherein the multiple of the longest averaging period is two.
10 . The method of claim 1 , wherein the first filtering of the output signal comprises filtering the output signal to reduce a vibration component within the output signal, and wherein the first filtering of the output signal to reduce the vibration component within the output signal comprises applying a low pass filter to the output signal.
11 . The method of claim 10 , wherein the first filtering of the output signal further comprises modifying the output signal to reduce a shock component within the output signal, and wherein modifying the output signal to reduce the shock component comprises limiting an amplitude of the output signal whenever it is above a predefined amplitude threshold.
12 . The method of claim 11 , wherein limiting the amplitude of the output signal comprises clamping an amplitude of the output signal to the predefined amplitude threshold or substituting the output signal with a zero signal.
13 . The method of claim 11 , further comprising dynamically modifying the first filtering, wherein the dynamically modifying comprises: modifying the predefined amplitude threshold for different sampling periods of the plurality of sampling periods; and adjusting a bandwidth of the low pass filter for different sampling periods of the plurality of sampling periods.
14 . The method of claim 13 , wherein the modifying and the adjusting correspond to a plurality of the increased sampling periods, and wherein the modifying and the adjusting are performed during at least some of the fixed sampling periods.
15 . The method of claim 14 , further comprising: determining whether the startup phase has elapsed; and based on the startup phase being elapsed, fixing the predefined amplitude threshold and the bandwidth for additional fixed sampling periods.
16 . The method of claim 1 , further comprising: adjusting, by a rate limiter, the offset compensation signal such that the offset compensation signal does not change by more than a rate; and downsampling the output signal from a first data rate prior to the processing.
17 . The method of claim 16 , further comprising, after the processing and before the compensating, increasing a data rate of the offset compensation signal to correspond to the first data rate.
18 . The method of claim 1 , wherein the output signal is based on one or more processing operations applied to a measured signal corresponding to a capacitance that is based on the movement of the proof mass, and wherein the processing operations comprise one or more of a capacitance to voltage converter, an integrator, a modification of a gain, a modification of scaling, a low pass filter, or a band pass filter.
19 . A method for removing offset from an output signal of a microelectromechanical system (MEMS) sensor, comprising: monitoring an output signal from the MEMS sensor to determine whether to start a fast offset compensation stage of operation, wherein an offset compensation signal that is used to compensate an offset in the output signal is not generated prior to the start of the fast offset compensation stage of operation; generating the offset compensation signal, during the fast offset compensation stage of operation, based on an increasing sampling period and an increasing averaging period; changing from the fast offset compensation stage of operation to a first slow offset compensation stage of operation; generating the offset compensation signal, during the first slow offset compensation stage of operation, based on a first fixed sampling period and a first fixed averaging period; changing from the first slow offset compensation stage of operation to a second slow offset compensation stage of operation; and generating the offset compensation signal, during the second slow offset compensation stage of operation, based on a second fixed sampling period and a second fixed averaging period, wherein the second fixed averaging period is different than the first fixed averaging period.
20 . A system for removing offset from an output signal of a (MEMS) sensor, comprising: amplitude limiting circuitry coupled to receive the output signal from the MEMS sensor and to limit an amplitude of the output signal to output an amplitude-limited signal; a filter coupled to the amplitude limiting circuitry to receive the amplitude-limited signal, wherein the filter removes a samples that are not within a pass band of the filter to output a filtered signal; averaging circuitry coupled to the filter to receive the filtered signal, wherein the averaging circuitry outputs an averaged signal based on a sampling period and an averaging period; processing circuitry coupled to the averaging circuitry to process the averaged signal to generate an offset compensation signal that is suitable for removing the offset from the output signal; and a bandwidth controller configured to modify a bandwidth of the filter, the sampling period, and the averaging period, wherein each of the bandwidth, the sampling period, and the averaging period are modified during an initial fast offset compensation stage of operation, and wherein the bandwidth, the sampling period, and the averaging period become fixed during at least a portion of a slow offset compensation stage that occurs after completion of the initial fast offset compensation stage.

Description

CROSS-REFERENCE TO RELATED APPLICATION This application claims priority to U.S. Provisional Patent Application No. 63/715,600, filed November 3, 2024, and entitled “Systems and Methods for Dynamic Bias Compensation,” which is incorporated by reference herein in its entirety for all purposes. BACKGROUND Numerous items such as smart phones, smart watches, tablets, automobiles, aerial drones, appliances, aircraft, exercise aids, and game controllers may utilize sensors such as microelectromechanical system (MEMS) sensors during their operation. In many applications, various types of motion sensors such as accelerometers and gyroscopes may be analyzed independently or together in order to determine varied information for particular applications. For example, gyroscopes and accelerometers may be used in gaming applications (e.g. , smart phones or game controllers) to capture complex movements by a user, drones and other aircraft may determine orientation based on gyroscope measurements (e.g., roll, pitch, and yaw), and vehicles may utilize measurements for determining direction (e.g., for dead reckoning) and safety (e.g., to recognizing skid or roll-over conditions). A MEMS sensor such as a MEMS accelerometer or gyroscope can have a bias (e.g., an offset) that corresponds to a non-motion output of sensor. This offset can be an inherent feature of the MEMS sensor design, can be a regular result of manufacturing processes and tolerances, and/or can be impacted in the field, based on wear, incurred stresses, temperature, or other factors. A large offset can impact measurements such as by incorrectly indicating motion or limiting bandwidth for accurate motion detection. Accordingly, techniques have been employed to identify or remove offset in MEMS sensors. These techniques are unable to quickly reach a steady state for offset removal while achieving accuracy and robustness to external shocks and vibrations. SUMMARY In an embodiment of the present disclosure, a method for dynamically compensating for offset in a MEMS sensor comprises receiving an output signal based on a movement of a proof mass of the MEMS sensor in response to a force and monitoring the output signal for at least one signal characteristic, wherein the output signal is discarded until the at least one signal characteristic is identified by the monitoring (e.g., an amplitude of the output signal not exceeding a predefined limit, or the output signal derivative not exceeding a predefined limit). The method can further comprise processing, once the at least one signal characteristic has been identified, the output signal to determine an offset compensation signal that compensates an offset in the output signal. The processing can further comprise performing a first filtering of the output signal, averaging a first set of samples of the first filtered output signal over an initial sampling period of a plurality of sampling periods to generate the offset compensation signal for the initial sampling period, and increasing from the initial sampling period to a first sampling period. The processing can further comprise averaging a second set of samples of the first filtered output signal over the first sampling period to generate the offset compensation signal for the first sampling period, repeating the increasing of the sampling period during a startup phase, and averaging an additional set of samples of the first filtered output signal over each increased sampling period to generate the offset compensation signal for each increased sampling period. The processing can further comprise determining that the startup phase is complete, sampling the first filtered output signal over a fixed sampling period once the startup phase is complete to generate a plurality of fixed sample sets, and averaging each of the fixed sample sets over a fixed averaging period once the startup phase is complete to generate the offset compensation signal for each fixed averaging period, wherein the fixed averaging period is greater than or equal to a longest sampling period of the increased sampling periods, and wherein the fixed sampling period is lower than or equal to the fixed averaging period. In an embodiment of the present disclosure, a method for removing offset from an output signal of a MEMS sensor comprises monitoring an output signal from the MEMS sensor to determine whether to start a fast offset compensation stage of operation, wherein an offset compensation signal that is used to compensate an offset in the output signal is not generated prior to the start of the fast offset compensation stage of operation. The method can further comprise generating the offset compensation signal, during the fast offset compensation stage of operation, based on an increasing sampling period and an increasing averaging period, and changing from the fast offset compensation stage of operation to a first slow offset compensation stage of operation. The method can further comprise gen