DE102022203430A1 - Method for self-calibration of a sensor system and sensor system - Google Patents

Abstract

A sensor system and a method for self-calibration of a sensor system are proposed, wherein the sensor system has at least one sensor designed as a multi-axis inertial sensor and is configured to apply a compensation matrix to sensor data in such a way that a transverse axis sensitivity of the at least one sensor is compensated, wherein Method for self-calibration in one measuring step at least one current value of a system parameter of the sensor system is determined; in a readout step, at least one reference value of the system parameter is read out from a storage element of the sensor system; in a calculation step, a change in the transverse axis sensitivity of the at least one sensor is calculated as a function of the reference value and the current value of the system parameter; and in an adaptation step the compensation matrix is adapted to the change in the transverse axis sensitivity.

Description

State of the art

The invention is based on a method according to the preamble of claim 1.

MEMS inertial sensor systems (MEMS, microelectromechanical system) typically consist of arrangements of two or more sensors to determine the rotation rate and acceleration applied to the sensor. Typically, each sensor has three measurement channels with which the accelerations and rotation rates can be determined along three spatial axes, so that all physical movements in three-dimensional space can be recorded. The axes of the respective axis system are usually at least approximately orthogonal to one another and the axis systems of different sensors are each aligned parallel to one another. A deviation from such an ideal configuration manifests itself as cross-axis sensitivity, through which, for example, a physical stimulus that should only produce a signal regarding a single sensor axis also results in a certain measurement component along one or more of the others Axes guides.

Such cross-axis sensitivities in MEMS inertial sensors for the consumer sector can reach 2 to 3% and do not represent a significant problem for many traditional areas of application. However, for certain applications, such as inertial indoor navigation or augmented reality, cross-axis sensitivities of this magnitude would result in a significant loss of accuracy with it (see for example the analysis in [1]). This results in the technical task of reducing transverse axis sensitivity effects as much as possible. In this context, a variety of options are known from the prior art for optimizing the transverse axis sensitivity by design. In the MEMS area, this is often achieved through the arrangement and geometric design of the micromechanical springs and masses of the sensor. Another possibility, which is used in particular with high-precision inertial sensors, is a signal correction to improve transverse axis sensitivities, in which an individual correction is determined for the sensor, which is then applied to the measurement signal during sensor operation (see, for example, [2] and [3 ]).

A common procedure is to set up transverse axis sensitivity matrices of dimension 3x3 (with three spatial axes) for the acceleration sensor and the yaw rate sensor by controlled application of physical stimuli and evaluation of the sensor signals. The diagonal elements, the transverse axis sensitivity matrices, represent the actual sensitivities of the axes (i.e. are ≈1), while the transverse sensitivities between two axes are given by the non-diagonal elements. The latter are significantly smaller than the diagonal elements and can be around 0.01, for example, which corresponds to 1% transverse axis sensitivity. If the two measured transverse axis sensitivity matrices are available, a 3x3 compensation matrix can be implemented at the acceleration and rotation rate sensor outputs, which compensates for the transverse axis effects. The simplest procedure for determining the two compensation matrices is a simple inversion of the measured transverse axis sensitivity matrices, as described, for example, in [4] or in Section II.D in [3].

• [1] L. Blocher, W. Mayer, M. Arena, D. Radovic, T. Hiller, J. Gerlach, O. Bringmann, „Purely inertial navigation with a low-cost MEMS sensor array,“ in IEEE Int. Symposium on Inertial Sensors and Systems 2021. IEEE, 2021.
• [2] L. Poletti, D. S. Sanchis, R. Siryani, „A direct approach for highquality MEMS based IMU/INS production,“ in 2020 DGON Inertial Sensors and Systems (ISS), 2020, pp. 1-19 .
• [3] J. Rohac, M. Sipos, J. Simanek, „Calibration of low-cost triaxial inertial sensors,“ IEEE I. & M. Mag., vol. 18, no. 6, pp. 32-38 , 2015.
• [4] T. Hiller, L. Blocher, M. Vujadinovic, Z. Pentek, A. Buhmann H. Roth, „Analysis and Compensation of Cross-Axis Sensitivity in Low-Cost MEMS Inertial Sensors,“ in IEEE International Symposium on Inertial Sensors and Systems 2021. IEEE, 2021.

The determination of the transverse axis sensitivity and the definition of the compensation matrices usually takes place in the form of a calibration as part of the manufacturing process. The finished and calibrated sensors are then further processed and, for example, applied to a circuit board together with other sensors and other components. However, since the sensors are exposed to considerable stress during production, handling, storage, transport and during the soldering process and are therefore subject to changes, the quality of the transverse axis compensation decreases compared to the newly calibrated state, so that the accuracy of the sensor is reduced over time .

• [1] L. Blocher, W. Mayer, M. Arena, D. Radovic, T. Hiller, J. Gerlach, O. Bringmann, “Purely inertial navigation with a low-cost MEMS sensor array,” in IEEE Int. Symposium on Inertial Sensors and Systems 2021. IEEE, 2021.
• [2] L. Poletti, DS Sanchis, R. Siryani, “A direct approach for high quality MEMS based IMU/INS production,” in 2020 DGON Inertial Sensors and Systems (ISS), 2020, pp. 1-19 .
• [3] J. Rohac, M. Sipos, J. Simanek, “Calibration of low-cost triaxial inertial sensors,” IEEE I. & M. Mag., vol. 18, no. 6, pp. 32-38 , 2015.
• [4] T. Hiller, L. Blocher, M. Vujadinovic, Z. Pentek, A. Buhmann H. Roth, “Analysis and Compensation of Cross-Axis Sensitivity in Low-Cost MEMS Inertial Sensors,” in IEEE International Symposium on Inertial Sensors and Systems 2021. IEEE, 2021.

Disclosure of the invention

Against this background, it is an object of the present invention to provide a method with which the signal correction for compensating for the transverse axis sensitivities can be recalibrated in order to compensate for the change compared to the original behavior in production.

The method according to claim 1 allows, largely independently of the actual sensor design, to carry out a subsequent, sensor-specific calibration using sensor-internal variables. In this way, the validity of the original calibration from production can be restored and transferred to the soldered product. In addition, the method can be used repeatedly over the lifetime of the sensor system in order to compensate for further drift effects, such as those caused by aging processes or environmental influences (humidity, temperature, etc.).

The method according to the invention takes advantage of the fact that, under the external influences mentioned at the beginning during production, further processing, etc., different system parameters of the sensor system generally do not change completely independently of one another, but rather have various direct and indirect relationships to one another. In particular, a statistical relationship between the transverse axis sensitivity and one or more system parameters (or their change compared to a reference state) can be used to calculate the change in the transverse axis sensitivity. Current values of the state variable or state variables can be determined in the sensor itself, in particular without using external devices, for example by measuring the corresponding variable using sensors arranged in the sensor system or obtaining it from a sensor signal. By comparing the current value and the reference value, the change in the system parameter can be determined and the associated shift in the transverse axis sensitivity can be determined. Various parameters that are particularly suitable for this are described below. For example, a comparison of the characteristics of a rotation rate sensor before and after soldering shows that the change in the transverse axis sensitivity and the change in the quadrature of the z-channel have a positive covariance. In the simplest case, this statistical relationship can be approximated by a linear relationship, which in turn serves as a calculation rule for the change in the transverse axis sensitivity. In the following, for the sake of brevity, the term transverse axis sensitivity is used for a corresponding parameter, for example a non-diagonal element of the transverse axis sensitivity matrix or the entirety of all non-diagonal elements. In addition to the at least one sensor, the sensor system can have further sensors, in particular further multi-axis inertial sensors, each of which has its own transverse axis sensitivity and for which the method is generalized accordingly.

Advantageous refinements and further developments of the invention can be found in the subclaims and the description with reference to the drawings.

According to a preferred embodiment of the method according to the invention, it is provided that in the measuring step current values of a plurality of system parameters are determined, in the readout step reference values of the plurality of system parameters are read out from a storage element of the sensor system and in the calculation step the change in the transverse axis sensitivity as a function of the reference values and the current values of the majority of system parameters is calculated. In this way, more complex relationships between the various system parameters and the transverse axis sensitivity can be recorded, or the accuracy and/or robustness of the calculation can be improved. The implementation options of the invention described with regard to a system parameter generalize directly to a plurality of system parameters.

According to a preferred embodiment of the method according to the invention, it is provided that the system parameter or the plurality of system parameters comprises one or more of the following variables: an amplitude of a quadrature signal, a (native, ie untrimmed) non-orthogonality of the sensor axes, a (native) misalignment between sensor axes of the at least one sensor and sensor axes of a further sensor, a phase position of a sensor signal, a drive frequency, a frequency difference between a drive mode and a detection mode, a raw sensitivity, a mechanical tension, a quantity detected by means of a built-in test equipment (BITE), in particular a controlled quantity induced change in a quadrature signal. If the at least one sensor is an acceleration sensor, the BITE can also be used to deflect a seismic mass of the acceleration sensor by applying a voltage and measure the response behavior of the sensor. The quadrature signal is the portion (in particular obtained through demodulation) of the detection signal of a yaw rate sensor that is 90° relative to that generated by the yaw rate th useful signal is out of phase. For the method according to the invention, for example, the quadrature signal of a specific detection channel (corresponding to a sensor axis) or a combination of several quadrature signals from different channels can be used. The misalignment between the sensor axes of two sensors (see, for example, section “accel-to-gyro triad alignment” in [4]) can be quantified in particular by the elements of the transformation matrix that maps the axes of one sensor to those of the other sensor. The sensor signal can be, for example, the rotation rate signal or the acceleration signal of a specific sensor channel or a combination of such signals. Mechanical stresses can be measured in particular by dedicated voltage sensors in the sensor. A BITE can cause a controlled change in the state of the sensor, for example by creating an artificial quadrature in the sensor by applying or changing an electrode voltage.

According to a preferred embodiment of the method according to the invention, it is provided that the change in the transverse axis sensitivity is calculated using a statistical model for the relationship between the system parameter and the transverse axis sensitivity, the statistical model being formed in particular by means of compensatory calculation or machine learning. In particular, a data set can be determined using a plurality of sensors by determining the system parameter and the transverse axis sensitivity of each sensor at different points in the manufacturing process (for example before and after soldering on the sensors) or by determining the sensor of a controlled change in state (for example by generating a mechanical voltage). A statistical relationship can then be obtained from the data set through analysis or by training a machine learning model, which is used in the calculation step to determine the change compared to the reference state.

According to a preferred embodiment of the method according to the invention, it is provided that the reference value of the system parameter is determined in a provision step and stored in the storage element of the sensor system, the provision step being carried out in particular during an end-of-line calibration following the production of the sensor system.

According to a preferred embodiment of the method according to the invention, it is provided that in the provision step at least two data points are determined by a controlled change in the system parameter and a measurement of the transverse axis sensitivity and the statistical model is formed depending on the data points. Instead of a behavior averaged over a statistical totality of sensors, a sensor-specific change behavior can be determined in this way. In the simplest case, two data points can be measured and a linear relationship between the system parameter and the transverse axis sensitivity can be obtained. Likewise, more than two data points can be measured and a linear or non-linear function can be adapted to the data points using compensation calculations. The controlled change in the system parameter can be carried out, for example, by applying a mechanical or thermal load to the sensor. The calculation rule can be optimized, for example, by applying two or more different loads during production (e.g. two identical measurements or two measurements at different temperatures) and using the change in sensor-internal parameters together with the change in the transverse axis sensitivities. This means that a sensor-specific calculation rule can be created for the self-calibration that takes place after soldering. In the simplest case, for example, the slope of a linear calculation rule is programmed individually for each sensor.

According to a preferred embodiment of the method according to the invention, it is provided that the sensor system has at least one further sensor designed as a multi-axis inertial sensor, the sensor system being configured to apply the compensation matrix to sensor data of the at least one sensor and a further compensation matrix to sensor data of the at least one further Apply the sensor, whereby the compensation matrix and the further compensation matrix are adapted in the adaptation step to the change in the transverse axis sensitivity calculated in the calculation step. Preferably, the at least one sensor is a rotation rate sensor, an acceleration sensor, a multi-axis magnetic field sensor or a rotational acceleration sensor and/or the at least one further sensor is a rotation rate sensor, an acceleration sensor, a multi-axis magnetic field sensor or a rotational acceleration sensor. For the sake of brevity, the at least one sensor is also referred to below as the first sensor and the at least one further sensor as the second sensor. The compensation matrices are referred to accordingly as the first and second compensation matrix.

In the manufacturing process, the determination of the transverse axis sensitivity matrices for the original calibration is usually carried out by the controlled application of physical stimuli along an axis system that is essentially predetermined by the measuring arrangement and is hereinafter referred to as a stimulus axis system to distinguish it from the intrinsic axes of the sensors. In the simplest case, the inverse cross-axis sensitivity matrices can be used as compensation matrices, which effectively results in the sensor axes of both sensors being aligned parallel to the stimulus axes, thereby eliminating both the non-orthogonality of the sensor axes and the rotation of the sensor axes relative to the stimulus. Axis system can be corrected. However, if the sensors are arranged slightly crooked in the measuring system, for example, the sensor axes are unnecessarily shifted by this compensation mechanism and “trimmed to the stimulus”. The improved determination of compensation matrices described below, however, allows the natural alignment of the axis system of the first sensor to be maintained and the axis system of the second sensor to be adapted to the first. This avoids unnecessary twisting in the sensor and the signals are not mixed through the compensation matrix any more than necessary. For this purpose, a first compensation matrix is formed depending on the first transverse axis sensitivity matrix, wherein the first transverse axis sensitivity matrix can be decomposed into a product of a rotation matrix and a triangular matrix and the first compensation matrix corresponds to an inverse of the triangular matrix; the second compensation matrix is formed as a function of the second transverse axis sensitivity matrix, the second compensation matrix corresponding to a product of the rotation matrix and an inverse of the second transverse axis sensitivity matrix. In the adaptation step, the formation regulations mentioned are then used to adapt the two compensation matrices to the changed transverse axis sensitivity matrix elements.

The underlying mathematical structure will be briefly illustrated below using a concrete example in which the first sensor is a three-axis acceleration sensor and the second sensor is a three-axis rotation rate sensor. The following relationship between the signal amplitudes and the external stimuli is assumed as a model: $$\left[\begin{array}{c}{\mathrm{\Omega }}_{\text{x}\text{,sen}}\\ {\mathrm{\Omega }}_{\text{y}\text{,sen}}\\ {\mathrm{\Omega }}_{\text{e.g}\text{,sen}}\end{array}\right]=\underset{{\text{s}}_{\text{cas}}^{\text{a}}}{\underbrace{\left[\begin{array}{ccc}{\text{s}}_{\text{xx}}^{\text{a}}& {\text{s}}_{\text{xy}}^{\text{a}}& {\text{s}}_{\text{xz}}^{\text{a}}\\ {\text{s}}_{\text{yx}}^{\text{a}}& {\text{s}}_{\text{yy}}^{\text{a}}& {\text{s}}_{\text{Y Z}}^{\text{a}}\\ {\text{s}}_{\text{zx}}^{\text{a}}& {\text{s}}_{\text{zy}}^{\text{a}}& {\text{s}}_{\text{currently}}^{\text{a}}\end{array}\right]}}\left(\left[\begin{array}{c}{\mathrm{\Omega }}_{\text{x}\text{,sti}}\\ {\mathrm{\Omega }}_{\text{y}\text{,sti}}\\ {\mathrm{\Omega }}_{\text{e.g}\text{,sti}}\end{array}\right]+\left[\begin{array}{c}{\text{b}}_{\text{x}}\\ {\text{b}}_{\text{y}}\\ {\text{b}}_{\text{e.g}}\end{array}\right]\right).$$

(a für „accelerometer“, cas für „cross-axis sensitivity“) multipliziert werden. Zusätzlich wird auf der rechten Seite ein Offset b berücksichtigt, der ebenfalls von der Querachsenempfindlichkeitsmatrix multipliziert wird. Die Querachsenempfindlichkeitsmatrix des Beschleunigungssensors lässt sich nun in drei Anteile bzw. Subkomponenten zerlegen: $${\text{s}}_{\text{cas}}^{\text{a}}={\text{M}}_{\text{scf}}\cdot {\text{M}}_{\text{non}}\cdot {\text{M}}_{\text{mis}},$$

The left side corresponds to the measured output values Ω _sen of the accelerometer along the three spatial axes, while on the right side the associated external stimuli Ω _sti through the cross-axis sensitivity matrix ${\text{s}}_{\text{cas}}^{\text{a}}$

(a for “accelerometer”, cas for “cross-axis sensitivity”) can be multiplied. In addition, an offset b is taken into account on the right-hand side, which is also multiplied by the transverse axis sensitivity matrix. The transverse axis sensitivity matrix of the acceleration sensor can now be broken down into three parts or subcomponents: $${\text{s}}_{\text{cas}}^{\text{a}}={\text{M}}_{\text{scf}}\cdot {\text{M}}_{\text{non}}\cdot {\text{M}}_{\text{mis}},$$

1. (i) Den Nicht-Orthogonalitäts- und Empfindlichkeitsfehler des Beschleunigungssensors zu korrigieren, nicht aber seine Verdrehung. Anders ausgedrückt behält das Achsensystem prinzipiell seine ursprüngliche Ausrichtung bei.
2. (ii) Der Drehratensensor wird ebenfalls bezüglich seiner Nicht-Orthogonalität und Empfindlichkeit korrigiert und so verdreht, dass die entsprechenden Sensorachsen eine identische Ausrichtung mit dem, mittels (i) kompensierten Beschleunigungssensor aufweisen.

The matrix M _scf (scf for “scale-factor”) is a pure diagonal matrix whose entries indicate the sensitivities with respect to the three axes and the matrix M _non (“non-orthogonality”) is a lower triangular matrix whose diagonal elements are all one and whose off-diagonal elements indicate the non-orthogonality of the sensor axes. The product M _scf · M _non corresponds to the triangular matrix, the inverse of which forms the compensation matrix for the first sensor (here the acceleration sensor). The matrix M _mis (“misalignment”) is ultimately a rotation matrix that reflects the rotation of the sensor axes relative to the stimulus axes and is used to compensate for the second sensor. The improved determination of the compensation matrices is based on the basic idea of not aligning the sensor with the stimulus, as is the case with the simple inversion of the sensitivity matrices, but rather:

1. (i) To correct the non-orthogonality and sensitivity error of the accelerometer, but not its twist. In other words, the axis system basically retains its original orientation.
2. (ii) The rotation rate sensor is also corrected for its non-orthogonality and sensitivity and rotated so that the corresponding sensor axes have an identical alignment with the acceleration sensor compensated by means of (i).

entfernt werden soll, lässt sich die Kompensationsmatrix ${\text{K}}_{\text{cmp}}^{\text{a}}$

für (i) beispielsweise als $${\text{K}}_{\text{cmp}}^{\text{a}}={\left({\text{M}}_{\text{scf}}\cdot {\text{M}}_{\text{non}}\right)}^{-1}={\left({\text{s}}_{\text{cas}}^{\text{a}}\cdot {\text{M}}_{\text{mis}}^{-1}\right)}^{-1}={\text{M}}_{\text{mis}}\cdot {\left({\text{s}}_{\text{cas}}^{\text{a}}\right)}^{-1}$$

berechnen, d.h. entweder als die Inverse der Diagonalmatrix M_scf · M_non oder als das Produkt der Rotationsmatrix M_mis und der Inversen der ersten Querachsenempfindlichkeitsmatrix ${\text{s}}_{\text{cas}}^{\text{a}}.$

Die Kompensationsmatrix für den Drehratensensor (d.h. den zweiten Sensor), mit der die Korrektur (ii) erzeugt wird, ergibt sich dann unter Benutzung der Rotationsmatrix M_mis als $${\text{K}}_{\text{cmp}}^{\text{g}}={\left({\text{s}}_{\text{cas}}^{\text{g}}\cdot {\text{M}}_{\text{mis}}^{-1}\right)}^{-1}={\text{M}}_{\text{mis}}\cdot {\left({\text{s}}_{\text{cas}}^{\text{g}}\right)}^{-1}$$

Since for the correction (i) the twist M _mis from the transverse axis sensitivity matrix ${\text{s}}_{\text{cas}}^{\text{a}}$

should be removed, the compensation matrix can be used ${\text{K}}_{\text{cmp}}^{\text{a}}$

for (i) for example as $${\text{K}}_{\text{cmp}}^{\text{a}}={\left({\text{M}}_{\text{scf}}\cdot {\text{M}}_{\text{non}}\right)}^{-1}={\left({\text{s}}_{\text{cas}}^{\text{a}}\cdot {\text{M}}_{\text{mis}}^{-1}\right)}^{-1}={\text{M}}_{\text{mis}}\cdot {\left({\text{s}}_{\text{cas}}^{\text{a}}\right)}^{-1}$$

calculate, ie either as the inverse of the diagonal matrix M _scf · M _non or as the product of the rotation matrix M _mis and the inverse of the first transverse axis sensitivity matrix ${\text{s}}_{\text{cas}}^{\text{a}}.$

The Compensation matrix for the rotation rate sensor (ie the second sensor), with which the correction (ii) is generated, is then obtained using the rotation matrix M _mis as $${\text{K}}_{\text{cmp}}^{\text{G}}={\left({\text{s}}_{\text{cas}}^{\text{G}}\cdot {\text{M}}_{\text{mis}}^{-1}\right)}^{-1}={\text{M}}_{\text{mis}}\cdot {\left({\text{s}}_{\text{cas}}^{\text{G}}\right)}^{-1}$$

und die Querachsempfindlichkeit ${\text{s}}_{\text{cas}}^{\text{g}}$

des Drehratensensors sind dabei mit dem Index g gekennzeichnet („gyroscope“).The compensation matrix ${\text{K}}_{\text{cmp}}^{\text{G}}$

and the transverse axis sensitivity ${\text{s}}_{\text{cas}}^{\text{G}}$

of the rotation rate sensor are marked with the index g (“gyroscope”).

If the sensor system has further axis systems, e.g. by another acceleration sensor with a different measuring range, a magnetic field sensor, another rotation rate sensor, etc., correction (ii) can also be applied to these. The example above is merely illustrative; for example, in principle (i) can also be done on the yaw rate sensor and (ii) on the acceleration sensor. It is also possible to apply the principle to axis systems that are not installed in the same sensor package, but only on the same circuit board or in the same housing.

The method according to the invention for self-calibration of the sensor system can be used in a method for compensating for a transverse axis sensitivity of a sensor system, wherein after carrying out an embodiment of the method for self-calibration according to the invention in a compensation step, the adapted compensation matrix is applied to sensor data of the at least one sensor or the at least one further sensor becomes.

A further object of the invention is a sensor system according to claim 10. The advantages and design options explained in relation to the method according to the invention are transferred analogously to the sensor system according to the invention and vice versa.

The invention relates to any form of use of MEMS inertial sensors (acceleration and rotation rate), such as in the automotive sector (automobiles, motorcycles, bicycles/e-bikes, trucks, off-road vehicles, etc.) and in the consumer electronics sector (smartphones, headphones, tablets, augmented reality Headsets and glasses, drones, game consoles, cameras, watches, fitness equipment, implants, toys, wearables, etc.).

The invention is not limited to the embodiments described above but rather can be used in a variety of applications for navigation, orientation and stabilization of objects based on inertial sensors. A computing unit in the sensor can be used to control the operation of the inertial sensor (e.g. energy saving mode, measuring ranges), to check the plausibility of sensor signals and their tolerances (e.g. for internal sensor monitoring), for signal processing (e.g. calculation of the position or orientation, filtering of the data) and for selecting communication protocols. Various self-learning AI-based algorithms can be used in the computer unit for the evaluation and signal processing of the data from the inertial sensors, the temperature sensors and also external data (e.g. GPS data, odometer data). The following applications are also conceivable: two-wheeler such as motorcycle, bicycle and scooter applications (e.g. ESP/AirBag, tilt detection, balancing); tricycles (e.g. TucTuc); avionics applications (e.g. flight stabilization and control); Industrial robot applications (e.g. position control of excavator buckets, drills, image stabilization, flight control, alignment of satellite antennas, fine motor skills for robotic gripping); Applications for home and garden (e.g. navigation of lawn mowers, position monitoring of doors); medical applications (e.g. fall detection, motion and posture detection); Sports and leisure activities (e.g. motion detection, posture detection, golf clubs, tennis tennis rackets, skis); numerous CE applications (e.g. smartphones, tablets, wearables, hearables, drones, toys). The invention can also be used in connection with smartphones and tablets for the following applications: screen orientation; significant movement; device orientation; activity, gesture and context recognition; image stabilization; Indoor SLAM (simultaneous localization and map creation); Impact and free fall detection; Motion control. In connection with wearables, hearables, AR and VR, the invention can be used for the following applications: display information; step counting; activity, gesture and context recognition; Count calories; in-ear detection; sleep monitoring; geriatric care; indoor navigation; positioning; Low-power sensing, real-time motion detection, head movement tracking; precise sensor data fusion. In the context of drones, games and toys, the invention can be used for the following applications: orientation; gimbal; altitude stabilization; flight controls; motion tracking, motion control, balance; Activity and gesture recognition. In connection with robots, the invention can be used for the following areas of application: navigation; border detection; dynamic path planning; Indoor SLAM; air quality monitoring; Detection of blockages. In connection with Smart home applications, the invention can be used for the following applications: burglary control; Air air quality monitoring, mold detection, climate control, soil soil detection; Indoor navigation. The invention can also be used in an industrial context for the following applications: water level detection; asset tracking; navigation and control; movement and position tracking; energy management; predictive maintenance. In addition, numerous modifications, variations, designs, arrangements and embodiments are possible, all of which fall within the scope of the invention.

Embodiments of the present invention are shown in the drawings and explained in more detail in the following description.

Brief description of the drawings

• 1 shows two axis systems of an inertial sensor system with slight transverse axis sensitivity.
• 2 illustrates in a schematic two-dimensional representation a rotation of the two MEMS dies, the sensor package, as well as a non-orthogonality of the axes within a die.
• 3 illustrates the method according to the invention.

Embodiments of the invention

In 1 Two axis systems 21, 22 of an inertial sensor system 1 are shown with slight transverse axis sensitivity. The sensor system 1 includes an acceleration sensor and a rotation rate sensor, the two sensors each having three measurement channels with which the accelerations and rotation rates can be determined along three spatial axes. The axis system 22 of the acceleration sensor and the axis system 21 of the rotation rate sensor are ideally aligned orthogonally and parallel to one another. The illustrated non-orthogonality of the axis system 22 represents a deviation from such an ideal configuration and manifests itself as cross-axis sensitivity, through which, for example, an acceleration that should only cause a signal with respect to the X-axis also occurs a certain measurement component leads along the Y and/or Z axis.

wobei die Matrix M_scf (scf für „scale-factor“) die Empfindlichkeiten bezüglich der drei Achsen angibt, M_non („non-orthogonality“) die Nicht-Orthogonalität der Sensorachsen und die Matrix M_mis („misalignment“) die Verdrehung der Sensorachsen gegenüber den Stimulus-Achsen. Dabei ist jedoch im Normalfall keine Zuordnung zwischen den Subkomponenten und einzelnen physikalischen Effekten möglich, da sich die Effekte gegenseitig überlagern. 2 zeigt schematisch wie beispielsweise der Beschleunigungs-MEMS-Die 18 und Drehraten-MEMS-Die 17 leicht verschoben zueinander sein können. Das gesamte Sensorsystem 1 kann ebenfalls zum (als perfekt angenommenen) äußeren Stimulus 2 verdreht sein, der entlang der Achsen 3, 4 wirkt. Innerhalb eines Die 17, 18 kann außerdem eine Nicht-Orthogonalität der Achsen auftreten (angedeutet durch die Achsen in 17 und 18), z.B. indem elektrisch ein Signal von einem Kanal in den anderen überkoppelt oder der physikalische Messeffekt selbst eine Querachsenempfindlichkeit hat. 2 uses a schematic, two-dimensional representation to show a rotation of the two MEMS dies, the sensor package, and the non-orthogonality of the axes within a die. The cross-axis sensitivity of a sensor can have various physical origins, whereby the cross-axis sensitivity matrix S _cas (cas for “cross-axis sensitivity”) can be mathematically decomposed into the following components: $${\text{s}}_{\text{cas}}={\text{M}}_{\text{scf}}\cdot {\text{M}}_{\text{non}}\cdot {\text{M}}_{\text{mis}},$$

where the matrix M _scf (scf for "scale-factor") indicates the sensitivities regarding the three axes, M _non ("non-orthogonality") the non-orthogonality of the sensor axes and the matrix M _mis ("misalignment") the rotation of the Sensor axes versus stimulus axes. However, normally no association between the subcomponents and individual physical effects is possible because the effects overlap each other. 2 shows schematically how, for example, the acceleration MEMS die 18 and rotation rate MEMS die 17 can be slightly shifted from one another. The entire sensor system 1 can also be rotated to the external stimulus 2 (assumed to be perfect), which acts along the axes 3, 4. Within a die 17, 18, a non-orthogonality of the axes can also occur (indicated by the axes in 17 and 18), for example by electrically coupling a signal from one channel to the other or by the physical measurement effect itself having a cross-axis sensitivity.

3 illustrates an embodiment of the method 10 according to the invention. Block 4 corresponds to the initial calibration in the manufacturing process of the sensor and block 5 corresponds to the application of the self-calibration according to the invention in the finished sensor system 1. First, the transverse axis sensitivity is measured in step 6. Depending on whether it is an acceleration or rotation rate sensor, the sensor can for this purpose be exposed to accelerations or rotations along three orthogonal stimulus axes and the elements of the transverse axis sensitivity matrix can be obtained by evaluating the sensor signals caused by the movements. The compensation matrix is derived from the transverse axis sensitivity matrix and stored in a memory element of the sensor in step 7. In step 8, one or more system parameters can be measured and in step 9, together with variables derived from them, stored as reference values. As indicated by the arrow 26, steps 6, 7, 8, 9 can be repeated under changed conditions, for example by changing another system parameter such as a mechanical or thermal load. As an optional additional step 23, a sensor-specific calculation rule can be created and saved.

This is followed by further manufacturing steps summarized in step 11, such as application to a blister belt (“tape & reel”), transport and soldering into the end product and conditioning. The transverse axis sensitivity changes, so that the compensation of the transverse axis sensitivity can no longer be carried out optimally in the end product 24.

In order to ensure effective compensation again, in the method according to the invention, a current value of the system parameter is determined in measuring step 12 and the reference value determined in step 9 is read out of the memory in readout step 13. In calculation step 14, a change in the transverse axis sensitivity is calculated depending on the reference value and the current value of the system parameter, and in adaptation step 15 the compensation matrix is adapted to the change in the transverse axis sensitivity. In this way, the result 25 of the method 10 is the newly calibrated sensor system 1 with correspondingly improved transverse axis sensitivity compensation.

QUOTES INCLUDED IN THE DESCRIPTION

This list of documents listed by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Non-patent literature cited

• L. Poletti, D. S. Sanchis, R. Siryani, “A direct approach for high quality MEMS based IMU/INS production,” in 2020 DGON Inertial Sensors and Systems (ISS), 2020, pp. 1-19 [0005]
• J. Rohac, M. Sipos, J. Simanek, “Calibration of low-cost triaxial inertial sensors,” IEEE I. & M. Mag., vol. 18, no. 6, pp. 32-38 [0005]

Claims (10)

Method (10) for self-calibration of a sensor system (1), wherein the sensor system (1) has at least one sensor designed as a multi-axis inertial sensor and is configured to apply a compensation matrix to sensor data in such a way that a transverse axis sensitivity of the at least one sensor is compensated, wherein the method has the following steps: -- in a measuring step (12), at least one current value of a system parameter of the sensor system (1) is determined; -- in a readout step (13), at least one reference value of the system parameter is read out from a storage element of the sensor system (1); -- in a calculation step (14), a change in the transverse axis sensitivity of the at least one sensor is calculated as a function of the reference value and the current value of the system parameter; -- in an adaptation step (15), the compensation matrix is adapted to the change in the transverse axis sensitivity. Procedure (10) according to Claim 1 , wherein in the measuring step (12) current values of a plurality of system parameters are determined, in the readout step (13) reference values of the plurality of system parameters are read out from a memory element of the sensor system and in the calculation step (14) the change in the transverse axis sensitivity depending on the reference values and the current values of the majority of system parameters is calculated. Procedure (10) according to Claim 1 or 2 , wherein the system parameter or the plurality of system parameters comprises one or more of the following variables: -- an amplitude of a quadrature signal, -- a non-orthogonality of the sensor axes (21, 22), -- a misalignment between sensor axes (21) of the at least one sensor and Sensor axes (22) of a further sensor, -- a phase position of a sensor signal, -- a drive frequency, -- a frequency difference between a drive mode and a detection mode, -- a raw sensitivity, -- a mechanical tension, -- one by means of a built-in Test Equipment, BITE, recorded quantity, in particular a controlled change in a quadrature signal. Method (10) according to one of the preceding claims, wherein the change in the transverse axis sensitivity is calculated by means of a statistical model for the relationship between the system parameter and the transverse axis sensitivity, the statistical model being formed in particular by means of compensation calculation or machine learning. Method (10) according to one of the preceding claims, wherein the reference value of the system parameter is determined in a provision step and stored in the storage element of the sensor system (1), the provision step in particular during an end-of-line calibration following the production of the sensor system (1) is carried out. Method (10) according to the Claims 4 and 5 , wherein in the provision step at least two data points are determined by a controlled change in the system parameter (1) and a measurement of the transverse axis sensitivity and the statistical model is formed depending on the data points. Method (10) according to one of the preceding claims, wherein the sensor system (1) has at least one further sensor designed as a multi-axis inertial sensor, the sensor system (1) being configured to apply the compensation matrix to sensor data from the at least one sensor and a further compensation matrix to apply to sensor data of the at least one further sensor, the compensation matrix and the further compensation matrix being adapted in the adaptation step (15) to the change in the transverse axis sensitivity calculated in the calculation step (14). Procedure (10) according to Claim 7 , wherein the at least one sensor is a rotation rate sensor, an acceleration sensor, a multi-axis magnetic field sensor or a rotational acceleration sensor and / or the at least one further sensor is a rotation rate sensor, an acceleration sensor, a multi-axis magnetic field sensor or a rotational acceleration sensor. Method for compensating for a transverse axis sensitivity of a sensor system (1), wherein a method (10) for self-calibration of the sensor system (1) is carried out according to one of the preceding claims and in a compensation step the adapted compensation matrix is based on sensor data of the at least one sensor or the at least one further sensor is applied. Sensor system (1) having at least one sensor designed as a multi-axis inertial sensor, the sensor system (1) being compatible with this Figured is to carry out a method (10) according to one of the preceding claims.

Images