Method for manufacturing a multiple range accelerometer - Patents.com

Abstract

The present disclosure relates to a MEMS inertial sensor device including a plurality of accelerometers formed in a semiconductor chip package, the accelerometers being formed in a conductive single crystal silicon SOI wafer bonded to a cap wafer including a hermetic encapsulant. The inertial sensor chip package may include an integrated circuit configured to process the inertial sensor data.

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to U.S. Provisional Application No. 63 / 333,360, filed April 21, 2022, the entire contents of which are incorporated herein by reference.

[0002] The general technical field relates to the manufacture of micro-electro-mechanical systems (MEMS). [Background technology]

[0003] Micro-electro-mechanical systems (MEMS) are an increasingly important assistive technology. MEMS inertial sensors are used to detect changes in the state of motion of an object, including changes in position, velocity, acceleration, or orientation, and include devices such as accelerometers, resonators, gyroscopes, vibrometers, and inclinometers. Broadly described, MEMS devices are integrated circuits (ICs) that contain tiny mechanical, optical, magnetic, electrical, chemical, biological, or other transducers or actuators. MEMS devices can be manufactured using high-volume silicon wafer fabrication techniques developed over the past 50 years for the microelectronics industry. The resulting small size and low cost of high-volume silicon wafer fabrication techniques make MEMS devices attractive for use in an ever-increasing number of applications in a wide variety of industries, including consumer, automotive, medical, aerospace, defense, green energy, industrial, and other markets. Summary of the Invention

[0004] Since the sensitivity of an inertial sensor can affect performance, MEMS inertial sensor devices can be manufactured that include multiple inertial sensor elements with different operating ranges to more accurately detect the motion of an object subjected to a wide range of acceleration. A preferred example can include multiple accelerometers, where a first accelerometer measures a lower range of acceleration and a second accelerometer measures a higher range of acceleration. It is further beneficial to manufacture multiple MEMS sensors using one manufacturing process, preferably in one chip package, thereby providing alignment of the device that has a smaller area and weight and operates with less power.

[0005] The preferred embodiment may be fabricated using multiple silicon-on-insulator (SOI) wafers that may be processed and fusion-bonded together to provide a hermetically sealed chip package. Preferably, there may be two, three, or four accelerometers fabricated on one wafer, for example. Further embodiments may include different types of inertial sensors, such as gyroscopes, or may further include non-inertial sensors, such as pressure sensors, time-of-flight (ToF) sensors, and magnetometers. The sensor data may be processed and used for navigation and / or platform and image stabilization, for example, using processing circuitry that may be included in the chip package in the preferred embodiment. The processor may be programmed with sensor data fusion software to process the sensor data for a particular application. This type of sensor fusion is widely used to obtain guaranteed position, navigation, and timing (A-PNT), especially where the combination of sensors serves to extend the system's ability to maintain accurate navigation without GPS assistance (i.e., "dead reckoning") over long periods of time. In this example, the inertial sensor further serves to support, assist, and improve the performance of other sensors that are part of the sensor fusion mix, most of which are imaging sensors (including cameras, lidar, radar, and sonar). Another example includes monitoring all of the accelerometer outputs and using the output of the low-range (high-sensitivity) accelerometer unless the output of the low-range (high-sensitivity) accelerometer exceeds a pre-set threshold. At this point, the output of the high-range accelerometer may be selected. In parallel, an impulse monitor (a high-pass accelerometer operating at resonance) may be monitored. When the impulse monitor output signal amplitude rises higher than the effective filtered accelerometer average acceleration, it is recorded as an impulse force that may be included in the dynamic calculations of velocity and position.For example, some special applications, such as Health and Usage Monitoring Systems (HUMS), also called "condition monitoring" for various kinds of equipment and autonomous vehicles exposed to high-shock environments, require lower noise for higher frequency ranges and bandwidths over 10 KHz, high accuracy and very small size, weight and power (SWaP) requirements. Sensor fusion aims to reduce the complexity of various sensor data by improving the signal-to-noise ratio, reducing uncertainty, and increasing reliability, resolution and accuracy over a longer period of time. Sensor fusion is used to compensate for the weaknesses of other sensors to improve reliability, especially when data must be processed "on the fly", as is the case in the case of dead reckoning involved in navigation. When computing resources are finite, artificial intelligence and machine learning can further support the identification of the best sensor data fusion techniques based on real-time operating conditions. This is especially true when realizing sensor fusion applications related to navigation, but also in standalone Industry 4.0, Internet of Things (IoT), Internet of Moving Things (IoMT), machine vision and image processing applications.

[0006] Further embodiments include inertial sensor chip packages in which the processing circuitry may be mounted on a circuit board or may be manufactured in a system-on-ship configuration with a system controller including a clock included on the MEMS chip or co-located with the processing circuitry to control timing operations within the chip package. Such embodiments may further include a neural processing unit to perform iterative computational analysis of the sensor data and may be further configured to perform sensor fusion operations.

[0007] It is to be noted that the present invention may admit of other equally effective embodiments, and therefore the attached drawings only depict exemplary embodiments of the invention and are not to be considered as limiting the scope of the invention. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is an isometric top view of a 3DS multiple range accelerometer that includes two accelerometers covering up to two acceleration ranges. [Diagram 2] FIG. 2 is an isometric view of a two-range accelerometer with the top cap removed to show the proof mass and support springs. [Figure 3A] 3 is a cross-sectional view of the dual-range accelerometer of FIG. 2 taken along line ABC through the support springs of the two proof masses. [Figure 3BC] 3 is a cross-sectional view of the dual-range accelerometer of FIG. 2 taken along line ABC through the support springs of the two proof masses. [Figure 4] FIG. 1 is an isometric view of a two-range accelerometer with the handle layer of the top cap removed to show the electrode placement. [Diagram 5] FIG. 1 illustrates an alternative embodiment of a multiple range accelerometer including four accelerometers covering up to four acceleration ranges. [Figure 6] FIG. 4 is a cross-sectional view of FIG. 3 showing two proof masses responsive to acceleration in the negative z direction that is greater than the mechanical range of one of the accelerometers. [Figure 7] 6A-6C illustrate an embodiment of the four accelerometers of FIG. 5 illustrating various modes of operation. [Figure 8A] 1 is a schematic, partially exploded perspective view of an integrated MEMS system according to an embodiment; [Figure 8B] FIG. 8B is a schematic exploded perspective view of the integrated MEMS system of FIG. 8A. [Figure 9A] 1 is a schematic cross-sectional view of an integrated MEMS system according to another embodiment. [Figure 9B] 1 is a schematic cross-sectional view of an integrated circuit wafer. [Figure 9C] 1 is a schematic cross-sectional view of a MEMS wafer stack. [Figure 9D]9C is a schematic cross-sectional view illustrating wafer-level inversion bonding of the integrated circuit wafer of FIG. 9B to the MEMS wafer stack of FIG. 9C. [Figure 9E] FIG. 9B is a schematic cross-sectional view of the integrated MEMS system of FIG. 9A bonded to a printed circuit board (PCB). [Figure 10A] 1 is a schematic diagram of an integrated MEMS system according to a preferred embodiment; [Figure 10B] FIG. 13 illustrates a schematic diagram of a further integrated MEMS system including on-board clocking and neural processing for computational analysis and / or feedback control of system operation. [Figure 11] FIG. 2 is a process flow diagram illustrating a method of operating a MEMS transducer device according to a preferred embodiment of the present invention. [Figure 12] FIG. 2 is a process flow diagram illustrating a method of operating a proof mass MEMS device according to a preferred embodiment of the present invention. [Figure 13] 1 is a schematic cross-sectional view of an integrated MEMS system according to a possible embodiment shown bonded to a PCB. [Figure 14] FIG. 1 is an exemplary process flow diagram illustrating a software program for processing sensor data from multiple inertial and optional non-inertial sensors for various applications. [Figure 15A] FIG. 15A illustrates an example neural network that may be implemented in a neural processor for an inertial sensor according to an embodiment of the present invention. [Figure 15B] FIG. 15B illustrates an example neural network that may be implemented in a neural processor for an inertial sensor according to an embodiment of the present invention. [Figure 16A] FIG. 1 illustrates the integration of an inertial measurement device for an autonomous vehicle according to various embodiments of the present invention. [Figure 16B] FIG. 1 illustrates the integration of an inertial measurement device for an autonomous vehicle according to various embodiments of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] The systems and methods described herein relate to the fabrication and use of inertial sensors capable of simultaneously detecting multiple acceleration ranges in a hermetic package using a silicon-on-insulator (SOI) wafer. As used herein, "g" refers to a magnitude of acceleration equal to the acceleration due to the Earth's gravity.

[0010] In many applications, especially those involving harsh or high shock environments, it is desirable to include a high resolution (e.g., navigation grade with drift <100ug) low range (e.g., range = 1-10g) accelerometer for navigation or tracking. However, shocks due to collisions or disturbances can easily overdrive these low range accelerometers, resulting in corruption or disruption in data used for, for example, navigation. To navigate through shocks, it is desirable to include at least one additional higher range (e.g., 10-100g or >100g) accelerometer. Integrating multiple accelerometers for use in one system introduces the complexity and cost of coaxially aligning multiple devices. In addition to these alignment issues, the use of multiple accelerometer chips adds additional area and weight to the sensor system footprint. In many modern navigation systems, such as those used in drones and autonomous vehicles, space can be at a premium. Finally, high precision low range accelerometers often use different sensing technology than high g accelerometers. For example, most low g accelerometers use capacitive sensing, whereas most high g accelerometers use piezoelectric or piezoresistive sensors. This can add complexity to navigation system design. It is therefore desirable to have multiple range accelerometers aligned and fabricated on the same integrated circuit substrate, preferably using a single process flow.

[0011] Many MEMS devices cannot withstand harsh or high shock environments. One of the main failure modes is the breakage of the proof mass from its support spring due to overstretching of the spring during an impact event. Spring overstretching is difficult to prevent using conventional packaging. Therefore, it would be further desirable to include a multiple range accelerometer packaged in a manner that prevents overstretching of the support spring.

[0012] In the following description, like features in the drawings are given like reference numerals, and some reference numerals may be omitted when already identified in previous drawings, so as to ensure the clarity of the drawings. It should also be understood that elements in the drawings are not necessarily drawn to scale, with an emphasis on clearly showing the elements and structures of the present embodiment.

[0013] Hermetically sealed MEMS inertial sensor embodiments According to an aspect, a method for manufacturing a multiple range accelerometer in a shock resistant wafer level package is provided.

[0014] FIG. 1 illustrates an exemplary embodiment of a 3DS (3D system) MEMS multiple range accelerometer 100 fabricated using the 3DS fabrication process described herein. This particular embodiment includes two accelerometers 200 and 300 fabricated together on the same silicon chip to enable measurement of acceleration in two different ranges. As shown in FIG. 1, the accelerometer 100 may include a top cap wafer 101, a MEMS wafer 110, and a bottom cap wafer 120. The embodiment of FIG. 1 is illustrative of a fabrication method that may be applied to any number of accelerometers, although other embodiments may be realized that include additional accelerometers with additional ranges. FIG. 2 illustrates the embodiment of FIG. 1 with the top cap wafer 101 removed to reveal the accelerometer proof masses 210, 310.

[0015] The 3DS process flow is now briefly described herein. Relevant process flows include those using multiple SOI wafers to fabricate MEMS inertial sensors, as described in U.S. Patent No. 10,407,299, issued September 10, 2019, and further described in U.S. Patent No. 10,273,147, issued April 30, 2019, suitable for use with the present system and method, the entire contents of each of which are incorporated herein by reference. Some of the key features of the 3DS architecture are shown in FIG. 3, which shows a cross section of FIGS. 1 and 2 taken along line ABC through the support spring. The 3DS device architecture comprises a MEMS wafer 110, a top cap wafer 101, and a bottom cap wafer 120. In a preferred embodiment, the process may include fabricating features in multiple SOI wafers that may be bonded together, for example by high temperature fusion bonding, to form one or more inertial sensors in a chip package. The MEMS wafer 110 includes one or more MEMS structures that may include or be embodied by any sensing and / or control element or combination thereof, such as, for example, but not limited to, a membrane, a diaphragm, a proof mass, an actuator, a transducer, a microvalve, a micropump, etc. The MEMS wafer 110 includes opposing first and second sides. The top cap wafer 101 and the bottom cap wafer 120 are bonded to the first and second sides of the MEMS wafer, respectively. The top cap wafer 101, the bottom cap wafer 120, and the MEMS wafer 110 are stacked along a stack axis, and together they may form one or more sealed cavities 201, 301 that surround the MEMS structures. The MEMS structures may comprise substructures or elements housed in a cavity or chamber of a device.At least one of the top cap wafer 101 and the bottom cap wafer 120 is a silicon-on-insulator (SOI) cap wafer comprising a cap device layer 102, a cap handle layer 104, and a cap insulating layer 103 interposed between the cap device layer 102 and the cap handle layer 104. An inner side of one of the cap device layer or the cap handle layer is bonded to the MEMS wafer, and an outer side of the other of the cap handle layer or the cap device layer is provided with an outer electrical contact 106 formed thereon. At least one electrical conductive path 105 extends through the cap handle layer and through the cap device layer of the SOI cap wafer to establish an electrical connection between one of the outer electrical contacts 106 and the MEMS structure. Although FIG. 3 shows the top cap wafer 101 as including a cap device layer, a cap insulating layer, and a cap handle layer, it should be understood that the bottom cap wafer 120 may be similarly constructed. Preferably, both the top and bottom cap wafers are SOI wafers.

[0016] The electrical conductive path 105 may include a conductive shunt 107 formed through the cap insulating layer 103 that electrically connects the cap handle layer 104 and the cap device layer 102. The conductive shunt may be formed by etching a via or a small area in the cap insulating layer 103 and depositing a conductive material therein to electrically connect the cap device layer 102 and the cap handle layer 104 of the SOI cap wafer 101, 120. The electrical conductive path 105 further comprises a pillar 108 formed in the cap handle layer 104, the pillar being represented by a closed loop trench patterned through the entire thickness of the cap handle layer 104. In this embodiment, one of the electrical contacts 106 is located on top of the pillar 108.

[0017] In some embodiments, the electrical conductive path 105 includes a pad 109 formed in the cap device layer 102, the pad being represented by a trench patterned through the entire thickness of the cap device layer 102, and the pad being aligned with the pillar. Note that the phrase "aligned with" means that the pad 109 and the pillar 108 are opposite (facing) each other along an axis parallel to the stack axis, such that at least a portion of a surface of the pad 109 faces at least a portion of the pillar 108.

[0018] In some embodiments, the MEMS wafer is a SOI MEMS wafer 110 comprising a MEMS device layer 111 bonded to a top cap wafer 101, a MEMS handle layer 113 bonded to a bottom cap wafer 120, and a MEMS insulating layer 112 interposed between the MEMS device layer 111 and the MEMS handle layer 113.

[0019] 2 and 3, the MEMS wafer 110 comprises an outer frame 201 and a MEMS structure comprising at least one proof mass 210, 310 suspended by springs 220, 320. The proof mass is patterned on both the MEMS handle layer 113 and the MEMS device layer 111, with the springs 220, 320 patterned on the MEMS device layer 111. The at least one proof mass includes a conductive shunt 117 electrically connecting the MEMS device and handle layers, and an electrical conductive path 105 connects one of the electrical contacts to the MEMS structure via at least one of the springs. In a possible embodiment, the electrical conductive paths connecting the outer electrical contacts located on the SOI cap wafer to the MEMS structure include posts patterned in the cap handle layer, pads patterned in the cap device layer and conductive shunts formed in the cap insulating layer to connect the posts to the pads, pads patterned in the MEMS device layer that are part of the outer frame, and springs patterned in the MEMS device layer that suspend the MEMS structure in a sealed cavity.

[0020] In a possible embodiment of the multiple range accelerometer, the cap device layer 102 includes cap electrodes 240, 340 patterned therein. In some embodiments, the 3D MEMS device includes further electrical conductive paths that are not connected to the MEMS structure, but are connected to electrodes provided on one of the caps. These further electrical conductive paths extend through the cap handle layer and through the cap device layer. Some of the paths that extend into the cap may be referred to as "cap feedthroughs." At least some of the further electrical conductive paths establish electrical connections between a subset of the outer electrical contacts and the cap electrodes. The cap electrodes may be located on any one of the cap wafers, and preferably on both caps.

[0021] Referring again to Figures 2 and 3, the multiple range accelerometer includes two or more proof masses 210, 310, each suspended by one or more springs or flexures 220, 320. The sensitivity of each proof mass to acceleration is the displacement of the proof mass in response to acceleration. For a motion Δz in response to acceleration in the z-direction perpendicular to the plane of the device, the sensitivity Sz = Δz / a = M / K, where a is the acceleration, M is the mass of the proof mass, and K is the effective spring constant of the support spring. The lateral dimensions of the springs and / or proof masses of each accelerometer can be designed to set the sensitivity of each of the proof mass and spring assemblies to different desired ranges without changing the thicknesses of the MEMS device layer and the MEMS handle layer, and therefore without changing the process flow. In other embodiments, the thicknesses of the device layer, the insulation layer, and the handle layer can be configured for a particular application. In one example, the proof mass of a MEMS sensor device may comprise a device layer having a thickness in the range of 1 micrometer to 100 micrometers that is patterned to form a moving mass when released from an underlying insulating layer. See, for example, the MEMS magnetometer shown and described in U.S. Patent No. 11,287,486, issued March 29, 2022, the entire contents of which are incorporated herein by reference. In a similar manner, for the embodiment shown in Figures 2 and 3, acceleration along x and y in the plane of the spring can be calculated as the angular rotation Δθ of the proof mass, and S x,y =Δθ / a=J / κMr 2 where J is the moment of inertia of the proof mass, r is the radius of gyration of the proof mass (i.e., the distance from the axis of rotation to the center of mass), and κ is the effective torsional spring constant of the support springs. Both J and κ can be adjusted in the design by changing the lateral dimensions of the springs or the proof masses without changing the process flow.

[0022] 4 shows a possible embodiment of the multiple range accelerometer 100 with the top cap handle layer 104 and the top cap insulating layer 103 removed to show an embodiment of the electrode arrangement 240, 340. A similar set of electrodes is present in the device layer of the bottom cap wafer 120. The vertical displacement Δz or angular displacement Δθ of each proof mass can be measured by measuring the differential capacitance between selected pairs of electrodes. These cap electrodes 240, 340 are electrically connected to leads 230, 330 that extend perpendicular to the stack axis, which form part of corresponding further electrical conductive paths.

[0023] The number of accelerometer units can be easily increased by adding them to the sensor design layout, and the acceleration detection range of each accelerometer unit can be easily adjusted by changing the lateral dimensions of the spring and mass. FIG. 5 shows an embodiment of a multi-range accelerometer 500 including four independent accelerometer units 510, 520, 530, 540, each covering a different range of acceleration values. The multi-range approach using the 3DS architecture has several advantages. The first advantage is that the accelerometers are aligned to photolithographic accuracy. Previously, multiple conventional accelerometers were used to cover multiple ranges, with each conventional accelerometer provided in an individual package, and therefore each conventional accelerometer must be mechanically aligned with the others and mounted in an additional package. This can result in alignment errors between the accelerometers, as well as additional size and cost to the assembly.

[0024] The resolution of an accelerometer is limited by its range. Due to electronic noise and limitations of A / D (analog-to-digital) conversion, an accelerometer may not be able to measure more than 10 5It is unusual to have a resolution much finer than a part in a second. Thus, a + / -1g accelerometer may be capable of achieving a resolution of 10ug, but if the measurement range is + / -1000g, the resolution is more like 10mg. Therefore, it is desirable to keep the range low (high accuracy), but if the sensor exceeds its range, the sensor may stop working correctly and acceleration data may be lost. In the extreme case of a very large acceleration or shock impulse, the proof mass may break off from the spring, permanently damaging the sensor.

[0025] FIG. 6 illustrates the embodiment of FIGS. 2, 3, and 4 under acceleration in the z-direction. In this embodiment, the springs 220 in one proof mass 210 are designed to be 100 times "softer" than the springs 320 in the proof mass 310. That is, the accelerometer 200 has a range of, for example, + / -1g, and the accelerometer 300 has a range of, for example, + / -100g. At low accelerations, the system may use the signal from the accelerometer 200 since it has a higher sensitivity. However, if an impact or high g acceleration occurs, the accelerometer 200 may exceed its range and provide an erroneous or disturbed signal output. In such an example, the system may use the signal from the accelerometer 300 during high g acceleration or impact. In some embodiments, the 3DS architecture may have built-in shock protection. Because the capacitor measurement gaps above and below the proof mass are only about 1-5 microns wide, the proof mass 210 cannot move more than a few microns (even when the proof mass 210 is overdriven) before it contacts the top cap wafer 101. This limited range of motion protects the sensor from damage due to impact by preventing overstretching of the springs or high speed collisions of the proof mass against surrounding features.

[0026] The 3DS multiple range accelerometer further allows for several accelerometer operating modes. Figure 7 is a block diagram of a multiple range accelerometer 500 with four accelerometers 510, 520, 530, 540 similar to the embodiment in Figure 5, where accelerometers A1 and A2 (510, 520) are low range accelerometers and A3 and A4 (530, 540) are high range accelerometers. In some embodiments, the accelerometers can be operated in one of three modes: low pass, high pass, and trimmed. In Figure 7, all four accelerometers are driven using an AC drive signal (505). The differential capacitance signal from each of the accelerometers is amplified by differential amplifiers DA1 511, DA2 521, DA3 531, and DA4 541. In the embodiment of Figure 7, the outputs of DA1 and DA3 are passed through respective low pass filters 512, 532. This signal, conditioned by a low pass filter, provides filtered and averaged values ​​513, 533 of the acceleration, respectively, in a manner used by most commercially available accelerometers. The low pass filtering filters out impulses and vibrations. The output of DA3 is run through a high pass filter 522. This allows the accelerometer to provide a higher frequency output response, which allows the measurement of vibrations, shocks, and impulses 523 that are longer than the resonant period of the accelerometer proof mass. This is a typical configuration for a vibration sensor or vibrometer. Accelerometer A4 540 is shown operated in impulse mode. Here, the output of DA4 541 is monitored at the resonant frequency of the proof mass where an amplitude of resonance 542 greater than a threshold is generated. A lock-in amplifier (LI) can be used to provide a signal, in this case the resonant frequency is used as a reference. A high, short duration impulse (frequency content higher than the resonant frequency of the proof mass) moves the proof mass, causing it to ring at the resonant frequency. The magnitude of the impulse can be calculated by monitoring the amplitude of the ringing 543. The embodiment shown in Figure 7 illustrates how the 3DS multiple range accelerometer architecture can accommodate multiple measurement modes as well as multiple ranges of acceleration.The specific association of each accelerometer in FIG. 7 to any type of block circuit diagram is not intended to be limiting, in that any circuit or combination of circuits may be used with any accelerometer in the multiple range accelerometer sensor depending on the measurement requirements.

[0027] The multiple accelerometer chip packages described herein may be used in combination with gyroscopes and other sensor components for inertial measurement units, such as those described in U.S. Patent No. 10,214,414, issued February 26, 2019, and U.S. Application No. 14 / 622,548, filed February 13, 2015, the entire contents of which are incorporated herein by reference. The chip components, which may include internal clocks for controlled timing of component operations, may be assembled in vertical stacks and / or lateral arrays and used for many applications for inertial navigation and stabilization of sensor platforms, autonomous vehicles, aircraft, unmanned aerial vehicles (UAVs), satellites, such as those described in International Application No. PCT / US2017 / 015393, filed January 27, 2017, published as WO2017 / 132539, the entire contents of which are incorporated herein by reference.

[0028] With reference to Figures 8A and 8B, a possible embodiment of an integrated MEMS system 1000 is shown. The system 1000 includes an architecture that may allow the MEMS sensor function to be integrated into a single MEMS chip, while the electronic function may be integrated into one IC chip. This architecture allows the transmission of auxiliary signals through the MEMS chip to be processed by the IC chip. The MEMS chip includes multiple insulated conductive paths, some of which run through the entire thickness of the MEMS chip, allowing for wirebond-free electrical connections to the IC chip. The auxiliary signals are non-MEMS signals, i.e., external to the MEMS chip, provided by another component, such as a PCB. The IC chip can be inverted chip bonded on top of the MEMS chip at the chip or wafer level, forming an integrated MEMS system, eliminating much of the cost of MEMS and IC integration, as well as the packaging complexity and costs mentioned above. The MEMS system may further include several single MEMS chips stacked vertically, and in some cases, more than one integrated circuit chip, each chip including an independent sensor element.

[0029] In this example, the integrated MEMS system 1000 comprises a single MEMS chip 1100 comprising a first cap layer 1120, a central MEMS layer 1160 and a second cap layer 1140. The layers 1120, 1160 and 1140 are made of a conductive material, for example silicon with a doping level sufficient to conduct electrical signals through the regions of the silicon that comprise the electrical contacts. The first cap layer 1120 is electrically bonded to a first side 1161 of the central MEMS layer 1160 and the second cap layer 1140 is electrically bonded to a second side 1162 of the central MEMS layer 1160 opposite the first side 1161. The central MEMS layer 1160 is located between the first cap layer 1120 and the second cap layer 1140 and is made from a silicon-on-insulator (SOI) wafer including a device layer 1164, a handle layer 1168 and an insulating layer 1166. The first cap layer 1120, the central MEMS layer 1160 and the second cap layer 1140 are fabricated from respective silicon-based wafers bonded at the wafer level as described in further detail below. The insulating layer 1166 is a SOI buried oxide layer between the SOI device layer 1164 and the SOI handle layer 1168. Conductive shunts are fabricated through the buried oxide layer 1166 to realize electrical connections between the SOI device layer 1164 and the SOI handle layer 1168 at specific desired locations, for example as part of an isolated conductive path.

[0030] At least one transducer 1170 is formed in the first cap layer 1120, the central MEMS layer 1160, and the second cap layer 1140 for generating motion or for detecting at least one parameter. The transducer can be a sensor, e.g., a motion sensor, or an actuator, e.g., a microswitch. The one or more transducers include a MEMS structure, e.g., a proof mass for a motion sensor, or a membrane for a pressure sensor or magnetometer. The architecture of the MEMS chip 1100, including two outer caps and a central MEMS layer interconnected and made of conductive materials, allows for the inclusion of several different types of transducers in a single MEMS chip. A MEMS structure is patterned on the central MEMS layer 1160, and a first electrode set 1180 and a second electrode set 1182 are patterned on the first layer 1120 and the second layer 1140 and are operatively linked (e.g., magnetically, capacitively, electrically, etc.) to the MEMS structure. Thus, a "single MEMS chip" is a chip incorporating one or more MEMS transducers patterned on the two cap layers 1120 and 1140 and the central MEMS layer 1160. As opposed to including multiple MEMS chips adhesively bonded side-by-side on a substrate with each chip containing a different MEMS sensor, the different MEMS features of the transducer (i.e., electrodes, proof masses, membranes, leads, etc.) are patterned in the same silicon wafer. For example, the present architecture allows for patterning MEMS features for measuring acceleration, angular velocity, and magnetic field along three different axes, along with other sensors, such as pressure sensors, in the same three wafer layers and therefore in the same MEMS chip.

[0031] 8A and 8B, the first cap layer 1120 includes electrical contacts 1124, 1126 on its exterior, preferably located around the periphery of the MEMS chip 1100. These electrical contacts 1124, 1126 of the first cap layer are referred to as first cap MEMS electrical contacts. The second cap layer 1140 further includes electrical contacts 1144 on its exterior, referred to as second cap MEMS electrical contacts. The first cap MEMS electrical contacts 1124, 1126 and second cap MEMS electrical contact 1144 are typically bond pads.

[0032] The single MEMS chip 1100 further includes a plurality of insulated conductive paths 1130, 1150 extending through one or more of the first cap layer 1120, the central MEMS layer 1160, and the second layer 1140. Thus, the MEMS chip includes electrically isolated "three-dimensional through-chip vias" (3DTCVs) for transmitting signals through the MEMS wafer layer 1160 to the cap layers 1120, 1140, and through the cap layers 1120, 1140 to bond pads 1124, 1144 on the outside of the MEMS chip. The insulated conductive paths 1130, 1150 may extend in one or more directions. The insulated conductive paths 1130, 1150 may be formed using a silicon "trench-and-fill" process. The insulated conductive paths 1130, 1150 are typically formed by an insulated closed-loop trench 28 surrounding a conductive wafer plug 26. The trenches 28 are filled with an insulating material and the conductive wafer plugs 26 allow the transmission of electrical signals. The insulated conductive paths 1130, 1150 include portions that extend into one or more layers and are aligned to the layer interfaces, allowing the conduction of electrical signals through the MEMS chip 1100. Some of the insulated conductive paths connect one or more transducers to the first cap MEMS electrical contacts, i.e., some of the first set of contacts 1124. These insulated conductive paths are referred to as first insulated conductive paths 1130. They conduct electrical MEMS signals between the transducer 1170 and the first cap MEMS electrical contacts of said first set 1124. More specifically, the insulated conductive paths 1130 connect the electrodes, leads, and / or MEMS structures of the transducer 1170 to the first cap MEMS electrical contacts 1124. Other insulated conductive paths extend through the entire thickness of the single MEMS chip 1100, i.e., through the first cap layer 1120, through the central MEMS layer 1160, and through the second cap layer 1140. These insulated conductive paths connect a second set of the first cap MEMS electrical contacts 1126 to some of the second cap MEMS electrical contacts 1144.They are referred to as second isolated conductive paths 1150 and function to conduct auxiliary signals, such as power or digital signals, through the MEMS chip 1100. The second isolated conductive paths 1150 provide an isolated path between the metallization and bond pads in the first cap layer 1120 and the bond pads in the second cap layer 1140 to transmit signals from the IC chip 1200, through the MEMS chip 1100, to another IC chip or to a PC board.

[0033] 8A and 8B, the integrated MEMS system 1000 further includes one IC chip 1200. The IC chip 1200 is typically an application specific integrated circuit (ASIC) chip fabricated using complementary metal oxide semiconductor (CMOS) technology, although other types of ICs are possible. The IC chip 1200 includes a MEMS signal processing circuit 1240 operably connected to the first isolated conductive path 1130 for processing the electrical MEMS signals of the one or more transducers 1170. The IC chip 1200 further includes an auxiliary signal processing circuit 1260 operably connected to the second isolated conductive path 1150 for processing auxiliary signals and for providing additional system functions. Management functions performed by the IC may include interpretation of sensor data, compensating for variations in sensor response due to temperature or other environmental variations, microcontroller management of system timing and functions, memory for storage of data such as calibration constants, sensor interpretation constants, and measured data, as well as wired and wireless data I / O interfaces to external devices such as drones or autonomous vehicles.

[0034] The MEMS signal processing circuit 1240 manages the data signals to and from the MEMS transducer 1170. The MEMS signal processing circuit 1240 controls and provides the analog drive and feedback signals required by the transducer, controls the timing of signal measurements, amplifies, filters, and digitizes the measured signals, and analyzes and interprets the incoming MEMS signals from the transducer 1170 to calculate different parameters, such as angular acceleration or environmental air pressure. The MEMS signal processing circuit 1240 typically includes at least A / D and D / A converters, a power supply, a system controller, memory, a calibration and composition module, and a data analysis module.

[0035] The auxiliary signal processing circuit 1260 processes signals other than those strictly required to operate the MEMS transducers and output measured MEMS signals. The auxiliary signal processing circuit 1260 may further provide additional system functions, such as monitoring sensor activity to minimize power usage, wirelessly transmit and receive data, receive and interpret GPS signals, integrate additional data from other sensors or the GPS for calibration or performance improvement, use the measured data to calculate additional system parameters of interest, or to trigger other system activities. When fully utilized, the auxiliary signal processing circuit 1260 enables the integrated 3D system 1000 to control, perform, and analyze measurements from the integrated MEMS sensors; act as a sensor hub between the 3DS system chip, other coupled external sensors, and a larger external system (e.g., a mobile phone, game controller, or display); and integrate all the data to make decisions or provide input to the larger system, since the larger system may receive, process, and transmit signals other than MEMS signals, for example, from / to a PCB board. The MEMS chip also acts as a "smart" interposer between the PCB and the IC chip. Digital and / or analog signals can pass through the MEMS chip for processing by the auxiliary circuitry 1260 for use by the MEMS transducer 1170 (e.g., for power signals), and can be sent back through the MEMS chip or transmitted wirelessly.

[0036] Thus, the IC chip 1200 includes IC electrical contacts bump-bonded to the MEMS electrical contacts of the first cap layer 1120. The IC electrical contacts are grouped into a first set 1228 and a second set 1230 that are bump-bonded to the first set 1124 and the second set 1126 of the first cap MEMS electrical contacts, respectively. In other words, the set 1128 of IC electrical contacts is connected to the set 1124 of MEMS electrical contacts, thereby connecting the first isolated conductive path 1130 to the MEMS signal processing circuitry 1240. The set 1230 of IC electrical contacts is connected to the set 1126 of MEMS electrical contacts, thereby connecting the second isolated conductive path 1150 to the auxiliary signal processing circuitry 1260. Typically, the MEMS electrical contacts of the first and second cap layers are bond pads.

[0037] 9A, there is shown another possible embodiment of the integrated MEMS system 2000. The exemplary 3DS MEMS chip 2100 is a hermetically sealed 9 degree of freedom (DOF) MEMS sensor chip including a 6DOF inertial sensor 2172 for measuring x, y, and z acceleration and angular rate as well as a 3-axis magnetometer 2176, all monolithically fabricated on the MEMS chip 2100.

[0038] The 6DOF inertial sensor 2172 detects three axes of linear acceleration and three axes of angular velocity. The 6DOF inertial sensor 2172 includes a first electrode set 2180 and a second electrode set 2182 provided on the first cap layer 2120 and the second cap layer 2140, respectively. One or more proof masses 2163, 2165 may be patterned on the central MEMS layer 2160, with the first electrode set 2180 and the second electrode set 2182 forming a capacitor with the proof mass. In FIG. 9A, only two proof masses 2163, 2165 are visible, but the 6DOF inertial sensor 2172 may include more proof masses. The ultimate resolution of the MEMS inertial sensor is set by the noise density to be shorter than the averaging period (<1 second) and longer by the bias stability, which is roughly proportional to the noise density. IMU noise density consists of two parts: electrical noise density, which mainly comes from the integrated circuit, and mechanical noise density, which comes from the MEMS sensor. Large MEMS sensor sensitivity, which is proportional to the Coriolis force 2MωΩ for gyroscopes (where M is the mass, ω is the drive frequency, and Ω is the angular velocity), or linear force Ma for accelerometers (where M is the mass and a is the acceleration), minimizes IC noise. Thermal noise of the MEMS sensor itself is inversely proportional to the mass. Thus, large mass is important to reduce overall noise. The 6DOF inertial sensor 2172 includes a large proof mass 2163, 2165, and a sensing capacitor 2180 that is hermetically vacuum sealed at the wafer level. It is important to keep the MEMS sensor area small for most applications, and therefore the disclosed sensor system maximizes the inertial mass by increasing its thickness. Using the disclosed architecture, the proof mass typically has a thickness of 400 μm, but can range from 100 μm to 1000 μm, compared to other MEMS inertial sensors that have a thickness of 40 μm or less.Large proof masses are typically fabricated on silicon-on-insulator (SOI) wafers that include a handle that can be 100 μm to 1000 μm thick, a buried oxide layer that is 1 μm to 5 μm thick, and a single crystal silicon (SCS) device layer that is 1 μm to 20 μm thick. The bulk of the proof mass is etched into the handle wafer using deep reactive ion etching (DRIE) of silicon. Alternatively, the proof mass can be formed with a thicker SCS device layer, for example in the range of 20 μm to 100 μm for specific applications.

[0039] Therefore, the mass of the proof varies depending on the lateral dimensions (e.g., 0.5 mm to 4 mm, or 1 mm 2 ~3mm 2 The resonant frequency (√(k / M) can be independently tuned by adjusting the spring constant k through the thickness of the device layer and the width and length of the spring. The spring constant k is determined by the area of ​​the wt 3 / L 3where w, t, and L are the width, thickness, and length of the spring, respectively. For accelerometers, lower frequencies (long, thin springs) of around 1000 Hz are desired, whereas for gyroscopes, higher frequencies (shorter, wider springs) are desired. Generally, resonant frequencies between 500 Hz and 1500 Hz are used for various applications. Capacitor electrodes and gaps are etched into the face of a cap wafer that is bonded to the MEMS wafer. The gaps are typically 1 μm to 5 μm thick, providing a sensing capacitor that can range from 0.1 picofarads to 5 picofarads. Further details regarding the fabrication and operation of MEMS transducer devices may be found in U.S. patent application Ser. No. 14 / 622,619, filed Feb. 13, 2015 (now U.S. Patent No. 9,309,106), and U.S. patent application Ser. No. 14 / 622,548, filed Feb. 13, 2015, the above-referenced patents and applications being incorporated herein by reference in their entireties.

[0040] For industrial and tactical grade applications involving high resolution motion capture and personal navigation, the thick mass and high quality factor (approximately 5000) of the fabrication results in a gyroscope noise density ranging from 0.005 deg / hr to 0.1 deg / hr. The resulting gyroscope bias stability is in the range between 0.05 deg / hr to 1 deg / hr. This noise is less than many fiber optic and ring laser gyroscopes that cost thousands of dollars more. Because existing consumer grade MEMS gyroscopes use inexpensive packaging and contain small inertial masses and sensing capacitors, they have low quality factors and low angular rate sensitivity, resulting in a large noise density of approximately 1 deg / hr and bias stability of approximately 10 deg / hr that is unsuitable for tactical and navigation use. Similarly, the accelerometer has a noise density ranging from 3 micro-g / Hz to 30 micro-g / Hz, and a bias stability ranging from 0.5 micro-g to 10 micro-g, much lower than consumer grade accelerometers. The platform further allows for the addition of other sensor types, such as pressure sensors (shown here as 3-axis magnetometer 2176) and magnetometers, to improve overall accuracy through sensor data fusion. The sensor data can be processed by a data processor circuit integrated with the MEMS and IC chips described herein or by an external processor. For navigation grade applications, including high performance unmanned vehicles and autonomous navigation, the two masses can be combined in anti-phase drive mode to increase the quality factor by reducing mechanical energy losses as well as increasing the effective mass by a factor of two. This approach can result in, for example, gyroscope noise density ranging from 0.002° / hr to 0.01° / hr and bias stability ranging between 0.01° / hr to 0.1° / hr, thereby providing improved gyroscope performance.

[0041] The MEMS chip 2100 includes a first insulated conductive path 2130 and a second insulated conductive path 2150 similar to those previously described. The first insulated conductive path 2130 connects the MEMS electrodes 2180, 2182 to a first set of MEMS electrical contacts 2124 in the first cap layer 2120. The second insulated conductive path 2150 extends through the entire thickness of the MEMS chip 2100 and allows for the transmission of auxiliary (or additional) signals through the MEMS chip 2100. The second insulated conductive path 2150 connects a second set of MEMS electrical contacts 2126 of the first cap layer 2120 to some of the MEMS electrical contacts 2144 of the second cap layer 2140. For clarity, only some of the first isolated conductive paths are shown in FIG. 9A, e.g., paths 2130a, 2130d extending between the second cap electrode 2182 and the MEMS electrical contact 2124 of the first cap layer 2120, and paths 2130b and 2130c connecting the first cap electrode 2180 patterned on the first layer 2120 to the MEMS electrical contact 2126 of the same layer 2120. Similarly, only some of the second isolated conductive paths are shown in FIG. 9A, e.g., paths 2150a and 2150b connecting the electrical contacts 2124, 2126 in the first cap layer 2120 to the electrical contact 2144 in the second cap layer 2140.

[0042] Referring again to FIG. 9A, the single MEMS chip may further include a transducer that is a non-inertial sensor. Examples of possible non-inertial sensors include pressure sensors, magnetometers, thermometers, microphones, microfluidic and micro-optical devices. Other types of non-inertial sensors are possible. The non-inertial sensor includes a non-inertial electrode patterned on at least one of the first layer and the second layer. The non-inertial sensor further includes at least one MEMS structure patterned on the central MEMS layer that may include the non-inertial electrode. Examples of MEMS structures in non-inertial sensors include membranes such as those used in pressure sensors, microphones, or magnetometers. Some of the first insulated conductive paths in the MEMS chip connect the non-inertial electrodes to at least some of the first cap MEMS electrical contacts to transmit signals from the non-inertial electrodes to bond pads of the first layer of the MEMS chip that are further connected to the IC chip.

[0043] 9A embodiment, the non-inertial sensor is a three-axis magnetometer 2176 that can be used to improve the accuracy of the inertial sensor 2172. The IC electrical contacts 2228, 2230 (e.g., IC I / O bond pads) of one IC chip 2200 are bonded directly to the MEMS electrical contacts 2126, 2124 (e.g., MEMS I / O bond pads) of the single MEMS chip 2100 to reduce electrical noise and eliminate wire bonding. The magnetometer 2176 includes non-inertial electrodes, such as electrode 2184, and resonant membranes 2167, 2169.

[0044] Analog data may be communicated between the MEMS sensors 2172, 2176 and the IC chip 2200 at an analog-to-digital converter (ADC) input / output mixed signal stage of the IC chip 2200. The MEMS signals generated by the sensors 2172, 2176 are analog signals, and therefore the MEMS signals are converted to digital by the ADC for further processing in the digital CMOS portion of the IC chip 2200. Data processing of the MEMS signals by the IC chip 2200 may include, for example, sensor calibration and compensation, navigation calculations, data averaging, or sensor data fusion. System control may be provided by an integrated microcontroller that may control data multiplexing, timing, calculations, and other data processing. Auxiliary (or additional) signals are sent to the IC chip via additional digital I / O. The IC chip 2200 may include auxiliary signal processing circuits, such as, for example, wireless communication or GPS (Global Positioning System) functions. GPS data can further be used to augment and combine with the MEMS sensor data to increase the accuracy of the MEMS sensor chip 2100. These are merely examples and more or less functionality may be present in any particular system embodiment. As can be appreciated, in addition to providing analog sensed data via MEMS signals, the MEMS chip 2100 can further provide electronic interfaces including power, analog, and digital I / O between the MEMS system 2000 and the outside world, such as a printed circuit board in a larger system.

[0045] According to the embodiment shown in Figure 9A, a single MEMS chip 2100 is integrated into a 3D MEMS system 2000 (3DS) and functions as both an active MEMS device and an interposer for signal distribution. One possible use of the 3DS architecture includes wafer-scale integration of MEMS and ICs, as represented diagrammatically in Figures 9B-9F.

[0046] FIG. 9B is a schematic diagram of an IC wafer 2001. The IC wafer may be constructed using any one of CMOS, gallium arsenide (GaAs) or other III-V compounds, indium phosphide (InP) or other II-VI compounds, silicon carbide, or other technologies. The IC wafer 2001 includes several IC chips 2200. Each IC chip includes MEMS signal processing circuitry 2240 and auxiliary processing circuitry 2260 formed by IC transistors. Functions included in the IC chips may include GPS, RF, logic, and / or memory. The IC wafer 2001 further includes interlevel metal interconnects and IC electrical contacts, typically bond pads. The IC electrical contacts are grouped into a first contact set 2228 and a second contact set 2230, with the IC contacts in the first set 2228 designed to connect to MEMS electrical contacts linked to a first isolated path and the second set 2230 designed to connect to MEMS electrical contacts linked to a second isolated path.

[0047] Figure 9C is a schematic diagram of a multi-wafer stack 1001 containing several single MEMS chips, such as the MEMS chip 2100 of Figure 9A. The ASIC wafer 2001 of Figure 9B and the MEMS multi-wafer stack 1001 of Figure 9C may be fabricated in separate MEMS and IC foundries to utilize existing processes to minimize costs and increase yields. In this example, two IC chips and two MEMS chips are shown before dicing.

[0048] During the manufacturing process of the MEMS stack 1001, channels are etched in the first and second layers to define the boundaries of the electrodes, leads, and feedthroughs on the inward facing surfaces of the first and second silicon wafers. The channels are then lined or filled with an insulating material, for example thermal oxide or CVD (chemical vapor deposition) silicon dioxide. Both sides of the central MEMS wafer, typically an SOI wafer, are patterned to include electrodes and MEMS structures, for example membranes and proof masses. Conductive shunts are formed at specific locations in the buried oxide layer to allow electrical signals to travel from the devices to the handle layer through what become isolated conductive paths. The central and cap MEMS wafers are further patterned to have respective frames surrounding the MEMS structures. The various conductive paths required by the devices are constructed by aligning the feedthrough structures at each level. Since the MEMS wafers are fully contained within the stack and the isolation trenches do not need to provide a seal against atmospheric leakage like the cap trenches, some of the isolated conductive paths in the central MEMS wafer can be isolated by channels filled with insulator or by etched open trenches. The frame is further bonded to form a hermetically sealed chamber around the MEMS structure. After the wafer stack 1001 is assembled, the cap wafer is ground and polished to expose the isolated conductive regions.

[0049] 9B-9D show a preferred approach for bonding the MEMS wafer 1001 and IC wafer 2001. Underfill 44 is applied to the top CMOS wafer 2001 and patterned to expose the IC electrical contacts (in this case the bond pads). Solder bumps 45 are deposited on the bond pads. The IC wafer 2001 is flipped over and aligned to the MEMS wafer 1001 so that the IC bond pads and solder bumps are aligned to the bond pads of the first cap wafer. The IC wafer 2001 is bonded to the MEMS wafer 1001 using temperature and pressure to produce a MEMS integrated system wafer.

[0050] The bonded 3DS wafer can now be diced (along the dotted lines in FIG. 9D) into individual integrated MEMS system components, also called 3D systems on chips (3DSoC). The exposed sides of the IC chips are protected by an oxide passivation layer applied to the silicon substrate, and the MEMS / ASIC interface is protected by underfill 44. The diced chips 2000 can be treated as packaged ICs, and the bottom cap bond pads provided on the second cap can be bump-bonded to bond pads on the PCB 3001 without further packaging, as shown in FIG. 9E. PCB underfill 44 is applied to the PCB and patterned to clear contacts on the PCB bond pads. Solder bumps 45 are applied to the exposed PCB bond pads, and the diced 3DS component chips 2000 can be flip-chip bonded to the PCB 3001. If further moisture protection is desired, a polymer encapsulant or other material 34 can be applied. No further capping and bond wires are required.

[0051] FIG. 10A is a block diagram representing a preferred embodiment of an integrated MEMS system, in this example a 10 degree of freedom (10DOF) IMU system 3000. It should be noted that a 6 degree of freedom (6DOF) system including three accelerometers, one or more gyroscopes, and additional sensors as previously described herein can further be fabricated using the circuit components configured in FIG. 10A. The system 3000 includes a single 10DOF MEMS chip 3100 and an IC chip 3200, where the MEMS chip and the IC chip have an architecture similar to that described for the system 2000 of FIG. 9A. The MEMS chip 3100 includes a top cap layer, a middle MEMS layer, and a bottom cap layer with transducers patterned on the layers. The transducers can include a 3-axis accelerometer (or three accelerometers), a gyroscope, and a magnetometer, as well as a pressure sensor. A first insulated conductive path 3130 and a second insulated conductive path 3150 are formed in the MEMS layer for transmitting MEMS signals and auxiliary signals. The insulated conductive paths 3130, 3150 are connected to MEMS electrical contacts in the first cap layer and / or the second cap layer. One IC chip 3200 is bump-bonded to the first layer of the MEMS chip and includes a MEMS signal processing circuit 3240 and an auxiliary processing circuit 3260. The MEMS signal processing circuit processes the I / O signals of the transducer, i.e., signals generated by the transducer and / or signals for controlling the transducer. The auxiliary processing circuit 3260 processes the auxiliary signals, i.e., signals traveling through the second insulating path of the MEMS chip 3100, e.g., signals for powering the transducer and / or digital signals for controlling the transducer.

[0052] In this embodiment, MEMS signal processing circuitry 3240 includes specialized digital CMOS circuit modules such as digital data analysis circuitry 3242, digital input / output circuitry 3244, memory 3246, system controller 3248, and calibration / compensation circuitry 3250. Auxiliary signal processing circuitry 3260 includes power management circuitry 3262, and high speed CMOS circuitry 3264, which may include wireless and / or GPS I / O modules. The digital components in MEMS signal processing circuitry 3240 and in auxiliary signal processing circuitry 3260 communicate through a digital bus 3272.

[0053] Because the transducer operates using analog signals, the IC chip 3200 includes mixed signal CMOS circuitry 3270 that allows the IC chip 3200 to interface to the inputs and outputs of the MEMS sensor 3100. The mixed signal CMOS circuitry 3270 includes ADCs for converting analog signals generated by the MEMS chip 3100 to digital signals for processing by the MEMS signal processing circuitry 3240. The mixed signal CMOS circuitry 3270 further includes DACs for converting digital signals received from the MEMS signal processing circuitry 3240 and / or the auxiliary signal processing circuitry 3260 to analog signals for controlling the MEMS chip 3100. The mixed signal CMOS circuitry 3270 communicates with other digital components of the IC chip 3200 through a digital bus 3272.

[0054] The 3DS subsystem is distributed among these various circuits. For example, consider a 3DS Inertial Navigation Unit (INU) based on a 10DOF MEMS sensor consisting of a 6DOF inertial sensor measuring angular velocity and acceleration, a pressure sensor, and a 3DOF magnetometer, as shown in FIG. 10A. Part of the 3DS system can act as a system sensor hub, as digital requests for position / attitude readings from the larger system come in through the PCB board digital I / O leads or wireless I / O 3264, which require high speed or RF CMOS running at a higher clock rate than the memory or logic sections. The requests travel through the digital bus 3272 and the digital I / O section 3244 to the system controller 3248. The system controller 3248 provides clock signals to trigger and time the measurement of each of the three angular velocity, three acceleration, three magnetic field, and one pressure readings. The analog / digital section 3270 provides the DC bias and gyroscope drive signals needed to measure the capacitance of the various sensors, as well as to amplify the signals and convert them into digital data representing angular velocity, acceleration, magnetic field, and pressure. The digital analysis circuit 3242 may take the raw digital sensor data and calculate real-time values ​​of acceleration and angular velocity (IMU output), as well as pressure and magnetic field, using algorithms and constants stored in memory 3246. However, if an inertial navigation output (e.g., position and attitude) is required, the digital data analysis section 3242 performs further calculations to integrate the 6DOF data, including pressure and magnetic field data, with external sensor readings (e.g., GPS) to provide instantaneous position and attitude. These "sensor fusion" algorithms and constants may be stored in memory 3246.Finally, the results are output through the digital I / O section 3244 via a digital bus, through the MEMS chip to the PCB board, or via RF wireless radio, using the 3DS chip to also act as a sensor hub to communicate with the larger system.

[0055] As shown, IC 3200 interfaces with MEMS chip 3100 via conductive paths 3130 and 3150. The first conductive path 3130 conducts transducer I / O signals and is therefore an analog channel. Thus, the first path 3130 extends through a mixed signal CMOS circuit 3270 interface before reaching digital bus 3272. The second conductive path 3150 conducts auxiliary signals. Because auxiliary signals can be analog or digital, they may take different routes into IC chip 3200 depending on their function. For example, analog auxiliary signals may interface with IC chip 3200 via mixed signal CMOS circuit 3270, whereas digital signals may interface directly with digital bus 3272. If the second conductive path 3150 carries a power signal, the second conductive path 3150 may function as a power bus 3274 and may interface directly with the power management circuit 3262, e.g., the power management circuit 3262 is further connected to a digital bus 3272 for communicating digital data.

[0056] A further example of an IC chip 950 configured for processing inertial sensor data is shown in FIG. 10B, which may further include feedback control of sensor operation or may be further configured to control the operation of an autonomous vehicle, such as a drone, satellite, or other propulsion system using the detected inertial data. A clock 970 controls the timing of control and signal processing operations in the IC chip 950. A system controller 960 transmits control signals for operation of the IMU operation. The IC chip 950 is connected to the MEMS inertial sensor 920, as well as to an external sensor 942 and one or more devices 944 that operate in response to the inertial sensor data generated by the inertial sensor 920. The IC chip may be connected to a neural processing unit (NPU) 980, which may be mounted on a circuit board containing the IC chip or may be manufactured by a system-on-chip design. The NPU 980 may comprise logic circuits, such as an FPGA, an application specific integrated circuit (ASIC), or a graphics processing unit (GPU), that perform computational processing of the inertial sensor data from the inertial sensor at 982, or may access data stored in memory 956.

[0057] Referring to FIG. 11, a process flow diagram illustrating a method 600 of operating a MEMS transducer device is shown. An analog electrical MEMS signal is generated using a MEMS transducer (step 602). The analog electrical MEMS signal is received through a first insulating conductive path at a mixed signal CMOS circuit in an IC chip (step 604). The mixed signal CMOS circuit converts the analog electrical MEMS signal to a digital electrical MEMS signal (step 606). The digital electrical MEMS signal is transmitted from the mixed signal CMOS circuit using a digital bus to a MEMS signal processing circuit including a digital CMOS circuit (step 608). The digital CMOS circuit processes the digital electrical MEMS signal (step 610). The digital CMOS circuit includes at least one of a digital data analysis circuit, a digital input / output circuit, a memory, a system controller, and a calibration / compensation circuit.

[0058] Referring to FIG. 12, a process flow diagram illustrating a method 700 of operating a proof mass MEMS device according to a preferred embodiment of the present invention is shown. Transducer data and sensor data are generated using a MEMS device (step 702). The MEMS device includes at least one movable mass with a thickness between 100 micrometers and 1000 micrometers. The area and thickness of the mass are selected to provide noise density and bias stability values ​​within a selected range. Optionally, the MEMS device including the first movable mass and the second movable mass is operated in an anti-phase drive mode (step 704). The multiple masses are selected to reduce noise. The transducer data and sensor data are processed using a MEMS IC processing circuit as described herein to generate a digital sensor data output (step 706). Thus, the device may transmit the sensor output data to an external application by a communication network via wired or wireless transmission.

[0059] To reduce the final device footprint, an alternative architecture for MEMS integrated systems allows multiple single MEMS wafers to be stacked vertically to form a 3DS MEMS wafer. In one embodiment, multi-wafer 3DS MEMS bonded IC wafers consisting of two MEMS wafers of different device types can be stacked and bonded together. By aligning the first and second insulated conductive paths (also called 3DTCV), MEMS and auxiliary signals can be transmitted throughout the stack of MEMS and ASIC chips, simplifying power bussing and minimizing lead routing between the various MEMS functions and electronics. FIG. 13 shows a diced 3DS component 4000 consisting of an IC chip 4200 and a stack of two single MEMS chips 4102, 4104 bump bonded to a printed circuit board 302. In this case, a second layer of single MEMS chips 4102 is bump bonded to a first layer of further single MEMS chips 4104. The second insulated conductive paths 4150′ of the further single MEMS chip 4104 are electrically connected to at least some of the second insulated conductive paths 4150 of the first single MEMS chip 4102 for conducting auxiliary signals through the first single MEMS chip and the further single MEMS chip to the auxiliary signal processing circuitry of the IC chip 4200. The interconnected second insulated conductive paths of the MEMS chips 4102 and 4104 are configured to transmit the auxiliary signals from the PCB to the IC chip for processing without the need for any wire bonding.

[0060] The MEMS signals for the MEMS chip 4104 may further travel through the MEMS chip 4102 to the IC chip 4200. The first MEMS chip 4102 comprises a third set of first cap MEMS electrical contacts and third insulated conductive paths 4170 for connecting the third set of first cap MEMS electrical contacts to at least some of the second cap MEMS electrical contacts of the second cap layer of the MEMS chip 4102 through the first cap layer, the central MEMS layer, and the second cap layer. These third insulated conductive paths 4170 are electrically connected to the MEMS signal processing circuitry 4240 of the IC chip 4200 and electrically connected to the insulated conductive paths 4130′ of the MEMS chip 4104. Thus, the MEMS signal processing circuitry 4240 may process the electrical MEMS signals of the first single MEMS chip and the at least one further single MEMS chip. Thus, the MEMS signal processing circuit 4240 can process the MEMS signals from both MEMS chips 4102 and 4104.

[0061] Of course, although in the embodiment shown in Figure 13 there are two MEMS chips, it is possible to stack more than two MEMS chips of the same or different types. Thus, an integrated MEMS system component may include a first single MEMS chip and a further single MEMS chip stacked vertically.

[0062] For an embodiment using a control system using an inertial measurement unit (IMU) including a controller, a processor, and a memory that causes the processor to receive sensor signals from one or more inertial and / or non-inertial sensors as previously described herein and convert the sensor signals into a number of components of a measurement vector, the processor may be programmed to: determine a first state vector using an IMU Kalman filter, update a first component subset of the system state vector, determine a second state vector using a spatial positioning Kalman filter, update the second component subset of the system state vector based on the second state vector, determine a third state vector using a system Kalman filter, update the third component subset of the number of components of the system state vector based on the third state vector, and control a system operating parameter based on at least one of the first state vector, the second state vector, the third state vector, and the system state vector.

[0063] In another embodiment, a tangible, non-transitory computer readable medium has stored thereon instructions executable by a processor, such that the instructions cause the processor to receive sensor signals from one or more inertial and / or non-inertial sensors as previously described herein, convert the sensor signals into components of a system measurement vector, determine a first state vector using an inertial measurement unit (IMU) Kalman filter based on the system measurement vector and a system state vector, update a first component subset of the many components of the system state vector based on the first state vector, determine a second state vector using a spatial positioning Kalman filter using the system measurement vector and the system state vector, update a second component subset of the many components of the system state vector based on the second state vector, determine a third state vector using a system Kalman filter based on the measurement vector and the system state vector, update a third component subset of the many components of the system state vector based on the third state vector, and control an action or movement based on at least one of the first state vector, the second state vector, the third state vector, and the system state vector.

[0064] The use of a Kalman filter for each of the inertial sensors may simplify computational complexity and improve processing speed. Thus, in an embodiment using three accelerometers, for example, three Kalman filters are used to process each accelerometer output. Other sensors, including magnetometers, pressure sensors, and light sensors, or cameras, may be further filtered as described herein to periodically update the system state vector. The spatial positioning device may include, for example, a Global Positioning System (GPS) or a Global Navigation Satellite System (GNSS), etc., which generates position data that is used with the filtered sensor data to enhance IMU operation. In certain embodiments, the spatial positioning device may be configured to determine the position of the system relative to a fixed point in the field (e.g., via a fixed radio transceiver). Thus, the spatial positioning device may be configured to determine the position of the system relative to a fixed global coordinate system or a fixed local coordinate system (e.g., via a GPS). In certain embodiments, the first transceiver is configured to broadcast a signal indicative of the position of the system to the transceiver of the base station.

[0065] The processor executes software, such as software implementing the Kalman filter described herein. The processor may include multiple microprocessors, one or more general-purpose microprocessors, and / or one or more application specific integrated circuits (ASICs), one or more field programmable gate array devices (FPGAs), or any combination thereof. In some embodiments, the subject matter described herein may be implemented by a tangible and non-transitory computer-readable medium having instructions stored thereon. For example, a commercially available software product incorporating a Kalman filter is available from Mathworks Inc. of Natick, Massachusetts, for estimating the state of a system using an inertial sensor using the products Matlab® and Simulink®.

[0066] The following contains the equations for the extended Kalman filter, which is iterated as follows:

[0067] Time-updated forecast (global): State Estimation Propagation

number

number

[0068] Measurement Updates (Global + Local): Initialization for state estimation

number

number

number

number

[0069] In the above, x is the state vector, F is the state transition matrix, P is the covariance matrix, Q is the dynamic disturbance noise covariance, R is the measurement noise covariance, H is the measurement sensitivity matrix, and K is the Kalman gain. The subscript "i" is used for iteration, and k is the time-related subscript. As can be determined from equations (1)-(6), for a local iterated extended Kalman filter implementation, only the measurement equations (4)-(6) are updated during an iteration. For an IMU that includes an iterated extended Kalman filter implementation, all inertial sensor data is used. Due to backpropagation of state estimates, a global iterated extended Kalman filter implementation is used, including for example use for gyroscope alignment. However, in preferred implementations of the devices described herein, one or more gyroscopes and / or accelerometers may be realized by fabrication on the same wafer or by stacking chips made using the same or similar photolithographic process steps to minimize alignment variations. The basic equations for the implementation of the global iterated extended Kalman filter use both the periodically updated time approximation equations (1)-(2) and the measurement equations (4)-(6). When convergence is achieved, the iterations can be stopped.

[0070] During the iteration, the residuals in the measurement equations may be used to identify error states that may indicate error reduction. Additionally, the Kalman gain associated with each Kalman filter is considered when identifying the size of the step for the next error state identification. In particular, the Kalman filter reduces gyroscope alignment time, for example, by iterating over both the time update equations and the measurement update equations. Due to the lower noise level of the MEMS devices described herein compared to data from various less expensive MEMS-based inertial systems, the time step for the extended Kalman filter may be adapted for a particular application. Smaller time steps may be used to measure nonlinearities to the sensor data during calibration.

[0071] As shown in process flow diagram 800 of FIG. 14, the method includes acquiring 802 system sensor data including data from at least a plurality of accelerometers in a sensor package, the plurality of accelerometers operating at different acceleration ranges. It determines 804 whether the sensor circuit exceeds one or more threshold conditions associated with each of the accelerometers during a selected time interval to select an accelerometer output value for the selected time period. A Kalman filter is applied 806 to the selected accelerometer value for the time period to generate a filtered accelerometer value for the time period. If additional sensor values ​​are further processed for the selected time interval, e.g., gyroscope data, temperature sensor data, pressure data, and / or position data, a Kalman filter is applied 808 to the additional sensor output during the selected time interval. This sequence of steps is iteratively applied 810 over subsequent time intervals to generate a filtered sensor data output, e.g., to generate inertial sensor measurement data for use by an inertial navigation unit. This system provides sensor data fusion in accordance with a preferred embodiment.

[0072] In an alternative embodiment, the Kalman filter is replaced by a neural network implemented in a digital signal processor (3242 in FIG. 10A) or by a neural processing unit (NPU) 980 (FIG. 10B) specifically designed for machine learning (ML) operations. ML is usually understood as allowing a system's ability to learn and improve its decision-making capabilities based on experience instead of relying on explicit programs. This approach allows for the processing of data without the need for rigid models. The learning process begins with some combination of observations, data, or instructions. The data is processed to derive patterns that can be used to improve future operations based on the observations and feedback received by the system. This technique allows the system to learn automatically with minimal human intervention, thus enabling adaptive behavior of the system using iterative computational methods to control the system's processing and operations.

[0073] ML allows a system to program itself and improve its performance through a process of continuous fine-tuning. In traditional systems, data and programs simply work together to produce a desired output, with any problems being caught or improvements applied by the programmer. In contrast, an ML system uses the data and resulting output to generate a program or model. This program can then be used in combination with a traditional program to operate one or more systems that include, for example, inertial sensors, or that use inertial sensor output to perform motion control operations.

[0074] There are several different approaches to ML, each best suited for a particular class of applications. Supervised learning is based on feeding the system "known" rules, characterized by expressing the association between input (e.g. position) and output (acceleration). In contrast, unsupervised learning only provides the system with inputs and lets the ML algorithm analyze the inputs to discover unique / separable classes and patterns.

[0075] Neural processing units are often realized using integrated circuit chips, which may include three classes: graphics processing units (GPUs), field programmable gate arrays (FPGAs), and application specific integrated circuits (ASICs). GPUs were originally designed for image processing applications that benefit from parallel computing. During 2012, GPUs began to see more use for training machine learning systems, and by 2017 they became dominant. GPUs are also sometimes used for inference, i.e., to predict results or outputs that will be used. Despite allowing for greater parallelism than CPUs, GPUs are still designed for general purpose computing.

[0076] FPGAs and ASICs have become more popular for inference due to their improved efficiency compared to GPUs. ASICs are increasingly being used for training as well. FPGAs contain logic blocks (i.e., modules that each contain a set of transistors) whose interconnects can be reconfigured by a programmer after manufacture to suit a particular algorithm, whereas ASICs contain word-wired circuits customized for a particular computing task or to execute an algorithm. While mainstream ASICs typically offer higher efficiency than FPGAs, FPGAs are more customizable than ASICs, facilitating design optimization as the computing method is adapted for a particular application. In contrast, ASICs become increasingly obsolete as new computing methods are developed.

[0077] Different neural processing chips may be used for training and inference, depending on the different demands on the chips imposed by each task. First, training requires additional computational steps in addition to those that training shares with inference, so different forms of data and model parallelization are suitable between training and inference. Second, training virtually always benefits from data parallelization, whereas inference often does not. For example, inference may be performed on one piece of data at a time. However, in some applications, especially when the application requires fast inference of many different pieces of data, inference may be performed in parallel on many pieces of data. Third, depending on the application, the relative importance of efficiency and speed for training and inference may differ. For training, both efficiency and speed are important. For inference, high inference speed may be essential, since many neural processing applications deployed in critical systems (e.g., autonomous vehicles) or with rushed users (e.g., mobile applications for classifying images) require fast real-time data classification. On the other hand, there may be a ceiling on useful inference speed. For example, inference never needs to be faster than the user's response time to a mobile application. A guess chip requires less research than a training chip, and it requires optimization for fewer operations than a training chip. ASICs require less development than GPUs and FPGAs because they are typically narrowly optimized for a particular algorithm, and design engineers consider far fewer variables. To design a circuit for just one computation, an engineer can simply replace the computation with a circuit optimized for that computation. But to design circuits for many kinds of computations, engineers must predict which circuits will work well for a wide variety of tasks, many of which are not known in advance.

[0078] The commercialization of neural processing chips depends on the extent of their general-purpose capabilities. GPUs have been widely commercialized for a long time, whereas FPGAs have been to a lesser extent. On the other hand, ASICs are more difficult to commercialize, given the high design costs and low volume production due to specialization. However, in an era of slower rates of improvement of general-purpose chips, specialized chips are relatively more economical, since they have a longer useful life until the next generation of CPUs achieve the same speed or efficiency. In an era of slower rates of improvement of CPUs, if a processing chip shows a 10-100 times speedup, sales volume of only 15,000-83,000 units may be sufficient to make the neural processing chip economical. The expected increase in market size for neural processing chips may provide the economic economies of scale necessary to make narrower capability ASICs profitable. Neural processing chips range in grades from powerful to weak. At the high end, server-grade neural processing chips are typically used in data centers. At the medium end, PC-grade neural processing chips are typically used by general consumers. At the low end, mobile neural processing chips are typically used for inference and are integrated into a system-on-chip that also includes a CPU. The mobile system-on-chip must be miniaturized to fit into mobile devices. At each of these grades, growth in chip market share has come at the expense of non-neural processing chips. Supercomputers have limited but increasing relevance for neural processing. Most commonly, server-grade chips are distributed in data centers and can run sequentially or in parallel in a setup called "grid computing". Supercomputers use server-grade chips, physically co-located and linking them together, as well as adding expensive cooling equipment to prevent overheating. This improves speed but significantly reduces efficiency. An acceptable trade-off for many applications that require fast analysis.Few current applications can justify the additional cost for faster speeds, but training or inferring large algorithms is sometimes too slow, so supercomputers are used as a last resort. Thus, while CPUs have traditionally been the supercomputing chips of choice, by 2018 GPUs were responsible for the majority of the computing power in the growing number of supercomputers worldwide.

[0079] There is no common industry scheme for evaluating CPU vs. neural processing chip benchmarks, since the chip speed and efficiency to be compared depends on the specific benchmark. However, for any given node, neural processing chips typically provide 10-1,000 times improvement in efficiency and speed compared to CPUs, using lower-end GPUs and FPGAs and higher-end ASICs. A neural processing chip that is 1,000 times more efficient compared to a CPU for a given node provides an improvement equivalent to 26 years of CPU improvements. Gains for GPUs, FPGAs, and ASICs compared to CPUs (normalized to 1x) for DNN training and inference at a given node. For example, GPUs are commercially available from NVidia Corporation that can be mounted on printed circuit board embodiments of IMUs that can be stacked vertically with inertial sensor devices as described herein. Alternatively, ASIC devices can be fabricated in the CMOS circuit layers of a system-on-chip integrated circuit as described herein.

[0080] An exemplary embodiment of a neural network configured to process inertial sensor data to correct for accumulated errors that may occur over time and to provide predictive modeling of position and heading, for example, for an autonomous vehicle, is shown in Figures 15A-15B. In this example, a recurrent neural network (RNN) comprises an input node 1500, a hidden layer 1502 of the network, and an output node 1504. As shown in Figure 15B, the input 1506 includes accelerometer and gyroscope data, each acquired across orthogonal x-, y-, and z-axes. After the two RNN layers process the data, a linear layer follows, scaling based on unit length to provide attitude data. Further details regarding such implementations are described in more detail by Weber et al., "Neural Networks Versus Conventional Filters for Inertial-Sensor based Attitude Estimation," arXiv:2005.06897[cs.LG], 2020, the entire contents of which are incorporated herein by reference. Further neural network configurations, such as deep neural networks (DNNs) and support vector machines (SVMs), have been implemented in connection with the operation of autonomous vehicles, for example as described in U.S. Patent Nos. 10,054,445, 11,562,231, and 11,508,049 (the entire contents of which are incorporated by reference into this specification), and further in convolutional neural networks (CNNs) as described in U.S. Application No. 2022 / 0332335 or U.S. Application No. 2021 / 0191424 (the entire contents of which are incorporated by reference into this specification).The above-described computational methods generally include an iterative sequence of steps whereby an error measure is defined for a particular application, and the iterative sequence minimizes the error measure to generate output values ​​that are used by the system to adjust operating parameters for the inertial sensors, e.g., drive frequency or sampling rate, to adjust the motion of a detected moving object, e.g., an autonomous vehicle, or to record or communicate revised position and heading information. Such feedback or closed-loop control may improve the accuracy of measurements of inertial sensors or other sensors used to monitor the conditions described herein.

[0081] The present invention is particularly adapted for use in object detection systems used to determine at least one of the range, angle, and velocity of an object. Broadly described, the present invention relates to the mounting of MEMS IMUs, and in particular 3DS MEMS IMUs, on individual sensor elements or subarrays of such sensor elements of a position detection system. Given their small size, weight, and lower power consumption, and given that they allow for minimal bias instability, e.g. less than 1 degree / hour, MEMS-based IMUs including accelerometers and angular velocity sensors (6DOF MEMS IMUs) can be mounted on several sensor elements, preferably directly on each sensor element. The measurement signals of the MEMS IMUs can be processed by MEMS processing circuits or by a sensor element processing unit directly in the sensor elements, or the measurement signals can be sent to a central processing unit assigned to a subset of the sensor elements.

[0082] Lidar (Light Detection and Ranging) is quickly becoming a key element in ADAS (Advanced Driver Assistance Systems) and autonomous vehicle navigation. Lidar was developed for surveying and mapping. Lidar can produce highly accurate 3D measurements of the local environment relative to the sensor. This accuracy is achieved by emitting thousands of pulses of laser light per second and measuring the time of flight (TOF) between the emission and the collection by the sensor of the reflected light from the environment.

[0083] A multi-model sensor system for an autonomous vehicle 1540 is shown in Figure 16A with forward facing sensors having a camera field of view 1545 for traffic sign recognition, a lidar field of view 1544, a longer range radar FOV 1542, and an optional shorter range radar FOV 1546, and ultrasonic 1549 looking both forward and rearward for very close range warnings. Side cameras 1547 and rear side looking radar 1548 and cameras 1551 may also be used.

[0084] Thus, the arrangement of modules in different sections of the wheeled ground vehicle or automobile may include selected combinations of sensors. The forward-facing modules are preferably configured with multiple sensors operating in different modes, such as, for example, radar emitters and detectors, one or more cameras, lidar sensors, and ultrasonic sensors that may operate to detect obstacles at different distances. A sensor fusion program may be used by the processor 1554 to simultaneously process data from each of the multiple sensors and to automatically send navigation and control commands to the braking and steering control systems (described below) of the vehicle 1540. Simultaneous Localization and Mapping (SLAM) programs have been extensively described in the art, for example, in U.S. Pat. Nos. 7,689,321 and 9,945,950, the entire contents of which are incorporated herein by reference.

[0085] A distribution of sensor arrays for an autonomous vehicle 1560 operated on wheels 1555 is shown in FIG. 16B, including an array of sensor modules 1561-1570 distributed around the vehicle and connected to a processor 1554. Each of the sensor modules 1561-1570 may perform one or more sensing functions as described above, including implementing a camera FOV 1545, 1547, 1551, a lidar FOV 1544, a long-range radar FOV 1542, 1548, a short / medium-range radar FOV 1546, or an ultrasonic 1549. Each sensor module 1561-1570 may transmit ranging data to or receive instructions from the processor 1554, in various embodiments. The wheels 1555 include brakes. The brakes may include sensors that detect data related to the wheels 1555 or brake status (e.g., wheel revolutions per minute, brake actuation status, rate of brake braking applied). The brake sensors are operably coupled to a brake application module 1556 in communication with the processor 1554. The brake application module 1556 may receive data related to the wheels 1555 or brake status and may selectively control the brakes to stop the autonomous vehicle 1560. The processor 1554 may control the brake application module 1556 to apply brakes based on an analysis of the ranging data received from the array of sensor modules 1561-1570 to enable the autonomous vehicle 1560 to avoid collisions with objects.

[0086] The autonomous vehicle 1560 may further include a steering module 1557 in communication with the processor 1554. The steering module 1557 may control a steering mechanism of the vehicle to change the direction or heading of the vehicle. The processor 1554 may control the steering module 1557 to steer the vehicle based on analysis of the ranging data received from the array of sensor modules 1561-1570 to enable the autonomous vehicle 1560 to avoid collisions with objects.

[0087] The scope of the claims should not be limited by the preferred embodiments set forth in the examples, but rather should be accorded the broadest interpretation consistent with the description as a whole, including any equivalents.

Claims

1. An inertial sensor device, A first proof mass formed on the device layer of a silicon-on-insulator (SOI) MEMS wafer, which includes an insulating layer and a handle layer, wherein the first proof mass is positioned relative to a first detection electrode to measure a first inertial sensor signal, A second proof mass formed on the device layer of the SOI MEMS wafer, the second proof mass being positioned relative to a second detection electrode to measure a second inertial sensor signal, wherein the first inertial sensor signal and the second inertial sensor signal measure a change in the position of the inertial sensor device installed in a single chip package, A processing circuit for receiving the first inertial sensor signal and the second inertial sensor signal, comprising a signal processor installed in one chip package and performing iterative calculations to generate corrected position data, and A device equipped with the following features.

2. The device according to claim 1, A device comprising a first proof mass, a first accelerometer aligned with a second proof mass having a second accelerometer, wherein the first accelerometer measures acceleration in a first range and the second accelerometer measures acceleration in a second range.

3. The device according to claim 2, A device further comprising a third accelerometer formed in the device layer.

4. The device according to claim 2, A device in which the first accelerometer and the second accelerometer are located on the same plane.

5. The device according to claim 1, A device comprising at least one proof mass formed in the device layer, comprising a gyroscope proof mass having three axes of motion, wherein the proof mass is suspended within a cavity by a spring.

6. The device according to claim 2, A device comprising a first accelerometer proof mass, a device layer, an insulating layer, and a handle layer, wherein the proof mass is movably mounted in a cavity by one or more springs.

7. The device according to claim 1, The device further comprises a low-pass filter, a high-pass filter, and a lock-in amplifier for detecting motion frequencies higher than the resonant frequency of at least one proof mass.

8. The device according to claim 1, A device in which the signal processor is connected to receive digitized sensor signals for processing inertial sensor data.

9. The device according to claim 1, A device further equipped with an inertial measurement unit (IMU).

10. The device according to claim 1, The present invention further comprises a first cap SOI wafer melt-bonded to the SOI MEMS wafer to form a sealed cavity, The first cap SOI wafer is a device having one or more electrodes in a cap device layer that perform at least one of detecting and driving the motion of at least one proof mass.

11. The device according to claim 9, A device in which the IMU further comprises a position sensor, such as a GPS or GNSS sensor.

12. A device according to any one of claims 1 to 11, A device in which the inertial sensor, including the MEMS SOI wafer, is vertically stacked in a single chip package together with the processing circuit, including an integrated circuit.

13. A device according to any one of claims 1 to 11, The device comprises an analog-to-digital converter and a power management circuit.

14. A device according to any one of claims 1 to 11, A device wherein the processing circuit comprises memory and / or shared memory, and the processing circuit further comprises a system control device connected to a data processor having a neural processor.

15. A device according to any one of claims 1 to 11, A device that further includes a connector for autonomous vehicles, such as aerial drones or ground vehicles.

16. A device according to any one of claims 1 to 11, A device in which the processing circuit is connected to further sensors, including a camera, or a pressure sensor, or a temperature sensor, or a magnetometer, or a lidar, or radar, or sonar, or a combination thereof.

17. A device according to any one of claims 1 to 11, A device in which at least one proof mass moves in a plane through the MEMS SOI wafer, and at least one proof mass optionally moves outside the plane extending through the MOS SOI wafer.

18. A method for detecting inertia using an inertial sensor device, A step of detecting the motion of a first proof mass formed on the device layer of a silicon-on-insulator (SOI) MEMS wafer, the device layer being located above an insulating layer and a handle layer, wherein the first proof mass is positioned relative to a first detection electrode to measure a first inertial sensor signal. A step of detecting the motion of a second proof mass formed on the device layer of the SOI MEMS wafer, wherein the second proof mass is positioned relative to a second detection electrode to measure a second inertial sensor signal, and the first inertial sensor signal and the second inertial sensor signal measure a change in the position of the inertial sensor device installed in a single chip package, The steps include detecting an image with an image sensor coupled to the inertial sensor device that generates image data, A step of processing a signal using a processing circuit that receives the image data, the first inertial sensor signal, and the second inertial sensor signal, wherein the processing circuit includes a signal processor that is installed in one chip package and performs iterative calculations to generate corrected position data. Methods that include...

19. The method according to claim 18, A method wherein the first proof mass comprises a first accelerometer aligned with the second proof mass which comprises a second accelerometer or gyroscope, the device comprises an inertial measuring unit (IMU), and a detection electrode and a drive electrode are formed in a SOI device layer of a cap wafer surrounding the at least one proof mass in a cavity.

20. A method according to any one of claims 18 or 19, A method further comprising controlling an autonomous vehicle, wherein a system control unit is connected to the autonomous vehicle and performs a closed-loop operation of the autonomous vehicle, the autonomous vehicle including an aerial drone or ground vehicle having position sensors such as GPS or GNSS.

Images