MEMS Sensor Self-Test via Deep Learning Signature Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
MEMS sensors in electronic devices can produce false positives or false negatives due to damage or degradation, leading to incorrect measurements and increased complexity, size, and cost in self-test processes, which reduce system performance and increase testing time.
Innovation Solution
An analysis method using a MEMS sensor with a control unit and sensing assembly that acquires data, performs self-tests, and processes signatures through a deep learning module to validate the operative state of the sensing assembly, allowing continuous device operation and improved reliability without suspending normal system functions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional self-test methods are used to validate MEMS sensor operation, then reliability is improved, but device complexity and testing time increase
Solution Approach 1:
The patent extracts the self-test function from the main sensor chip and implements it through a separate test structure with dedicated test stimuli and measurement paths. This allows the sensing element to be tested independently without requiring complex modifications to the main sensor functionality, thereby reducing overall device complexity while maintaining reliability.
Solution Approach 2:
The self-test process is segmented into distinct phases: applying test stimuli through the test structure, acquiring test signals, and comparing results against reference values. This segmentation allows for simplified control logic and easier implementation compared to integrated self-test approaches, reducing the complexity of the test management system.
2Measurement precision
If self-test is performed to validate sensing element operation, then measurement accuracy is improved, but system performance and throughput are reduced
Solution Approach 1:
The patent implements periodic self-test execution where the sensing element is tested at predetermined time intervals rather than continuously. This allows the system to maintain normal operation and high throughput during most of the time, while periodically validating measurement accuracy through self-test to ensure reliability without significantly impacting overall system performance.
Solution Approach 2:
The test structure is pre-configured with known test stimuli and reference values before actual measurement operations. This preliminary preparation allows for rapid self-test execution without requiring complex real-time analysis, thereby maintaining high system throughput while ensuring measurement accuracy through pre-planned validation procedures.
3Reliability
If comprehensive self-test is implemented to detect sensor damage, then reliability is improved, but testing time and cost increase
Solution Approach 1:
The patent applies local quality by using a dedicated test structure with specific test stimuli tailored to detect particular types of sensor damage or degradation. Rather than implementing a comprehensive test of all possible failure modes, the test structure is locally optimized to detect the most critical sensor anomalies, thereby reducing testing time and cost while maintaining adequate reliability for damage detection.
Data Source
AI summary
An analysis method of a device through a MEMS sensor is provided in which the MEMS sensor includes a control unit and a sensing assembly coupled to the device. The analysis method includes acquiring, through the sensing assembly, first data indicative of an operative state of the device. Testing is performed for the presence of a first abnormal operating condition of the device. If the first abnormal operating condition of the device is confirmed, a self-test of the sensing assembly is performed to generate a quantity indicative of an operative state of the sensing assembly. The self-test includes acquiring, through the sensing assembly, second data indicative of the operative state of the sensing assembly, generating a signature according to the second data, and processing the signature through deep learning techniques to generate said quantity.