Sensor Vector Calibration Without Known Stimulus
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current calibration methods for systematic errors in data vectors are inconvenient as they require generating a known input stimulus, which can be time-consuming and difficult to apply, especially in operational environments.
Innovation Solution
A method involving binning and optimizing successive data vectors using an initial calibration estimate, where bins are defined based on the direction of calibrated vectors, and a gradient descent routine minimizes a cost function representing deviation from unit magnitude, allowing for continuous calibration without a predefined stimulus.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a known input stimulus is generated for calibration, then calibration accuracy is improved, but calibration time and operational interruption increase
Solution Approach 1:
The system performs preliminary calibration actions by continuously binning and storing data vectors in the background during normal operation. When calibration is needed, the pre-collected binned data is immediately available for optimization, eliminating the need to interrupt operation for stimulus generation and calibration execution.
Solution Approach 2:
The calibration process is made continuous through background binning of incoming data vectors. Instead of stopping normal data processing to perform calibration, the system continuously accumulates binned data during operational periods, allowing calibration to proceed without interruption while maintaining measurement accuracy.
2Measurement precision
If a known input stimulus is applied for calibration, then calibration accuracy is improved, but ease of operation deteriorates
Solution Approach 1:
The system performs self-calibration by automatically binning incoming data vectors and optimizing calibration parameters using the collected data. No external stimulus generation or manual intervention is required - the system uses its own operational data to calibrate itself, making the process as easy as normal operation.
Solution Approach 2:
Instead of applying external stimuli to the sensor system for calibration, the system inverts the approach by using the sensor's own operational data outputs. The calibration is derived from analyzing the natural data vectors produced during normal operation, rather than forcing calibration inputs into the system.
3Measurement precision
If traditional calibration methods are used, then systematic errors are corrected, but device complexity increases
Solution Approach 1:
The calibration process is segmented into two independent modules: (1) background binning of data vectors that occurs continuously during normal operation, and (2) optimization of calibration parameters using the binned data. This segmentation allows each module to be simple and independent, reducing overall system complexity while maintaining correction capability.
Data Source
AI summary
A method of data calibration, and in particular sensor calibration, which involves gathering an initial first estimate and then binning the data samples, so that calibration can be performed without the need for a known reference stimulus. The present disclosure relates to calibration of vectors in a measurement system, and in particular to calibration of a correction function for systematic errors in successive data vectors. There is provided a method of determining a vector calibration function comprising: binning successive data vectors; and optimising the binned data vectors once data vectors allocated to a minimum number of unique bins have been observed. The method comprises establishing an initial calibration estimate and where the binning and optimising are performed based on said initial calibration estimate.