ML-Based ADC Calibration for Dynamic Nonlinearity Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing calibration algorithms for analog-to-digital converters (ADCs) struggle to address dynamic nonlinearities without prior knowledge of impairment sources, limiting their effectiveness in correcting for unknown errors.
Innovation Solution
A trained machine-learning model, such as an artificial neural network, is used to correct ADC digital outputs by learning from a reference ADC's ground-truth data, enabling it to address both known and unknown nonlinearities without requiring a priori knowledge of impairment sources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional calibration algorithms with lookup tables are used, then static nonlinearities can be corrected, but dynamic nonlinearities cannot be addressed effectively
Solution Approach 1:
The calibration engine uses machine learning models that automatically learn and adapt to the ADC's nonlinearity characteristics through training data, enabling self-calibration without requiring external knowledge of impairment sources. The system serves itself by identifying and correcting both static and dynamic nonlinearities through the trained model's inherent pattern recognition capabilities.
Solution Approach 2:
The patent transitions from fixed lookup table parameters to dynamic machine learning model parameters that can adapt to different operating conditions. The trained model's weights and biases are adjusted based on training data to optimize correction performance across varying input signals, frequencies, and amplitude levels, enabling effective dynamic nonlinearity correction.
2Measurement precision
If calibration algorithms require knowledge of impairment sources, then correction can be targeted, but unknown errors cannot be corrected
Solution Approach 1:
The calibration engine incorporates feedback mechanisms where the trained machine learning model continuously receives ADC output data and reference data, processes this feedback information, and adjusts its internal parameters to improve correction accuracy. This closed-loop feedback enables the system to learn from actual performance and adapt to unknown impairment sources that manifest in the data.
Solution Approach 2:
The patent replaces traditional calibration approaches that rely on mechanical or algorithmic knowledge of impairment sources with a data-driven machine learning approach. Instead of requiring explicit knowledge of how impairments occur, the system uses the trained model's ability to recognize patterns in the data, substituting mechanistic understanding with statistical learning to correct unknown errors.
3Ease of manufacture
If lookup tables are used for calibration, then implementation is simple, but dynamic nonlinearities remain uncorrected
Solution Approach 1:
The trained machine learning model acts as an intermediary layer between the raw ADC output and the final corrected signal. This intermediary processed the ADC outputs through learned transformation functions, enabling effective correction of dynamic nonlinearities while maintaining a relatively simple overall system architecture. The model serves as a bridge that translates imperfect ADC data into accurate corrected values.
Data Source
AI summary
A system includes a primary analog-to-digital converter (ADC) having an input electrically coupled to an input voltage, the primary ADC configured to sample the input voltage at a frequency and convert sampled input voltages to respective primary ADC digital outputs; and a trained calibration engine having an input electrically coupled to an output of the primary ADC, the trained calibration engine including a trained machine-learning (ML) model configured to correct each primary ADC digital output to a respective corrected digital output, the trained ML model having been trained with reference digital outputs from a reference ADC and training digital outputs from the primary ADC, the reference digital outputs representing ground-truth data for modeling the training digital outputs from the primary ADC.