IC Device Model Data Selection for Reusable Parameter Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting parameters from integrated circuit device models consume significant time and computing resources due to repeated data extraction steps based on flexible bias condition settings, lacking a reusable data selection process.
Innovation Solution
A data selection method and apparatus that utilizes a user-defined condition filtering setting interface to configure filtering conditions and bias attributes, allowing for primary fixed filtering and secondary customized filtering, generating a mapping form to fit the device model with customized test datasets, thereby avoiding repeated data extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If flexible bias condition settings are applied to meet different extraction process requirements, then the adaptability of the parameter extraction process is improved, but the data extraction step must be repeatedly performed, increasing computational complexity and time consumption
Solution Approach 1:
The patent pre-divides test data into multiple data sets according to different behavior regions before the actual parameter extraction process. This preliminary classification allows the system to quickly select and apply appropriate data sets for different bias conditions, avoiding the need to re-process all test data repeatedly when extraction requirements change, thus reducing computational complexity while maintaining process adaptability
Solution Approach 2:
The patent segments the entire test data into multiple distinct data sets based on different behavior regions (e.g., weak inversion, strong inversion, saturation, triode regions). This segmentation enables the extraction process to selectively use only the relevant data sets for specific bias conditions, rather than processing all data, thereby reducing computational burden while preserving the ability to adapt to different extraction scenarios
2Adaptability or versatility
If flexible bias condition settings are applied to meet different extraction process requirements, then the adaptability of the parameter extraction process is improved, but the time consumption for repeated data extraction increases
Solution Approach 1:
The patent performs data classification and segmentation into behavior-region-specific data sets before the actual parameter extraction. This preliminary action ensures that when different bias conditions are required, the system can directly retrieve and use the pre-prepared relevant data sets without performing time-consuming re-extraction, thus reducing time loss while maintaining adaptability
Solution Approach 2:
By dividing test data into multiple behavior region-based data sets in advance, the patent enables rapid selection and application of specific data sets for different extraction scenarios. This segmentation eliminates the need to re-process all test data when bias conditions change, significantly reducing extraction time while preserving process flexibility
3Manufacturing precision
If all test data are processed to ensure comprehensive model fitting, then the manufacturing precision of model parameters is improved, but the computational resources consumed increase significantly
Solution Approach 1:
The patent applies different data sets with specific behavior region characteristics to different parameter extraction tasks. Each data set is optimized for particular operating regions (e.g., weak inversion for threshold voltage extraction, saturation for current modeling), ensuring high precision for each parameter while avoiding the waste of computational resources that would result from using all test data uniformly for all extraction purposes
Data Source
AI summary
A data selection method (and system and apparatus) for extracting parameters from an integrated circuit device model includes receiving a test dataset of an integrated circuit device; configuring at least three different filtering conditions, and setting bias attributes for each filtering condition; filling the filtering conditions and the bias attributes in the form of labels to generate a mapping form; performing primary fixed filtering classification on a plurality of test data according to the filtering conditions, and then performing secondary customized filtering by utilizing the bias attributes; mapping filtering results to corresponding labels in the mapping form; constructing association relationships of the customized test data in different filtering results; storing the customized test data screened each time in the form of a set; fitting a device model of the integrated circuit device to perform condition instantiation; and adjusting the bias attributes to realize the purpose of data selection.