Neural Network Parameter Search for Semiconductor Model Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Users face challenges in determining whether model parameters for semiconductor elements in a netlist satisfy required characteristics, as multiple parameters may satisfy these characteristics, and the judgment often relies on user experience, which can lead to overlooking better parameters and failing to adapt to varying required characteristics of different circuits.
Innovation Solution
A parameter search method utilizing a classification model, neural network, parameter extraction portion, circuit simulator, and control portion, which involves extracting model parameters, performing simulations, learning from results, classifying parameters, and using reinforcement learning to update parameters and find the best candidate satisfying the required characteristics of the netlist.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If users manually judge simulation results to select model parameters, then user experience and expertise are utilized, but the process is time-consuming and may overlook better parameters
Solution Approach 1:
The system performs automatic parameter extraction and classification without requiring manual user intervention. The classification model autonomously processes simulation results and identifies suitable parameters, enabling the system to serve itself rather than relying on user expertise for each evaluation.
Solution Approach 2:
The patent replaces manual mechanical judgment with automated computer-based classification. Instead of users manually evaluating simulation results, a classification model automatically processes and categorizes parameters based on learned patterns from training data, substituting human cognitive processes with computational algorithms.
2Adaptability or versatility
If multiple model parameters satisfy required characteristics, then parameter flexibility increases, but it becomes difficult to determine the best candidate
Solution Approach 1:
The classification model uses feedback from simulation results to continuously improve its parameter classification accuracy. By analyzing the output of circuit simulations and comparing it with required characteristics, the model refines its ability to identify the best parameters among multiple candidates.
Solution Approach 2:
The classification model acts as an intermediary between the simulation results and the final parameter selection. It processes the raw simulation data, categorizes parameters based on their performance, and presents the best candidates, making the evaluation process more manageable and accurate.
3Adaptability or versatility
If the netlist includes multiple semiconductor elements, then circuit functionality increases, but the number of model parameters to manage increases
Solution Approach 1:
The patent segments the parameter management process by separately extracting and classifying parameters for each semiconductor element. Instead of handling all parameters simultaneously, the system processes them individually through the classification model, reducing the complexity of managing multiple parameters across different elements.
Solution Approach 2:
The classification model serves multiple functions: it extracts parameters from simulation results, classifies them based on required characteristics, and identifies the best candidates. This multi-functional approach consolidates what would otherwise be multiple separate processes into a single unified system.
Data Source
AI summary
A parameter candidate for a semiconductor element is provided. A data set of measurement data is provided to a parameter extraction portion, and a model parameter is extracted. A first netlist is provided to a circuit simulator, simulation is performed using the first netlist and the model parameter, and a first output result is output. A classification model learns the model parameter and the first output result and classifies the model parameter. A second netlist and a model parameter are provided to the circuit simulator. A variable to be adjusted is supplied to a neural network, an action value function is output, and the variable is updated. The circuit simulator performs simulation using the second netlist and the model parameter. When a second output result to be output does not satisfy conditions, a weight coefficient of the neural network is updated. When the second output result satisfies the conditions, the variable is judged to be the best candidate.