Circuit Failure Prediction via Sensitivity Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for simulating integrated circuit variability, such as Monte Carlo methods, are excessively time-consuming and often fail to accurately reflect actual operating conditions, as they either randomly vary component characteristics or assume worst-case variations.
Innovation Solution
A method that involves measuring component characteristics, defining a figure of merit, determining sensitivities to these variables, and using simulation deviations to predict the likelihood of circuit failure based on a predetermined maximum tolerable failure probability, allowing for efficient simulation and design optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Monte Carlo methods are used to simulate circuit operation with random variations in component characteristics, then the accuracy of variability determination is improved, but the simulation time increases excessively
Solution Approach 1:
The patent transforms the simulation approach by changing from random parameter variation (Monte Carlo) to sensitivity-based parameter analysis. Instead of randomly varying component characteristics across millions of simulations, the method identifies key parameters and their sensitivities, then uses targeted simulations with controlled parameter deviations to achieve the same accuracy with far fewer simulation runs.
Solution Approach 2:
The patent creates a simplified model that copies the essential variability behavior without requiring full Monte Carlo simulation. By using sensitivity analysis to identify dominant parameters and their impact on circuit performance, the method creates a reduced-order model that replicates variability effects with minimal computational effort.
2Loss of time
If worst-case variation assumptions are used for component characteristics, then the simulation time is reduced, but the accuracy of reflecting actual operating conditions deteriorates
Solution Approach 1:
The patent applies partial action by focusing simulations only on the most critical parameters identified through sensitivity analysis, rather than simulating all possible parameter variations. This selective approach achieves sufficient accuracy for reliability assessment without the computational burden of comprehensive worst-case analysis across all parameters.
Solution Approach 2:
The method changes from static worst-case parameter assumptions to dynamic parameter deviations based on measured statistical data. By using actual measured mean and standard deviation values from fabricated components, the simulation reflects real operating conditions more accurately while maintaining computational efficiency through targeted sensitivity-based analysis.
3Reliability
If the number of simulations is increased to accurately determine circuit variability, then the reliability assessment accuracy is improved, but the productivity of the design process deteriorates
Solution Approach 1:
The patent performs preliminary sensitivity analysis to identify which parameters most significantly affect circuit performance before conducting full reliability simulations. This preliminary action allows the design process to focus computational resources on the most critical parameters, achieving accurate reliability assessment with minimal simulation effort and maintaining high design productivity.
Solution Approach 2:
The method transforms the simulation strategy by changing from exhaustive parameter sampling to sensitivity-guided parameter selection. By measuring actual component parameters and using sensitivity analysis to weight their importance, the approach achieves accurate reliability assessment with a fraction of the simulations required by traditional methods, thereby preserving design process productivity.
Data Source
AI summary
A method, implemented in a processor, of determining a likelihood of failure of a circuit to be made in accordance with a circuit design, and a computer-readable storage medium storing instructions to the processor for carrying out the method. A sensitivity of a figure of merit to each variable of a plurality of variables is determined by simulating operation of the circuit using the processor. Determining the sensitivity is based on a departure of each of the variables from a respective mean value, where the variables include at least one variable derived from measurements of a fabricated component or component combination to be included in the circuit. Results from the simulation are used to predict a failure probability of the circuit to be made in accordance with the circuit design.