Monte Carlo IC Cell Library for Frequency and Power Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current integrated circuit (IC) design and testing methods struggle to accurately predict the frequency and power distribution of manufactured ICs, leading to inefficiencies in production and yield loss.
Innovation Solution
A method utilizing a Monte Carlo simulation to create a library of IC cell types and their behavioral values for various process variations, allowing for the prediction of frequency and power distribution in ICs by correlating IC designs with this library and applying path selection algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Monte Carlo simulation is used to create a library of IC cell types for predicting frequency and power distribution, then prediction accuracy is improved, but computational time and complexity increase
Solution Approach 1:
The patent applies preliminary action by performing Monte Carlo simulations during the library creation phase to pre-calculate behavioral values for various process variations. This allows the simulation results to be stored in a library that can be quickly referenced during actual IC design and testing, avoiding the need to perform time-consuming simulations at every design stage.
Solution Approach 2:
The patent segments the IC design process into distinct phases: library creation with full Monte Carlo simulations, and actual design analysis using pre-computed behavioral values. This segmentation allows computationally intensive simulations to be performed once during library preparation, while subsequent design work uses lightweight lookups from the pre-built library.
2Reliability
If comprehensive Monte Carlo simulation is performed for all IC cell types, then prediction reliability is improved, but device complexity increases
Solution Approach 1:
The patent creates simplified behavioral models that copy the essential characteristics of complex IC cell behaviors under various process variations. Instead of performing full Monte Carlo simulations during design analysis, the system uses pre-computed behavioral values from the library that replicate the effects of comprehensive simulations, maintaining reliability while reducing complexity.
Solution Approach 2:
The patent changes parameters by pre-calculating behavioral values for multiple process variations during library creation. This allows the system to account for process variations through parameter variations in the behavioral models, rather than performing complex simulations during actual design work, thus maintaining reliability while simplifying the design process.
3Productivity
If path selection algorithms are applied to identify critical paths, then productivity is improved, but measurement precision requirements increase
Solution Approach 1:
The patent introduces behavioral values as an intermediary between the complex physical processes and the path selection algorithm. These behavioral values serve as simplified representations that capture the essential frequency and power characteristics, allowing the path selection algorithm to operate with reduced precision requirements while still identifying critical paths accurately.
Data Source
AI summary
A method comprising using at least one hardware processor for: running a Monte Carlo simulation of possible integrated circuit (IC) process variations of each of a plurality of IC cell types, wherein each of the plurality of IC cell types is defined by multiple specific transistors and multiple specific interconnects; based on the results of the Monte Carlo simulation, creating a library of IC cell types and their corresponding behavioral values for each of the possible IC process variations, and storing the library in a non-transient memory; receiving an IC design embodied as a digital file; correlating the received IC design with the library; and predicting a frequency distribution and a power distribution of ICs manufactured according to the IC design.