Leakage Current Estimation Using Global and Local Variation Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating leakage current in semiconductor devices, such as the corner-based estimation technique and statistical analysis, are either pessimistic due to ignoring local variations or resource-intensive, failing to accurately consider both global and local variations in system-on-chip (SoC) designs.
Innovation Solution
A method that calculates a leakage current cumulative distribution function (CDF) by determining mean shifts around global process corners and operating voltages, incorporating both global and local variations to estimate and compare designs, thereby selecting the optimal design for reduced leakage and improved power management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If corner based estimation technique is used, then leakage estimation is simplified, but the estimation becomes pessimistic and inaccurate due to ignoring local variations
Solution Approach 1:
The patent segments the leakage estimation process into two independent parts: global process corner analysis and local variation analysis. By dividing the transistor set into groups and analyzing global and local variations separately, the method achieves accurate leakage estimation without the pessimism of traditional corner-based techniques while maintaining computational simplicity.
Solution Approach 2:
The patent introduces mean shift values as an intermediary parameter that bridges global process corners and local variations. These mean shifts represent the average impact of local variations on leakage current and are used to adjust the global corner-based estimation, thereby achieving accurate results without performing exhaustive simulations.
2Measurement precision
If statistical analysis process is used, then local variations are considered, but the process becomes computation intensive and resource wasteful
Solution Approach 1:
The patent segments the transistor population into multiple groups and performs statistical analysis only on representative samples from each group rather than all transistors. This segmentation allows local variations to be captured while dramatically reducing the computational burden compared to exhaustive statistical analysis.
Solution Approach 2:
The patent applies partial statistical analysis by calculating mean shifts for local variations only for a subset of transistors or representative cells, rather than performing complete statistical analysis on all transistors. This partial action achieves sufficient accuracy for leakage estimation while maintaining high computational efficiency.
Data Source
AI summary
A method of designing a semiconductor device includes preparing a first design for a semiconductor device and estimating leakage current for the first design. The method also includes determining a leakage current cumulative distribution function (CDF) for the first design. The method further includes preparing a second design for the semiconductor device based on determination of the leakage current CDF for the first design. Further, the method includes estimating leakage current for the second design. The method also includes determining a leakage current CDF for the second design in accordance to the determination of the leakage current CDF for the first design. Moreover, the method includes selecting one of the first design and the second design based on a comparison of the leakage current CDF for the first design and the leakage CDF for the second design.