Pattern Matching Parasitic Extraction with Database Reuse
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In advanced semiconductor technology nodes, accurate RC parasitic modeling is necessary for device modeling and timing analysis, but existing methods like pre-characterization and 2.5D extraction are inadequate due to increased complexity and computational intensity, especially with the dominance of RC parasitics and impracticality of pure 3D extraction for large-scale designs.
Innovation Solution
A method utilizing a pattern database and pattern-matching tool to partition designs into patterns, where 3D extraction parameters are stored and reused, allowing for efficient extraction and storage of patterns, reducing computational time by applying extraction results multiple times across designs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pure 3D extraction is used for accurate RC parasitic modeling, then extraction accuracy is improved, but computational time and resource consumption increase significantly
Solution Approach 1:
The system performs preliminary 3D extraction on representative patterns and stores the results in a database before actual design extraction. This pre-computed data is then reused during pattern matching, eliminating the need to perform full 3D extraction on every design instance while maintaining accuracy.
Solution Approach 2:
The system creates copies of extraction results from representative patterns and applies them to matching patterns in the actual design. Instead of performing extraction on every instance, the system copies pre-computed extraction data from the database to matched patterns, significantly reducing computational time while maintaining accuracy.
2Productivity
If pre-characterization and 2.5D extraction are used for large-scale designs, then computational time is reduced, but extraction accuracy deteriorates due to increased RC parasitic complexity
Solution Approach 1:
The system segments the design into discrete patterns that can be individually matched against the database. By dividing the large-scale design into manageable pattern units, the system can apply accurate 3D extraction results to each segment while maintaining overall computational efficiency.
Solution Approach 2:
The system changes the extraction parameters from approximate 2.5D methods to accurate 3D extraction parameters for matched patterns. When patterns are identified in the database, the system retrieves and applies the precise 3D extraction parameters associated with those patterns, thereby improving accuracy without sacrificing productivity.
Data Source
AI summary
The present disclosure relates to a method and apparatus for accurate RC extraction. A pattern database is configured to store layout patterns and their associated 3D extraction parameters. A pattern-matching tool is configured to partition a design into a plurality of patterns, and to search the pattern database for a respective pattern and associated 3D extraction parameters. If the respective pattern is already stored in the pattern database, then the associated 3D extraction parameters stored in the database are assigned to the respective pattern without the need to extract the respective pattern. If the respective pattern is not stored in the pattern database, then the extraction tool extracts the pattern and stores its associated 3D extraction parameters in the pattern database for future use. In this manner a respective pattern is extracted only once for a given design or plurality of designs. Moreover, the extraction result may be applied multiple times for a given design simultaneously, speeding up computation time. The extraction result may also be applied to a plurality of designs simultaneously.