ML Pattern Recognition for IC Die Connector Assignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for assigning connectors to pad pins on integrated circuit (IC) dies face challenges in optimizing routability and wirelength metrics, especially with large numbers of connectors and pad pins, leading to long runtime and poor quality of results due to sequential processing and inability to accurately measure quality until after redistribution layer routing.
Innovation Solution
The use of machine learning-based image recognition to identify patterns within a design image of an IC die, associating mappings of connections between connectors and pad pins, and assigning these connections concurrently for groups of identified patterns, allowing for faster runtime and improved quality of results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional sequential processing methods are used to assign connectors to pad pins, then the assignment can be completed with detailed optimization, but the runtime becomes excessively long and productivity decreases
Solution Approach 1:
The design image is divided into multiple tiles, and pattern recognition is performed independently on each tile. This segmentation allows parallel processing of different regions, significantly reducing overall runtime while maintaining assignment optimization quality through localized pattern matching and connection assignment.
2Manufacturing precision
If detailed optimization of connector-pad pin assignments is performed sequentially, then assignment quality can be optimized, but the complexity of the process increases and productivity decreases
Solution Approach 1:
The patent replaces complex sequential mechanical-like processing with machine learning-based pattern recognition. The ML model automatically identifies patterns and makes assignments based on learned features, eliminating the need for complex sequential optimization algorithms while maintaining or improving assignment quality.
3Measurement precision
If conventional methods are used to measure assignment quality, then detailed analysis can be performed, but the measurement can only be done after redistribution layer routing, causing loss of time
Solution Approach 1:
The patent performs preliminary pattern recognition and assignment quality assessment before completing the full redistribution layer routing. The ML model evaluates assignment quality based on recognized patterns and spatial relationships, enabling early quality measurement and feedback without waiting for the complete routing process.
Data Source
AI summary
A method for assigning connections between IO pad pins and connectors on an integrated circuit (IC) die. A pattern (300) including a physical layout of connectors (302) and pad pins (304) is associated with a mapping of connections between the connectors (302) and the pad pins (304). A processor (204) identifies instances (402, 404) of the pattern (300) within a design image (400) of an integrated circuit (IC) die using a machine learning model. The design image (400) includes a physical layout of connectors (414) and pad pins (416). For each identified instance (402, 404) of the pattern (300) within the design image (400), the mapping of connections is assigned to respective connectors (414) and pad pins (416) in the identified instance (402, 404).