Semiconductor Die Offset Compensation Using Predictive Shift Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for compensating die offset variations in integrated circuit production are limited, failing to accurately account for local variations due to substrate shape, mold deposition, adhesive thickness, and other factors, leading to increased defect rates and misalignment issues.
Innovation Solution
An automated semiconductor device placement method using a processor to measure, store, and predict shift measurements, applying mathematical models such as polynomial, transcendental, and Thin-Plate Spline models to adjust die placement accurately, minimizing stochastic effects and improving placement accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If current compensation methods (mean offset calculation or linear regression models) are used, then the process is simple to implement, but the manufacturing precision deteriorates due to inability to account for local variations
Solution Approach 1:
The patent implements local quality by dividing the wafer into multiple regions and applying region-specific compensation values rather than a single global offset. The system measures actual die positions in different areas and calculates localized compensation values that account for regional variations in substrate shape, mold deposition, and adhesive thickness. This allows precise correction of local placement errors while maintaining overall system effectiveness.
Solution Approach 2:
The system performs preliminary action by measuring die positions and calculating compensation values before the actual packaging process. Historical measurement data is stored and analyzed to predict and pre-correct for expected offset variations. This advance preparation enables the placement system to proactively compensate for known variations rather than reacting to errors after they occur.
2Productivity
If manual adjustment methods are used, then the device complexity is low, but the productivity deteriorates due to time-consuming manual intervention
Solution Approach 1:
The system implements self-service by automatically measuring die positions, analyzing measurement data, calculating compensation values, and applying corrections without manual intervention. The automated system performs all compensation operations independently, from data collection through final placement adjustment, eliminating the need for operators to manually calculate and apply offset corrections while maintaining high productivity.
Solution Approach 2:
The patent incorporates feedback by continuously measuring actual die positions and using this information to refine compensation calculations. The system stores historical measurement data and uses it to improve future placement accuracy. This closed-loop feedback mechanism enables the system to learn from past performance and progressively enhance placement precision while operating at high speed.
3Adaptability or versatility
If multiple models are generated and superimposed to account for different placing modules, then the adaptability improves, but the device complexity and error accumulation increase
Solution Approach 1:
The system achieves universality by implementing a unified compensation framework that can handle multiple placing modules (hands and spindles) through a single integrated model rather than requiring separate models for each module. The measurement and compensation system is designed to accommodate various placing configurations while maintaining consistent accuracy standards across all modules, eliminating the need for complex multi-model superposition.
Data Source
AI summary
A method and system for improving an automated pick and place apparatus semiconductor device placement process is provided. The method includes automatically executing a shift measurement associated with an offset from an original placement of a plurality of semiconductor die of a semiconductor wafer for processing. An associated shift measurement value is retrieved and stored in a database that includes previously retrieved shift measurement values of previously measured shift measurements. Specified models are executed with respect to all shift measurement values and a predicted shift measurement value associated with a future offset for a new plurality of semiconductor die on a new semiconductor wafer for processing is determined. Placement hardware of the pick and place apparatus is placed in multiple positions for generating the new plurality of semiconductor die on the new semiconductor wafer in accordance with the predicted shift measurement value.