Characteristic Prediction System for Semiconductor Electrical Testing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semiconductor device manufacturing, it is challenging to measure electrical characteristics under high temperature environments or with large currents, as standard measurement instruments are not capable of operating within these conditions.
Innovation Solution
A characteristic prediction system that acquires electrical characteristics from semiconductor devices using a measurement instrument within its measurable range and employs a prediction model to forecast out-of-range characteristics, such as those at high temperatures or with large currents.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If measurement instruments are used to measure electrical characteristics under high temperature or large current conditions, then measurement capability is improved, but measurement cost and system complexity increase significantly
Solution Approach 1:
The patent uses a prediction model as an intermediary to bridge the gap between measurable characteristics (within instrument range) and unmeasurable characteristics (high temperature, large current conditions). Instead of directly measuring difficult-to-obtain data, the system measures easy characteristics and uses the prediction model to infer the difficult characteristics, thereby avoiding the need for complex high-temperature measurement equipment.
Solution Approach 2:
The patent creates a virtual copy of the measurement process through machine learning models. By training the model on limited actual measurement data and simulating additional measurement scenarios computationally, the system generates prediction results that mimic what would be obtained from direct high-temperature measurements, eliminating the need for physical high-temperature measurement instruments.
2Measurement precision
If special measurement equipment is acquired to measure out-of-range characteristics, then measurement range is improved, but measurement cost increases
Solution Approach 1:
The patent creates a virtual copy of the measurement process through machine learning models. By training the model on limited actual measurement data and simulating additional measurement scenarios computationally, the system generates prediction results that mimic what would be obtained from direct high-temperature measurements, eliminating the need for physical high-temperature measurement instruments.
Solution Approach 2:
The patent replaces the physical measurement system (requiring expensive high-temperature equipment) with a computational system. Instead of using mechanical/physical means to create high-temperature measurement conditions, the system uses information processing and machine learning algorithms to predict the characteristics, substituting computational resources for expensive physical measurement infrastructure.
3Reliability
If burn-in test is performed on all packages, then reliability is improved, but productivity decreases due to increased processing time
Solution Approach 1:
The patent performs preliminary classification of packages into high-risk and low-risk groups using prediction models before the burn-in test process. By predicting which packages are likely to fail based on manufacturing data and electrical characteristics, the system identifies only the necessary packages for burn-in testing, performing the reliability screening action in advance rather than universally applying it to all packages.
Solution Approach 2:
The patent applies burn-in test only to the subset of packages predicted to be high-risk, rather than applying it excessively to all packages. This partial action approach maintains adequate reliability screening for the critical packages while avoiding the excessive processing time and resource consumption that would result from testing every package, thereby optimizing the balance between reliability and productivity.
Data Source
AI summary
A characteristic prediction system acquires an electrical characteristic of a semiconductor device measured by a measurement instrument, and predicts an out-of-range characteristic, which is the electrical characteristic of the semiconductor device beyond a measurable range of the measurement instrument, from at least the electrical characteristic acquired and using a prediction model stored in a memory unit.