Data Processing Apparatus for Semiconductor Error Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Managing and analyzing high-resolution time series data from semiconductor manufacturing processes, such as light emission spectroscopy data, is costly and challenging due to the large amount of data, making it difficult to determine error occurrences directly, and compression can lead to loss of feature data.
Innovation Solution
A data processing apparatus that normalizes time series data, divides it into regions, extracts outlier values, and converts these values into color data to generate image data, reducing data loss and management costs while enabling error detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If time series data of multiple wavelengths is imaged and displayed, then the user can easily determine the presence or absence of error occurrence, but the data management cost increases due to the enormous amount of data
Solution Approach 1:
The patent extracts only the necessary feature data from the time series data by dividing it into regions and selecting representative values (such as maximum, minimum, or average values) from each region. This extraction process retains the essential information needed for error detection while discarding redundant data, thereby reducing the data amount required for management and display.
Solution Approach 2:
The patent segments the time series data into multiple regions based on time ranges and wavelength ranges. By dividing the continuous data stream into discrete segments and selecting representative values from each segment, the system reduces the overall data volume while preserving the key features necessary for error detection.
2Quantity of substance
If the imaged data is compressed and managed, then the management cost may be reduced, but there is a possibility that feature data indicative of the presence or absence of error occurrence may be lost
Solution Approach 1:
The patent applies different processing strategies to different regions of the data. By dividing the time series data into regions and selecting representative values based on local characteristics (such as maximum, minimum, or average values within each region), the system preserves the essential feature data needed for error detection while achieving compression. Each region is processed independently to maintain local quality and prevent feature data loss.
3Measurement precision
If high resolution measurement data is collected from semiconductor manufacturing processes, then the measurement precision is improved, but the data management cost increases due to the enormous amount of data
Solution Approach 1:
The patent extracts representative values from high-resolution time series data by dividing it into regions and selecting key data points (such as maximum, minimum, or average values) from each region. This extraction process maintains the measurement precision needed for detecting manufacturing errors while significantly reducing the data volume that requires management and processing.
Solution Approach 2:
The patent segments the high-resolution measurement data into multiple regions based on time and wavelength characteristics. By processing each segment independently and selecting representative values, the system preserves the detailed measurement information necessary for precision error detection while reducing the overall data burden for management.
Data Source
AI summary
A data processing apparatus comprises processing circuitry configured to generate normalized data by normalizing time series data of multiple wavelengths, measured from a substrate, by using predetermined reference data; divide the normalized data into a plurality of regions for each predetermined time range and predetermined wavelength range; for each respective region of the plurality of regions, extract an outlier value in the respective region as a representative value; and convert the representative values from the plurality of regions into color data to generate image data from the color data.