Unlock instant, AI-driven research and patent intelligence for your innovation.

A method for improving semiconductor chip yield using machine learning classifiers

A machine learning and classifier technology, applied in machine learning, semiconductor/solid-state device testing/measurement, instruments, etc., to achieve the effect of improving yield rate and accuracy

Active Publication Date: 2022-02-11
普赛微科技(杭州)有限公司
View PDF10 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The ensemble method mainly has two layers of algorithm hierarchy

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • A method for improving semiconductor chip yield using machine learning classifiers
  • A method for improving semiconductor chip yield using machine learning classifiers
  • A method for improving semiconductor chip yield using machine learning classifiers

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0051] Embodiment 1: After the semiconductor chip wafer is manufactured, it will undergo a wafer acceptance test (WAT), and then the wafer will be sent to a packaging factory for classification testing and packaging. The packaging factory will classify the wafers by quality according to the WAT data results of the wafers, so as to carry out the next step of the bare chip CP test. Different quality bare chips will undergo different CP test procedures. These pre-test data of the die will be used to predict the final test (FT) results of the die. Using machine learning algorithms to predict wafer die FT results mainly includes the following two parts:

[0052] (1) Classifier training. Usually, the proportion of unqualified die on the ex-factory wafer is relatively small, so there is a great imbalance between the data sets of unqualified and qualified die. Using such a data set will affect the accuracy of machine learning model training. come to have a big impact. One of the em...

Embodiment 2

[0054] Embodiment 2: After the semiconductor chip wafer is manufactured, it will undergo a wafer acceptance test (WAT), and then the wafer will be sent to a packaging factory for classification testing and packaging. The packaging factory will classify the dies according to the quality according to the WAT data results of the wafer, so as to carry out the next step of the die CP test. Different quality dies will undergo different CP test procedures. These pre-test data of the die will be used to predict the final test (FT) results of the die. Using machine learning algorithms to predict wafer die FT results mainly includes the following two parts:

[0055] (1) Classifier training. Usually, the proportion of unqualified die on the ex-factory wafer is relatively small, so there is a great imbalance between the data sets of unqualified and qualified die. Using such a data set will affect the accuracy of machine learning model training. come to have a big impact. One of the emb...

Embodiment 3

[0057] Embodiment 3: After the semiconductor chip wafer is manufactured, a wafer acceptance test (WAT) will be performed, and then the wafer will be sent to a packaging factory for classification testing and packaging. The packaging factory will classify the dies according to the quality according to the WAT data results of the wafer, so as to carry out the next known qualified die (CP) test, and different quality dies will undergo different CP test procedures. These pre-test data of the die will be used to predict the finished product test (FT) results of the die. Using machine learning algorithms to predict wafer die FT results mainly includes the following two parts:

[0058] (1) Classifier training. Usually, the proportion of unqualified die on the ex-factory wafer is relatively small, so there is a great imbalance between the data sets of unqualified and qualified die. Using such a data set will affect the accuracy of machine learning model training. come to have a big ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a method for improving the yield rate of semiconductor chips by using a machine learning classifier. The machine learning classifier is used to predict the finished product of the wafer die after packaging by analyzing the early data of the die before packaging. Test (Final Test, FT) results, and then classify the quality of the wafer die according to the predicted FT results, and finally package according to the quality of the die. The method can effectively improve the accuracy of quality classification before wafer die packaging, thereby improving the yield rate of semiconductor chips after packaging.

Description

technical field [0001] The invention relates to the field of development / manufacturing of semiconductor chip products, in particular to a method for improving the accuracy of wafer bare chips in the quality classification stage by using machine learning algorithms, thereby improving the yield rate of chips after packaging. Background technique [0002] Wafers of semiconductor chip products (such as memory chips, SOC chips, etc.) need to undergo a series of various tests from the beginning of manufacturing to delivery to customers, such as figure 1 (100) mainly includes: (1) Step 102, the wafer acceptance test (Wafer Acceptance Test, WAT) carried out after the manufacturing is completed, mainly tests the special test pattern (Test Key), and checks it by electrical parameters Whether the manufacturing process of each step is normal and stable; (2) Step 104, the performance and function test (Circuit Probe, CP) of the chip before packaging is mainly to perform different levels ...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06N20/00H01L21/66
CPCG06N20/00H01L22/20
Inventor 刘瑞盛蒋信喻涛
Owner 普赛微科技(杭州)有限公司