Method for realizing automatic design and implementation of GDBC screening production link

Abstract

The invention provides a method for realizing automatic design and implementation of GDBC screening production links, which is characterized by comprising the following steps of: compiling a processing flow executed by a GDBC automatic script by using a computer language; and setting and calling the script to automatically run at regular time every day by utilizing a built-in task planning program of the FTP server system, and managing through a file structure of the FTP server. The method does not need manual processing to participate in the production link, and ensures the accuracy and real-time performance of the production link. According to the method, automation is achieved, the result is accurate, the cost performance is higher, people do not need to find die conforming to the GDBC rule with eyes, marking processing is manually conducted on the wafermap file of the text type, and mistakes and omissions are avoided.

Description

Technical Field

[0001] This invention belongs to the field of semiconductor chip technology, and specifically relates to a method for automating the production process of GBBC screening. Background Technology

[0002] Semiconductor manufacturing is a highly complex process, but it can be broken down into four steps: wafer fabrication, probe testing (CP), assembly, and final testing (FT). After CP testing, the CP site generates a wafer map text file, which is provided to the assembly site as a basis for selecting the bare dies that have passed CP testing and are ready for packaging on the wafer. After dicing and packaging at the assembly site, the wafer undergoes FT testing to screen out defects introduced during the manufacturing process. Even though a chip passes CP and FT testing from wafer to finished product shipment, potential defects still exist. These defects will manifest later in the chip's lifespan and trigger quality issues for end-user customers. In certain reliability-critical applications, such as automotive or medical devices, extremely low defect rates are required.

[0003] Based on the fact that defects on wafers tend to cluster, dies can be divided into different reliability sets according to the number of neighboring failures. Dies that pass all CP tests but belong to a region with many defective dies have a higher probability of potential defects. These potentially defective dies surrounded by defective dies are called "Good Die in a Bad Cluster" (GDBC). Marking these potentially defective dies to prevent them from being subsequently packaged and tested using FT is called the GDBC screening production stage.

[0004] Therefore, based on industry needs, some factories have already invested in R&D to post-process test data generated by wafer CP testing equipment to improve chip quality. This involves performing GBBC processing on the test data generated by the CP testing equipment and automatically generating a processed wafermap text file, which is provided to the assembly site of the packaging and testing factory as a basis for wafer dicing. This prevents potentially defective dies from entering the customer market and causing customer complaints, reduces chip DPPM, and avoids wasting packaging and testing costs.

[0005] However, many factories currently cannot support GDBC processing for all projects and test models. Therefore, many fabless chip design companies purchase third-party data analysis software such as KLA. When the back-end equipment of the packaging and testing plant's CP station automatically uploads STDF test data to the FTP, fabless's production engineers will download the STDF data from the FTP to a PC with the software installed. Then, they will use the software to parse and process the raw STDF format data generated during the test, generate a text-format wafer map file that has undergone GDBC filtering, and then supply it to the packaging and testing plant's assembly site as the basis for wafer dicing.

[0006] However, existing technologies also have the following drawbacks:

[0007] Because many packaging and testing (OSA) plants lack the capability to perform GBBC filtering on data generated from certain projects and equipment, many fabless chip design companies purchase third-party data analysis software for GBBC processing. This requires production engineers to download STDF data from FTP to a PC with the software installed, then parse and process the STDF data using the software to generate a text-based wafer map file after GBBC processing. This disrupts automation in the factory production process, making it partially manual. The time between CP testing and FT testing depends entirely on the engineer's ability to process the software promptly. Furthermore, STDF data is large, making processing time-consuming and labor-intensive. Additionally, the format of the wafer map file generated from STDF data may not match the factory's requirements, necessitating secondary conversion by the factory.

[0008] Some companies even abandon purchasing software due to high initial development costs, time costs, and annual license fees. Production engineers are forced to manually inspect the wafer maps generated by the CP testing plant using GBBC, and then manually mark the die tags that conform to GBBC in the text-format wafer map file generated after CP. Finally, this is provided to the assembly site of the packaging and testing plant as the basis for sharding. This inspection method cannot guarantee the effectiveness and accuracy of GBBC screening results, nor can it guarantee real-time performance.

[0009] In addition, commonly used technical terms include:

[0010] In semiconductor chip manufacturing, a whole silicon wafer is generally called a wafer. After going through the process, each unit is diced and packaged. The bare die of a single unit before packaging is called a die. It is the body of a small, unpackaged integrated circuit made of semiconductor materials.

[0011] A wafer map is an important tool for recording the test results of individual chips (dies) on a lens, and it is crucial for the semiconductor packaging and testing stage.

[0012] GBBC / GDBN (Good Die Bad Cluster / Good Die Bad Neighbor) technology analyzes the quality of adjacent chips on a wafer for each good chip. If all adjacent chips of a good chip are defective, the good chip is considered to have a high quality risk and is marked to avoid subsequent packaging and testing. Figure 1 As shown, an example of the GBBC screening rules of this invention is: CP test pass die, F is CP test fail die, and M is a die with no edge or a CP test failure. DPPM (Defects Parts Per Million) refers to the number of defective chips per million chips, and is an indicator of product defect rate. In chip manufacturing, DPPM is commonly used to measure the quantity and distribution of defects during the manufacturing process. This indicator can help manufacturers understand the problems existing in the production process and take corresponding measures to improve product quality. Summary of the Invention

[0013] To address the aforementioned issues, the purpose of this application is to: for test projects and equipment where CP (Content Processing) plants cannot support automated GBBC, manage FTP file structures and schedule daily calls to a lightweight script written in free, open-source Python. This allows for direct batch GBBC processing of text-based wafer maps uploaded to the FTP server by the CP plant without handling large volumes of STDF data. The resulting GBBC-filtered wafer map files are provided to the assembly site of the packaging and testing plant as a basis for wafer partitioning. The entire batch of wafer map file processing takes less than 1 second. Furthermore, the wafer map text files after GBBC filtering are in the same format as those generated by the factory's CP station, eliminating the need for secondary format conversion at the factory.

[0014] Specifically, this invention provides a method for automating the design and implementation of a GBBC screening production process. The method uses a computer language to write a GBBC automated script execution flow; it utilizes the task scheduler built into the FTP server system to set up timed automatic execution of the script, and manages the file structure through the FTP server; further comprising:

[0015] Using a PC as an FTP server, set up daily automatic scheduled execution of GDBC scripts on the PC. Create a directory containing four paths on this FTP server: Path0 is used to store the script execution records, path1 is the upload path folder for the wafermap generated by the CP test station, path2 is the storage path for the new wafermap after the GDBC filtering process is completed, which is specifically provided for the assembly station to download, and path3 is a temporary storage path for files that are being processed by GDBC.

[0016] Three accounts and passwords (usernames and passwords) are set up, each with different operation permissions for the four FTP paths. Username 0 is used by production engineers to manage, maintain, and update GBBC scripts and check GBBC processing records, and can operate on paths 0, 1, 2, and 3. Username 1 is used when the CP test station automatically uploads data, and can only operate on path 1. Username 2 is used when the assembly station automatically downloads data, and can only operate on path 2.

[0017] After the factory CP site test is completed, configure the test machine to automatically upload the wafermap text file to the FTP path1 using account password 1;

[0018] Using a computer language to design a GDBC automation script, combined with a spatial screening method for potential defects in wafers, the GDBC automation script is set to run automatically at fixed times on an FTP server.

[0019] The GDBC automation script automatically obtains the latest list of wafer map files uploaded by the CP factory that need to be processed by acquiring and comparing the differences between the file lists in path1 and path2; then it automatically copies the files in the list to path3 and performs GDBC filtering processing in sequence to prevent unfinished files from being downloaded by the user with FTP account 2; finally, it generates a new wafer map text file after GDBC filtering and moves these files to path2, while continuously writing the processing information to the log file in path0.

[0020] The assembly site automatically downloads the wafermap text file from path2 in the FTP server based on the file name, using account password 2 as the basis for partitioning;

[0021] The above methods automate the GBBC screening process.

[0022] The method further includes:

[0023] First, set the FTP file structure and account / password permissions for each path as follows:

[0024] Path0 is used to store records of script execution;

[0025] Path1 is the path to the file containing the map from CP in the local directory;

[0026] Path2 is the path in the local directory where the map for assembly file regenerated after the GDBC filtering is performed;

[0027] Path3 is a temporary storage path for files that are being processed by GDBC;

[0028] Account password 0: Used for mass production engineer management, with access to all folders under the project;

[0029] Account Password 1: Used for uploading wafermaps data to the CP site, and only has permissions to the path1 folder;

[0030] Account Password 2: Used for downloading wafermap data from the Assembely site; only has permissions to the path2 folder.

[0031] Configure the FTP server's built-in task scheduler to periodically run GDBCautomation.py, a script written in computer language.

[0032] The method further includes:

[0033] Whenever a batch of wafers from a product passes through the CP station, the testing machine at the CP station automatically uploads the wafer map text file generated after the test to the specified directory path1 on the FTP server using the specified account and password 1. During the timed script execution period, the assembly station automatically downloads the wafer map text file that has been processed and filtered and marked by GBBC from the specified directory path2 on the FTP server using the specified account and password 2.

[0034] The spatial screening method for potential defects in wafers includes:

[0035] The wafermap text file produced by the CP testing plant is inspected and processed using GBBC. By extracting data from a fixed number of lines in the file, the row and column coordinate distribution of all dies on the wafermap and information on whether they are good dies are obtained.

[0036] Next, check each die according to its coordinates to see if it is a good die. If it is, extract the information of the 8 surrounding dies in the 3×3 matrix and check if there is a good die. If not, this good die belongs to the die in the area with many defective dies, and it meets the GBBC rule. Mark the die at this coordinate. If there is a good die, continue to check if the die at the next coordinate position is a good die. Continue until all dies in this wafer map have been checked and marked.

[0037] Generate a wafermap file after filtering by GDBC, which will be used as the basis for wafer partitioning at the assembly site of the packaging and testing plant; finally, obtain the current time, extract the wafer ID, count the final number of good dies, and generate records of the defective dies filtered by GDBC in a specified format.

[0038] The method described uses a lightweight GBBC automation script written in the free and open-source Python programming language to execute the processing flow.

[0039] The method for designing the execution flow of the GBBC automation script further includes the following steps:

[0040] S0, Start;

[0041] S1, obtain the file lists of path1 and path2 and compare them to get the file list to be processed final_filenames1;

[0042] S2, copy the files in the final_filenames1 list to the path3 path, where path3 contains the filename list file_names3;

[0043] S3, determine if the traversal of file in file_name3 has ended? If not, proceed to step S4; if yes, proceed to step S11.

[0044] S4, extract die distribution information and row and column lengths ROW_len and COL_len from the wafer map text file and store them in map_dic;

[0045] S5, determine whether Y is within the range of COL_len, whether X is within the range of ROW_len, and whether the traversal has ended. If not, proceed to step S6; if yes, proceed to step S13.

[0046] S6, check die information one by one according to coordinates;

[0047] S7. Determine if die is a good die. If not, return to step S6; if yes, proceed to step S8.

[0048] S8, extract the information list of the remaining 8 dies in the 3×3 matrix;

[0049] S9. Determine if the list contains "good die". If not, proceed to step S10; if yes, increment Y and X, and return to step S5.

[0050] S10, mark, update map_dic, record GDBC die to record_dic; further, Y++, X++, further return to step S5;

[0051] S11, Move all wafermap files in path3 to path2;

[0052] S12, End;

[0053] S13, calculate the number of good dies in map_dic and write it to the end of the first line of the wafe map text file to form a GDBC flag;

[0054] S14, write map_dic into the WAFE map text file to form the updated file; further, the WAFE map text file will serve as the basis for sharding the assembly site of the packaging and testing plant;

[0055] S15, obtain time information, extract the wafer ID from the wafer map text file, extract the record_dic record, and write them all to record_msg;

[0056] S16, Write all information in the record_msg list into the record file according to the format; further, the record text file will be used as a wafer to execute GBBC records for traceability;

[0057] S17, return to step S3.

[0058] The GDBCautomation script is

[0059] The specific execution steps of the internal processing flow in GDBCautomation.py are as follows:

[0060] Get the file lists in directories path1 and path2, compare the differences, and obtain the latest wafermap file list final_filenames1 produced and uploaded by the factory;

[0061] Copy all wafermap text files in the final_filenames1 list to the path3 directory; obtain the file list file_names3 under path3, and perform GDBC filtering on each wafermap text file in the list. The GDBC filtering process adopts the spatial filtering method for potential wafer defects.

[0062] By extracting data from a fixed number of lines in the wafer map text file, the row and column coordinate distribution of all dies on the wafer map, as well as the row and column lengths ROW_len and COL_len, are obtained and stored in the created dictionary map_dic. Then, each die is checked one by one according to its coordinates to determine whether it is a good die. If it is, the information of the 8 surrounding dies within the 3x3 matrix is ​​extracted and it is checked whether there are any good dies. If not, this good die belongs to the die in the region with many defective dies and conforms to the GBBC rules. The die information at this coordinate is marked as X and updated in map_dic and stored as a record in the created dictionary record_dic. If there is a good die, the process continues to check whether the die at the next coordinate position is a good die until all dies in this wafer map have been checked and marked. The number of good dies in map_dic is calculated and written to the end of the first line of the wafer map text file to form a GBBC flag. The updated map_dic is then written back to the processed wafer map text file according to the coordinates. At the same time, time information is obtained, wafer ID information is extracted from the wafer map text file, record_dic records are extracted and written to the created list record_msg. Then, all information in the list record_msg is written to the GBBC_record file under the path0 directory as wafer execution GBBC records for traceability.

[0063] After all wafer map text files in the file_names3 list have been filtered by GDBC, move all wafer map text files to the path 2 directory;

[0064] This completes the steps in GDBCautomation.py.

[0065] Therefore, the advantage of this application is:

[0066] This invention provides an automated GBBC processing method that eliminates the need for manual intervention in the production process, ensuring accuracy and real-time performance. This invention achieves automation, provides accurate results, and offers better cost-effectiveness. It avoids the need for manual identification of dies conforming to GBBC rules and the manual marking of text-based wafer map files, thus preventing errors and omissions. Attached Figure Description

[0067] The accompanying drawings, which are provided to further illustrate the invention and form part of this application, are not intended to limit the scope of the invention.

[0068] Figure 1 This is a diagram illustrating an example of a GBBC filtering rule.

[0069] Figure 2 This is a schematic diagram of the GBBC automated script execution process designed in this application.

[0070] Figure 3 This is a schematic diagram of the FTP file structure and account / password permissions in this application.

[0071] Figure 4 This is a schematic diagram showing a comparison example of the GBBC screening process before and after in this application. Detailed Implementation

[0072] To better understand the technical content and advantages of the present invention, the present invention will now be described in further detail with reference to the accompanying drawings.

[0073] This invention provides a method for automating the production process of GBBC screening. A PC is used as an FTP server, and a GBBC script processing program is set to run automatically daily. A directory containing four paths is created on this FTP server: Path0 stores script execution records; Path1 is the upload path folder for the wafermap generated by the CP test station; Path2 is the storage path for new wafermaps after GBBC screening, specifically provided for download by the assembly station; and Path3 is a temporary storage path for files currently being processed by GBBC.

[0074] Set up 3 accounts and passwords to have different operation permissions for the 4 FTP paths. Set account 0 as the production engineer to manage, maintain and update GDBC scripts and check GDBC processing records. It can operate on paths 0, 1, 2 and 3.

[0075] When setting up account password 1 for automatic data upload on the CP test site, it can only be operated on path1;

[0076] When setting up account password 2 for automatic downloading from the assembly site, only path2 can be operated on.

[0077] The above settings prevent CP sites from overwriting wafermap files processed by GBBC, and also prevent assembly from downloading unprocessed wafermap files.

[0078] After the factory CP site test is completed, configure the test machine to automatically upload the wafer map text file to the FTP path 1 using account password 1.

[0079] A spatial screening and processing method for potential defects in wafers is combined with a free and open-source Python-based automated script for GBBC. The script is set to run automatically at a fixed time (e.g., 7:00 AM daily) on an FTP server. The spatial screening and processing method for potential defects in wafers includes:

[0080] The wafermap text file produced by the CP testing plant is inspected and processed using GBBC. By extracting data from a fixed number of lines in the file, the row and column coordinate distribution of all dies on the wafermap and information on whether they are good dies are obtained.

[0081] Next, check each die according to its coordinates to see if it is a good die. If it is, extract the information of the 8 adjacent dies in the 3x3 matrix and check if there is a good die. If not, the good die belongs to the area with many defective dies and meets the GBBC rule. Mark the die at this coordinate. If there is a good die, continue to check if the die at the next coordinate is a good die. Continue until all dies in this wafermap have been checked and marked.

[0082] Generate a wafermap file after filtering by GDBC, which will be used as the basis for wafer partitioning at the assembly site of the packaging and testing plant; finally, obtain the current time, extract the wafer ID, count the final number of good dies, and generate records of the defective dies filtered by GDBC in a specified format.

[0083] The GDBC script automatically obtains the latest list of wafermap files uploaded by the CP plant that need to be processed by obtaining and comparing the differences between the file lists in path1 and path2. It then automatically copies the files in the list to path3 for sequential GDBC filtering to prevent incomplete files from being downloaded by the user with FTP account 2. This automatic copying of files to path3 for sequential GDBC filtering is done by the GDBC automation script, rather than directly copying them to path2. This is to prevent incomplete files from being downloaded by the user with FTP account 2 during the GDBC script's execution and marking process, as the GDBC automation script is set to be executed by the system's built-in task scheduler on the FTP machine and does not require user intervention. Finally, a new wafermap text file filtered by GDBC is generated, and these files are moved to path2, while processing information is continuously written to a log file in path0.

[0084] The Assembly site automatically downloads the wafermap text file from the FTP path 2 using the username and password 2, based on the file name, as the basis for wafer partitioning. For example... Figure 2 The diagram below illustrates the design of the GBBC automated script execution flow:

[0085] S0, Start;

[0086] S1, obtain the file lists of path1 and path2 and compare them to get the file list to be processed final_filenames1;

[0087] S2, copy the files in the final_filenames1 list to the path3 path, where path3 contains the filename list file_names3;

[0088] S3, determine if the traversal of file in file_name3 has ended? If not, proceed to step S4; if yes, proceed to step S11.

[0089] S4, extract die distribution information and row and column lengths ROW_len and COL_len from the wafer map text file and store them in map_dic;

[0090] S5, determine whether Y is within the range of COL_len, whether X is within the range of ROW_len, and whether the traversal has ended. If not, proceed to step S6; if yes, proceed to step S13.

[0091] S6, check die information one by one according to coordinates;

[0092] S7. Determine if die is a good die. If not, return to step S6; if yes, proceed to step S8.

[0093] S8, extract the information list of the remaining 8 dies in the 3×3 matrix;

[0094] S9. Determine if the list contains "good die". If not, proceed to step S10; if yes, increment Y and X, and return to step S5.

[0095] S10, mark, update map_dic, record GDBC die to record_dic; further, Y++, X++, further return to step S5;

[0096] S11, Move all wafermap files in path3 to path2;

[0097] S12, End;

[0098] S13, calculate the number of good dies in map_dic and write it to the end of the first line of the wafe map text file to form a GDBC flag;

[0099] S14, write map_dic into the WAFE map text file to form the updated file; further, the WAFE map text file will serve as the basis for sharding the assembly site of the packaging and testing plant;

[0100] S15, obtain time information, extract the wafer ID from the wafer map text file, extract the record_dic record, and write them all to record_msg;

[0101] S16, Write all information in the record_msg list into the record file according to the format; further, the record text file will be used as a wafer to execute GBBC records for traceability;

[0102] S17, return to step S3.

[0103] The above methods automate the GBBC screening process. Taking the GBBC automation application in the connectivity product project at Beijing Junzheng's ASE factory as an example, the details are as follows:

[0104] First, set the FTP file structure and permissions for each path as follows:

[0105] Path0 is used to store records of script execution;

[0106] Path1 is the path to the file containing the map from CP in the local directory;

[0107] Path2 is the path in the local directory where the map for assembly file regenerated after the GDBC filtering is performed;

[0108] Path3 is a temporary storage path for files that are being processed by GDBC;

[0109] Account password 0: Used for mass production engineer management, with access to all folders under the project;

[0110] Account Password 1: Used for uploading wafermaps data to the CP site, and only has permissions to the path1 folder;

[0111] Account Password 2: Used for downloading wafermap data from the Assembely site; only has permissions to the path2 folder.

[0112] like Figure 3 As shown, this displays the FTP file structure and account / password permissions.

[0113] Configure the FTP server's built-in task scheduler to run the Python script GDBCautomation.py once a day at 7:00 am.

[0114] Within the ASE factory, whenever a batch of wafers passes through the CP station, the testing machine at the CP station automatically uploads the generated wafermap text file to the specified directory path1 on the FTP server using a designated account and password 1. Within 24 hours (depending on the settings of the scheduled script), the assembly station can automatically download the processed wafermap text file, which has been filtered and marked by GBBC, from the specified directory path2 on the FTP server using an account and password 2.

[0115] The specific execution steps of the processing flow within the GDBCautomation.py script are as follows:

[0116] Get the file lists in directories path1 and path2, compare the differences, and obtain the latest wafermap file list final_filenames1 produced and uploaded by the factory;

[0117] Copy all wafer map text files in the final_filenames1 list to the path3 directory;

[0118] Get the file list file_names3 under path3, and perform GDBC filtering on each wafermap text file in the list. The GDBC filtering process is a spatial filtering method for potential defects in wafers.

[0119] The data from the fixed number of lines in the wafer map text file is extracted to obtain the row and column coordinate distribution of all dies on the wafer map, as well as the row and column lengths ROW_len and COL_len, which are stored in a created dictionary map_dic. Then, each die is checked against its coordinates to determine if it is a good die. If it is, the information of the eight adjacent dies within the 3x3 matrix is ​​extracted, and it is checked whether there are any good dies there. If not, this good die belongs to a region with many defective dies and conforms to the GBBC rules. The die information at this coordinate is marked as X, updated in map_dic, and stored as a record in the created dictionary record_dic. If a good die exists, the process continues to check whether the die at the next coordinate position is a good die, until all dies in the wafer map have been checked and marked. The number of good dies in map_dic is calculated and written to the end of the first line of the wafer map text file to form a GBBC flag. The updated map_dic is then written back to the processed wafer map text file according to the coordinates. Simultaneously, time information is acquired, wafer ID information is extracted from the wafer map text file, record_dic records are extracted and written to the created list record_msg, and then all information in the list record_msg is written to the GDBC_record file under the path0 directory as wafer execution GDBC records for traceability.

[0120] After all wafermap text files in the file_names3 list have been filtered using GDBC, move all wafermap text files to the path 2 directory.

[0121] This completes the steps in GDBCautomation.py.

[0122] like Figure 4 The image shows a comparison example of wafermap GDBC filtering before and after processing.

[0123] In summary, this invention provides a method for automating the GBBC production process, and also provides a processing flow that can be executed using a GBBCautomation script written in the free and open-source Python language. By utilizing the task scheduler built into the FTP server system to set up the automatic execution of the GBBCautomation.py script at regular intervals, and by managing the file structure of the FTP server, the data security and accuracy of the GBBC production process are ensured, thus guaranteeing the automation and de-management of the GBBC production process.

[0124] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for automating the production process of GBBC screening, characterized in that, The method uses computer languages ​​to write GBBC automated scripts to execute the processing flow. The script is scheduled to run automatically daily using the task scheduler built into the FTP server system, and file structure management is implemented through the FTP server; further including: Using a PC as an FTP server, set up a daily scheduled automatic execution of GDBC scripts on the PC. Create a directory containing four paths on this FTP server: Path0 is used to store the script execution records, path1 is the upload path folder for the wafermap generated by the CP test station, path2 is the storage path for the new wafermap after the GDBC filtering process is completed, which is specifically provided for the assembly station to download, and path3 is a temporary storage path for files that are being processed by GDBC. Three accounts and passwords (usernames and passwords) are set up, each with different operation permissions for the four FTP paths. Username 0 is used by production engineers to manage, maintain, and update GBBC scripts and check GBBC processing records, and can operate on paths 0, 1, 2, and 3. Username 1 is used when the CP test station automatically uploads data, and can only operate on path 1. Username 2 is used when the assembly station automatically downloads data, and can only operate on path 2. After the factory CP site test is completed, configure the test machine to automatically upload the wafermap text file to the FTP path1 using account password 1; Using a computer language to write a GDBC automation script, combined with a spatial screening method for potential defects in wafers, the GDBC automation script is set to run automatically at fixed times on an FTP server. The GDBC automation script automatically obtains the latest list of wafer map files uploaded by the CP factory that need to be processed by acquiring and comparing the differences between the file lists in path1 and path2; then it automatically copies the files in the list to path3 and performs GDBC filtering processing in sequence to prevent unfinished files from being downloaded by the user with FTP account 2; finally, it generates a new wafer map text file after GDBC filtering and moves these files to path2, while continuously writing the processing information to the log file in path0. The assembly site automatically downloads the wafermap text file from path2 in the FTP server based on the file name, using account password 2 as the basis for partitioning; The above methods automate the GBBC screening process.

2. The method for automating the production process of GBBC screening according to claim 1, characterized in that, The method further includes: First, set the FTP file structure and account / password permissions for each path as follows: Path0 is used to store records of script execution; Path1 is the path to the file containing the map from CP in the local directory; Path2 is the path in the local directory where the map for assembly file regenerated after the GDBC filtering is performed; Path3 is a temporary storage path for files that are being processed by GDBC; Account password 0: Used for mass production engineer management, with access to all folders under the project; Account Password 1: Used for uploading wafermaps data to the CP site, and only has permissions to the path1 folder; Account Password 2: Used for downloading wafermap data from the Assembely site; only has permissions to the path2 folder. Configure the FTP server's built-in task scheduler to run GDBCautomation.py, a script written in computer language, once a day.

3. The method for automating the production process of GBBC screening according to claim 2, characterized in that, The method further includes: Whenever a batch of wafers from a product passes through the CP station, the testing machine at the CP station automatically uploads the wafer map text file generated after the test to the specified directory path1 on the FTP server using the specified account and password 1. During the timed script execution period, the assembly station automatically downloads the wafer map text file that has been processed and filtered and marked by GBBC from the specified directory path2 on the FTP server using the specified account and password 2.

4. The method for automating the production process of GBBC screening according to claim 2, characterized in that, The spatial screening method for potential defects in wafers includes: The wafermap text file produced by the CP testing plant is inspected and processed using GBBC. By extracting data from a fixed number of lines in the file, the row and column coordinate distribution of all dies on the wafermap and information on whether they are good dies are obtained. Next, check each die according to its coordinates to see if it is a good die. If it is, extract the information of the 8 surrounding dies in the 3×3 matrix and check if there is a good die. If not, this good die belongs to the die in the area with many defective dies, and it meets the GBBC rule. Mark the die at this coordinate. If there is a good die, continue to check if the die at the next coordinate position is a good die. Continue until all dies in this wafer map have been checked and marked. Generate a wafermap file after filtering by GDBC, which will be used as the basis for wafer partitioning at the assembly site of the packaging and testing plant; finally, obtain the current time, extract the wafer ID, count the final number of good dies, and generate records of the defective dies filtered by GDBC in a specified format.

5. The method for automating the production process of GBBC screening according to claim 2, characterized in that, The method described uses a lightweight GBBC automation script written in the free and open-source Python programming language to execute the processing flow.

6. The method for automating the production process of GBBC screening according to claim 5, characterized in that, The method for designing the execution flow of the GBBC automation script further includes the following steps: S0, Start; S1, obtain the file lists of path1 and path2 and compare them to get the file list to be processed final_filenames1; S2, copy the files in the final_filenames1 list to the path3 path, where path3 contains the filename list file_names3; S3, determine if the traversal of file in file_name3 has ended? If not, proceed to step S4; if yes, proceed to step S11. S4, extract die distribution information and row and column lengths ROW_len and COL_len from the wafer map text file and store them in map_dic; S5, determine whether Y is within the range of COL_len, whether X is within the range of ROW_len, and whether the traversal has ended. If not, proceed to step S6; if yes, proceed to step S13. S6, check die information one by one according to coordinates; S7. Determine if die is a good die. If not, return to step S6; if yes, proceed to step S8. S8, extract the information list of the remaining 8 dies in the 3×3 matrix; S9. Determine if the list contains "good die". If not, proceed to step S10; if yes, increment Y and X, and return to step S5. S10, mark, update map_dic, record GDBC die to record_dic; further, Y++, X++, further return to step S5; S11, Move all wafermap files in path3 to path2; S12, End; S13, calculate the number of good dies in map_dic and write it to the end of the first line of the wafe map text file to form a GDBC flag; S14, write map_dic into the WAFE map text file to form the updated file; further, the WAFE map text file will serve as the basis for sharding the assembly site of the packaging and testing plant; S15, obtain time information, extract the wafer ID from the wafer map text file, extract the record_dic record, and write them all to record_msg; S16, Write all information in the record_msg list into the record file according to the format; further, the record text file will be used as a wafer to execute GBBC records for traceability; S17, return to step S3.

7. The method for automating the production process of GBBC screening according to claim 6, characterized in that, The GDBCautomation script is GDBCautomation.py, and its internal processing steps are as follows: Get the file lists in directories path1 and path2, compare the differences, and obtain the latest wafermap file list final_filenames1 produced and uploaded by the factory; Copy all wafermap text files in the final_filenames1 list to the path3 directory; obtain the file list file_names3 under path3, and perform GDBC filtering on each wafermap text file in the list. The GDBC filtering process adopts the spatial filtering method for potential wafer defects. The data after a fixed number of lines in the wafer map text file is extracted to obtain the row and column coordinate distribution of all dies on the wafer map, as well as the row and column lengths ROW_len and COL_len, which are stored in the created dictionary map_dic. Then, each die is checked for good dies according to its coordinates. If it is a good die, the information of the 8 surrounding dies in the 3x3 matrix is ​​extracted and it is checked for good dies. If it is not a good die, the good die belongs to the die in the area with many defective dies and conforms to the GBBC rules. The information of the die at this coordinate is marked as X and updated in map_dic and stored as a record in the created dictionary record_dic. If a good die exists, the process continues to check whether the die at the next coordinate position is a good die until all dies in the wafer map have been checked and marked. The number of good dies within map_dic is calculated and written to the end of the first line of the wafer map text file to form a GDBCflag. The updated map_dic is then written back to the processed wafermap text file according to the coordinates. At the same time, time information is obtained, wafer ID information is extracted from the wafer map text file, and record_dic records are extracted and written to the created list record_msg. Then, all information in the list record_msg is written to the GDBC_record file under the path0 directory as wafer execution GDBC records for traceability. After all wafer map text files in the file_names3 list have been filtered by GDBC, move all wafer map text files to the path 2 directory; This completes the steps in GDBCautomation.py.

Images