Automated Low-Yield Processing System (Part 1)

Abstract

This invention relates to an automated low-yield processing system. The system primarily connects information from this invention to a testing device and uses a data acquisition module to connect to its testing station to read the yield test results of multiple test objects. Then, a calculation module calculates the average and standard deviation to obtain outlier data. A comparison module then compares the outlier data with standard test data in a test data database to identify the individual outliers among the test objects. Finally, a processing recommendation module automatically selects high-success-rate processing solutions from a processing case database for the specific test item, providing recommended solutions for handling the low-yield problem of the test object. This allows for rapid identification of the causes of low-yield problems and automatic provision of high-success-rate recommended solutions, thereby improving repair efficiency.

Description

Technical Field

[0001] This invention provides an automated low-yield processing system that can quickly identify the causes of low-yield problems and automatically provide recommended solutions with high success rates to improve repair efficiency. Prior Technology

[0002] Note that yield control on the production line is a crucial quality management process, aiming to ensure that products manufactured during the process meet expected quality standards. A production line typically involves more than one type of machine, and each machine uses different hardware, such as testing machines, carriers (handlers and probers), and fixtures. A production line with stable, high yields usually requires clearly defined quality standards, an effective production process, real-time monitoring of testing parameters at each stage, implementation of product quality testing procedures, and setting yield targets. If a product is found to be non-compliant with quality standards, an anomaly handling procedure is immediately initiated, and the production line is regularly reviewed and evaluated. However, besides yield, production volume is also a major focus; as production volume increases, the difficulty of quality control also increases.

[0003] In mass production, low yield rates may be directly related to hardware failures. Given the increasing number of testing items and the simultaneous testing of multiple test sites (Multi-site), the variables in fixing low yield problems increase significantly. In order to improve factory production efficiency, R&D engineers used to spend a lot of time training online engineers to interpret data. However, due to the testing items and the requirements of Multi-site testing, the product of the number of test sites and the number of testing items is the amount of data that engineers need to interpret. The larger the production volume, the larger the amount of data that needs to be interpreted, and the more content that needs to be tested. In the past, engineers could only rely on their experience or the knowledge passed down from senior technicians to shorten the interpretation time and speed up problem elimination, but this obviously has great instability.

[0004] Therefore, how to solve the aforementioned problems and shortcomings of conventional methods is the direction that the applicant of this invention and related manufacturers in this industry urgently want to study and improve. Summary of the Invention

[0005] Therefore, in view of the above-mentioned deficiencies, the applicant of this invention collected relevant information, conducted multiple evaluations and considerations, and, based on years of experience accumulated in this industry, and through continuous trial and modification, designed this automated low-yield processing system that can quickly identify the causes of low-yield problems and automatically provide recommended solutions with high success rates to improve repair efficiency.

[0006] The main objective of this invention is to quickly identify the different individuals from multiple yield test results of test objects by utilizing the design of the calculation module and the comparison module, thereby shortening the time required to find the root cause of the problem.

[0007] Another main objective of this invention is to automatically select recommended solutions to solve the low yield problem based on the statistical success rate of the processing solutions by utilizing the design of the processing recommendation module and the success rate statistics module.

[0008] To achieve the above objectives, the present invention provides an information link to at least one testing device, which has multiple test stations for multiple test objects to simultaneously perform at least one test item. The automated low-yield processing system mainly includes: a test data database, a processing case database, a success rate statistics module, a data acquisition module, a calculation module, a comparison module, and a processing recommendation module. The test data database stores standard test data for the test item, the processing case database stores processing solutions for the test item, and the success rate statistics module is located in the processing case database and operates according to the processing solution. The success rate is calculated based on the number of successful tests. The data acquisition module is linked to the test station to read multiple yield test results. The calculation module is linked to the data acquisition module to calculate outlier data from the yield test results using the mean and standard deviation. The comparison module is linked to the calculation module and the test data database to compare the outlier data with the standard test data to filter out individuals with discrepancies. The processing recommendation module is linked to the comparison module and the processing case database to automatically select recommended solutions for individuals with discrepancies and for the test item based on the success rate of the processing solution.

[0009] When users utilize this invention to address low yield issues on production lines, they connect the information of this invention to a testing device and use a data acquisition module to connect to its testing station to read the yield test results of multiple test items. Then, a calculation module calculates the average and standard deviation to obtain outlier data. A comparison module then compares the outlier data with standard test data in a test data database to identify the individual outliers. Finally, a processing recommendation module automatically selects high-success-rate processing solutions from a processing case database for the specific test item, providing a recommended solution for addressing the low yield problem of the test item. This allows for rapid identification of the cause of the low yield problem and automatic provision of high-success-rate recommended solutions, thereby improving repair efficiency.

[0010] The above-mentioned technologies can overcome the problems of low yield in conventional production lines, which rely solely on experience and inheritance to shorten the time for problem discovery and find the correct solutions, and still require a longer processing time when the amount of data increases, thus achieving the practical progress mentioned above. Simple Explanation of the Diagram

[0011] The first figure is a block diagram of the structure of the first preferred embodiment of the present invention. The second figure is a flowchart of the action block of the first preferred embodiment of the present invention. The third figure is a calculation comparison diagram of the first preferred embodiment of the invention. The fourth figure is a schematic diagram of the recommended scheme of the first preferred embodiment of the invention. The fifth figure is a structural block diagram of the second preferred embodiment of the present invention. The sixth figure is a schematic diagram of the second preferred embodiment of the present invention. Figure 7 is a block diagram of the structure of the third preferred embodiment of the present invention. Figure 8 is a block diagram of the structure of the fourth preferred embodiment of the present invention. Figure 9 is a schematic diagram illustrating the implementation of the fourth preferred embodiment of the present invention. Implementation

[0012] To achieve the above objectives and effects, the technical means and structure adopted in this invention are described in detail below with reference to the preferred embodiments of the invention, so as to facilitate a complete understanding.

[0013] Please refer to Figures 1 and 2, which are structural block diagrams and operational flowcharts of a first preferred embodiment of the present invention. As can be clearly seen from the figures, the present invention provides information connection to at least one test device 9. This test device 9 has multiple test stations 91, allowing multiple test objects P to simultaneously perform at least one test item 911. The automated low-yield processing system 100 mainly includes:

[0014] A test data database 1 is used to store the standard test data 11 for this test item 911;

[0015] Case database 2 is used to store the solutions 21 for this test item 911;

[0016] A success rate statistics module 22 is set in the case database 2 and calculates the success rate based on the number of successful cases of the processing solution 21.

[0017] A data acquisition module 3 is connected to the test station 91 to read multiple yield test results 31;

[0018] A calculation module 4 is linked to the data acquisition module 3 and uses the mean and standard deviation to calculate outlier data 41 from the yield test results 31.

[0019] A comparison module 5 is an information link between the calculation module 4 and the test data database 1, used to compare the outlier data 41 with the standard test data 11, thereby identifying the dissimilar individuals 51; and

[0020] A processing recommendation module 6 is an information link between the comparison module 5 and the processing case database 2, which automatically selects a recommended solution 61 for the individual 51 with the difference and the test item 911 based on the success rate of the processing solution 21.

[0021] The testing equipment 9 is a testing machine connected to the production equipment. Depending on the type of testing equipment 9, it can perform DC IV testing (voltage and current testing), digital logic testing, gain testing, and output power testing, each with its own test conditions that meet the good product standard. For example, the test conditions for DC IV testing are 5mV-10mV, the test conditions for digital logic testing are Pass / Fail, the test conditions for gain testing are 25dB-30dB, and the test conditions for output power testing are 15.6dBm-16.0dBm. In other words, regardless of the test performed by testing equipment 9, as long as the test data falls within the range of the test conditions or matches the expected feedback value, it is judged as a good product; otherwise, it is judged as a defective product. The testing machine is exemplified by a wafer testing machine or a packaging testing machine. The wafer testing machine is exemplified by the Cohu Diamond-X, and its test station 91 is exemplified by the TSK UF-3000 carrier. The packaging testing machine can be exemplified by NI. Taking STS as an example, its test station 91 is based on the HT-7045 classification tester; the automated low-yield processing system 100 is a software program that can be installed in the test equipment 9 or a mobile electronic device. This embodiment uses installation in the test equipment 9 as an example. Therefore, the success rate statistics module 22, a data acquisition module 3, a calculation module 4, a comparison module 5, and a processing recommendation module 6 are all part of the software program. The test data database 1 and the processing case database 2 are either computer storage media such as a hard drive or network storage space such as Google Drive. This embodiment uses a hard drive as an example. However, the corresponding forms of the above components are only examples of preferred embodiments. Any form with the same function is within the scope of this invention and is not limited to the above examples.

[0022] The above explanation has provided an understanding of the structure of this technology. Based on the corresponding combination of this structure, the cause of low yield problems can be quickly identified, and a recommended solution with a high success rate can be automatically provided, thus improving the efficiency of repair.

[0023] Please refer to Figures 1 to 4 for structural block diagrams and schematic diagrams of the preferred embodiment of this invention. As can be clearly seen from the figures, when a user uses this invention to handle low yield issues on the production line, the automated low yield handling system 100 installed in the testing equipment 9 is activated, and the data acquisition module 3 is connected to its testing station 91 to read the yield test results 31 of each test object P at each testing station 91. This embodiment uses multiple test objects P to perform the same test item 911 as an example.

[0024] For example, in gain testing, the test conditions are within the range of 25dB-30dB. If the detected data is 28dB within the range, it is considered "Pass"; if the detected data is 20dB outside the range, it is considered "Fail". Similarly, in digital logic testing, the test conditions are Pass / Fail. If the feedback data matches the expectation, it is considered a good product; otherwise, it is considered a defective product. Therefore, quantity can be used as the yield calculation standard. For example, if test station A 91 performs 10 tests, with 9 passing and 1 failing, the yield test result 31 is 90%. Thus, the yield test result 31 can come from the calculation function built into the test equipment 9, or from the calculation function performed when the data acquisition module 3 reads the output data of test station 91. The meaning of the yield test result 31 can be the degree of compliance of a single test item P, or the proportion of compliance of multiple test items P or a single test item P after multiple tests.

[0025] Next, the calculation module 4 calculates the mean and standard deviation of the yield test results 31, obtaining an outlier data 41. Then, the comparison module 5 compares the outlier data 41 with the standard test data 11 in the test data database 1, thus filtering out the differential individuals 51 in the test item P. For example, when the calculated mean deviates from the standard test data 11, an anomaly can be determined. Based on the magnitude of the standard deviation, it can be seen whether the overall data deviates slightly or a few data deviates significantly. In statistics, the mean is the equilibrium point of a set of data, calculated by adding up a set of numbers or quantities and then dividing by the number of numbers in the set. The standard deviation reflects the dispersion among individuals within a group, calculated by squarening the differences between each value and the mean, summing them, and then dividing by the number of values ​​to obtain the variance. Taking the square root of the variance gives the standard deviation. Outliers are data points that differ significantly from other observations.

[0026] For example, when the calculated average deviates from the standard test data 11, an anomaly can be identified. The magnitude of the standard deviation indicates whether the overall data deviates slightly or a few data points deviate significantly from the average. Referring to Figures 2 and 3, which show the yield test results 31 for 10 test items P, assuming the standard test data 11 is 100 (i.e., 100% compliance), the average calculated by module 4 is 96.3, and the standard deviation is 7.18. This indicates that the overall data deviation relative to the standard test data 11 is approximately 100 - 96.3 = 3.7 (%). The magnitude of this deviation depends on the user's definition or expectations. However, since the standard deviation is 7.18, the calculation module 4 will calculate the difference between the yield test result 31 of each test item P and the average value. If most of these differences are close to or less than the standard deviation, the calculation module 4 will regard the yield test results 31 with these deviations as outliers 41. The comparison module 5 will compare each outlier 41 with the standard test data 11. When the difference between the standard test data 11 and the yield test result 31 is greater than twice the standard deviation, it indicates that the deviation is too large, and it will be screened out and regarded as a different individual 51. However, under the condition of this small deviation, the comparison module 5 cannot screen out the different individual 51, and the system will determine that this deviation is a normal phenomenon. However, if only a few of these differences are close to or greater than the standard deviation, it indicates that only a few test items P have significant deviations. The system will determine this as an anomaly and extract the data of the test items P with such significant deviations, defining them as outlier data 41. This embodiment uses this situation as an example and uses the above comparison method to determine the degree of deviation of each outlier data 4. The difference (100-75=25) of the fourth test item P is nearly 3.5 times greater than the standard deviation, so the fourth test item P is defined as a differential individual 51. However, the above data is only for illustrative purposes to facilitate understanding of the technology and is not based on actual test data.

[0027] However, minor deviations in overall data are generally not caused by equipment malfunctions and may require reassessment of the production process or resetting of inspection standards. In this case, the processing recommendation module 6 will not select a processing solution 21 from the processing case database 2, but will directly suggest a recommended solution 61 for reassessing the production process or resetting inspection standards. If a few data points deviate significantly, it is generally caused by equipment malfunctions. Therefore, the processing recommendation module 6 will automatically select a high-success-rate processing solution 21 from the processing case database 2 for this test item 911 as the recommended solution 61 for handling the low yield problem of the test item P. The high-success-rate processing solution 21 is obtained by statistically analyzing the relationship between the number of executions and the number of successes of each processing solution 21 in the processing case database 2 for this equipment malfunction situation, and then providing the processing recommendation module 6 with the selection and recommendation data. For example, as shown in Figures 2 and 4, processing solution 21 includes air gun cleaning of the connector, connector judgment, cleaning of the board under test, checking and judging the XY position, checking and judging the height, and executing E-cal. The number of executions and successes for each step are recorded. The success rate of each processing solution 21 is calculated by the success rate statistics element 22 as (number of successes / number of executions) * 100%. Executing E-cal has the highest success rate, so the processing recommendation module 6 automatically selects processing solution 21 as the recommended solution 61. This allows for quick identification of the source of low yield problems, rapid determination of the cause of low yield issues, and automatic provision of a high-success-rate recommended solution 61, thereby reducing the time engineers spend judging data and repeating tests, and significantly improving repair efficiency.

[0028] Please refer to Figures 5 and 6 for a structural block diagram and implementation schematic diagram of the second preferred embodiment of the present invention. As can be clearly seen from the figures, this embodiment is largely the same as the above embodiment, except that the calculation module 4 has a setting unit 42 and an alarm unit 43. The setting unit 42 is used to set the ideal value and allowable error value of the calculation method. The alarm unit 43 provides an alert when the deviation between the outlier data and the ideal value exceeds the allowable error value. In addition, the automated low yield processing system 100 has a monitoring module 7 that is connected to the calculation module 4 for real-time monitoring of the relationship between the outlier data and the ideal value. The monitoring module 7 has an early warning unit 71 that issues an alert when the deviation increases. The automated low yield processing system 100 in this embodiment is exemplified by being installed in a mobile electronic device, and the setting element 42, the alarm element 43, and the monitoring module 7 are also exemplified by the circuit structure of ICs and electronic components on the control board.

[0029] Compared to the first preferred embodiment, where the user typically activates the automated low-yield processing system 100 to quickly identify the cause of the problem when the test equipment 9 detects a decrease in yield, this embodiment, due to the design of the setting unit 42 and the monitoring module 7, allows the automated low-yield processing system 100 to be permanently executed in the test equipment 9. This enables the monitoring module 7 to periodically or irregularly drive the data acquisition module 3, and the calculation module 4 to automatically calculate outlier data to monitor the relationship between the outlier data and the ideal value in real time. Furthermore, the design of the early warning unit 71 issues an early warning when the deviation increases. In this way, since the monitoring focuses on whether the yield test result 31 deviates from the standard test data, potential problems can be detected in advance before the final product yield decreases, thus preventing low-yield problems from occurring. As for the design of the setting unit 42, it allows engineers to set a tolerance value. This means that the warning unit 71 will only issue a warning when the deviation of the yield test result 31 from the standard test data exceeds this tolerance value, preventing overreaction and the resulting workload for engineers. Simultaneously, the setting unit 42 also allows engineers to quickly identify which yield test result 31 in the outlier data deviates from the ideal value beyond the tolerance value through the alert unit 43 when reviewing outlier data. This facilitates inspection by engineers. Generally, the tolerance value can be set to 1 to 2 times the standard deviation. The warning method of the warning unit 71 is illustrated by the alarm sound of the testing equipment 9, while the alert method of the warning unit 43 is illustrated by making a prominent mark in the outlier data and displaying it on the testing equipment 9.

[0030] For example, as shown in Figure 6, the yield test results 31 of 10 test objects were obtained in three separate tests. The ideal values ​​of the mean and standard deviation are 100 and 0, respectively, and the allowable error values ​​of the mean and standard deviation are 5 and 3, respectively. The calculation module 4 calculates the mean and standard deviation each time the data acquisition module 3 obtains the yield test results 31. When the monitoring mode 7 finds that the deviation of the mean or standard deviation from the ideal value is greater than the allowable error value, the warning element 71 will issue a warning (such as the warning message WARNING on the screen), and the warning element 43 will indicate the abnormal yield test results 31.

[0031] Please also refer to Figure 7, which is a block diagram of the third preferred embodiment of the present invention. As can be clearly seen from the figure, this embodiment is largely similar to the previous embodiments, except that the data acquisition module 3 has a classification unit 32, which is used to classify similar test objects P according to the test item 911. When the testing device 9 tests multiple test objects P simultaneously, and the test items 911 are multiple, in the operation of the first preferred embodiment, it is necessary to distinguish different test items 911 to select corresponding standard test data and processing schemes from the test data database 1 and the processing case database 2 respectively, and to perform calculations and comparisons one by one. In this embodiment, by utilizing the design of the classification unit 32, even if the test objects P are not the same, if they have the same or similar test items 911, they can be regarded as similar test objects P and classified into the same category of test objects P. This not only simplifies the testing process but also automatically organizes similar test content, which can both speed up the testing process and simplify the data.

[0032] Please refer to Figures 8 and 9 for a structural block diagram and implementation schematic diagram of the fourth preferred embodiment of the present invention. As can be clearly seen from the figures, this embodiment is largely the same as the above embodiments, except that the processing recommendation module 6 has a time judgment unit 62 and a record judgment unit 63. The record judgment unit 63 adds the processing record parameter 631 of the processing scheme 21 when filtering the recommended scheme, and the time judgment unit 62 adds the processing time parameter 621 of the processing scheme 21 when filtering the recommended scheme. In this way, in addition to using the success rate statistics module 22 to count the number of executions and successes, this embodiment also adds the judgment of processing record parameter 631 and processing time parameter 621. The processing record parameter 631 refers to the fact that there are usually multiple processing solutions 21 for the same test item 911. When the problem occurs again, the record judgment unit 63 can be used to read the processing solutions 21 that have been used to exclude the solutions that have been used and have poor effects. The figure shows the current processing record. The processing time parameter 621 refers to reading the execution time of each processing solution in the past. For example, if the same problem has recently been successfully solved using a certain solution 21, the weight of solution 21 will be increased. Conversely, if the problem has been handled before but is old and has little reference value, the weight of solution 21 will be decreased. The figure shows the last processing time. In this way, the processing recommendation module 6 is prevented from blindly selecting solution 21 as the recommended solution simply because of its high success rate, and only making other choices when the success rate decreases. The intelligence of the processing recommendation module 6 is also improved by evaluating the processing record parameter 631 and the processing time parameter 621 at the same time.

[0033] For example, as shown in Figure 9, among the six processing options 21, although option 6 has the highest success rate, its last processing time was 3 months ago, making it less valuable for reference. Option 2 has the second highest success rate, but its current processing record is invalid, indicating that option 2 has already been used and the result was invalid, so this option can be excluded. As for options 1 and 5, their success rates are similar, and neither has a processing record yet. However, the last processing time is closer to that of option 1, so the processing recommendation module 6 will automatically prioritize option 1 as the recommended option 61.

[0034] However, the above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Therefore, any simple modifications and equivalent structural changes made based on the description and drawings of the present invention should also be included within the patent scope of the present invention and are hereby stated.

[0035] In conclusion, the automated low-yield processing system of this invention can reliably achieve its intended function and purpose when in use. Therefore, this invention is indeed a highly practical invention. In order to meet the application requirements for an invention patent, this application is hereby filed in accordance with the law. We hope that the Examination Board will grant this invention as soon as possible to protect the applicant's hard work. If the Examination Board has any questions, please do not hesitate to write to us for guidance. The applicant will do its best to cooperate. Thank you for your help.

[0036] 100: Automated Low-Yield Handling System 1: Test Data Library 11: Standard Test Data 2: Processing the case database 21: Solution 22: Success Rate Statistics Module 3: Data Acquisition Module 31: Yield Test Results 32: Classification Unit 4: Computing Module 41: Outlier Data 42: Setting Unit 43: Warning Unit 5: Comparison Module 51: Individual Differences 6: Processing Recommendation Modules 61: Recommended Solution 62: Time Judgment Unit 621: Processing time parameter 63: Record Judgment Unit 631: Processing Record Parameters 7: Monitoring Module 71: Early Warning Unit 9: Testing equipment 91: Test Station 911: Test Project P: Test substance

Claims

1. An automated low-yield processing system, which provides information connection to at least one testing device, the testing device having multiple testing stations for multiple test objects to simultaneously perform at least one test item, and the automated low-yield processing system mainly includes: A test data database stores standard test data for the test item; a processing case database stores processing solutions for the test item; a success rate statistics module, located in the processing case database, calculates the success rate based on the number of successful processing solutions; a data acquisition module connects to the test station to read multiple yield test results; a classification unit, located in the data acquisition module, classifies similar test items according to the test item; a calculation module, connected to the data acquisition module, calculates outlier data from the yield test results using the mean and standard deviation calculation method; a setting unit, located in the calculation module, sets the ideal value and allowable error value for the calculation method; and an alert unit, located in the calculation module, provides an alert when the deviation of the outlier data from the ideal value exceeds the allowable error value. A monitoring module is connected to the calculation module to monitor the relationship between the outlier data and the ideal value in real time; an early warning unit is located in the monitoring module and issues an early warning when the deviation of the ideal value increases; a comparison module is connected to the calculation module and the test data database to compare the outlier data with the standard test data and filter out the individuals with discrepancies; a processing recommendation module is connected to the comparison module and the processing case database to automatically filter recommended solutions for the individuals with discrepancies and the test item based on the success rate of the processing solutions; a time judgment unit is located in the processing recommendation module and adds the processing time parameter of the processing solution when filtering the recommended solution; and a record judgment unit is located in the processing recommendation module and adds the processing record parameter of the processing solution when filtering the recommended solution.

Images