Crystal grain testing method based on neighborhood information, terminal, system and storage medium

Abstract

The invention relates to the technical field of integrated circuit testing, and provides a crystal grain testing method based on neighborhood information, a terminal, a system and a storage medium, and the method comprises the steps: determining a fault crystal grain according to first test data in a test data table, and the test data table is obtained through the first test of all crystal grains on a wafer; calculating a fault crystal grain neighborhood rate of the fault crystal grain, wherein the fault crystal grain neighborhood rate is a ratio of the fault crystal grains in a crystal grain neighborhood; and according to the neighborhood rate of the fault crystal grains, judging whether the fault crystal grains are tested for the second time or not. According to the method, the fault crystal grains are subjected to classification testing, compared with a traditional retesting method, higher average prediction accuracy can be achieved, average testing items are saved, and the retesting cost can be effectively reduced on the premise that the testing quality is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of integrated circuit testing, and in particular to a neighborhood information-based die testing method, terminal, system, and storage medium. Background Art

[0002] Dies are the basic unit of a wafer. A single wafer can contain tens of thousands of dies, which are then tested and packaged to form integrated circuit chips.

[0003] Generally speaking, test data correlation can be categorized into three types: spatial correlation, correlation between test items, and temporal correlation. By exploring correlations between test data, we can shorten test time, predict chip quality, identify and diagnose outliers, uncover weaknesses in the manufacturing process, and improve design robustness.

[0004] In the process of exploring test data for spatial correlation, X. Li of Duke University proposed a virtual probe testing method. Based on compressed sensing technology and combined with discrete cosine transform (DCT), this method maps spatial statistical data into the frequency domain, solves the optimal solution for the DCT coefficients in the frequency domain, and finally uses the inverse discrete cosine transform (IDCT) to restore the original image. The wafer map predicted by the virtual probe is similar to the original wafer map in terms of overall potential. Samsung Semiconductor proposed a yield prediction model based on deep learning. This model utilizes the spatial relationship between the positions of the die on the wafer and the yield variation of the die collected from wafer testing without process parameters. By modeling these spatial features, the accuracy of yield prediction is significantly improved.

[0005] At the 2021 International Testing Conference, scholars proposed a wafer-level performance prediction method for multi-site testing. This method is based on Gaussian processes and improves prediction accuracy by extending hierarchical modeling and combining test site information provided by test engineers. In addition, wafer map fault pattern recognition provides useful clues for yield learning, learning different failure modes on wafer maps, guiding process manufacturing, and improving wafer yield. Although this type of spatial modeling method significantly reduces testing costs, it comes at the cost of increasing the risk of test escape (TE) and yield loss (YL). In addition, this type of method relies too much on the characteristics of the wafer itself. When there are no clustered faults on the wafer, the prediction method based on spatial correlation will become unreliable.

[0006] In the context of test data analysis to reduce test costs and improve test quality, some methods based on test item correlation have been proposed. For example, an adaptive method for multi-site testing uses device-to-device correlation and neighborhood statistics. Based on statistical and machine learning methods, these methods can predict the pass / fail of a set of test items for test compaction. Alternative test frameworks are used to capture a set of strongly correlated feature values ​​from simpler test configurations to predict circuit performance values. Researchers at Sorbonne University use the dynamic partial average test (DPAT) method, which compares the test value of a chip with the average value of a normal chip. If the difference exceeds 6𝜎, the chip is considered to have failed the test.

[0007] However, the DPAT test method only considers a single test item and ignores the multivariate relationship between test items. To address this issue, several other studies have proposed multivariate screening methods based on the correlation of test items. For example, a hybrid feature screening method that combines the mRMR algorithm and the genetic algorithm (GA) is used, and the BP neural network is combined to predict the wafer yield. Professor Chakrabarty (IEEE Fellow) of Duke University proposed a fine-grained adaptive testing method based on random forests. This method grades the chip quality by evaluating the quality index (QI) based on machine learning, and then establishes a probability model for adaptive test selection of chips of different grades.

[0008] Furthermore, a two-tiered indirect testing approach for RF chip predictions utilizes machine learning algorithms in the first tier to predict chip performance and provide a confidence interval. Chips outside this confidence interval are then sent to the second tier for standardized testing. Adaptive testing methods leveraging temporal correlation are more focused on identifying process fluctuations and enabling quality prediction. For example, at the 2012 European Test Symposium (ETS), a proposal was made to leverage correlations between wafer batch data to mitigate the impact of process fluctuations. Intel and Texas Instruments proposed a spatiotemporal wafer-level correlation modeling approach at the 2014 ITC conference, demonstrating that three consecutively produced wafers exhibit similar spatial trends, with this trend exhibiting a linear gradient over time. Freescale Semiconductor, while studying the impact of industrial drift on test content, found that process statistics from wafer to wafer are generally correlated, with only minor variations within the same batch. Furthermore, their research demonstrated that, given a mature process, this inter-wafer process drift is predictable. In a 2018 study, researchers from Tsinghua University used partial test data to implement general device modeling and statistical modeling based on standard models. The research results showed that the IV curves of grains on different wafers in the same batch can be approximately overlapped through linear transformations such as translation and scaling, indicating that the performance of grains in the same batch is similar.

[0009] In summary, existing research methods for adaptive testing fall into three main categories: those based on spatial correlation, those based on test item correlation, and those based on temporal correlation. However, these methods generally have shortcomings. While methods based on spatial correlation significantly reduce testing costs, this comes at the cost of increased risk of test escapes and yield loss. While methods based on test item correlation can ensure a low rate of test escapes, they still have significant room for improvement in terms of test cost savings. Summary of the Invention

[0010] The purpose of the present invention is to provide a grain testing method, terminal, system and storage medium based on neighborhood information to achieve higher average prediction accuracy and average test item savings, which can effectively reduce the cost of retesting while ensuring test quality.

[0011] To achieve the above objectives, in a first aspect, the present invention provides a method for testing a die based on neighborhood information, comprising: confirming the faulty die according to the first test data in the test data table, wherein the test data table is obtained by performing the first test on all the dies on the wafer; Calculating a faulty grain neighborhood rate of the faulty grain, where the faulty grain neighborhood rate is a ratio of faulty grains within a grain neighborhood; Whether to perform a second test on the faulty die is determined according to the faulty die neighborhood rate.

[0012] Further, judging whether to perform a second test on the faulty die according to the faulty die neighborhood rate includes: If the faulty die neighbor rate falls within a first threshold range, performing a second test on the faulty die; If the faulty die neighbor rate falls within a second threshold range, not performing a second test on the faulty die; and If the faulty die neighborhood rate falls within a third threshold range, the first test data of the faulty die is input into a preset secondary test prediction model for secondary test prediction to obtain a prediction result, which includes whether to perform a second test or not.

[0013] Furthermore, the first threshold range is [0%-20%); The second threshold interval is [80%-100%]; The third threshold interval is [20%-80%].

[0014] Furthermore, before the first test data of the faulty die is input into a preset second test prediction model for second test prediction, the BIN value in the first test data is set to a predetermined value.

[0015] Furthermore, the method for establishing the secondary test prediction model includes: Construct a training set; The training set is used to train the preset machine learning model to establish a secondary test prediction model.

[0016] Furthermore, the method for constructing the training set includes: Extracting first test data and second test data of a target die, setting a BIN value in the first test data to a predetermined value, and using the first test data and second test data of the target die as an initial training set; wherein the target die is a faulty die whose faulty die neighborhood rate falls within a first threshold range; The initial training set is filled with the first test data of the fault-free dies in the neighborhood of the target die to construct the training set.

[0017] In a second aspect, the present invention provides a die test terminal based on neighborhood information, comprising: a faulty die confirmation module, configured to confirm the faulty die according to the first test data in the test data table, wherein the test data table is obtained by performing the first test on all the dies on the wafer; a faulty grain neighborhood rate calculation module, configured to calculate the faulty grain neighborhood rate of the faulty grain, wherein the faulty grain neighborhood rate is a ratio of faulty grains within a grain neighborhood; The second test judgment module is used to judge whether to perform a second test on the faulty die according to the faulty die neighborhood rate.

[0018] Furthermore, the second test judgment module is specifically used to: If the neighboring rate of the faulty die falls within the first threshold range, performing a second test on the faulty die; If the faulty die neighbor rate falls within a second threshold range, not performing a second test on the faulty die; and If the faulty die neighborhood rate falls within a third threshold range, the first test data of the faulty die is input into a preset secondary test prediction model for secondary test prediction to obtain a prediction result, which includes whether to perform a second test or not.

[0019] In a third aspect, the present invention provides a grain testing system based on neighborhood information, comprising a processor, an input device, an output device and a memory, wherein the processor, input device, output device and memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the method described in the first aspect.

[0020] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes the method as described in the first aspect.

[0021] The grain testing method based on neighborhood information proposed in this application can achieve higher average prediction accuracy and average test item savings by classifying and testing faulty grains compared to traditional retesting methods. It can effectively reduce the cost of retesting while ensuring test quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A diagram showing the relationship between wafer test duration and yield provided by an embodiment of the present invention; Figure 2 A schematic flow chart of a method for testing a die based on neighborhood information provided by an embodiment of the present invention; Figure 3 A schematic diagram of the distribution structure of neighboring grains on a wafer provided by an embodiment of the present invention; Figure 4 A schematic diagram of the purpose of a secondary test and interference information provided by an embodiment of the present invention; Figure 5 A schematic diagram of a process for filling a training set with neighborhood grain information provided by an embodiment of the present invention; Figure 6 A schematic structural diagram of a die test terminal based on neighborhood information provided by an embodiment of the present invention; Figure 7 A schematic structural diagram of a die testing system based on neighborhood information provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The following is a more detailed description of specific embodiments of the present invention with reference to schematic diagrams. The advantages and features of the present invention will become more apparent from the following description. It should be noted that the drawings are greatly simplified and not to exact ratios, and are intended solely to facilitate and clearly illustrate the embodiments of the present invention.

[0024] Figure 1A relationship diagram between wafer test time and yield provided for an embodiment of the present invention, in which the horizontal axis represents the test time and the vertical axis represents the yield. The first data point at time mark a represents the time spent in the first test. The data point at time mark c represents the total test time assuming that all the grains that failed the first test are retested, and the maximum yield can be achieved in this way. This process usually wastes a lot of unnecessary costs. The ultimate goal of retesting is to find out which faulty grains are caused by problems in the test process. However, the vast majority of failures are caused by defects in the production process. Therefore, in order to save unnecessary testing costs, the present application proposes a grain testing method based on neighborhood information. The data point at time mark b represents the time point of the present application scheme. Although the accuracy of the prediction may result in a small loss of yield, it is effective in reducing the total test time. The total test time reduced by the present application scheme is cb.

[0025] First, as Figure 2 As shown, a flowchart of a method for testing a die based on neighborhood information provided by an embodiment of the present invention includes the following steps: Step S110: confirming the faulty die according to the first test data in the test data table, wherein the test data table is obtained by performing the first test on all the dies on the wafer.

[0026] Specifically, during the first test, all dies on the wafer are tested, generating the first test data for each die. Based on this first test data, faulty dies on the wafer can be distinguished from non-faulty dies. Dies identified as faulty during the first test require further processing.

[0027] Step S120: calculating the faulty grain neighborhood rate of the faulty grain, where the faulty grain neighborhood rate is the ratio of faulty grains within a grain neighborhood.

[0028] like Figure 3 As shown in the figure, there are n dies distributed on a wafer. The neighborhood surrounding faulty die a is composed of eight dies, namely b1-b8. From the test data table, we can count the number of faulty dies among these eight dies and calculate the faulty die ratio. For example, if two of these eight dies are faulty, the faulty die neighborhood ratio is 25%; if four of these eight dies are faulty, the faulty die neighborhood ratio is 50%.

[0029] Step S130: determining whether to perform a second test on the faulty die according to the faulty die neighborhood rate.

[0030] Specifically include: If the neighboring rate of the faulty die falls within the first threshold range, performing a second test on the faulty die; If the faulty die neighbor rate falls within a second threshold range, not performing a second test on the faulty die; and If the faulty die neighborhood rate falls within a third threshold range, the first test data of the faulty die is input into a preset secondary test prediction model for secondary test prediction to obtain a prediction result, which includes whether to perform a second test or not.

[0031] In this embodiment, the first threshold interval is [0%-20%]; the second threshold interval is (80%-100%]; and the third threshold interval is [20%-80%].

[0032] To avoid retesting every faulty die, a sample of die is tested, while the remaining die are predicted. Currently, random sampling and interval sampling are commonly used in machine learning, but neither approach is interpretable, leading to unstable prediction model performance.

[0033] To fully utilize the characteristics of each die, this application proposes a sampling method based on the faulty die neighbor rate (BNR). This method first calculates the proportion of faulty die around each die. This calculation process uses data from the first test, so it does not increase the cost of the second test.

[0034] For faulty dies with a BNR in the range of [0%-20%], the probability of them being faulty is low, and therefore these dies are likely to have failed during the testing process. Therefore, all these dies will undergo secondary testing, and the secondary test data will be used as a training set for subsequent prediction models. For faulty dies with a BNR in the range of [80%-100%], the probability of them still being faulty after secondary testing is high, and these dies will not undergo additional secondary testing and will be directly classified as faulty. Faulty dies in the range of [20%-80%] do not have fixed characteristics, so subsequent processes will use prediction methods to determine whether secondary testing is required for these dies.

[0035] The sampling method based on the faulty grain neighborhood rate makes full use of the fact that the grains in the middle of the bad neighborhood are more likely to be faulty grains. The grains in the low BNR area are retested, while the grains in the high BNR area do not need to be tested. This is the key to saving testing costs in this application scheme.

[0036] In order to solve the secondary testing problem of faulty grains with BNR values ​​between [20%-80%], the modeling and prediction methods for these grains will be further discussed later.

[0037] In this embodiment, before the first test data of the faulty die is input into a preset second test prediction model for second test prediction, the BIN value in the first test data is set to a predetermined value.

[0038] Traditionally, chip test results are categorized as pass or fail. However, in actual testing, die are classified into multiple failure types based on the values ​​of the faulty test items and the specific test items in which the failure occurred. Because some types of BINs are very rare, this can easily lead to sample imbalance, and the classifier can easily overlook BINs of smaller types. Furthermore, this solution aims to predict which of the faulty die require retesting, thereby reducing the number of secondary tests. In other words, the die to be found must have failed the first test but passed the second.

[0039] For example, in this embodiment, Figure 4 As shown, BIN 1 indicates no fault, and all values ​​other than BIN 1 are faulty. When a BIN value changes from 0 to 1, machine learning records this change as label 1. However, different BINs may change after secondary testing. For example, BIN 5 becomes BIN 7 in the figure. Although these are still faulty dies, the label is recorded as 1 due to the change in BIN value. However, the faults of these dies are not actually caused by the testing process. Therefore, to avoid this interference, the BIN values ​​of the faulty dies are all set to "0" before the first test result is used as a label. In other words, the default value is 0.

[0040] In this embodiment, the method for establishing the secondary test prediction model includes: Construct a training set; The training set is used to train the preset machine learning model to establish a secondary test prediction model.

[0041] Specifically, the method for constructing the training set includes: extracting first test data and second test data of a target die, setting a BIN value in the first test data to a predetermined value, and using the first test data and the second test data of the target die as an initial training set; wherein the target die is a faulty die whose faulty die neighborhood rate falls within a first threshold range; The initial training set is filled with the first test data of the fault-free dies in the neighborhood of the target die to construct the training set.

[0042] When training the model, if only the grains with BNR between [0%-20%) are used as the training set, two problems will arise: (1) the training set samples are too small, which may cause the trained model to underfit; (2) the training set is entirely composed of faulty grain data, and the trained model may not be able to predict the results of fault-free grains, which is exactly the opposite of the actual purpose. Considering these two points, this solution will use neighborhood information to fill the training set. The specific filling method is as follows Figure 5 As shown in Figure 2, test data for fault-free die surrounding a die with a BNR range of [0%-20%) is added to the training set. This method effectively increases the amount of sample data and the number of fault-free die in the training set. Furthermore, since fault-free die are only tested once, adding these fault-free die to the training set does not increase the cost of the second test.

[0043] In this embodiment, the preset machine learning model is a gradient boosting decision tree model.

[0044] In summary, the grain testing method based on neighborhood information proposed in this application can achieve higher average prediction accuracy and average test item savings compared to traditional retesting methods by performing classification tests on faulty grains. It can effectively reduce the cost of retesting while ensuring the test quality.

[0045] Based on the same inventive concept, in the second aspect, Figure 6 As shown, the present invention also provides a die test terminal based on neighborhood information. The terminal may include: A faulty die confirmation module 201 is configured to confirm a faulty die based on first test data in a test data table obtained by first testing all the dies on the wafer; A faulty grain neighborhood rate calculation module 202 is configured to calculate a faulty grain neighborhood rate of the faulty grain, wherein the faulty grain neighborhood rate is a ratio of faulty grains within a grain neighborhood; The second test judgment module 203 is configured to judge whether to perform a second test on the faulty die according to the faulty die neighborhood rate.

[0046] Furthermore, the second test judgment module 203 is specifically configured to: If the neighboring rate of the faulty die falls within the first threshold range, performing a second test on the faulty die; If the faulty die neighbor rate falls within a second threshold range, not performing a second test on the faulty die; and If the faulty die neighborhood rate falls within a third threshold range, the first test data of the faulty die is input into a preset secondary test prediction model for secondary test prediction to obtain a prediction result, which includes whether to perform a second test or not.

[0047] In this embodiment, the first threshold interval is [0%-20%]; the second threshold interval is (80%-100%]; and the third threshold interval is [20%-80%].

[0048] To avoid retesting every faulty die, a sample of die is tested, while the remaining die are predicted. Currently, random sampling and interval sampling are commonly used in machine learning, but neither approach is interpretable, leading to unstable prediction model performance.

[0049] To fully utilize the characteristics of each die, this application proposes a sampling method based on the faulty die neighbor rate (BNR). This method first calculates the proportion of faulty die around each die. This calculation process uses data from the first test, so it does not increase the cost of the second test.

[0050] For faulty dies with a BNR in the range of [0%-20%], the probability of them being faulty is low, and therefore these dies are likely to have failed during the testing process. Therefore, all these dies will undergo secondary testing, and the secondary test data will be used as a training set for subsequent prediction models. For faulty dies with a BNR in the range of [80%-100%], the probability of them still being faulty after secondary testing is high, and these dies will not undergo additional secondary testing and will be directly classified as faulty. Faulty dies in the range of [20%-80%] do not have fixed characteristics, so subsequent processes will use prediction methods to determine whether secondary testing is required for these dies.

[0051] The sampling method based on the faulty grain neighborhood rate makes full use of the fact that the grains in the middle of the bad neighborhood are more likely to be faulty grains. The grains in the low BNR area are retested, while the grains in the high BNR area do not need to be tested. This is the key to saving testing costs in this application scheme.

[0052] In order to solve the secondary testing problem of faulty grains with BNR values ​​between [20%-80%], the modeling and prediction methods for these grains will be further discussed later.

[0053] In this embodiment, before the first test data of the faulty die is input into a preset second test prediction model for second test prediction, the BIN value in the first test data is set to a predetermined value.

[0054] Traditionally, chip test results are categorized as pass or fail. However, in actual testing, die are classified into multiple failure types based on the values ​​of the faulty test items and the specific test items in which the failure occurred. Because some types of BINs are very rare, this can easily lead to sample imbalance, and the classifier can easily overlook BINs of smaller types. Furthermore, this solution aims to predict which of the faulty die require retesting, thereby reducing the number of secondary tests. In other words, the die to be found must have failed the first test but passed the second.

[0055] For example, in this embodiment, Figure 4 As shown, BIN 1 indicates no fault, and all values ​​other than BIN 1 are faulty. When a BIN value changes from 0 to 1, machine learning records this change as label 1. However, different BINs may change after secondary testing. For example, BIN 5 becomes BIN 7 in the figure. Although these are still faulty dies, the label is recorded as 1 due to the change in BIN value. However, the faults of these dies are not actually caused by the testing process. Therefore, to avoid this interference, the BIN values ​​of the faulty dies are all set to "0" before the first test result is used as a label. In other words, the default value is 0.

[0056] In this embodiment, the method for establishing the secondary test prediction model includes: Construct a training set; The training set is used to train the preset machine learning model to establish a secondary test prediction model.

[0057] Specifically, the method for constructing the training set includes: extracting the first test data and the second test data of the target grain, setting the BIN value in the first test data to a predetermined value, and using the first test data and the second test data of the target grain as the initial training set; and filling the initial training set with the first test data of the fault-free grains in the neighborhood of the target grain to construct the training set.

[0058] When training the model, if only the grains with BNR between [0%-20%) are used as the training set, two problems will arise: (1) the training set samples are too small, which may cause the trained model to underfit; (2) the training set is entirely composed of faulty grain data, and the trained model may not be able to predict the results of fault-free grains, which is exactly the opposite of the actual purpose. Considering these two points, this solution will use neighborhood information to fill the training set. The specific filling method is as follows Figure 5As shown in Figure 2, test data for fault-free die surrounding a die with a BNR range of [0%-20%) is added to the training set. This method effectively increases the amount of sample data and the number of fault-free die in the training set. Furthermore, since fault-free die are only tested once, adding these fault-free die to the training set does not increase the cost of the second test.

[0059] In this embodiment, the preset machine learning model is a gradient boosting decision tree model.

[0060] In a third aspect, the present invention also provides a die testing system based on neighborhood information, such as Figure 7 As shown, the system may include: one or more processors 101, one or more input devices 102, one or more output devices 103, and a memory 104. The processors 101, input devices 102, output devices 103, and memory 104 are interconnected via a bus 105. The memory 104 is used to store a computer program, which includes program instructions. The processor 101 is configured to call the program instructions to execute the method of the embodiment of the die testing method based on neighborhood information.

[0061] It should be understood that in the embodiment of the present invention, the processor 101 may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0062] The input device 102 may include a keyboard, etc., and the output device 103 may include a display (LCD, etc.), a speaker, etc.

[0063] The memory 104 may include a read-only memory and a random access memory, and provides instructions and data to the processor 101. A portion of the memory 104 may also include a non-volatile random access memory. For example, the memory 104 may also store device type information.

[0064] In a specific implementation, the processor 101, input device 102, and output device 103 described in the embodiment of the present invention can execute the implementation method described in the embodiment of the grain testing method based on neighborhood information provided by the embodiment of the present invention, which will not be repeated here.

[0065] It should be noted that, regarding the specific workflow of the die testing system based on neighborhood information, reference may be made to the aforementioned method embodiment section, which will not be described in detail here.

[0066] Furthermore, an embodiment of the present invention also provides a readable storage medium storing a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the above-mentioned grain testing method based on neighborhood information is implemented. The computer-readable storage medium may be the internal storage unit of the backend server described in the aforementioned embodiment, such as the system's hard disk or memory. The computer-readable storage medium may also be an external storage device of the system, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc., equipped on the system. Furthermore, the computer-readable storage medium may include both the system's internal storage unit and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by the system. The computer-readable storage medium may also be used to temporarily store data that has been output or is about to be output.

[0067] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the composition and steps of each example according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0068] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0069] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0070] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. The systems disclosed in the embodiments are described briefly because they correspond to the methods disclosed in the embodiments. For relevant details, refer to the method description.

[0071] It should also be noted that although the present invention has been disclosed above with reference to preferred embodiments, the above embodiments are not intended to limit the present invention. For any person skilled in the art, without departing from the scope of the technical solution of the present invention, the technical content disclosed above can be used to make many possible changes and modifications to the technical solution of the present invention, or to modify it into an equivalent embodiment with equivalent changes. Therefore, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of protection of the technical solution of the present invention.

[0072] It should also be understood that, unless otherwise specified or indicated, the terms "first", "second", "third", etc. in the specification are only used to distinguish the various components, elements, steps, etc. in the specification, and are not used to indicate the logical relationship or sequential relationship between the various components, elements, steps, etc.

[0073] It should also be understood that the terms described herein are intended to describe particular embodiments only and are not intended to limit the scope of the invention. It should be noted that the singular forms "a" and "an" as used herein and in the appended claims include plural references unless the context clearly indicates otherwise. For example, a reference to "a step" or "a device" means a reference to one or more steps or devices, and may include secondary steps as well as secondary devices. All conjunctions used should be understood in their broadest sense. Also, the word "or" should be understood to have the definition of a logical "or" rather than a logical "exclusive or" unless the context clearly indicates otherwise. Furthermore, implementation of the methods and / or apparatus in embodiments of the present invention may include performing selected tasks manually, automatically, or in combination.

Claims

1. A method for testing a die based on neighborhood information, characterized in that: include: confirming the faulty die according to the first test data in the test data table, wherein the test data table is obtained by performing the first test on all the dies on the wafer; Calculating a faulty grain neighborhood rate of the faulty grain, where the faulty grain neighborhood rate is a ratio of faulty grains within a grain neighborhood; Whether to perform a second test on the faulty die is determined according to the faulty die neighborhood rate.

2. The method for testing a die based on neighborhood information according to claim 1, wherein: Determining whether to perform a second test on the faulty die according to the faulty die neighborhood rate includes: If the neighboring rate of the faulty die falls within the first threshold range, performing a second test on the faulty die; If the faulty die neighbor rate falls within a second threshold range, not performing a second test on the faulty die; and If the faulty die neighborhood rate falls within a third threshold range, the first test data of the faulty die is input into a preset secondary test prediction model for secondary test prediction to obtain a prediction result, which includes whether to perform a second test or not.

3. The method for testing a die based on neighborhood information according to claim 2, wherein: The first threshold interval is [0%-20%); The second threshold interval is [80%-100%]; The third threshold interval is [20%-80%].

4. The method for testing a die based on neighborhood information according to claim 2, wherein: Before inputting the first test data of the faulty die into a preset second test prediction model for second test prediction, the BIN value in the first test data is set to a predetermined value.

5. The method for testing a die based on neighborhood information according to claim 2, wherein: The method for establishing the secondary test prediction model includes: Construct a training set; The training set is used to train the preset machine learning model to establish a secondary test prediction model.

6. The method for testing a die based on neighborhood information according to claim 5, wherein: The method of constructing the training set includes: Extracting first test data and second test data of a target die, setting a BIN value in the first test data to a predetermined value, and using the first test data and second test data of the target die as an initial training set; wherein the target die is a faulty die whose faulty die neighborhood rate falls within a first threshold range; The initial training set is filled with the first test data of the fault-free dies in the neighborhood of the target die to construct the training set.

7. A chip test terminal based on neighborhood information, characterized in that: include: a faulty die confirmation module, configured to confirm the faulty die according to the first test data in the test data table, wherein the test data table is obtained by performing the first test on all the dies on the wafer; a faulty grain neighborhood rate calculation module, configured to calculate the faulty grain neighborhood rate of the faulty grain, wherein the faulty grain neighborhood rate is a ratio of faulty grains within a grain neighborhood; The second test judgment module is used to judge whether to perform a second test on the faulty die according to the faulty die neighborhood rate.

8. The chip test terminal based on neighborhood information according to claim 7, characterized in that: The second test judgment module is specifically used for: If the neighboring rate of the faulty die falls within the first threshold range, performing a second test on the faulty die; If the faulty die neighbor rate falls within a second threshold range, not performing a second test on the faulty die; and If the faulty die neighborhood rate falls within a third threshold range, the first test data of the faulty die is input into a preset secondary test prediction model for secondary test prediction to obtain a prediction result, which includes whether to perform a second test or not.

9. A die testing system based on neighborhood information, characterized in that: The method comprises a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the method according to any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions. When the program instructions are executed by a processor, the processor is caused to perform the method according to any one of claims 1 to 6.

Images