Test method and test system for semiconductor device testing

Abstract

The invention provides a test method for testing a semiconductor device, which comprises the following steps of: executing at least one test item in a group of test items on the semiconductor device under a first environment condition; generating, by the machine learning model, an identification result based on the test results of the at least one test item, the identification result predicting at least one test result of performing one or more test items of the set of test items on the semiconductor device under a second environmental condition; and determining whether to perform one or more test items in the set of test items on the semiconductor device under a second environmental condition based on the identification result. According to the invention, based on the test result under the first environment condition, the machine learning model is utilized to predict and obtain the corresponding identification result, so that whether the test needs to be executed again under the second environment or not is determined according to the identification result.

Description

Technical Field

[0001] This invention relates to the field of semiconductor technology, and in particular to test methods and test systems for testing semiconductor devices. Background Technology

[0002] In semiconductor manufacturing, chip probing is a crucial process for ensuring the reliability and functionality of semiconductor devices. Traditionally, to verify the operational stability of semiconductor devices under different environmental conditions, comprehensive testing is required under various environmental conditions, typically including normal and extreme conditions. For example, semiconductor devices can be tested at room temperature, high temperature, and low temperature. This traditional testing method requires performing a complete set of tests under each environmental condition.

[0003] Traditional testing methods face significant challenges in terms of testing efficiency and cost-effectiveness. For example, when testing semiconductor devices at different temperatures, many tests need to be repeated, even though these tests are relatively insensitive to temperature changes. This approach leads to unnecessary testing overhead, especially when some tests show consistent results under different temperature conditions, or when a particular semiconductor device exhibits stable performance characteristics.

[0004] In view of the above, the semiconductor testing field needs a self-adjusting testing method that can optimize the testing process by identifying and eliminating redundant test operations, while maintaining the reliability and quality of semiconductor devices. Summary of the Invention

[0005] In view of this, the present invention provides a test method and test system for testing semiconductor devices, which can be adaptively configured to optimize the test process across multiple environmental conditions based on predictive analysis of test results.

[0006] Therefore, the purpose of this invention is to provide an adaptive testing method and system for optimizing the testing of semiconductor devices under various environmental conditions while maintaining quality assurance. This invention employs a machine learning model to analyze test results obtained under initial environmental conditions and generate identification results predicting the necessity of repeating tests under other environmental conditions. This intelligent prediction mechanism significantly reduces test redundancy while ensuring the reliability of semiconductor devices under various environmental conditions.

[0007] According to one embodiment, a testing method for testing a semiconductor device is provided. The testing method includes: performing at least one test item from a set of test items on the semiconductor device under a first environmental condition; generating an evaluation result based on the test results of the at least one test item using a machine learning model, the evaluation result predicting at least one test result for performing one or more test items from the set of test items on the semiconductor device under a second environmental condition; and determining, based on the evaluation result, whether to perform one or more test items from the set of test items on the semiconductor device under the second environmental condition.

[0008] Furthermore, this set of test items includes at least one classification test item and at least one parameter test item. This allows for more refined test classification, thereby more accurately determining whether certain test items need to be performed on the semiconductor device again under a second environmental condition.

[0009] Furthermore, each category test item is configured to generate a test result selected from a limited set of predefined results. Each predefined result indicates whether one or more specific parameters or functions of the at least one semiconductor device meet predetermined criteria through one or more test iterations. This determines whether the test results meet the requirements.

[0010] Furthermore, it also includes: when the test results of the classification test items show that the semiconductor device fails to meet the predetermined standard through one or more test iterations under the first environmental condition, at least one semiconductor device is discarded and the execution of that set of test items is skipped under the second environmental condition. This can screen out problematic chips and avoid repetitive and useless testing processes.

[0011] Furthermore, each parameter test item is configured to generate a test result containing a measured value that quantifies a specific electrical, physical, or operational characteristic of the at least one semiconductor device. The chip's functionality and characteristics are thus determined based on the test results of the parameter test items, which can also serve as a reference for determining whether certain test items need to be performed again on the semiconductor device under a second environmental condition.

[0012] Furthermore, the evaluation result includes an overall evaluation score, and the step of determining whether to perform one or more tests in the set of tests includes: skipping the execution of the set of tests on the at least one semiconductor device under the second environmental condition when the overall evaluation score exceeds a predetermined threshold; and executing the set of tests on the at least one semiconductor device under the second environmental condition when the overall evaluation score is lower than or equal to the predetermined threshold. This allows for a more intuitive determination of whether certain tests need to be performed on the semiconductor device again under the second environmental condition using the score. The score can be assigned and divided based on, for example, all test results of a batch of chips, according to the expected rejection rate.

[0013] Furthermore, the evaluation result includes an evaluation score for at least one specific item, and the step of determining whether to execute one or more test items from the set of test items includes: when the evaluation score for the specific item exceeds a predetermined threshold, skipping the execution of that specific test item from the set of test items on the at least one semiconductor device under a second environmental condition; and when the evaluation score for the specific item is lower than or equal to the predetermined threshold, executing the specific test item from the set of test items on the at least one semiconductor device under the second environmental condition. This allows for a more refined and accurate determination that, for example, only one or more specific test items may need to be tested, without having to test each test item, further optimizing the selection of test items.

[0014] Furthermore, the at least one semiconductor device includes multiple semiconductor devices, and the testing method further includes: performing at least one test item from the set of test items on the multiple semiconductor devices under the first environmental condition; obtaining multiple test results for the at least one test item, wherein each test result corresponds to a corresponding semiconductor device among the multiple semiconductor devices; generating multiple identification results through the machine learning model and based on the multiple test results, wherein each identification result can predict at least one test result to be generated when performing one or more test items from the set of test items on the corresponding semiconductor device under the second environmental condition; and if at least one identification result indicates that the corresponding semiconductor device needs to be tested, performing one or more test items from the set of test items on the multiple semiconductor devices under the second environmental condition; and if all identification results indicate that the corresponding semiconductor device does not need to be tested, skipping the performance of one or more test items from the set of test items on the multiple semiconductor devices under the second environmental condition. This allows for more accurate and reliable testing of multiple semiconductor devices.

[0015] According to one embodiment, a test system for testing a semiconductor device is provided. The test system includes a test device and a test control module. The test device is configured to perform at least one test item from a set of test items on the semiconductor device under a first environmental condition. The test control module is configured to: generate an evaluation result using a machine learning model and based on the test results of at least one test item, the evaluation result predicting the test result for performing at least one test item from the set of test items on the semiconductor device under a second environmental condition; and, based on the evaluation result, determine whether to control the test device to perform one or more test items from the set of test items on the semiconductor device under the second environmental condition.

[0016] Furthermore, this set of test items includes at least one classification test item and at least one parameter test item.

[0017] Furthermore, each category test item is configured to generate a test result selected from a limited set of predefined results, each predefined result indicating whether one or more specific parameters or functions of at least one semiconductor device meet predetermined criteria through one or more test iterations.

[0018] Furthermore, when the test results of the classification test items indicate that the at least one semiconductor device fails to meet the predetermined standard through one or more iterations of testing under the first environmental condition, the test control module is configured to discard the at least one semiconductor device and control the test equipment to skip performing the set of test items on the at least one semiconductor device under the second environmental condition.

[0019] Furthermore, each parameter test item is configured to generate a test result containing a measurement that quantifies a particular electrical, physical, or operational characteristic of the at least one semiconductor device.

[0020] Furthermore, the identification result includes an overall identification score, and the test control module is configured to control the test equipment to perform the following operations: when the overall identification score exceeds a predetermined threshold, skip performing the set of test items on the at least one semiconductor device under the second environmental condition; and when the overall identification score is lower than or equal to the predetermined threshold, perform the set of test items on the at least one semiconductor device under the second environmental condition.

[0021] Furthermore, the identification results include at least one specific item identification score, and the test control module is configured to control the test equipment to perform the following operations: when the specific item identification score corresponding to the specific test item exceeds a predetermined threshold, skip performing the specific test item in the group of tests on at least one semiconductor device under a second environmental condition; and when the specific item identification score corresponding to the specific test item is lower than or equal to the predetermined threshold, perform the specific test item in the group of tests on at least one semiconductor device under a second environmental condition.

[0022] Furthermore, the at least one semiconductor device includes a plurality of semiconductor devices, and the test equipment is further configured to perform at least one test item from the set of test items on the plurality of semiconductor devices under a first environmental condition, and accordingly obtain a plurality of test results for the at least one test item, wherein each test result corresponds to a corresponding semiconductor device among the plurality of semiconductor devices; the test control module is further configured to generate a plurality of identification results through the machine learning model and based on the plurality of test results, wherein each identification result predicts at least one test result when performing one or more test items from the set of test items on the corresponding semiconductor device under a second environmental condition; and the test control module is further configured to control the test equipment: if at least one identification result indicates that its corresponding semiconductor device needs to be tested, then perform one or more test items from the set of test items on the plurality of semiconductor devices under the second environmental condition; if all identification results indicate that their respective semiconductor devices do not need to be tested, then skip performing one or more test items from the set of test items on the plurality of semiconductor devices under the second environmental condition.

[0023] The testing method for semiconductor devices in this invention includes: performing at least one test item from a set of test items on the semiconductor device under a first environmental condition; generating an evaluation result based on the test results of the at least one test item using a machine learning model, the evaluation result predicting the test result of performing at least one test item from the set of test items on the semiconductor device under a second environmental condition; and determining, based on the evaluation result, whether to perform the set of test items on the semiconductor device under the second environmental condition. Therefore, this invention can predict the corresponding evaluation result based on the test results under the first environmental condition using a machine learning model, and then determine whether to perform the test again under the second environment based on the evaluation result. When the corresponding evaluation result meets expectations, it is unnecessary to perform the test again under the second environment, thereby greatly saving testing time and steps, avoiding redundant testing operations, improving testing efficiency, and saving testing costs, thus having higher industrial practicality and wider applicability. Attached Figure Description

[0024] Figure 1 A schematic diagram of a test system for testing semiconductor devices according to an embodiment of the present invention is shown.

[0025] Figure 2 A flowchart of a test method according to an embodiment of the present invention is shown.

[0026] Figure 3 An implementation of a test control module according to an embodiment of the present invention is shown.

[0027] Figure 4 A test method implemented by a test control module for controlling test equipment according to an embodiment of the present invention is shown.

[0028] Figure 5 A test method implemented by a test control module for controlling a test device according to another embodiment of the present invention is shown.

[0029] Figure 6 It shows how to determine predetermined thresholds for evaluating overall identification scores and specific item identification scores.

[0030] Figure 7 A partial structure of a machine learning model implemented in one embodiment of the present invention is shown. Detailed Implementation

[0031] The following description is for illustrative purposes only and should not be construed as limiting. The scope of the invention is best determined by reference to the appended claims. In embodiments of the invention, when a component or layer is referred to as being “located,” “connected to,” or “coupled to” another component or layer, it may be directly located, connected to, or coupled to that other component or layer, or there may be intermediate components or layers. Conversely, when a component is referred to as being “directly located,” “directly connected to,” or “directly coupled to” another component or layer, there are no intermediate components or layers. The same numbers always refer to the same component. The term “and / or” as used herein includes any and all combinations of one or more of the associated listed items. “Directly above” or “directly below” may indicate that the projections of two or more of them at least partially overlap, while “not directly above” or “directly below” may indicate that the projections of two or more of them do not overlap at all.

[0032] In this specification, the reference to "an embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with that embodiment or example is included in at least one embodiment of the invention. Therefore, the phrases "in one embodiment" or "in an embodiment" appearing throughout this specification do not necessarily refer to the same embodiment. Furthermore, particular features, structures, or characteristics may be combined in any suitable combination and / or sub-combination in one or more embodiments.

[0033] Please see Figure 1 This diagram illustrates a test system for testing a semiconductor device according to an embodiment of the present invention. As shown, the test system 100 includes a test device 110 and a test control module 120. The test device 110 is used to perform test items on the semiconductor device. As used herein, the term "test item" refers to a specific test procedure designed to evaluate a particular characteristic, function, or parameter of the semiconductor device under test. As used herein, the term "semiconductor device" refers to a semiconductor chip configured to be tested at various manufacturing stages, including: 1) a formation stage, in which semiconductor dies exist on a wafer prior to separation; 2) a post-separation stage, in which the semiconductor dies are separated from the wafer and remain in an unpackaged state; and 3) a packaging stage, in which the semiconductor dies are integrated into a packaged semiconductor device.

[0034] In some embodiments, test items may comprise two categories: 1) categorical test items and 2) parametric test items. Specifically, each categorical test item may be configured to generate a test result selected from a finite set of predefined outcomes, where each predefined outcome indicates whether a specific parameter or function of the semiconductor device meets a predetermined criterion through one or more test iterations. For example, such categorical test items may include logic test items, such as automatic test pattern generation (ATPG) for detecting manufacturing defects, or memory test items, such as memory built-in self-test (MBIST), both of which typically produce pass or fail results. Predefined outcomes may also specify the result obtained through one or more iterations of testing, such as meeting or failing a predetermined criterion in a single iteration (e.g., pass or fail), meeting a predetermined criterion in all iterations (e.g., all pass), meeting a predetermined criterion in some iterations but not in others (e.g., fail first, then pass), or not meeting a predetermined criterion in any iteration (e.g., all fail). As used herein, "iteration" testing refers to a single execution of a specific test item under identical conditions. Therefore, performing multiple iterations can verify the stability of the results or identify specific result patterns. Furthermore, each parameter test item can be configured to generate a test result containing a measurement that quantifies a specific electrical, physical, or operational characteristic of the semiconductor device. In some embodiments, examples of parameter test items include leakage current measurements in the transistor's off-state state or the determination of the transistor's threshold voltage (Vth), both of which provide continuous values ​​crucial for evaluating device performance and reliability.

[0035] To optimize the efficiency of the testing process, the test control module 120 is configured to control the test equipment 110 based on the test method provided by this invention. Please refer to... Figure 2 The diagram illustrates a flowchart of a test method according to one embodiment. In step S101, the test equipment 100 is configured to perform at least one test item from a set of test items on a semiconductor device (at least one semiconductor device) under a first environmental condition, wherein the set of test items may include at least one of classification test items and parameter test items.

[0036] In step S102, the test control module 120 is configured to generate a qualification result using a machine learning model and based on the test results of at least one test item. This qualification result predicts at least one test result to be generated when one or more test items from a set of test items (or a group of test items) are performed on the semiconductor device (at least one semiconductor device) under a second environmental condition. As used herein, the “qualification result” is a predictive output generated by the machine learning model, rather than a direct measurement from the semiconductor device (at least one semiconductor device). It (the qualification result) originates from the initial test results of classification tests and parametric tests, which serve as input features to the machine learning model. The qualification result can be represented as one or more numerical scores or data pointers. For example, as detailed in later embodiments, it can be a single “overall qualification score” evaluating the overall reliability of the semiconductor device (at least one semiconductor device), or a set of “item-specific qualification scores” corresponding to specific future test items. Ultimately, this result will serve as the basis for determining whether further testing under the second environmental condition is necessary.

[0037] According to various embodiments of the present invention, environmental conditions include one or more environmental parameters selected from the following: temperature, voltage, humidity, pressure, or other environmental parameters that may affect the performance or reliability of the semiconductor device (at least one semiconductor device). Furthermore, the first environmental condition and the second environmental condition differ in at least one aspect of these environmental parameters. For example, the first environmental condition may be room temperature (i.e., normal operating environment), while the second environmental condition may be a high temperature of 80 degrees Celsius or a low temperature of -40 degrees Celsius (i.e., extreme operating environment).

[0038] In step S103, the test control module 120 is configured to determine, based on the evaluation results, whether to control the test equipment 110 to perform one or more test items on the semiconductor device (at least one semiconductor device) under the second environmental conditions. Please refer to... Figure 3-5 To gain a further understanding of the test control module 120.

[0039] Figure 3An implementation of a test control module 120 according to an embodiment of the present invention is illustrated. As shown, the test control module 120 typically includes at least one processor 121, one or more memory devices 122, a graphics processing unit (GPU) 123, and a storage device 124. The processor 121 may be a high-performance multi-core central processing unit (CPU) capable of handling the complex computations required for model training and inference. Alternatively, for specialized machine learning tasks, the test control module 120 may employ application-specific integrated circuits (ASICs) or field-programmable gate arrays (FPGAs) designed specifically for model computation. The memory device 122 stores model parameters and intermediate computation results, and may implement a hierarchical memory architecture, including: static random access memory (SRAM) for supporting ultra-high-speed data transfer; high bandwidth memory (HBM) for supporting large-scale high-speed data transfer; and dynamic random access memory (DRAM) for supporting large-scale data transfer. In some embodiments, test control module 120 may utilize GPU 123, which excels at handling parallel processing tasks common in machine learning workloads. Storage device 124 may include a non-volatile memory express (NVMe) solid-state drive for fast storage and retrieval of model data. By integrating these hardware components, test control module 120 is configured to perform machine learning operations and control test equipment 110 to perform adaptive testing on the semiconductor device based on the generated evaluation results.

[0040] In one embodiment of the present invention, the test control module 120 is configured to be based on Figure 4 The test method shown controls the test equipment 110. For example... Figure 4 As shown, the testing method includes two consecutive phases: testing under a first environmental condition CP1 and testing under a second environmental condition CP2. Furthermore, under the first environmental condition CP1, the test equipment 110 performs at least classification test items J1 and J2 and at least parameter test items P1, P2, and P3 on the semiconductor devices D1-D9. For example, the plurality of semiconductor devices includes semiconductor devices D1 to D9, and at least one semiconductor device refers to one or more of semiconductor devices D1 to D9.

[0041] During testing under the first environmental condition CP1, each semiconductor device D1-D9 will receive test results, including the results of classification test items J1-J2 (e.g., "V" indicates pass, "X" indicates failure) and the measured values ​​of parameter test items P1-P3. Please note that... Figure 4 The values ​​for parameter test items P1-P3 displayed may not be the original measurements. Instead, these values ​​may represent quantifications derived from the original measurements of parameter test items P1-P3, or scores calculated based on the original measurements of parameter test items P1-P3. Parameter test items P1-P3.

[0042] Based on the test results generated under the first environmental condition CP1 (i.e., the results of each category test item and each parameter test), the test control module 120 inputs the generated test results into a machine learning model to generate an identification result containing an overall identification score. In this embodiment, the overall identification score of the semiconductor device evaluates the overall reliability of the semiconductor device under the second environmental condition CP2. The overall identification score serves as a reliability indicator to determine whether the semiconductor device needs to undergo comprehensive testing under the second environmental condition CP2.

[0043] For example, a machine learning model can use the test results of an initial set (or group) of test items (including the results of classification test items (e.g., J1 and J2) and the measurements of parametric test items (e.g., P1, P2, and P3)) as a multidimensional input vector to derive an aggregated overall identification score. When the overall identification score exceeds a predetermined threshold (e.g., 60), it indicates a high level of confidence in the performance stability of the semiconductor device under different environmental conditions, meaning that testing under the second environmental condition CP2 can be safely skipped. On the other hand, when the overall identification score is below the predetermined threshold of 60, it indicates a low level of confidence, meaning that the semiconductor device should be fully tested under the second environmental condition CP2.

[0044] Based on the judgment generated by the overall identification score and a predetermined threshold, the test control module 120 is configured to control the test equipment 110 to repeatedly execute all test items in the test item set or group of test items (e.g., J1-J2 and P1-P3) under the second environmental condition CP2, or to skip all test items in the test item set (or group) executed on the corresponding semiconductor device.

[0045] Furthermore, for semiconductor device D7, since the test result of its classification test item J2 indicates failure under the first environmental condition CP1, the test control module 120 is configured to discard semiconductor device D7 and accordingly skip generating any qualification results for semiconductor device D7 so that subsequent testing can be performed under the second environmental condition CP2.

[0046] Figure 4 The self-adjusting testing method shown significantly improves efficiency by eliminating unnecessary testing while maintaining quality assurance standards.

[0047] In another embodiment of the invention, a more refined test optimization method is provided by generating an overall assessment score and a specific item assessment score in the assessment results. This testing method allows for the selective execution of individual test items under the second environmental condition CP2. Please refer to... Figure 5 To gain a deeper understanding.

[0048] and Figure 4 Similar to the embodiments described above, the testing method includes testing under a first environmental condition CP1 and testing under a second environmental condition CP2. Furthermore, under the first environmental condition CP1, the test equipment 110 performs at least classification test items J1 and J2, and at least parameter test items P1, P2, and P3 on the semiconductor devices D1-D9 (a plurality of semiconductor devices). For example, the plurality of semiconductor devices includes semiconductor devices D1 to D9, and at least one semiconductor device refers to (or includes) one or more semiconductor devices D1 to D9.

[0049] Based on the test results of the test items performed under the first environmental condition CP1, the machine learning model is configured to generate specific item identification scores IS_1-IS_3 for each semiconductor device D1-D9, corresponding to the necessity of parameter test items P1-P3 under the second environmental condition CP2, and an overall identification score corresponding to the necessity of classification test items J1-J2 under the second environmental condition CP2.

[0050] First, based on the overall qualification scores of semiconductor devices D1-D9 and a predetermined threshold 60, the test control module 120 is configured to determine whether classification test items J1 and J2 need to be repeatedly performed on each semiconductor device D1-D9 under the second environmental condition CP2. For example, since the overall qualification scores of semiconductor devices D1, D2, D3, D4, D5, and D9 are higher than 60, the test control module 120 is configured to control the test equipment 110 to skip performing classification test items J1 and J2 on semiconductor devices D1, D2, D3, D4, D5, and D9 under the second environmental condition CP2. On the other hand, since the overall qualification scores of semiconductor devices D6 and D8 are lower than 60, the test control module 120 is configured to control the test equipment 110 to repeat classification test items J1 and J2 on semiconductor devices D6 and D8 under the second environmental condition CP2.

[0051] Secondly, based on the specific item identification score corresponding to each parameter test P1-P3 and a predetermined threshold 10 (for illustrative purposes only), the test control module 120 is configured to determine whether it is necessary to repeat each parameter test P1-P3 for each semiconductor device D1-D9 under the second environmental condition CP2. For example, when the specific item identification score IS_1 of semiconductor device D6 exceeds the predetermined threshold 10 (where a higher score indicates lower confidence in the performance stability of the semiconductor device, and a lower score indicates higher confidence in the performance of the semiconductor device), the test control module 120 controls the test equipment 110 to repeat parameter test item P1 for semiconductor device D6 under the second environmental condition CP2. For example, when the specific item identification score IS_2 of semiconductor device D5 exceeds the predetermined threshold 10, the test control module 120 controls the test equipment 110 to repeat parameter test item P2 for semiconductor device D5 under the second environmental condition CP2. For example, when the specific item identification score IS_3 of semiconductor devices D1, D2 and D8 exceeds a predetermined threshold 10, the test control module 120 is configured to control the test equipment 110 to repeat the parameter test item P3 of semiconductor devices D1, D2 and D8 under the second environmental condition CP2.

[0052] Furthermore, this testing method maintains stringent quality control by immediately halting the testing process for semiconductor device D7 that failed classification test item J2 under the first environmental condition CP1. Semiconductor device D7 is marked "N / A" in all subsequent predictions and tests to ensure proper identification and isolation of defective devices at an early stage of the testing process.

[0053] While the above embodiments describe predicting the necessity of testing under a second environmental condition CP2 based solely on test results obtained under a first environmental condition CP1, the invention is not limited to such embodiments. In alternative embodiments, the necessity of testing under a third environmental condition CP3 can be predicted based on test results obtained under the first environmental condition CP1 and the second environmental condition CP2. Furthermore, the necessity of testing under both the second environmental condition CP2 and the third environmental condition CP3 can be predicted based on test results obtained under the first environmental condition CP1.

[0054] Furthermore, although in the above embodiments, the identification results are generated based on specific classification test items J1-J2 and parameter test items P1-P3 performed under the first environmental condition CP1, the number and type of test items do not limit the scope of the invention. In various embodiments, the identification results may be generated based on different numbers of classification test items and / or parameter test items performed under the first environmental condition CP1. Moreover, the identification results may be determined based on the test results of individual classification test items or individual parameter test items, wherein one or more test items are of the selected type.

[0055] In summary, Figure 5 The testing methodology shown achieves a balance between comprehensiveness and efficiency by assessing the necessity of categorical testing based on the overall identification score and independently assessing the necessity of testing each parameter based on the identification score of specific items.

[0056] As illustrated in the above embodiments, while the overall identification score and the identification score of specific items provide predictive indicators, the selection of a predetermined threshold plays a crucial role in determining whether the item needs to be retested under the second environmental condition CP2. The choice of threshold directly affects testing efficiency and quality assurance.

[0057] Figure 6 This paper illustrates how predetermined thresholds are determined for evaluating overall qualification scores and specific item qualification scores. Specifically, in this invention, the predetermined thresholds are determined through statistical analysis of the qualification scores of semiconductor devices that have passed basic testing. The qualification score distributions of these semiconductor devices typically exhibit the following characteristics: as shown in the figure, the threshold determination is based on a typical pattern. Furthermore, threshold determination needs to achieve an optimal balance between testing efficiency and quality assurance, measured in parts per million (DPPM). Setting a stricter threshold requires repeating more tests under different environmental conditions, which reduces DPPM but decreases testing efficiency. Conversely, setting a more lenient threshold increases testing efficiency but leads to an increase in DPPM. Therefore, the predetermined thresholds need to be dynamically adjusted based on the actual score distribution and target DPPM requirements to maintain the optimal balance between efficiency and quality during testing.

[0058] As described above, the test control module 120 uses a machine learning model to generate identification results to assess the necessity of repeating test items under different environmental conditions. Figure 7 The diagram illustrates a partial structure of a machine learning model implemented in one embodiment of the present invention. It should be noted that the model structure shown represents only one possible implementation and does not limit the scope of the invention. Furthermore, the structure shown depicts only a portion of the complete model architecture.

[0059] As shown in the figure, the machine learning model used by the test control module 120 includes a decision tree structure for generating scores (which can be the overall identification score as described above, or a project-specific identification score). The training process of the decision tree of the machine learning model may include: collecting an initial dataset (or a set of initial data) containing the comprehensive test results of semiconductor devices tested under different environmental conditions (e.g., first environmental condition CP1 and second environmental condition CP2); cleaning and filtering the initial dataset to remove outlier data points, thereby obtaining a refined training dataset (or a set of initial data); and finally, building a machine learning model based on the refined training dataset (or set of initial data).

[0060] In practical applications, the test control module 120 inputs the test results obtained under the first environmental condition CP1 into the machine learning model to generate predictions about the test results under the second environmental condition CP2. For example, when the test results obtained by the semiconductor device D1 under the first environmental condition CP1 are: the result of test item IT1 (which can be a classification test item) is 0, the result of test item IT2 (which can be a parametric test item) is 0.7, the result of test item IT4 (which can be a parametric test item) is 50, and the result of test item IT8 (which can be a classification test item or a parametric test item) is 2, the machine learning model shown in the figure is configured to generate an inference score of 8 to evaluate the necessity of further testing of the semiconductor device D1 under the second environmental condition CP2.

[0061] In this embodiment of the invention, to improve testing efficiency, the test equipment 110 typically employs parallel testing, simultaneously performing one or more test items on multiple semiconductor devices during each grounding operation. Therefore, the characteristics of this group-based testing method must be considered when evaluating the necessity of repeatedly performing test items under different environmental conditions.

[0062] This invention implements the following group-based testing strategy: When any one semiconductor device in a group (or multiple semiconductor devices) needs to repeatedly perform a specific test item (test item) under a second environmental condition, all semiconductor devices in that group (or multiple semiconductor devices) must undergo the same test (test item). Conversely, testing of that group (or multiple semiconductor devices) under the second environmental condition can only be skipped if all semiconductor devices in the group (or multiple semiconductor devices) meet the criteria for skipping the test. However, if only one semiconductor device in the group (or multiple semiconductor devices) does not meet the criteria for skipping the test, all semiconductor devices in the group (or multiple semiconductor devices) must be tested under the second environmental condition. This strategy maintains an optimal balance between testing efficiency and quality control while taking into account the practical constraints of parallel testing operations.

[0063] This invention proposes an innovative semiconductor device testing optimization method that improves testing efficiency while maintaining quality assurance. The core innovation lies in the intelligent prediction mechanism, which uses a machine learning model to assess the necessity of repeating tests under various environmental conditions. This invention provides both a holistic optimization method and a granular optimization method. The holistic optimization method uses an overall qualification score to determine whether all test items need to be repeated, while the granular optimization method uses item-specific qualification scores to selectively execute individual test items. In summary, this invention effectively reduces test redundancy while ensuring quality assurance under various operating conditions. Based on the test results under a first environmental condition, this invention uses a machine learning model to predict the corresponding qualification results, thereby determining whether to repeat the test under a second environment. When the corresponding qualification results meet expectations, there is no need to repeat the test under the second environment, thus significantly saving testing time and steps, avoiding redundant testing operations, improving testing efficiency, and saving testing costs, resulting in greater industrial applicability and wider applicability.

[0064] The embodiments of this invention can be implemented as an apparatus, method, or computer program product. Therefore, this invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects, all of which are collectively referred to herein as a "module" or "system." Furthermore, this invention can take the form of a computer program product embodied in any tangible medium containing computer-usable program code. In terms of hardware, the invention can be implemented by applying any of the following techniques or related combinations: independently operating logic having logic gates capable of performing logical functions according to data signals, and application-specific integrated circuits (ASICs), programmable gate arrays (PGAs), or field-programmable gate arrays (FPGAs) with suitable combinational logic. In some embodiments, the test control module (e.g., test control module 120) in this invention can be a calculator device, including a storage medium storing program code for the calculator device (e.g., test control module 120) to read and execute. By reading this program code from the storage medium, the calculator device (e.g., test control module 120) can execute the control method described above (e.g., refer to...). Figures 2-7 (The test method or a method related to the test method). The storage medium may be, for example, flash memory, solid-state memory, DRAM, SRAM (static random-access memory), hard disk storage, or the above-mentioned storage device 124.

[0065] The flowcharts and block diagrams in the flowcharts illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, code segment, or portion of code, containing one or more executable instructions for implementing a specified logical function. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware system that performs the specified function or action, or a combination of dedicated hardware and computer instructions. These computer program instructions may be stored in a computer-readable medium that instructs a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable medium produce an article of manufacture including instruction means that implement the function / action specified in the flowchart and / or block diagram blocks.

[0066] While the invention has been described by way of example and according to preferred embodiments, it should be understood that the invention is not limited to the disclosed embodiments. Rather, it is intended to cover various modifications and similar arrangements (as will be apparent to those skilled in the art). Therefore, the scope of the appended claims should be given the broadest interpretation to cover all such modifications and similar arrangements.

Claims

1. A test method for testing semiconductor devices, characterized in that, include: Under a first environmental condition, at least one test item from a set of test items is performed on at least one semiconductor device; Using a machine learning model, an identification result is generated based on the test results of at least one test item in the set of test items. The identification result predicts the test results of at least one test item in the set of test items when the at least one semiconductor device is subjected to a second environmental condition. as well as Based on the identification results, it is determined whether, under the second environmental conditions, one or more of the test items in the set of test items should be performed on the at least one semiconductor device.

2. The test method as described in claim 1, characterized in that, This set of test items includes at least one classification test item and at least one parameter test item.

3. The test method as described in claim 2, characterized in that, Each classification test item is configured to generate a test result selected from a limited set of predefined results, each predefined result indicating whether one or more specific parameters or functions of the at least one semiconductor device meet predetermined criteria through one or more test iterations.

4. The test method as described in claim 3, characterized in that, Also includes: If the test results of the classification test items show that the semiconductor device fails to meet the predetermined standard through one or more test iterations under the first environmental condition, the at least one semiconductor device is discarded and the test items of that group are skipped under the second environmental condition.

5. The test method as described in claim 2, characterized in that, Each parameter test item is configured to generate a test result containing a measurement that quantifies a particular electrical, physical, or operational characteristic of the at least one semiconductor device.

6. The test method as described in claim 1, characterized in that, The evaluation results include an overall evaluation score, and the steps to determine whether to perform one or more test items from this set of test items include: When the overall identification score exceeds a predetermined threshold, skip performing the set of tests on the at least one semiconductor device under the second environmental condition; and When the overall identification score is lower than or equal to the predetermined threshold, the set of tests is performed on the at least one semiconductor device under the second environmental conditions.

7. The test method as described in claim 1, characterized in that, The assessment results include an assessment score for at least one specific item, and the steps to determine whether to perform one or more test items from this set of test items include: When the specific test score for a particular test item exceeds a predetermined threshold, the execution of that particular test item in the set of tests on at least one semiconductor device under the second environmental condition is skipped; and When the identification score of a specific test item corresponding to a specific test item is lower than or equal to the predetermined threshold, under the second environmental condition, the specific test item in the set of test items is performed on the at least one semiconductor device.

8. The test method as described in claim 1, characterized in that, The at least one semiconductor device includes multiple semiconductor devices, and the test method further includes: Under these first environmental conditions, at least one of the test items in the set of test items is performed on the plurality of semiconductor devices; Acquire multiple test results for the at least one test item, wherein each test result corresponds to a corresponding semiconductor device among the multiple semiconductor devices; The machine learning model generates multiple identification results based on the multiple test results, wherein each identification result can predict at least one test result generated when one or more of the test items in the set of test items are performed on the corresponding semiconductor device under the second environmental condition; and If at least one qualification result indicates that the corresponding semiconductor device requires testing, then under the second environmental conditions, one or more tests from that set of test items are performed on the plurality of semiconductor devices; and If all qualification results indicate that the corresponding semiconductor device does not require testing, then one or more test items from that set of test items are skipped for the multiple semiconductor devices under the second environmental conditions.

9. A test system for testing semiconductor devices, characterized in that, include: The test equipment is configured to perform at least one test item from the set of test items on at least one semiconductor device under a first environmental condition; as well as The test control module is configured as follows: Based on the test results of the at least one test item, an identification result is generated using a machine learning model. This identification result predicts at least one test result of the set of one or more test items when the at least one semiconductor device is subjected to one or more test items in the set of test items under a second environmental condition. Based on the identification results, it is determined whether to control the test equipment to perform one or more of the test items in the set of test items on the at least one semiconductor device under the second environmental conditions.

10. The testing system as described in claim 9, characterized in that, The at least one semiconductor device includes a plurality of semiconductor devices, and wherein The testing equipment is also configured to perform at least one of the set of test items on the plurality of semiconductor devices under the first environmental conditions, and accordingly obtain a plurality of test results for the at least one test item, wherein each test result corresponds to a corresponding semiconductor device among the plurality of semiconductor devices; The test control module is also configured to generate multiple qualification results using the machine learning model and based on the multiple test results, wherein each qualification result predicts at least one test result when one or more test items from the set of test items are performed on the corresponding semiconductor device under the second environmental condition; and The test control module is also configured to control the test equipment: If at least one qualification result indicates that its corresponding semiconductor device requires testing, then one or more of the test items in the set of test items are performed on the plurality of semiconductor devices under the second environmental conditions; if all qualification results indicate that their respective semiconductor devices do not require testing, then one or more of the test items in the set of test items are skipped under the second environmental conditions.

Images