Chip test parameter abnormality detection method, storage medium, terminal
By calculating the difference between the edge chip ratio and the reference ratio of the chip test parameters, automatic detection and early warning of abnormal chip test parameters is realized, the problem of inability to warning in the existing technology is solved, and the chip reliability and stability of the production process are improved.
Patent Information
- Application Number
- CN202010898440.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-08-31
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2040-12-03
AI Technical Summary
The prior art cannot be early warning before abnormal chip test parameters, and manpower involves abnormal numerical screening and failure analysis, and cannot play an early warning role before Bin fails.
By obtaining the test parameters of the current batch of chips, calculate the difference between the edge chip ratio and the reference edge chip ratio. If the difference is greater than the preset prompt value, abnormal chip test parameters are detected and an early warning is issued when the change trend continues to exceed the control range. .
Automatic detection and early warning of abnormal chip test parameters is realized, reliability problems and production process deviations are identified in advance, potential problems before Bin failure are avoided, and reliability of production and testing is improved.
Smart Images

Figure CN112180230B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of chip manufacturing and testing, and in particular to a method, storage medium and terminal for detecting abnormal chip test parameters. Background Art
[0002] When testing a chip, the test machine will perform hundreds or even thousands of tests on the chip in sequence according to the test program, and the test results are called test parameters. The value of each test parameter measured by the test program (i.e., the test result) is compared with the chip test specification (specification, referred to as spec) of the corresponding test item specified in the chip specification to determine whether it is good or bad. If the value of the test parameter obtained by testing a certain test item of the chip does not meet the requirements of the chip test specification, then according to the correspondence between the test parameter and the Bin, the chip is judged as the Bin corresponding to the test parameter failed. Among them, Bin can be used to characterize the classification of a chip and to distinguish whether the chip is a functional chip. Furthermore, when the chip cannot work normally, Bin can also be used to distinguish which specific part of the chip's function cannot be implemented normally.
[0003] Regarding chip test parameters, due to the extremely large amount of test data, the compressed test data file of each wafer is as high as hundreds of megabytes. Currently, many companies are basically in the stage of engineering analysis. When a Bin failure occurs, the test parameter value of this Bin is analyzed to provide direction for failure cause analysis.
[0004] Current chip test parameters are also often used for lot or wafer disposition. Specifically, based on manual experience and the failure reasons obtained based on chip test parameter analysis, it is decided whether to release the product, whether to notify the customer for processing, and whether to mark some good chips as bad chips and not ship them.
[0005] In summary, the use of chip test parameters by existing technologies is still limited to failure analysis-related scenarios. On the one hand, human intervention is required to screen abnormal values and conduct failure analysis. On the other hand, the existing chip test parameters only take effect after a Bin failure occurs and cannot serve as a warning before a Bin failure. Summary of the invention
[0006] The technical problem solved by the present invention is how to realize automatic detection of abnormal chip test parameters.
[0007] To solve the above technical problems, an embodiment of the present invention provides a method for detecting abnormal chip test parameters, including: obtaining a first test parameter of a current batch of chips, wherein the first test parameter is a test result of the current batch of chips in a preset test item; calculating the edge chip ratio of the current batch of chips according to the first test parameter, wherein the edge chip ratio is the proportion of edge chips in the current batch of chips, and the edge chip is a chip whose value of the first test parameter falls within a qualified range but exceeds an edge range; when the difference between the edge chip ratio and a benchmark edge chip ratio is greater than a first preset prompt value, it is determined that a chip test parameter abnormality is detected, wherein the benchmark edge chip ratio corresponds to the preset test item.
[0008] Optionally, the step of calculating the benchmark edge chip ratio includes: obtaining a second test parameter of a historical batch of chips, wherein the second test parameter is a test result of the historical batch of chips in the preset test item; obtaining the number of qualified chips in the historical batch of chips and yield test buffer data according to the second test parameter, wherein the yield test buffer data is used to characterize the numerical distribution of the second test parameter; narrowing the qualified range based on the yield test buffer data to obtain the edge range; based on the edge range and the qualified range, screening a first number of chips from the historical batch of chips, wherein the first number of chips are chips whose values of the second test parameter fall within the qualified range but exceed the edge range; and determining the ratio of the first number to the number of qualified chips as the benchmark edge chip ratio.
[0009] Optionally, obtaining the yield test buffer data of the historical batch of chips according to the second test parameter includes: calculating the interquartile range of the second test parameter to obtain the yield test buffer data; or, calculating the standard deviation of the second test parameter to obtain the yield test buffer data.
[0010] Optionally, calculating the edge chip ratio of the current batch of chips based on the first test parameter includes: obtaining the number of qualified chips in the current batch of chips; screening the edge chips from the current batch of chips based on the qualified range and the edge range; and determining the ratio of the number of the screened edge chips to the number of qualified chips in the current batch of chips as the edge chip ratio of the current batch of chips.
[0011] Optionally, the detection method further includes: calculating an average edge chip ratio of all batches of chips detected within a preset detection period; and determining that a chip test parameter abnormality is detected when a difference between the average edge chip ratio and the benchmark edge chip ratio is greater than a second preset prompt value.
[0012] Optionally, both the first preset prompt value and the second preset prompt value are determined according to the return rate of the chip, and the second preset prompt value is smaller than the first preset prompt value.
[0013] Optionally, the detection method also includes: obtaining the value change trend of the third test parameter of multiple batches of chips within a preset time period over time, wherein the third test parameter is the test results of the multiple batches of chips in the preset test items; when the change trend continues to exceed the control range, it is determined that a chip test parameter abnormality is detected.
[0014] Optionally, the process of determining the control range includes the following steps: obtaining a fourth test parameter of multiple historical batches of chips, wherein the fourth test parameter is the test results of the multiple historical batches of chips under the preset test items; calculating a key statistical value of the fourth test parameter; and determining the control range based on the key statistical value and the standard deviation of the key statistical value.
[0015] Optionally, obtaining the temporal variation trend of the value of the third test parameter of multiple batches of chips within a preset time period includes: calculating the key statistical value of the third test parameter; and plotting the temporal variation trend of the key statistical value.
[0016] Optionally, the key statistical value is selected from: median, 95th quantile, 5th quantile, 99th quantile and 1st quantile.
[0017] To solve the above technical problems, an embodiment of the present invention further provides a device for detecting abnormal chip test parameters, comprising: an acquisition module, used to acquire a first test parameter of a current batch of chips, wherein the first test parameter is a test result of the current batch of chips in a preset test item; a calculation module, used to calculate the edge chip ratio of the current batch of chips according to the first test parameter, wherein the edge chip ratio is the proportion of edge chips in the current batch of chips, and the edge chip is a chip whose value of the first test parameter falls within a qualified range but exceeds an edge range; a detection module, which determines that a chip test parameter abnormality is detected when the difference between the edge chip ratio and a benchmark edge chip ratio is greater than a first preset prompt value, wherein the benchmark edge chip ratio corresponds to the preset test item.
[0018] In order to solve the above technical problem, an embodiment of the present invention further provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are executed.
[0019] To solve the above technical problem, an embodiment of the present invention further provides a terminal, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor executes the steps of the above method when running the computer program.
[0020] Compared with the prior art, the technical solution of the embodiment of the present invention has the following beneficial effects:
[0021] An embodiment of the present invention provides a method for detecting abnormal chip test parameters, comprising: obtaining a first test parameter of a current batch of chips, wherein the first test parameter is a test result of the current batch of chips in a preset test item; calculating the edge chip ratio of the current batch of chips according to the first test parameter, wherein the edge chip ratio is the proportion of edge chips in the current batch of chips, and the edge chip is a chip whose value of the first test parameter falls within a qualified range but exceeds an edge range; when the difference between the edge chip ratio and a benchmark edge chip ratio is greater than a first preset prompt value, determining that a chip test parameter abnormality is detected, wherein the benchmark edge chip ratio corresponds to the preset test item.
[0022] The use of this implementation scheme can realize automatic detection and push of chip test parameter anomalies, so as to warn chip reliability problems and possible problems in production processes and testing links in advance before Bin failure actually occurs. Specifically, by analyzing the proportion of edge chips in the current batch of chips, the distribution of test results of the current batch of chips in the preset test items can be obtained more accurately. Furthermore, if the proportion of edge chips in the current batch increases to a difference greater than the first preset prompt value from the proportion of benchmark edge chips, it indicates that the numerical distribution of the first test parameter deviates from the benchmark distribution, that is, the chip test parameter is abnormal. Therefore, this implementation scheme can realize automatic detection of test parameter anomalies. When the numerical distribution of the first test parameter deviates from the benchmark distribution, even if the numerical change of the first test parameter has not yet caused Bin failure, the use of this implementation scheme can still accurately and automatically detect this abnormal deviation and issue an early warning in time, which is conducive to solving chip reliability problems.
[0023] Furthermore, the detection method also includes: obtaining the change trend of the value of the third test parameter of multiple batches of chips over time within a preset time period, wherein the third test parameter is the test results of the multiple batches of chips in the preset test items; when the change trend continues to exceed the control range, it is determined that a chip test parameter abnormality is detected.
[0024] As a result, potential production process and test problems can be detected as early as possible, making it possible to detect process and test defects in advance before Bin failure. Specifically, when the value of the third test parameter of multiple batches of chips in the same preset test item continues to increase or decrease over a period of time, even if the test results still fall within the qualified range specified by the chip test specifications, it actually very likely means that a process parameter of the production line has deviated. If it is not resolved in advance, a large number of low yield problems may occur later. Therefore, this implementation scheme sets a control range and cooperates with the changing trend of the test results over time to achieve automatic detection. When the changing trend continues to exceed the control range, it indicates that the value of the third test parameter has approached the upper or lower limit of the qualified range, and an early warning needs to be issued. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] Figure 1 It is a numerical distribution diagram of chip test parameters;
[0026] Figure 2 It is a trend chart of the value of a chip test parameter changing over time;
[0027] Figure 3 It is a flow chart of a method for detecting abnormal chip test parameters according to an embodiment of the present invention;
[0028] Figure 4 It is the trend of the difference between the edge chip ratio and the benchmark edge chip ratio in the test item IDDQ_GPU of multiple batches of chips monitored over a period of time;
[0029] Figure 5 yes Figure 4 The numerical distribution diagram of the test parameters of the batch A chip in the test item IDDQ_GPU;
[0030] Figure 6 It is the change trend of the difference between the edge chip ratio and the benchmark edge chip ratio in the test item Vmin_CPU of multiple batches of chips monitored over a period of time;
[0031] Figure 7 yes Figure 6 The numerical distribution diagram of the test parameters of the test item Vmin_CPU for all batches of chips obtained in the 20th week of testing;
[0032] Figure 8 It is a structural schematic diagram of a chip test parameter abnormality detection device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0033] As mentioned in the background technology, the use of chip test parameters in existing technologies is still limited to failure analysis-related scenarios. On the one hand, human intervention is required to screen abnormal values and conduct failure analysis. On the other hand, existing chip test parameters only take effect after a Bin failure occurs and cannot serve as a warning before a Bin failure.
[0034] The inventor of the present application has found through analysis that for the same test item of the same product, the numerical distribution range of the test results (i.e., test parameters) of all chips for the test item has a reference distribution. This reference distribution can be at the wafer level, Fab manufacturing batch level, or packaging batch level. Even if the numerical change of the test parameter does not cause Bin failure, if the numerical distribution deviates from the reference distribution, it is hoped that the wafer or batch with abnormal test parameters will be detected with as few false alarms as possible.
[0035] According to the classification by the inventor, chip test parameter abnormalities usually include: Figure 1 and Figure 2 The four cases shown are Figure 1 Each curve represents the numerical distribution of a certain test parameter of a wafer or a batch of chips. Figure 2 The curve in the figure represents the changing trend of the value of a certain test parameter of multiple wafers or multiple batches of chips over time. Figure 1 and Figure 2 The numerical distribution and change trend of the test parameters of a specific test item are taken as an example for specific display.
[0036] Four types of chip test parameter abnormalities are listed below:
[0037] 1. The numerical distribution of the test parameters is shifted, such as Figure 1 As shown in the middle curve 11;
[0038] 2. The numerical distribution of test parameters becomes wider, such as Figure 1 As shown in the middle curve 12; 3. The numerical distribution tail (long tail) of the test results of the test parameters, such as Figure 1 As shown in the middle curve 13;
[0039] 4. The value of the test parameter becomes higher or lower continuously. Figure 2 The example of continuously increasing is used for illustrative demonstration, and the same is true for continuously decreasing.
[0040] For the above four abnormal chip test parameter situations, since the test parameter values still fall within the qualified range specified by the specification (spec), the prior art still determines the chip as a "good" chip when these test parameters are obtained. Figure 1For example, Figure 1 The qualified line 14 shown defines the qualified upper limit of the chip in this test item. The chip whose test parameters are greater than the value represented by the qualified line 14 will be judged as a "bad" chip, and the chip whose test parameters are less than the value represented by the qualified line 14 will be judged as a "good" chip. Figure 2 For example, Figure 2 The upper specification line 21 and the lower specification line 22 shown together define the qualified range. Chips whose test results are greater than the numerical value represented by the upper specification line 21 and chips whose test results are less than the numerical value represented by the lower specification line 22 will be judged as "bad" chips, and chips whose test results fall between the upper specification line 21 and the lower specification line 22 will be judged as "good" chips.
[0041] Taking curve 13 as an example, since the values of the test parameters always fall to the left of the qualified line 14, the wafer or batch of chips represented by curve 13 is a "good" chip in the usual sense. However, it can be seen from the figure that the specific values of the test parameters of this batch of chips are actually very close to the qualified line 14, especially the circled part in the figure. These chips may have reliability problems after long-term use. The existing technology, whether based on manual or computer, cannot detect and identify such situations.
[0042] by Figure 2 The value of the test parameter shown in the figure changes over time as an example. Although the value of the test parameter Figure 2 The plotted time period is always between the upper specification line 21 and the lower specification line 22, but in fact, the value of the test parameter is continuously increasing as time goes by. This means that a process parameter of the production line has deviated. If it is not resolved in advance, a large number of low yield problems may occur later. Similarly, the existing technology, whether based on manual or computer, cannot detect and identify such situations.
[0043] An embodiment of the present invention provides a method for detecting abnormal chip test parameters, comprising: obtaining a first test parameter of a current batch of chips, wherein the first test parameter is a test result of the current batch of chips in a preset test item; calculating the edge chip ratio of the current batch of chips according to the first test parameter, wherein the edge chip ratio is the proportion of edge chips in the current batch of chips, and the edge chip is a chip whose value of the first test parameter falls within a qualified range but exceeds an edge range; when the difference between the edge chip ratio and a benchmark edge chip ratio is greater than a first preset prompt value, determining that a chip test parameter abnormality is detected, wherein the benchmark edge chip ratio corresponds to the preset test item.
[0044] The use of this implementation scheme can realize automatic detection and push of chip test parameter anomalies, so as to warn chip reliability problems and possible problems in production processes and testing links in advance before Bin failure actually occurs. Specifically, by analyzing the proportion of edge chips in the current batch of chips, the distribution of test results of the current batch of chips in the preset test items can be obtained more accurately. Furthermore, if the proportion of edge chips in the current batch increases to a difference greater than the first preset prompt value from the proportion of benchmark edge chips, it indicates that the numerical distribution of the first test parameter deviates from the benchmark distribution, that is, the chip test parameter is abnormal. Therefore, this implementation scheme can realize automatic detection of test parameter anomalies. When the numerical distribution of the first test parameter deviates from the benchmark distribution, even if the numerical change of the first test parameter has not yet caused Bin failure, the use of this implementation scheme can still accurately and automatically detect this abnormal deviation and issue an early warning in time, which is conducive to solving chip reliability problems.
[0045] In order to make the above-mentioned objects, features and beneficial effects of the present invention more obvious and easy to understand, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
[0046] Figure 3 The present invention is a flowchart of a method for detecting abnormal chip test parameters according to an embodiment of the present invention.
[0047] This embodiment can be applied to chip manufacturing scenarios, and can timely and automatically detect abnormal chip test parameters, so that chip manufacturing companies can more accurately handle batches or wafers and ship early warning. The detection method can be performed by a device that performs or assists chip testing, such as an automatic test equipment (ATE) or a computing device locally or remotely coupled to the ATE.
[0048] Specifically, refer to Figure 3 The method for detecting abnormal chip test parameters in this embodiment may include the following steps:
[0049] Step S101, obtaining a first test parameter of a current batch of chips, wherein the first test parameter is a test result of the current batch of chips in a preset test item;
[0050] Step S102, calculating the marginal chip ratio of the current batch of chips according to the first test parameter, wherein the marginal chip ratio is the proportion of marginal chips in the current batch of chips, and the marginal chip is a chip whose value of the first test parameter falls within a qualified range but exceeds a marginal range;
[0051] Step S103, when the difference between the edge chip ratio and the reference edge chip ratio is greater than a first preset prompt value, it is determined that a chip test parameter abnormality is detected, wherein the reference edge chip ratio corresponds to the preset test item.
[0052] More specifically, the preset test item can be any test item that the test machine needs to perform on the chip according to the test program. This embodiment can be repeatedly executed, and the test results of different preset test items are obtained each time to determine whether there is an abnormality in the first test parameter of the preset test item of the current batch of chips.
[0053] Furthermore, the first test parameter may include the test result of each chip in the current batch of chips in the preset test item. For ease of description, the test result of the preset test item of a single chip in the current batch is also collectively referred to as the first test parameter.
[0054] Furthermore, the present embodiment can be implemented in batches to detect whether the test parameters of a specific batch of chips in a preset test item are abnormal. It should be noted that in actual applications, the present embodiment can also be implemented in wafers.
[0055] In a specific implementation, the current batch of chips refers to the batch of chips for which test data needs to be detected for abnormality during the current execution of this implementation.
[0056] In the step S101, the test result may be obtained in real time by the test machine during the process of testing the current chip according to the test program.
[0057] Alternatively, the test result may be obtained from a preset database, wherein the preset database is used to store the test results obtained by the test machine after testing one or more batches of chips according to the test procedure in the past.
[0058] In a specific implementation, the marginal range and the qualified range correspond to preset test items one by one, that is, each preset test item is provided with a corresponding marginal range and qualified range.
[0059] The qualified range is a pre-set chip test specification, that is, the qualified range refers to the upper and / or lower limits of the specifications that the chip should meet in the preset test items as specified in the chip specification, such as Figure 1 The qualified line 14 is shown. Figure 2 An upper specification line 21 and a lower specification line 22 are shown.
[0060] The marginal range is an updated chip test specification obtained by narrowing down the pre-set chip test specification, that is, the marginal range is an upper limit and / or lower limit obtained by narrowing down the qualified range. For chips whose test parameter values fall within the qualified range but exceed the marginal range, it means that the test results of such chips in the preset test items are actually very close to the critical line specified in the qualified range (such as Figure 1 The circled part in the middle). Experiments show that such chips are very likely to have reliability problems during long-term use. Therefore, this embodiment automatically detects chips with reliability risks by adding an edge range, so that Figure 1 Automatic detection of shifting, broadening and tailing of the test parameter distribution is now possible.
[0061] Furthermore, the degree of narrowing of the edge range compared to the qualified range can be determined by statistical analysis of historical data, wherein the historical data includes test parameters of multiple batches of chips in the past for the preset test items. By setting the edge range reasonably, it is possible to effectively detect that the numerical distribution of the test parameters is approaching the boundary of the qualified range while avoiding false alarms.
[0062] In a specific implementation, the second test parameters of multiple batches of chips whose yields have reached the benchmark yield in history can be selected, and the edge range can be determined based on the second test parameters, wherein the second test parameters are the test results of the historical batches of chips in the preset test items. For example, 20 batches of chips whose yields are close to the benchmark yield in history can be selected, and the second test parameters of these batches of chips can be obtained.
[0063] Specifically, all recent batches of the chip product (for example, the last quarter) can be selected to calculate their average yield (average) and standard deviation (sigma) of the yield, and the average value obtained by recalculating the yield after removing wafers with a yield lower than average-6×sigma is used as the benchmark yield of the chip product.
[0064] The average yield rate may be the sum of the yield rates of all the batches of chips and the average of them. For any batch of chips, the yield rate of the batch of chips refers to the ratio of the number of qualified chips to the number of valid chips in the batch of chips.
[0065] The standard deviation of the yield rate can be the overall standard deviation Where n is the total number of all recent batches of chip products, x i is the yield rate of the i-th batch of chips, is the average yield.
[0066] The standard deviation of the yield rate can also be the sample standard deviation
[0067] Furthermore, the second test parameters of the qualified chips (pass dies) in the plurality of batches of chips can be obtained. Among them, the qualified chips refer to chips whose test results of various test parameters obtained by the test procedure fall within the qualified range specified in the chip specification, that is, "good" chips in the usual sense.
[0068] Further, the yield test buffer data of the historical batch of chips can be obtained according to the second test parameter, and the yield test buffer data is used to characterize the numerical distribution of the second test parameter. Specifically, the numerical distribution can be used to characterize the numerical fluctuation and distribution of the second test parameter. For example, the numerical distribution can be the inter quartile range (IQR) of the second test parameter.
[0069] Furthermore, a reduction amount may be determined according to the yield test buffer data, and the qualified range of the preset test item may be reduced according to the reduction amount to obtain a marginal range of the preset test item.
[0070] Assuming that the qualified range of the preset test item only includes the lower spec, the corresponding edge range can be the lower spec plus the IQR multiplied by the preset coefficient, wherein the preset coefficient can be 0.5.
[0071] Assuming that the qualified range of the preset test item only includes the upper limit (high spec), the corresponding edge range can be the upper limit minus the IQR multiplied by the preset coefficient. The preset coefficient can be 0.5.
[0072] Assuming that the qualified range of the preset test item includes a lower limit and an upper limit, that is, [low spec, high spec], the corresponding edge range may be [low spec+0.5×IQR, high spec-0.5×IQR].
[0073] Only the test results of qualified chips are considered because this embodiment focuses on the abnormal detection of test parameters whose test results are within the qualified range. Further, combined with the resistance of IQR to abnormal values, this specific implementation can objectively and accurately characterize the numerical distribution of test parameters whose test results are within the qualified range.
[0074] In a variation, the interquartile range may be replaced by a standard deviation (sigma). That is, the standard deviation of the second test parameter may be calculated to obtain the yield test buffer data.
[0075] In a specific implementation, chips with test parameter values between a marginal range and a qualified range may be recorded as marginal chips to indicate that these chips are chips close to chip test specifications, that is, the aforementioned chips that may have reliability issues.
[0076] In the step S102, the marginal chip ratio can be calculated based on the formula marginal die ratio=marginal_die_count / good_die_count, wherein marginal die ratio is the marginal chip ratio, marginal_die_count is the number of marginal chips in the current batch of chips, and good_die_count is the number of qualified chips in the current batch of chips.
[0077] Specifically, the qualified chips in the current batch of chips are the good chips that are actually shipped.
[0078] Furthermore, for the current batch of chips, marginal_die_count is the number of chips in the current batch of chips whose values of the first test parameter fall within the qualified range but exceed the marginal range.
[0079] In a specific implementation, the benchmark edge chip ratio can be determined by statistical analysis of historical data, wherein the historical data includes the second test parameters of multiple batches of chips in history. Specifically, the benchmark edge chip ratio can be used to describe the proportion of edge chips when the numerical distribution of the chip test parameters is the benchmark distribution. The values of the test parameters in the benchmark distribution fall within the qualified range and are far from the edge of the qualified range, such as Figure 1 shown.
[0080] For example, after determining the marginal range, for the multiple batches of chips in history used to determine the marginal range, a first number of chips can be screened based on the second test results of these batches of chips, and the first number of chips are chips whose values of the second test parameters fall within the qualified range but exceed the marginal range, that is, marginal chips in the multiple batches of chips in history. Further, the first number is divided by the number of good chips that were finally shipped in the multiple batches of chips in history (that is, the number of qualified chips in the multiple batches of chips in history), and the marginal chip ratio of baseline can be obtained, which is denoted as R MD_Baseline .
[0081] In a specific implementation, the step S102 may include the steps of: obtaining the number of qualified chips in the current batch of chips; screening the edge chips from the current batch of chips according to the qualified range and the edge range; and determining the ratio of the number of the screened edge chips to the number of qualified chips in the current batch of chips as the edge chip ratio of the current batch of chips.
[0082] In a specific implementation, in step S103, by formula i_ΔR MD-Lotx =i_R MD_Lotx -i_R MD_baseline The difference between the edge chip ratio and the reference edge chip ratio is calculated. MD-Lotx is the difference between the edge chip ratio of the x-th batch of chips in the i-th preset test item and the corresponding benchmark edge chip ratio, i_R MD_Lotx is the ratio of marginal chips in the x-th batch of chips in the ith preset test item, i_R MD_baseline is the benchmark edge chip ratio of the i-th preset test item.
[0083] In a specific implementation, the first preset prompt value can be determined based on the return rate of the chip, such as based on parts per million (ppm). Specifically, ppm can be used to represent the return rate of the chip due to premature failure after the chip is shipped to the terminal.
[0084] Furthermore, different chip products have different requirements for ppm, and the return rate of shipped chips in each fixed period (such as weekly, monthly or quarterly) cannot exceed the agreed ppm.
[0085] For example, the first preset prompt value may be 1000 ppm.
[0086] In a specific implementation, the detection method described in this embodiment may also include: calculating the average edge chip ratio of all batches of chips detected within a preset detection period; when the difference between the average edge chip ratio and the benchmark edge chip ratio is greater than a second preset prompt value, determining that a chip test parameter abnormality is detected.
[0087] Specifically, the second preset prompt value may be determined according to the return rate of the chip, and the second preset prompt value is less than the first preset prompt value. For example, the second preset prompt value may be 500 ppm.
[0088] For example, the average marginal chip ratio of all batches of chips manufactured in this cycle may be counted every 7 days. If the average marginal chip ratio is greater than the second preset prompt value, it is determined that the chips produced in this cycle have test parameter abnormalities.
[0089] In a specific implementation, when it is determined that the chip test parameter is detected to be abnormal, a prompt message can be sent out to prompt the detection result.
[0090] Specifically, the prompt information can be sent to quality inspectors, quality control personnel and other personnel of the chip manufacturing company who need to obtain abnormal information on chip test parameters.
[0091] Furthermore, the prompt information can be sent to a server logged in by the above-mentioned personnel, such as a computer, etc., and can also be sent to a personal terminal, such as a mobile phone, IPAD, etc.
[0092] For example, when the prompt information is triggered, the prompt information may be automatically generated by the device executing this embodiment and sent to the above-mentioned server and / or personal terminal.
[0093] Furthermore, the prompt information may include batches with abnormal test parameters, numerical distribution of abnormal test parameters, analysis of causes of abnormalities, etc.
[0094] This specific implementation converts the numerical distribution of the test parameters from a graphic to a digital description. That is, the huge volume of data results formed by one chip and one value for each test parameter is converted into a calculated value (i.e., marginal range) that can be used to describe the reliability of each test parameter. In this way, the amount of data required to be stored is reduced by thousands of times, while the detection of test parameter anomalies is achieved.
[0095] Furthermore, this specific implementation can achieve anomaly detection while minimizing false alarms, ensuring the accuracy of the system running the solution described in this embodiment. That is, even if the numerical distribution of a test parameter of a batch of chips shows abnormalities such as distribution widening and tailing, as long as it is still far from the qualified range or does not cause an increase in the difference between the marginal chip ratio and the corresponding benchmark marginal chip ratio, no prompt information will be issued to avoid causing trouble to the user.
[0096] Furthermore, this specific implementation can be run regularly by the system to achieve automatic calculation and user push of anomaly detection.
[0097] Furthermore, by entering the concepts of edge chips and ppm, flexible feedback correction can be achieved. Specifically, it can be flexibly optimized according to the actual situation of the product, combined with the feedback of the classification of the reasons for the return rate of shipments (that is, how many ppm return rates are contributed by different test failures), and the preset coefficients, first preset prompt values, and second preset prompt values of each test parameter used to determine the edge range can be corrected. In this way, it is ensured that the ppm calculated by each test parameter is as close as possible to the actual value of the final chip return rate, so as to accurately know the early warning of batch disposal and shipment.
[0098] In a specific implementation, the detection method described in this embodiment may also include the steps of: obtaining the value change trend of the third test parameter of multiple batches of chips over time within a preset time period, wherein the third test parameter is the test result of the multiple batches of chips in the preset test item; when the change trend continues to exceed the control range, it is determined that a chip test parameter abnormality is detected.
[0099] Specifically, the control range may be an upper limit and / or a lower limit obtained by narrowing the qualified range. When the value of the third test parameter continues to exceed the control range over time, it indicates that the test results of multiple batches of chips manufactured within a period of time are approaching the edge of the qualified range. Figure 2 Experiments show that this kind of trend is likely to mean that a process parameter of the chip production line has deviated. If no intervention is made, the value of the corresponding test parameter of the chip manufactured according to the current process in the future is likely to exceed the qualified range. Therefore, this embodiment automatically detects the deviation of process parameters by adding a control range, so that Figure 2 Automatic detection of situations in which the value of the indicated test parameter continuously increases or decreases becomes possible.
[0100] Furthermore, the degree of narrowing of the control range compared to the qualified range can be determined by statistical analysis of historical data, wherein the historical data includes the fourth test parameter of multiple batches of chips in history, and the fourth test parameter is the test result of the multiple batches of chips in history under the preset test item. By setting the control range reasonably, it is possible to effectively detect that the numerical distribution of the test parameter is continuously and gradually approaching the boundary of the qualified range while avoiding false alarms.
[0101] In a specific implementation, the fourth test parameter of qualified chips from multiple batches of chips whose yields historically reach the benchmark yield can be selected.
[0102] Furthermore, a key statistic value of the fourth test parameter may be calculated. For example, the key statistic value may be selected from: median, 95th quantile, 5th quantile, 99th quantile and 1st quantile.
[0103] Furthermore, the control range may be determined according to the key statistical value and the standard deviation of the key statistical value.
[0104] Assuming that the qualified range of the test parameter only includes the lower limit (low spec), the corresponding control range can be statistic-3×statistic value's sigma, where statistic is the key statistical value and statistic value's sigma is the standard deviation of the key statistical value. This control range is the lower limit. In the scenario where only the lower limit is considered, the high-side percentiles such as the 95th percentile and the 99th percentile can be ignored.
[0105] Assuming that the qualified range of the test parameter only includes the upper limit (high spec), the corresponding control range can be statistic+3×statistic value's sigma, where statistic is the key statistical value and statistic value's sigma is the standard deviation of the key statistical value. This control range is the upper limit. In the scenario where only the upper limit is considered, the low-side percentiles such as the 1st quantile and the 5th quantile can be ignored.
[0106] Assuming that the qualified range of the test parameter includes the lower limit and the upper limit, that is, [low spec, high spec], the corresponding control range can be [statistic-3×statistic value's sigma, statistic+3×statisticvalue's sigma].
[0107] The value 3 of the preset coefficient used in the above assumption is only used as an example. In actual applications, the specific value of the preset coefficient can be adjusted as needed.
[0108] In a specific implementation, the key statistical value of the third test parameter may be calculated. Further, the changing trend of the key statistical value over time may be plotted.
[0109] Specifically, if any key statistical value exceeds the control range for a preset number of batches in a row, it can be determined that a chip test parameter abnormality is detected.
[0110] For example, the preset number may be 8. In practical applications, the preset number may be optimized according to the actual situation of the product.
[0111] From the above, the scheme of this embodiment can realize automatic detection and push of abnormal chip test parameters, so as to warn chip reliability problems and possible problems in production process and testing links in advance before Bin failure actually occurs. Specifically, by analyzing the proportion of edge chips in the current batch of chips, the distribution of test results of the current batch of chips in the preset test items can be obtained more accurately. Further, if the proportion of edge chips in the current batch increases to a difference greater than the first preset prompt value from the proportion of benchmark edge chips, it indicates that the numerical distribution of the first test parameter deviates from the benchmark distribution, that is, the chip test parameter is abnormal. Therefore, this implementation scheme can realize automatic detection of test parameter abnormalities. When the numerical distribution of the first test parameter deviates from the benchmark distribution, even if the numerical change of the first test parameter has not yet caused Bin failure, the use of this implementation scheme can still accurately and automatically detect this abnormal deviation and issue a warning in time, which is conducive to solving chip reliability problems.
[0112] Furthermore, potential production process and test problems can be detected as early as possible, making it possible to detect process and test defects in advance before Bin failure. Specifically, when the test result values of multiple batches of chips under the same test parameter continue to increase or decrease over a period of time, even if the test results still fall within the qualified range specified by the specification, it actually very likely means that a process parameter of the production line has deviated. If it is not resolved in advance, a large number of low yield problems may occur later. Therefore, this embodiment sets a control range and cooperates with the changing trend of the test results over time to achieve automatic detection. When the changing trend continues to exceed the control range, it indicates that the test results of the corresponding test parameters of the multiple batches of chips have approached the upper or lower limit of the qualified range, and an early warning needs to be issued.
[0113] Furthermore, failure conditions outside the test specifications can be characterized by a dedicated Bin failure rate.
[0114] In the first typical application scenario, Figure 4 The difference between the edge chip ratio of the test item IDDQ_GPU and the benchmark edge chip ratio (denoted as IDDQ_GPU_ΔR) of multiple batches of chips monitored over a period of time is shown. MD , unit ppm) over time, Figure 5 yes Figure 4 The numerical distribution of test parameters of batch A chips in the test item IDDQ_GPU.
[0115] Figure 4 The 1000 ppm shown is a first preset prompt value of the test item IDDQ_GPU, and 500 ppm is a second preset prompt value.
[0116] refer to Figure 4It can be seen that the marginal chip ratio of batch A in the test item IDDQ_GPU is 1200ppm, which has exceeded the first preset prompt value of 1000ppm. Figure 5 It can be seen that the numerical distribution of the test parameters of batch A in the test item IDDQ_GPU did not trigger the pass line, but in fact, the numerical distribution of this batch has seriously deviated from the benchmark distribution of the test item IDDQ_GPU.
[0117] Therefore, the present embodiment is used to automatically detect, by setting edge lines, the situation where the numerical distribution of the test parameters of the test item IDDQ_GPU of batch A is tailing compared to the benchmark distribution, and a prompt is issued.
[0118] In the second typical application scenario, Figure 6 The difference between the ratio of edge chips and the ratio of benchmark edge chips in the test item Vmin_CPU (denoted as Vmin_CPU_ΔR) of multiple batches of chips monitored over a period of time is shown. MD , unit ppm) over time, where the marginal chip ratio in the 20th week is 800ppm, which has exceeded the second preset warning value of 500ppm. Figure 7 yes Figure 6 The numerical distribution of test parameters of the test item Vmin_CPU for all batches of chips obtained in the 20th week of testing. Figure 7 The solid line represents the benchmark distribution, and the dashed line represents Figure 6 Numerical distribution of test parameters of test item Vmin_CPU for all test batches in week 20.
[0119] Figure 8 1 is a schematic diagram of a chip test parameter abnormality detection device according to an embodiment of the present invention. A person skilled in the art will appreciate that the chip test parameter abnormality detection device 8 described in this embodiment can be used to implement the above Figure 3 The method and technical solution described in the embodiments.
[0120] Specifically, refer to Figure 8The chip test parameter abnormality detection device 8 described in this embodiment may include: an acquisition module 81, used to acquire a first test parameter of a current batch of chips, wherein the first test parameter is a test result of the current batch of chips in a preset test item; a calculation module 82, used to calculate the edge chip ratio of the current batch of chips according to the first test parameter, wherein the edge chip ratio is the proportion of edge chips in the current batch of chips, and the edge chip is a chip whose value of the first test parameter falls within a qualified range but exceeds an edge range; a detection module 83, when the difference between the edge chip ratio and the benchmark edge chip ratio is greater than a first preset prompt value, determines that a chip test parameter abnormality is detected, wherein the benchmark edge chip ratio corresponds to the preset test item.
[0121] For more information about the working principle and working method of the chip test parameter abnormality detection device 8, please refer to the above Figure 3 The relevant description in will not be repeated here.
[0122] Furthermore, an embodiment of the present invention also discloses a storage medium having a computer program stored thereon, wherein the computer program is executed by a processor Figure 3 The method and technical solution described in the illustrated embodiment. Preferably, the storage medium may include a computer-readable storage medium such as a non-volatile memory or a non-transitory memory. The storage medium may include a ROM, a RAM, a magnetic disk or an optical disk, etc.
[0123] Furthermore, an embodiment of the present invention also discloses a terminal, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor executes the above-mentioned Figure 3 The method and technical solution described in the illustrated embodiment. Specifically, the terminal may be a computer, a server or other equipment.
[0124] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the scope defined by the claims.
Claims
1. A method for detecting abnormal chip test parameters, characterized in that: include: Acquire a first test parameter of a current batch of chips, wherein the first test parameter is a test result of the current batch of chips in a preset test item; Calculating the marginal chip ratio of the current batch of chips according to the first test parameter, wherein the marginal chip ratio is the proportion of marginal chips in the current batch of chips, and the marginal chips are chips whose values of the first test parameter fall within a qualified range but exceed a marginal range, and the marginal range is an upper limit and / or a lower limit obtained by narrowing the qualified range; When the difference between the edge chip ratio and the benchmark edge chip ratio is greater than a first preset prompt value, it is determined that a chip test parameter abnormality is detected, wherein the benchmark edge chip ratio corresponds to the preset test item, and the benchmark edge chip ratio is used to describe the proportion of edge chips when the numerical distribution of the chip test parameter is the benchmark distribution.
2. The detection method according to claim 1, characterized in that: The step of calculating the benchmark edge chip ratio includes: Acquire a second test parameter of a historical batch of chips, wherein the second test parameter is a test result of the historical batch of chips in the preset test item; Acquire the number of qualified chips in the historical batch of chips and yield test buffer data according to the second test parameter, wherein the yield test buffer data is used to characterize the value distribution of the second test parameter; Narrowing the qualified range based on the yield test buffer data to obtain the marginal range; According to the marginal range and the qualified range, a first number of chips are screened from the historical batch of chips to obtain the chips, wherein the first number of chips are chips whose values of the second test parameter fall within the qualified range but exceed the marginal range; A ratio of the first number to the number of the qualified chips is determined as the reference edge chip ratio.
3. The detection method according to claim 2, characterized in that: The step of acquiring the yield test buffer data of the historical batch of chips according to the second test parameter includes: calculating the interquartile range of the second test parameter to obtain the yield test buffer data; or, The standard deviation of the second test parameter is calculated to obtain the yield test buffer data.
4. The detection method according to any one of claims 1 to 3, characterized in that: Calculating the marginal chip ratio of the current batch of chips according to the first test parameter includes: Get the number of qualified chips in the current batch of chips; According to the qualified range and the marginal range, the marginal chip is screened from the current batch of chips; The ratio of the number of the edge chips obtained by screening to the number of qualified chips in the current batch of chips is determined as the edge chip ratio of the current batch of chips.
5. The detection method according to claim 1, characterized in that: Also includes: Calculate the average marginal chip ratio of all batches of chips detected within a preset detection period; When the difference between the average marginal chip ratio and the reference marginal chip ratio is greater than a second preset prompt value, it is determined that a chip test parameter abnormality is detected.
6. The detection method according to claim 5, characterized in that: The first preset prompt value and the second preset prompt value are both determined according to the return rate of the chip, and the second preset prompt value is smaller than the first preset prompt value.
7. The detection method according to any one of claims 1 to 3 or claim 5 or 6, characterized in that: Also includes: Obtaining a change trend of a value of a third test parameter of multiple batches of chips over time within a preset time period, wherein the third test parameter is a test result of the multiple batches of chips in the preset test item; When the change trend continues to exceed the control range, it is determined that the chip test parameter is detected to be abnormal. The control range is used to detect that the value distribution of the test parameter continues to and gradually approaches the boundary of the qualified range while avoiding false alarms.
8. The detection method according to claim 7, characterized in that: The process of determining the control range includes the following steps: Acquire a fourth test parameter of multiple batches of chips in history, wherein the fourth test parameter is a test result of the multiple batches of chips in history on the preset test item; Calculating key statistical values of the fourth test parameter; A control range is determined according to the key statistical value and a standard deviation of the key statistical value.
9. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are performed.
10. A terminal comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor runs the computer program, the following steps are performed: Acquire a first test parameter of a current batch of chips, wherein the first test parameter is a test result of the current batch of chips in a preset test item; Calculating the marginal chip ratio of the current batch of chips according to the first test parameter, wherein the marginal chip ratio is the proportion of marginal chips in the current batch of chips, and the marginal chips are chips whose values of the first test parameter fall within a qualified range but exceed a marginal range, and the marginal range is an upper limit and / or a lower limit obtained by narrowing the qualified range; When the difference between the edge chip ratio and the benchmark edge chip ratio is greater than a first preset prompt value, it is determined that a chip test parameter abnormality is detected, wherein the benchmark edge chip ratio corresponds to the preset test item, and the benchmark edge chip ratio is used to describe the proportion of edge chips when the numerical distribution of the chip test parameter is the benchmark distribution.
11. The terminal according to claim 10, characterized in that: The step of calculating the benchmark edge chip ratio includes: Acquire a second test parameter of a historical batch of chips, wherein the second test parameter is a test result of the historical batch of chips in the preset test item; Acquire the number of qualified chips in the historical batch of chips and yield test buffer data according to the second test parameter, wherein the yield test buffer data is used to characterize the value distribution of the second test parameter; Narrowing the qualified range based on the yield test buffer data to obtain the marginal range; According to the marginal range and the qualified range, a first number of chips are screened from the historical batch of chips to obtain the chips, wherein the first number of chips are chips whose values of the second test parameter fall within the qualified range but exceed the marginal range; A ratio of the first number to the number of the qualified chips is determined as the reference edge chip ratio.
12. The terminal according to claim 11, characterized in that: The step of acquiring the yield test buffer data of the historical batch of chips according to the second test parameter includes: calculating the interquartile range of the second test parameter to obtain the yield test buffer data; or, The standard deviation of the second test parameter is calculated to obtain the yield test buffer data.
13. The terminal according to any one of claims 10 to 12, characterized in that: Calculating the marginal chip ratio of the current batch of chips according to the first test parameter includes: Get the number of qualified chips in the current batch of chips; According to the qualified range and the marginal range, the marginal chip is screened from the current batch of chips; The ratio of the number of the edge chips obtained by screening to the number of qualified chips in the current batch of chips is determined as the edge chip ratio of the current batch of chips.
14. The terminal according to claim 10, characterized in that: When the processor runs the computer program, the processor also performs the following steps: Calculate the average marginal chip ratio of all batches of chips detected within a preset detection period; When the difference between the average marginal chip ratio and the reference marginal chip ratio is greater than a second preset prompt value, it is determined that a chip test parameter abnormality is detected.
15. The terminal according to claim 14, characterized in that: The first preset prompt value and the second preset prompt value are both determined according to the return rate of the chip, and the second preset prompt value is smaller than the first preset prompt value.
16. The terminal according to any one of claims 10 to 12 or claim 14 or 15, characterized in that: When the processor runs the computer program, the processor also performs the following steps: Obtaining a change trend of a value of a third test parameter of multiple batches of chips over time within a preset time period, wherein the third test parameter is a test result of the multiple batches of chips in the preset test item; When the change trend continues to exceed the control range, it is determined that the chip test parameter is detected to be abnormal. The control range is used to detect that the value distribution of the test parameter continues to and gradually approaches the boundary of the qualified range while avoiding false alarms.
17. The terminal according to claim 16, characterized in that: The process of determining the control range includes the following steps: Acquire a fourth test parameter of multiple batches of chips in history, wherein the fourth test parameter is a test result of the multiple batches of chips in history on the preset test item; Calculating key statistical values of the fourth test parameter; A control range is determined according to the key statistical value and a standard deviation of the key statistical value.
Citation Information
Patent Citations
Method and device for statistical process control for discrete samples data
CN101290517A
Method of determining boundary of wafer test data standard and crystal grain marking method
CN104183511A
Wafer quality management and control method and device
CN104900551A
Test scanning method
CN109585323A
Method and device for determining electrical characteristics of chip, and computer readable storage medium
CN111579961A
