Discrete risk crystal grain determination method and device, crystal grain marking method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a discrete risk crystal grain determination method and device, a crystal grain marking method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring a test item data sequence of a wafer; calculating a slope difference value of the test item target data according to the cumulative probability distribution value corresponding to each test item target data in the test item data sequence; obtaining test item target data meeting a first preset condition based on the slope difference value, and obtaining a discrete critical point; the first preset condition comprises that the slope difference value of the target data of each test item is greater than a slope difference threshold value; taking the discrete critical point meeting a second preset condition as a target discrete risk crystal grain; the second preset condition comprises that data does not exist in the next interval of the interval to which the discrete critical point belongs; the interval is determined based on equidistant division of test item target data. By adopting the method, the accuracy is considered while the universality of determining the discrete risky crystal grains is improved.

Description

Technical Field

[0001] The present application relates to the field of semiconductor technology, and particularly to a method for determining discrete risk dies, a die marking method, a device, a computer device, a computer-readable storage medium, and a computer program product. Background Art

[0002] Wafer probe testing (CP testing) can comprehensively and accurately measure the key performance parameters of dies (such as operating voltage, current consumption, signal delay, etc.) to ensure that these parameters meet the design requirements. If risk dies in the CP test results can be identified, the reliability of the product will be greatly increased.

[0003] In traditional methods, the Dynamic Part Average Testing (DPAT) method is used to identify risk dies, which has specific requirements for the distribution characteristics of test data, requiring the test data to be continuous variables or overall distribution variables.

[0004] However, due to the specific requirements for the distribution characteristics of test data in traditional methods, the versatility of the methods is poor; if traditional methods are used without considering the data distribution, the accuracy will be poor. Summary of the Invention

[0005] Based on this, in view of the above technical problems, it is necessary to provide a method for determining discrete risk dies, a die marking method, a device, a computer device, a computer-readable storage medium, and a computer program product that can balance versatility and accuracy.

[0006] In a first aspect, the present application provides a method for determining discrete risk dies, the method including:

[0007] Obtain a test item data sequence of a wafer;

[0008] Calculate a slope difference value of each test item target data according to the cumulative probability distribution value corresponding to each test item target data in the test item data sequence;

[0009] Based on the slope difference value, obtain the test item target data that meets the first preset condition to obtain a discrete critical point; the first preset condition includes that the slope difference value of each test item target data is greater than a slope difference threshold;

[0010] Use the discrete critical point that meets the second preset condition as the target discrete risk die; the second preset condition includes that there is no data in the next interval of the interval to which the discrete critical point belongs; the interval is determined by equally dividing based on the test item target data.

[0011] In one embodiment, the first preset condition further includes that the number of grains at the preset position of each piece of target data of the test item is less than the quantity threshold of the total number of grains.

[0012] In one embodiment, the preset position includes at least one of the following: the position on the right side of the target data of the test item; and / or the position on the left side of the target data of the test item.

[0013] In one embodiment, calculating the slope difference value of the target data of the test item according to the cumulative probability distribution value corresponding to each piece of target data of the test item in the test item data sequence includes:

[0014] When the target data of the test item is not the first data in the test item data sequence, obtain the cumulative probability distribution value corresponding to the order of the target data of the test item in the test item data sequence as the first cumulative probability distribution value, and obtain the cumulative probability distribution value corresponding to the previous order of the order of the target data of the test item in the test item data sequence as the second cumulative probability distribution value;

[0015] Based on the first cumulative probability distribution value, the target data of the test item corresponding to the corresponding order, the second cumulative probability distribution value, and the target data of the test item of the previous order of the order, determine the slope of each order where the target data of the test item is located;

[0016] Based on the slope of the order where the obtained target data of the test item is located, the slope of the previous order of the order where the target data of the test item is located, and the slope of the next order of the order where the target data of the test item is located, determine the slope difference value of each order where the target data of the test item is located.

[0017] In one embodiment, obtaining the test item data sequence of the wafer includes:

[0018] Obtain the original sequence of test items of the wafer;

[0019] Remove duplicates from the original sequence of test items and sort them in ascending order to obtain the initial data sequence of test items;

[0020] Perform smoothing processing on the initial data sequence of test items to determine the test item data sequence.

[0021] In one embodiment, the method for determining the slope difference threshold includes:

[0022] Obtain the experimental data of the normal distribution sequence of the wafer;

[0023] Calculate the slope difference value of each to-be-experimented data according to the cumulative probability distribution value corresponding to each to-be-experimented data in the to-be-experimented data sequence of the normal distribution;

[0024] Take the mode of the slope difference values of the to-be-experimented data as the reference slope difference value;

[0025] Obtain the candidate slope difference value corresponding to the reference slope difference value; wherein, the candidate slope difference value is greater than the reference slope difference value;

[0026] Obtain the average good product loss rate corresponding to each candidate slope difference value, and select the candidate slope difference value corresponding to the maximum average good product loss rate as the slope difference threshold.

[0027] In one embodiment, the interval division method includes:

[0028] Obtain the first quantile, the second quantile and the target distance factor of the test item data sequence, and determine the division distance; wherein, the first quantile is the 25% quantile; the second quantile is the 75% quantile; the target distance factor is a positive integer;

[0029] Perform equidistant division on the test item data sequence according to the division distance.

[0030] In one embodiment, the determination method of the target distance factor includes:

[0031] Obtain each candidate distance factor; wherein, the candidate distance factor is a positive integer;

[0032] Calculate the candidate distance corresponding to each candidate distance factor according to the first quantile, the second quantile and each candidate distance factor of the test item data sequence;

[0033] Take the candidate distance factor corresponding to the candidate distance that meets the third preset condition as the target distance factor, wherein the third preset condition includes that the number of target discrete risk grains meets the number threshold of risk grains.

[0034] In a second aspect, the present application provides a grain marking method, and the method includes:

[0035] Obtain the target discrete risk grains determined according to any of the above methods;

[0036] Mark the target discrete risk grains on the wafer map.

[0037] In a third aspect, the present application provides a discrete risk grain determination device, and the device includes:

[0038] An acquisition module, configured to acquire a test item data sequence of a wafer;

[0039] A calculation module, configured to calculate a slope difference value of the target test item data according to a cumulative probability distribution value corresponding to each target test item data in the test item data sequence;

[0040] A first screening module, configured to obtain the target test item data that meets a first preset condition based on the slope difference value to obtain a discrete critical point; the first preset condition includes that the slope difference value of each target test item data is greater than a slope difference threshold;

[0041] A second screening module, configured to use the discrete critical point that meets a second preset condition as the target discrete risk die; the second preset condition includes that there is no data in the next interval of the interval to which the discrete critical point belongs; the interval is determined by equally dividing based on the target test item data.

[0042] In a fourth aspect, the present application provides a computer device, including a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of the above method are implemented.

[0043] In a fifth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0044] In a sixth aspect, the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0045] The above discrete risk grain determination method, device, computer device, computer-readable storage medium, and computer program product. First, by calculating the slope difference value of the target data of each test item in the test item data sequence based on the cumulative probability distribution value corresponding to the target data of each test item, the subsequent obtained target discrete risk grains have the meaning of weighting, and it is not limited to the data in the test item data sequence being continuous variables. Second, based on the slope difference value, the target data of the test item that meets the first preset condition is obtained to obtain the discrete critical point. The first preset condition includes that the slope difference value of the target data of each test item is greater than the slope difference threshold. The preliminary discrete critical point is determined according to the comparison between the slope difference value and the slope difference threshold. Finally, the discrete critical point that meets the second preset condition is used as the target discrete risk grain. The second preset condition includes that there is no data in the next interval of the interval to which the discrete critical point belongs. The interval is determined by equally dividing based on the target data of the test item. Since there is no data in the next interval of the interval to which the discrete critical point belongs, the true discrete meaning of the target discrete risk grain is ensured, and the problem of misjudging the target discrete risk grain due to the sharp increase in the slope ratio caused by the continuous and subtle fluctuations of the data is avoided. While taking into account the accuracy, the generality of the discrete risk grain determination is also improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments of the present application or related technologies. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0047] Figure 1 It is a flowchart of the discrete risk grain determination method in an embodiment;

[0048] Figure 2 It is a flowchart of calculating the slope difference value of the target data of the test item in an embodiment;

[0049] Figure 3 It is a flowchart of obtaining the test item data sequence of the wafer in an embodiment;

[0050] Figure 4 It is a schematic diagram of the wafer mapping diagram for marking the target discrete risk grains in an embodiment;

[0051] Figure 5 It is a flowchart of the discrete risk grain determination method in another embodiment;

[0052] Figure 6 It is a comparison diagram of the method of the present application and the traditional method on the cumulative probability distribution diagram in an embodiment;

[0053] Figure 7 It is a comparison chart between the method of the present application and the traditional method in a probability density distribution chart in an embodiment;

[0054] Figure 8 It is a comparison chart between the method of the present application and the traditional method based on the normal distribution in a cumulative probability distribution chart in an embodiment;

[0055] Figure 9 It is a comparison chart between the method of the present application and the traditional method based on the heavy-tailed distribution in a cumulative probability distribution chart in an embodiment;

[0056] Figure 10 It is a comparison chart between the method of the present application and the traditional method based on the tail-compact distribution in a cumulative probability distribution chart in an embodiment;

[0057] Figure 11 It is a comparison chart between the method of the present application and the traditional method based on the skewed distribution in a cumulative probability distribution chart in an embodiment;

[0058] Figure 12 It is a structural block diagram of a discrete risk grain determination device in an embodiment;

[0059] Figure 13 It is an internal structure diagram of a computer device in an embodiment. Detailed implementation manners

[0060] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0061] In one embodiment, as Figure 1 shown, a discrete risk grain determination method is provided. In this embodiment, taking the application of this method to a server as an example, it can be understood that this method can also be applied to a terminal, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps S102 to step S108:

[0062] Step S102, obtaining a test item data sequence of a wafer.

[0063] Optionally, the server obtains a test item data sequence of a wafer: , where: , the test item data sequence is determined by processing the original test item data sequence.

[0064] Step S104: Calculate the slope difference value of each test item target data according to the cumulative probability distribution value corresponding to each test item target data in the test item data sequence.

[0065] Among them, the slope difference value can be the slope ratio of adjacent segments. In discrete data, the slope difference value can be determined by calculating the difference ratio; in continuous data, the slope difference value can be determined by calculating the derivative.

[0066] Optionally, for each test item target data in the test item data sequence, the server calculates the cumulative probability distribution value corresponding to each test item target data through formula (1).

[0067] Formula (1)

[0068] Among them, the cumulative probability distribution value cdf corresponding to each test item target data: count represents the quantity; Y represents the total quantity of test item target data existing in the test item data sequence.

[0069] For example, if there are n test item target data xi in the test item data sequence, i = 1...n, as shown in Table 1.

[0070] Table 1 Specific values of the test item data sequence

[0071]

[0072] Among them, for the cumulative probability distribution value of the first test item target data X1, there is one that is less than or equal to 1.1, cdf1: 1 / n; for the cumulative probability distribution value of the second test item target data X2, there are two that are less than or equal to 1.2, cdf2: 2 / n; for the cumulative probability distribution value cdfn of the nth test item target data Xn: 1.

[0073] Optionally, the server determines the slope of each test item target data according to the cumulative probability distribution value corresponding to each test item target data in the test item data sequence, and then calculates the slope calculation difference value of the test item target data through formula (3), such as the slope ratio.

[0074] Formula (2)

[0075] Formula (3)

[0076] Among them, Vxi represents the slope of the i-th test item target data; Avi represents the slope ratio of the i-th test item target data; cdf i represents the cumulative probability distribution value of the i-th test item target data.

[0077] Principle of calculating slope and slope ratio through cumulative probability distribution values:

[0078] If only the specific values of the test item data are used to calculate the slope, then continuous variable test data is required. However, in reality, due to the test accuracy of the testing machine, the specific values of the same test data may appear multiple times, which means there are test data of discrete variables. For example, for the test item data: there are 10 points with a value of 0.1 and 20 points with a value of 0.2, etc. But 0.1 and 0.2 are not real data. Instead, due to the accuracy problem of the testing machine, 0.1001 is also regarded as 0.1, or 0.09999 is also regarded as 0.1. If only the specific values of the test item data are used to calculate the slope, then the occurrence times of the same test item data are not considered, and the critical point where the discrete variable starts to be discrete cannot be found. By using the cumulative probability distribution value to calculate the slope, the problem that the test item data may be a discrete variable is solved, making the sequence used to find the mutation point have the weighted meaning of the number of times. Therefore, it is not limited to the test data of continuous variables.

[0079] Step S106, obtain the target test item data that meets the first preset condition based on the slope difference value to obtain the discrete critical point.

[0080] Among them, the first preset condition includes that the slope difference value of each target test item data is greater than the slope difference threshold.

[0081] Optionally, the server obtains the target test item data that meets the first preset condition based on the slope difference value. For example, if the slope difference threshold is determined to be 2, and the slope ratio Avi of the i-th target test item data of the server is greater than 2, then the target test item data that meets the condition is used as the discrete critical point. Among them, the slope difference threshold can also be other values.

[0082] Step S108, use the discrete critical point that meets the second preset condition as the target discrete risk die.

[0083] Among them, the second preset condition includes that there is no data in the next interval of the interval to which the discrete critical point belongs; the interval is determined by equally spaced division based on the target test item data. In wafer probe testing, discrete risk dice refer to those dice that may cause electrical performance or functional abnormalities due to various reasons (such as manufacturing defects, material problems, design problems, etc.). These dice may cause problems in subsequent packaging and testing processes, thus reducing the overall production yield and quality of the chip.

[0084] For example, the interval is equally spaced based on the target test item data. For example, the group distance is ( ), and equally spaced division will obtain several intervals. By means of interval segmentation, it is ensured that the data density after the discrete critical point is significantly reduced.

[0085] The server further determines whether the discrete critical point satisfies that there is no data in the next interval of the interval to which each discrete critical point belongs. If there is data in the next interval, it indicates that the discrete critical point is not a true discrete critical point; if there is no data in the next interval, it indicates that there is no other data near the discrete critical point, and the discrete critical point is a true discrete critical point, which can also be called the target discrete risk grain.

[0086] In the discrete risk grain determination method, first, by calculating the slope difference value of the target data of each test item corresponding to the cumulative probability distribution value in the test item data sequence, the target discrete risk grain obtained subsequently has the meaning of weighting, and it is not limited to the data in the test item data sequence being continuous variables; secondly, based on the slope difference value, the target data of the test item that satisfies the first preset condition is obtained to get the discrete critical point; the first preset condition includes that the slope difference value of the target data of each test item is greater than the slope difference threshold; according to the comparison between the slope difference value and the slope difference threshold, the preliminary discrete critical point is determined; finally, the discrete critical point that satisfies the second preset condition is used as the target discrete risk grain; the second preset condition includes that there is no data in the next interval of the interval to which the discrete critical point belongs; the interval is determined by equally dividing based on the target data of the test item. Since there is no data in the next interval of the interval to which the discrete critical point belongs, the true discrete meaning of the target discrete risk grain is ensured, and the problem of misjudging the target discrete risk grain due to the sharp increase in the slope ratio caused by continuous and subtle fluctuations of the data is avoided. While taking into account the accuracy, the generality of the discrete risk grain determination is also improved.

[0087] In an exemplary embodiment, the first preset condition further includes that the number of grains at the preset position of the target data of each test item is less than the quantity threshold of the total number of grains.

[0088] Among them, the quantity threshold is determined according to the actual application, and the quantity threshold is to ensure that the killing of the yield by a single parameter is below a certain quantity.

[0089] Optionally, the number of grains at the preset position of the target data of each test item is less than the quantity threshold of the total number of grains, so as to ensure that the killing of the yield of the grains is below a certain value. Since this application determines the discrete risk grains among the grains passing the CP test, it is necessary to set the killing quantity of the yield of the grains to avoid considering all the grains passing the CP test as discrete risk grains.

[0090] In this embodiment, by setting the number of killed grains, the yield can be increased as much as possible on the basis of ensuring the reliability of "zero defect" products, and excessive grains are avoided from being killed.

[0091] In an exemplary embodiment, the preset position includes at least one of the following: the position on the right side of the test item target data; and / or the position on the left side of the test item target data.

[0092] Optionally, the first preset condition further includes that the number of grains at the preset position of each test item target data is less than the quantity threshold of the total number of grains. Assuming the quantity threshold is set to 10%, for example, the number of grains at the preset position of each test item target data, such as the number of grains on the right side of the test item data sequence where the test item target data is located; such as on the right side of Avi , that is, the number of grains at the back < 10% of the total number of grains.

[0093] The server may also be that the number of grains on the left side of the test item data sequence where the test item target data is located is less than the quantity threshold of the total number of grains, such as the number of grains on the left side of the test item data sequence where the test item target data is located is less than the quantity threshold of the total number of grains; such as on the left side of Av i on the right side of Av1 - Av i-1 , that is, the number of grains at the front < 20% of the total number of grains.

[0094] The server may also be to simultaneously set that the number of grains on the left side of the test item data sequence where the test item target data is located is less than the quantity threshold of the total number of grains, and the number of grains on the left side of the test item data sequence where the test item target data is located is less than the quantity threshold of the total number of grains.

[0095] In this embodiment, by adjusting the preset position of the test item target data of different wafer test items according to the actual situation, it can be more in line with the actual situation, thereby improving the accuracy of subsequent determination of discrete risk grains.

[0096] In an exemplary embodiment, as Figure 2 shown, according to the cumulative probability distribution value corresponding to each test item target data in the test item data sequence, calculating the slope difference value of the test item target data includes steps S202 to S206. Wherein:

[0097] Step S202, in the case where the test item target data is not the first data in the test item data sequence, obtain the cumulative probability distribution value corresponding to the order of the test item target data in the test item data sequence as the first cumulative probability distribution value, and obtain the cumulative probability distribution value corresponding to the previous order of the order of the test item target data in the test item data sequence as the second cumulative probability distribution value.

[0098] Optionally, when the target test item data is not the first data in the test item data sequence, it is because the slope of the first target test item data in the test item data sequence cannot be calculated. Therefore, when it is not the first data in the test item data sequence, obtain the cumulative probability distribution value corresponding to the order of the target test item data in the test item data sequence as the first cumulative probability distribution value. For example, for the i-th target test item data X i The corresponding first cumulative probability distribution value cdf i ; obtain the cumulative probability distribution value corresponding to the previous order of the order of the target test item data in the test item data sequence as the second cumulative probability distribution value. For example, for the (i - 1)-th target test item data X i-1 The corresponding second cumulative probability distribution value cdf （i-1） .

[0099] Step S204: Based on the first cumulative probability distribution value, the target test item data corresponding to the order where it is located, the second cumulative probability distribution value, and the target test item data of the previous order of the order where it is located, determine the slope of the order where each target test item data is located.

[0100] Optionally, the server is based on the first cumulative probability distribution value cdf i And the corresponding target test item data X i , the second cumulative probability distribution value cdf （i-1） And the corresponding target test item data X i-1 , determine the slope of the order where the i-th target test item data is located. i can be any number other than 1. Therefore, the slope of each target test item data in the test item data sequence can be calculated .

[0101] Furthermore, the server subtracts the second cumulative probability distribution value cdf i From the first cumulative probability distribution value cdf （i-1） To obtain the first value; the server also subtracts the target test item data X i From the target test item data X i-1 To obtain the second value; divide the first value by the second value to obtain the slope of the target test item data X i . The calculation formula is as shown in formula (2).

[0102] Formula (2)

[0103] Step S206: Based on the slope of the order where the obtained target test item data is located, the slope of the previous order of the order where the target test item data is located, and the slope of the next order of the order where the target test item data is located, determine the slope difference value of the order where each target test item data is located.

[0104] Optionally, the server defines a 1×3 rectangular box, which requires three data. Therefore, based on the slope of the order where the target test item data is located, the slope of the previous order of the order where the target test item data is located, and the slope of the next order of the order where the target test item data is located, these are used as the three filling data in the rectangular box. The slope difference value of the order where each target test item data is located is determined through formula (3), that is, the slope ratio of adjacent front and rear segments.

[0105] Formula (3)

[0106] In this embodiment, by calculating the slope and slope ratio through the cumulative probability distribution value, it is not necessary to be limited to test data where the test item data must be a continuous variable, solving the problem that the test item data may be a discrete variable. At the same time, the cumulative probability distribution value also has the weighted meaning of the number of times. Therefore, the distribution followed by the data itself can be not restricted, and the accuracy can be ensured.

[0107] In an exemplary embodiment, as Figure 3 shown, obtaining the test item data sequence of the wafer includes steps S302 to S306. Among them:

[0108] Step S302, obtaining the original sequence of test items of the wafer.

[0109] Optionally, according to the test item dimension of the wafer probe test (CP test) and the wafer, obtain the original sequence of test items of the wafer for the latest time, denoted as: .

[0110] Step S304, removing duplicates from the original sequence of test items and sorting them in ascending order to obtain the initial data sequence of test items.

[0111] Optionally, the server removes duplicates from the original sequence of test items and arranges them in ascending order to obtain the order statistic, that is, the initial data sequence of test items.

[0112] Step S306, performing a smoothing process on the initial data sequence of test items to determine the test item data sequence.

[0113] Optionally, the server performs a smoothing process on the initial data sequence of test items through the spline smoothing method to determine the test item data sequence: , where: . The purpose of the smoothing process is to eliminate the influence of the small fluctuations in the initial data sequence of test items on the subsequent determination of the target discrete risk grains.

[0114] In this embodiment, by removing duplicates and sorting the original sequence of test items and performing smoothing processing, a test item data sequence is obtained, which can eliminate the influence of small fluctuations in the initial data sequence of test items on the subsequent determination of target discrete risk grains, thereby improving the accuracy of subsequent determination of target discrete risk grains.

[0115] In an exemplary embodiment, a method for determining a slope difference threshold includes: obtaining experimental data to be processed of a normal distribution sequence of a wafer; calculating slope difference values of each experimental data to be processed according to the cumulative probability distribution value corresponding to each experimental data to be processed in the normal distribution sequence of experimental data to be processed; using the mode of the slope difference values of the experimental data to be processed as a reference slope difference value; obtaining candidate slope difference values corresponding to the reference slope difference value; wherein the candidate slope difference values are greater than the reference slope difference value; obtaining the average good product loss rate corresponding to each candidate slope difference value, and selecting the candidate slope difference value corresponding to the maximum average good product loss rate as the slope difference threshold.

[0116] The reference slope difference value refers to the slope difference value corresponding to the stable value of the slope difference values of each experimental data to be processed.

[0117] Optionally, numpy randomly generates 1000 groups of experimental data to be processed of a normal distribution sequence with random mean and standard deviation; taking the midpoint in the normal distribution sequence as the center point, calculating the slope difference value of the cumulative probability distribution value corresponding to each experimental data to be processed by dividing the slope between the center point and the previous point of the center point by the slope between the center point and the next point of the center point; using the mode of the slope difference values of the experimental data to be processed as the reference slope difference value, for example, 1.6, and obtaining candidate slope difference values corresponding to the reference slope difference value; wherein the candidate slope difference values are greater than the reference slope difference value; for example, the candidate slope difference values are 2, 2.5, and 3 respectively, obtaining the average good product loss rate corresponding to each candidate slope difference value, the average good product loss rate (loss) corresponding to the candidate slope difference value of 2 is 0.54%, the average good product loss rate (loss) corresponding to the candidate slope difference value of 2.5 is 0.21%, and the average good product loss rate (loss) corresponding to the candidate slope difference value of 3 is 0.02%. Select the candidate slope difference value corresponding to the maximum average good product loss rate as the slope difference threshold, such as 2. Reviewing the discrete critical points (high-risk grains die) after card control, when the slope difference threshold is set to 2, about 96% of the wafers are reasonably card-controlled, and the remaining wafers are card-controlled more tightly. Therefore, it is reasonable to set the slope difference threshold to 2 as a whole.

[0118] Principle for determining the slope difference threshold: Taking the midpoint of the sequence as the center point, divide the slope between the center point and the previous point by the slope between the center point and the next point. If it is greater than 1, it indicates that the degree of inclination becomes milder. The slope difference threshold to be found is how mild it should be considered that the data is truly significantly discretized, that is, the data that makes the experimental data of the normal distribution sequence truly discrete. Since the data distribution of the normal distribution is relatively uniform and the cumulative probability distribution graph is a relatively uniform "S" shape, the change trend of the slope difference threshold is first increasing and then decreasing. Therefore, by calculating the cdf slope ratio using stable data of multiple simulated normal distributions, the slope difference threshold to be determined should be greater than the maximum value of the reference slope difference value.

[0119] In this embodiment, the reference slope difference value obtained through multiple experiments is used to verify and obtain an accurate slope difference threshold. The precise setting of the slope difference threshold can improve the accuracy of subsequent determination of target discrete risk grains.

[0120] In an exemplary embodiment, the interval division method includes: obtaining the first quantile, the second quantile, and the target distance factor of the test item data sequence, and determining the division distance; performing equidistant division on the test item data sequence according to the division distance.

[0121] Among them, the first quantile is the 25% quantile; the second quantile is the 75% quantile; the target distance factor is a positive integer.

[0122] Optionally, the 25% quantile Q1, the 75% quantile Q3 of the server test item data sequence, and the positive integer target distance factor. The target distance factor is determined through multiple experiments. For example, it is 20. According to Q1, Q3, and the target distance factor, the division distance can be determined, which can also be called the group distance ( ), and the server performs equidistant division on the test item data sequence according to the division distance.

[0123] In this embodiment, through the division distance, that is, by equidistantly dividing the test item data sequence according to the group distance, the data density after the discrete critical point is significantly reduced by means of interval segmentation, thereby improving the accuracy of subsequent determination of target discrete risk grains.

[0124] In an exemplary embodiment, the method for determining the target distance factor includes: obtaining each candidate distance factor; calculating the candidate distance corresponding to each candidate distance factor according to the first quantile, the second quantile, and each candidate distance factor of the test item data sequence; and using the candidate distance factor corresponding to the candidate distance that satisfies the third preset condition as the target distance factor.

[0125] Among them, the candidate distance factor is a positive integer; the first quartile Q1 is the 25% quartile to the second quartile Q3 which is the 75% quartile, representing the overall distribution of the test item data sequence. Q2 is the 50% quartile. The candidate distance factor is a positive integer; the third preset condition includes that the number of target discrete risk grains meets the number threshold of risk grains. The third preset condition is set according to the number of risk grains visible to the naked eye.

[0126] Optionally, the server obtains each candidate distance factor. For example, each data from 10 to 30 is a candidate distance factor. The candidate distance factor can also be other positive integers, which are not limited here. According to the first quartile Q1, the second quartile Q3 of the test item data sequence, and each candidate distance factor, calculate the candidate distance corresponding to each candidate distance factor: (Q3 - Q1) / candidate distance factor. Take the candidate distance factor corresponding to the candidate distance that meets the condition that the number of target discrete risk grains meets the number threshold of risk grains as the target distance factor, for example, it is 20.

[0127] In this embodiment, by using the first quartile Q1, the second quartile Q3, each candidate distance factor, and the third preset condition, the target distance factor is determined, making the interval division more reasonable to improve the accuracy of subsequent determination of target discrete risk grains.

[0128] In an exemplary embodiment, obtain the target discrete risk grains determined by the method in any of the above embodiments; mark the target discrete risk grains on the wafer map.

[0129] Optionally, the wafer map is as Figure 4 shown in (a). The red area represents the grains that fail the CP test; the green area represents the grains that pass the CP test. The server marks the target discrete risk grains determined in any of the above embodiments as the blue area on the wafer map, and obtains as shown in Figure 4 (b). According to the wafer map marked with the target discrete risk grains, the type of grains can be intuitively judged, such as the grains that pass the CP test, the grains that fail the CP test, and the target discrete risk grains, so as to intuitively observe which grains are risk grains, which can effectively improve the reliability of the chip product and meet the "zero defect" shipping requirement. The correspondence between the grain type and the color in the wafer map table is shown in Table 2.

[0130] Table 2 Correspondence table of grain type and color

[0131]

[0132] In this embodiment, by marking the target discrete risk dies on the wafer map, it is possible to visually observe which dies are risk dies, which can effectively improve the reliability of chip products and meet the shipping requirement of "zero defects".

[0133] In an exemplary embodiment, as Figure 5 shown. The server obtains the test item data of the wafer at different time points, and the server obtains the original sequence of the test items of the wafer for the latest time, denoted as: . The server removes duplicates from the original sequence of test items and arranges them in ascending order to obtain order statistics, that is, the initial data sequence seq of test items. The server performs a smoothing process on the initial data sequence of test items through a spline smoothing method to determine the test item data sequence: , where: . The purpose of the smoothing process is to eliminate the influence of the small fluctuations in the initial data sequence of test items on the subsequent determination of the target discrete risk dies.

[0134] For each test item target data in the test item data sequence, the server calculates the cumulative probability distribution value corresponding to each test item target data through formula (1).

[0135] Formula (1)

[0136] where the cumulative probability distribution value cdf corresponding to each test item target data: count represents the quantity; Y represents the total quantity of the test item target data existing in the test item data sequence.

[0137] In the case of not being the first data in the test item data sequence, obtain the cumulative probability distribution value corresponding to the order of the test item target data in the test item data sequence as the first cumulative probability distribution value. For example, for the i-th test item target data X i corresponding first cumulative probability distribution value cdf i ; obtain the cumulative probability distribution value corresponding to the previous order of the order of the test item target data in the test item data sequence as the second cumulative probability distribution value. For example, for the (i - 1)-th test item target data X i-1 corresponding second cumulative probability distribution value cdf （i-1） . The server is based on the first cumulative probability distribution value cdf i and the corresponding test item target data X i , the second cumulative probability distribution value cdf （i-1） and the corresponding test item target data X i-1 to determine the slope of the order where the i-th test item target data is located. i can be any number other than 1. Therefore, the slope of each test item target data in the test item data sequence can be calculated 。The server subtracts the second cumulative probability distribution value cdf i from the first cumulative probability distribution value cdf （i-1） to obtain a first value; the server also subtracts the test item target data X i from the test item target data X i-1 to obtain a second value; dividing the first value by the second value can obtain the slope of the test item target data X i . The calculation formula is as shown in formula (2).

[0138] Formula (2)

[0139] The server defines a 1*3 rectangular box, that is, three data are required. Therefore, based on the slope of the test item target data in the obtained order, the slope of the previous order of the test item target data in the order, and the slope of the next order of the test item target data in the order as the three filling data in the rectangular box. The slope difference value of each test item target data in the order is determined by formula (3), that is, the adjacent front and back segment slope ratio, which can also be called the slope difference ratio sequence.

[0140] Formula (3)

[0141] The first preset condition includes that the slope difference value of each test item target data is greater than the slope difference threshold, and also includes that the number of grains at the preset position of each test item target data is less than the quantity threshold of the total number of grains. The server obtains the test item target data that meets the first preset condition based on the slope difference value. For example, if the slope difference threshold is determined to be 2, the slope ratio Avi of the i-th test item target data of the server is greater than 2, and on the right side of Avi , that is, the number of grains behind < 10% of the total number of grains is used as the first preset condition to screen and exclude the test item target data in the test item data sequence, and the test item target data that meets the first preset condition is retained as the discrete critical point. A new list list can also be created, and the discrete critical points are added to the list list. Exclude the test item target data that does not meet the first preset condition.

[0142] The interval is equally divided based on the test item target data. For example, the group distance is ( ), equidistant division will result in several intervals. The server further determines whether the discrete critical point satisfies that there is no data in the next interval of the interval to which each discrete critical point belongs. If there is data in the next interval, it indicates that the discrete critical point is not a real discrete critical point, and the corresponding discrete critical point is removed from the list list; if there is no data in the next interval, it indicates that there is no other data near the discrete critical point, and the discrete critical point is a real discrete critical point, which can also be called the target discrete risk grain, and the target discrete risk grain is retained in the list list. Then the server determines whether there is data in the list list by judging the length length(list) of the list. If there is data in the list list, it is defined as an inflection point, that is, the target discrete risk grain. If there is no data in the list list, that is, length(list)=0, it indicates that no obvious inflection point is found. The server can mark the position of the target discrete risk grain in the cumulative probability distribution graph. The server can also mark the obtained target discrete risk grain on the wafer mapping graph, resulting in Figure 4 (b), According to the wafer mapping graph marked with the target discrete risk grain, the type of the grain can be intuitively judged, such as the grain passing the CP test, the grain failing the CP test, and the target discrete risk grain, so as to intuitively observe which grains are risk grains, which can effectively improve the reliability of the chip product and meet the shipping requirement of "zero defect".

[0143] To ensure the feasibility of the method of this application, the method of this application and the existing DPAT method are used to verify on the actual test item data. In the verification effect, the following are all the results of the method of this application with a red dotted line, and the results of the existing DPAT method with a blue dotted line.

[0144] The DPAT (Dynamic Part Average Testing) method is a statistical method proposed by the AEC (Automotive Electronics Council) Automotive Electronics Council in the AEC-Q100 standard for the Parametric Part Average Testing (PPAT) method, which is used to detect abnormal characteristics of outer semiconductor components. DPAT calculates the upper and lower limits according to the data of the wafer and the test item dimensions, and the calculation formula is shown in formula (4).

[0145] Formula (4)

[0146] Formula (5)

[0147] Among them, Median represents the median; Q3 is the 75% quantile, and Q1 is the 25% quantile.

[0148] The method of DPAT assumes that the data conforms to a normal distribution, while the actual distribution of test data is not necessarily a normal distribution, which may lead to inaccurate discrete critical points calculated by this method.

[0149] Verification on the cumulative probability distribution graph, as Figure 6 shown. Verification on the probability density graph, as Figure 7 shown. Among them, the cumulative probability distribution graph and the probability density graph are different forms of expression. It can be clearly obtained that the sensitivity of the method of this application is greater than that of the traditional method.

[0150] Furthermore, the server can also calculate the number of peaks of the probability density on the entire wafer data. If the number of peaks is greater than 1, it means that there are multiple failure modes in the wafer. Then highlight it to the product engineer for further analysis of the wafer to avoid the loss of the factory caused by the shipment of risky wafers, thereby improving product reliability.

[0151] This application has the property of generalization and does not limit the distribution followed by the test data itself. The following are the verification effects on the normal distribution, heavy-tailed distribution, tight-tailed distribution, and skewed distribution. Among them, the red dashed line is the result of the method of this application, and the blue dashed line is the result of the existing DPAT method.

[0152] For the normal distribution, the verification of the two methods on the cumulative probability distribution value graph is as Figure 8 shown. Obviously, the sensitivity of the method of this application is much greater than that of the traditional method.

[0153] For the heavy-tailed distribution, the verification of the two methods on the cumulative probability distribution value graph is as Figure 9 shown. Although the sensitivity of the traditional method is greater than that of the method of this application, the difference in sensitivity between the two is very small.

[0154] For the tight-tailed distribution, the verification of the two methods on the cumulative probability distribution value graph is as Figure 10 shown. Obviously, the sensitivity of the method of this application is much greater than that of the traditional method.

[0155] For the skewed distribution, the verification of the two methods on the cumulative probability distribution value graph is as Figure 11 shown. Although the sensitivity of the traditional method is greater than that of the method of this application, the difference in sensitivity between the two is very small, and their values are very close.

[0156] Based on the verification of the above four data distributions (normal distribution, heavy-tailed distribution, tight-tailed distribution, skewed distribution), under the normal distribution and the tight-tailed distribution, the sensitivity of the method of this application is much greater than that of the traditional method. Under the heavy-tailed distribution and the skewed distribution, the method of this application is close to the sensitivity of the traditional method, which shows that the method of this application does not limit the distribution followed by the test data itself and has the property of generalization.

[0157] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, there is no strict order limit for the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0158] Based on the same inventive concept, an embodiment of the present application further provides a discrete risk grain determination device for implementing the discrete risk grain determination method described above. The solution provided by this device to solve the problem is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the discrete risk grain determination device provided below can refer to the limitations on the discrete risk grain determination method in the above text, and will not be repeated here.

[0159] In an exemplary embodiment, as Figure 12 shown, a discrete risk grain determination device is provided, including: an acquisition module 1201, a calculation module 1202, a first screening module 1203, and a second screening module 1204, where:

[0160] The acquisition module 1201 is configured to acquire the test item data sequence of the wafer.

[0161] The calculation module 1202 is configured to calculate the slope difference value of the test item target data according to the cumulative probability distribution value corresponding to each test item target data in the test item data sequence.

[0162] The first screening module 1203 is configured to obtain the test item target data that meets the first preset condition based on the slope difference value to obtain the discrete critical point; the first preset condition includes that the slope difference value of each test item target data is greater than the slope difference threshold.

[0163] The second screening module 1204 is configured to use the discrete critical point that meets the second preset condition as the target discrete risk grain; the second preset condition includes that there is no data in the next interval of the interval to which the discrete critical point belongs; the interval is determined by equally dividing the test item target data.

[0164] In an exemplary embodiment, the first preset condition further includes that the number of grains at the preset position of each test item target data is less than the quantity threshold of the total number of grains.

[0165] In an exemplary embodiment, the preset position includes at least one of the following: the position on the right side of the test item target data; and / or the position on the left side of the test item target data.

[0166] In an exemplary embodiment, the calculation module 1202 includes:

[0167] A cumulative probability distribution value calculation unit, configured to, when the test item target data is not the first data in the test item data sequence, obtain the cumulative probability distribution value corresponding to the order of the test item target data in the test item data sequence as the first cumulative probability distribution value, and obtain the cumulative probability distribution value corresponding to the previous order of the order of the test item target data in the test item data sequence as the second cumulative probability distribution value.

[0168] A slope calculation unit, configured to determine the slope of the order where each test item target data is located based on the first cumulative probability distribution value and the test item target data corresponding to the order where it is located, the second cumulative probability distribution value, and the test item target data of the previous order of the order where it is located.

[0169] A slope difference value calculation unit, configured to determine the slope difference value of the order where each test item target data is located based on the slope of the order where the obtained test item target data is located, the slope of the previous order of the order where the test item target data is located, and the slope of the next order of the order where the test item target data is located.

[0170] In an exemplary embodiment, the acquisition module 1201 is further configured to acquire the original sequence of test items of the wafer; remove duplicates from the original sequence of test items and sort them in ascending order to obtain the initial test item data sequence; perform a smoothing process on the initial test item data sequence to determine the test item data sequence.

[0171] In an exemplary embodiment, a die marking device is provided, configured to acquire the target discrete risk die determined by any of the above methods; mark the target discrete risk die on the wafer map.

[0172] Each module in the above discrete risk die determination device can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in the processor of the computer device in hardware form or be independent of it, or be stored in the memory of the computer device in software form so that the processor can call and execute the operations corresponding to each of the above modules.

[0173] In an exemplary embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 13As shown in the figure. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the test data of the wafers. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, it implements a method for determining discrete risk grains.

[0174] Those skilled in the art can understand that Figure 13 the structure shown in the figure is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0175] In one embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0176] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0177] In one embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by the processor, the steps in the above method embodiments are implemented.

[0178] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logics, data processing logics based on quantum computing, artificial intelligence (AI) processors, etc., without limitation.

[0179] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this application.

[0180] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A discrete risk grain determination method, characterized in that, The method includes: Obtaining a test item data sequence of a wafer; Calculating a slope difference value of each target test item data according to a cumulative probability distribution value corresponding to each target test item data in the test item data sequence; Obtaining the target test item data that meets the first preset condition based on the slope difference value to obtain a discrete critical point; the first preset condition includes that the slope difference value of each target test item data is greater than a slope difference threshold; Taking the discrete critical point that meets the second preset condition as the target discrete risk die; the second preset condition includes that there is no data in the next interval of the interval to which the discrete critical point belongs; the interval is determined by equally dividing based on the target test item data.

2. The method according to claim 1, wherein The first preset condition further includes that the number of dies at a preset position of each target test item data is less than a quantity threshold of the total number of dies.

3. The method according to claim 2, wherein The preset position includes at least one of the following: a position on the right side of the target test item data; and / or a position on the left side of the target test item data.

4. The method according to any one of claims 1 to 3, characterized in that, The calculating the slope difference value of each target test item data according to the cumulative probability distribution value corresponding to each target test item data in the test item data sequence includes: When the target test item data is not the first data in the test item data sequence, obtaining the cumulative probability distribution value corresponding to the order of the target test item data in the test item data sequence as a first cumulative probability distribution value, and obtaining the cumulative probability distribution value corresponding to the previous order of the order of the target test item data in the test item data sequence as a second cumulative probability distribution value; Determining the slope of each order where the target test item data is located based on the first cumulative probability distribution value, the target test item data corresponding to the order where it is located, the second cumulative probability distribution value, and the target test item data of the previous order where it is located; Determining the slope difference value of each order where the target test item data is located based on the slope of the order where the obtained target test item data is located, the slope of the previous order where the target test item data is located, and the slope of the next order where the target test item data is located.

5. The method according to any one of claims 1 to 3, characterized in that The obtaining the test item data sequence of the wafer includes: Obtaining an original test item sequence of the wafer; Removing duplicates from the original test item sequence and sorting it in ascending order to obtain an initial test item data sequence; Performing a smoothing process on the initial test item data sequence to determine the test item data sequence.

6. The method according to claim 1, characterized in that, The method for determining the slope difference threshold includes: Obtaining experimental data of a normal distribution sequence of the wafer; Calculating the slope difference value of each experimental data according to the cumulative probability distribution value corresponding to each experimental data in the normal distribution sequence of the experimental data; Taking the mode of the slope difference values of the experimental data as a reference slope difference value; Obtaining a candidate slope difference value corresponding to the reference slope difference value; wherein, the candidate slope difference value is greater than the reference slope difference value; Obtain the average good product loss rate corresponding to each of the candidate slope difference values, and select the candidate slope difference value corresponding to the maximum average good product loss rate as the slope difference threshold.

7. The method according to claim 1, characterized in that, The interval division method includes: Obtain the first quantile, the second quantile, and the target distance factor of the test item data sequence, and determine the division distance; wherein, the first quantile is the 25% quantile; the second quantile is the 75% quantile; the target distance factor is a positive integer; Perform equidistant division on the test item data sequence according to the division distance.

8. The method according to claim 7, characterized in that The determination method of the target distance factor includes: Obtain each candidate distance factor; wherein, the candidate distance factor is a positive integer; According to the first quantile, the second quantile, and each candidate distance factor of the test item data sequence, calculate the candidate distance corresponding to each candidate distance factor; Use the candidate distance factor corresponding to the candidate distance that satisfies the third preset condition as the target distance factor, wherein the third preset condition includes that the number of target discrete risk grains satisfies the risk grain number threshold.

9. A method for marking crystal grains, characterized in that, The method includes: Obtain the target discrete risk grains determined according to any one of the methods of claims 1 to 8; Mark the target discrete risk grains on the wafer map.

10. A discrete risk grain determination device, characterized in that, The device includes: An acquisition module, configured to acquire the test item data sequence of the wafer; A calculation module, configured to calculate the slope difference value of the test item target data according to the cumulative probability distribution value corresponding to each test item target data in the test item data sequence; A first screening module, configured to obtain the test item target data that satisfies the first preset condition based on the slope difference value, and obtain the discrete critical point; the first preset condition includes that the slope difference value of each test item target data is greater than the slope difference threshold; A second screening module, configured to use the discrete critical point that satisfies the second preset condition as the target discrete risk grain; the second preset condition includes that there is no data in the next interval of the interval to which the discrete critical point belongs; the interval is determined based on equidistant division of the test item target data.

11. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 9 are implemented.

13. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images