A semiconductor device test probe and a semiconductor device test method thereof

Abstract

The present application relates to the technical field of semiconductor device testing and its testing probe, in particular to a kind of semiconductor device testing probe and its semiconductor device testing method.The present application first obtains the contact stability coefficient of testing probe at each test monitoring moment, further determines all abnormal contact moments under each abnormal contact type in its preset historical test period, and then according to the total number of abnormal contact type and contact stability coefficient, and the change distribution of contact resistance in the preset time domain of abnormal contact moment under different abnormal contact types, obtains the pressure application abnormal coefficient of testing probe and adjusts the applied pressure.The present application evaluates the contact state by analyzing the change characteristics of contact resistance, and then evaluates the abnormal situation of the applied pressure of testing probe, to adjust the applied pressure, so as to ensure the good contact between testing probe and device test point, and improve the testing accuracy of semiconductor device.

Description

Technical Field

[0001] This invention relates to the field of semiconductor device testing and testing probe technology, specifically to a semiconductor device testing probe and a semiconductor device testing method. Background Technology

[0002] A test probe is a high-precision, high-reliability miniature conductive component that serves as a bridge connecting semiconductor devices and testing equipment. Test probes can form a stable electrical connection with test points, used to verify and test the performance of semiconductor devices. During the testing process of semiconductor devices, the contact condition between the test probe and the test point directly affects the accuracy of the test results.

[0003] Currently, contact quality is mainly ensured by controlling the pressure applied to the test probe. However, applying the right pressure is crucial. Excessive pressure may ensure contact, but it may damage the test probe or semiconductor device. Insufficient pressure will result in unstable contact between the test probe and the test point on the semiconductor, i.e., poor contact, which will affect the test accuracy of the semiconductor device. Summary of the Invention

[0004] To address the technical problem of poor accuracy in semiconductor device testing caused by inappropriate pressure applied by the test probe, the present invention aims to provide a semiconductor device test probe and a semiconductor device testing method thereof. The specific technical solution adopted is as follows:

[0005] A semiconductor device testing method using a semiconductor device test probe, the method comprising:

[0006] During semiconductor device testing, the contact resistance between the test probe and the device test point is collected in real time to obtain the pressure applied by the test probe at each test monitoring moment.

[0007] At each test monitoring moment, based on the distribution characteristics of contact resistance at all historical acquisition moments within its preset historical test period, all suspected abnormal contact moments within its preset historical test period and the contact stability coefficient of the test probe are obtained. Based on the contact resistance at each suspected abnormal contact moment and the change characteristics of the contact resistance within its preset time domain, all abnormal contact moments under each abnormal contact type are determined from all suspected abnormal contact moments.

[0008] Within the preset historical test period at each test monitoring moment, based on the total number of abnormal contact types and the distribution of contact resistance variation characteristics within the preset time domain at each abnormal contact moment under each abnormal contact type, and combined with the contact stability coefficient, the pressure application abnormality coefficient of the test probe at each test monitoring moment is obtained.

[0009] At each test monitoring moment, the applied pressure of the test probe is adjusted according to the pressure anomaly coefficient of the test probe.

[0010] Furthermore, the method for obtaining the suspected abnormal contact time includes:

[0011] The preset historical test period is the corresponding time period between each test monitoring time and the previous adjacent test monitoring time; at each test monitoring time, based on the fluctuation characteristics of the contact resistance at all historical acquisition times within its preset historical test period, all contact resistances to be analyzed are determined, and all contact resistances to be analyzed are density-clustered to obtain all clusters;

[0012] The cluster with the highest density is taken as the target cluster, and all the historical acquisition times corresponding to the contact resistance to be analyzed in the remaining clusters outside the target cluster are taken as suspected abnormal contact times.

[0013] Furthermore, the method for obtaining the contact resistance to be analyzed includes:

[0014] Based on the contact resistance fitting resistance change curves at all historical acquisition times within a preset historical test period, the contact resistance corresponding to each maximum value in the resistance change curve is taken as the contact resistance to be analyzed.

[0015] Furthermore, the method for obtaining the contact stability coefficient includes:

[0016] Using the second densest cluster as a reference cluster, the contact stability coefficient of the test probe is obtained based on the density difference between the target cluster and the reference cluster, as well as the difference in the number of contact resistors to be analyzed within the corresponding cluster.

[0017] Furthermore, the methods for obtaining all abnormal contact moments under each abnormal contact type include:

[0018] For each suspected abnormal contact moment, the time interval between the corresponding moment of the previous adjacent minimum contact resistance and the next adjacent minimum contact resistance is taken as a preset time domain. Within the preset time domain, the time interval to the left of the suspected abnormal contact moment time sequence is taken as the abnormal start time interval, and the time interval to the right of the suspected abnormal contact moment time sequence is taken as the abnormal fallback time interval.

[0019] Based on the contact resistance at each suspected abnormal contact moment, the changes in contact resistance during the abnormal start period and the abnormal fallback period, a resistance change feature vector is determined for each suspected abnormal contact moment.

[0020] Based on the similarity features of the resistance change feature vector, all abnormal contact moments are selected from all suspected abnormal contact moments, and all abnormal contact moments are clustered to obtain all moment clusters. All abnormal contact moments in the same moment cluster correspond to the same abnormal contact type.

[0021] Furthermore, the method for obtaining the resistance change feature vector includes:

[0022] The rate of change of contact resistance during the initial period of an anomaly is used as the characteristic parameter of resistance abrupt change; the product of the difference between the previous and next adjacent minimum contact resistances within a preset time domain and the duration of the anomaly fallback period is used as the characteristic parameter of resistance fallback; the contact resistance, the characteristic parameter of resistance abrupt change, and the characteristic parameter of resistance fallback at the suspected abnormal contact moment are normalized and then used as vector elements to construct a resistance change characteristic vector.

[0023] Furthermore, the method for obtaining the abnormal contact time includes:

[0024] Based on the resistance change feature vector, the LOF value of each suspected abnormal contact moment is obtained; suspected abnormal contact moments with LOF values ​​greater than a preset threshold are taken as abnormal contact moments.

[0025] Furthermore, the method for obtaining the pressure application anomaly coefficient includes:

[0026] The first abnormal parameter is obtained by combining the negative correlation mapping result of the contact stability coefficient with the total number of abnormal contact types;

[0027] Based on the discreteness of the resistance drop characteristic parameter at all abnormal contact times under each abnormal contact type, and the similarity of the resistance distribution between the contact resistance to be analyzed at all abnormal contact times under each abnormal contact type and the contact resistance to be analyzed in the target cluster, the second abnormal parameter is obtained.

[0028] By combining the first abnormal parameter and the second abnormal parameter, the pressure application abnormality coefficient is obtained.

[0029] Furthermore, adjusting the applied pressure of the test probe includes:

[0030] At each test monitoring time, the pressure adjustment weight is determined according to the pressure application anomaly coefficient, and the difference between the applied pressure at each test monitoring time and the previous adjacent test monitoring time is taken as the pressure change amplitude.

[0031] The pressure change amplitude is weighted using the pressure adjustment weight, and the weighted result is used as the pressure adjustment value; the pressure adjustment value is added to the applied pressure at the test monitoring time to obtain the applied pressure at the next test monitoring time.

[0032] A semiconductor device test probe includes a test probe body and a test control system, wherein the system implements the steps of the semiconductor device test method using the semiconductor device test probe.

[0033] The present invention has the following beneficial effects:

[0034] This invention, during semiconductor device testing, real-time acquisition of the contact resistance between the test probe and the device test point, and acquisition of the applied pressure of the test probe at each test monitoring moment, prepares for subsequent analysis of the pressure application effect and adjustment of the applied pressure; then, at each test monitoring moment, based on the distribution characteristics of the contact resistance at all historical acquisition moments within a preset historical test period, all suspected abnormal contact moments within that preset historical test period are acquired, along with the contact stability coefficient reflecting the contact state of the test probe. Finally, the contact resistance at each suspected abnormal contact moment and its variation characteristics within a preset time domain are analyzed to evaluate the contact resistance at each suspected abnormal contact moment. The invention analyzes the contact state to determine all abnormal contact moments under each abnormal contact type. Then, within a preset historical test period at each test monitoring moment, it analyzes the variability or instability of abnormal contact states based on the total number of abnormal contact types. Furthermore, it analyzes the distribution of contact resistance variation characteristics within a preset time domain for each abnormal contact type, assesses the systematic nature of contact defects by analyzing the similarity of contact resistance variation characteristics between different abnormal contact states, and combines this with a contact stability coefficient to obtain a contact state stability reference. Finally, it adjusts the applied pressure of the test probe based on this pressure application anomaly coefficient. This invention assesses the contact state by analyzing the variation characteristics of contact resistance, and then evaluates abnormalities in the applied pressure of the test probe to adjust the applied pressure, thereby ensuring a good contact state between the test probe and the device test point and improving the testing accuracy of semiconductor devices. Attached Figure Description

[0035] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1This is a flowchart of a semiconductor device testing method using a semiconductor device test probe, provided in one embodiment of the present invention.

[0037] Figure 2 This is a flowchart illustrating a method for obtaining all abnormal contact moments under each abnormal contact type, as provided in an embodiment of the present invention.

[0038] Figure 3 This is a flowchart illustrating a method for obtaining an anomaly coefficient of pressure application, as provided in one embodiment of the present invention. Detailed Implementation

[0039] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a semiconductor device test probe and its semiconductor device testing method according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0040] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0041] The following description, in conjunction with the accompanying drawings, details a specific scheme for a semiconductor device test probe and its semiconductor device testing method provided by the present invention.

[0042] In the semiconductor device testing process, this invention collects the contact resistance between the test probe and the device test point in real time, and then analyzes the historical changes in the contact resistance at each test monitoring moment to evaluate the contact effect. In this way, the pressure applied by the test probe is adjusted at each test monitoring moment to ensure the test effect.

[0043] Please see Figure 1 The diagram illustrates a semiconductor device testing method using a semiconductor device test probe according to an embodiment of the present invention, specifically including:

[0044] Step S1: During the semiconductor device testing process, the contact resistance between the test probe and the device test point is collected in real time to obtain the pressure applied by the test probe at each test monitoring moment.

[0045] To facilitate understanding of the solution, the configuration of the semiconductor device testing equipment and the testing procedures are briefly described below:

[0046] The test probe is mounted on the test head of the test device, and the semiconductor device to be tested is placed on the test stage. The test head and test probe are moved by the motion platform of the test device so that the bottom of the test probe is aligned with the test point of the semiconductor device. Then the test probe is moved down to gradually approach the test point of the semiconductor device. When the test probe first contacts the test point, the power-on test begins.

[0047] To analyze whether the test probe and the device test point are in good contact, i.e. whether the test results are accurate, one embodiment of the present invention collects the contact resistance between the test probe and the device test point in real time. The higher the contact resistance, the greater the possibility of poor contact, which prepares for evaluating the contact effect at each test monitoring moment. At the same time, the pressure applied to the test probe is also collected at each test monitoring moment so that the applied pressure can be adjusted periodically to ensure good contact.

[0048] It should be noted that the test duration varies for different semiconductor devices, and implementers can adjust the sampling frequency of contact resistance and applied pressure according to the actual situation. Since the test process is the same for each semiconductor device, the analysis and description will be based on any semiconductor device to be tested. Considering the large amount of data collected, implementers can upload the real-time collected contact resistance and applied pressure to the data management platform for analysis and calculation via wired or wireless communication.

[0049] In this example, a semiconductor device with a total test duration of 30 minutes (since power-on testing) is used as an example. A single probe test is performed on it, and the contact resistance acquisition frequency is set to 10Hz to obtain the contact resistance at each acquisition moment. The test monitoring frequency is set to once per minute to obtain the applied pressure at each test monitoring frequency, and the applied pressure of the test probe is analyzed and adjusted at each test monitoring moment.

[0050] Among them, a pressure sensor is installed between the probe rod and the probe head of the test probe, which can be used to collect the pressure applied by the test probe to the test point of the device in real time; the contact resistance between the test probe and the test point of the device can be read in real time by the test device; the acquisition of contact resistance and applied pressure are existing technologies and will not be described in detail.

[0051] Meanwhile, in one embodiment of the present invention, after power-on testing, the pressure applied to the test point of the device by the test probe is controlled by the test device to gradually increase from 0 until a preset pressure is reached; wherein the preset pressure is the critical pressure value that the test probe will damage the semiconductor device. The implementer can adjust the rate of change of the applied pressure according to the test duration of the semiconductor device, for example, so that the critical pressure value is reached when the test duration is more than halfway through and then remains unchanged, wherein the applied pressure remains unchanged during the time period between two adjacent test monitoring moments, and the implementer can also adjust the pressure application scheme by himself.

[0052] It should be noted that the pressure application scheme is only a preset scheme before the test, and the applied pressure may be adjusted in real time according to the actual situation during the subsequent test.

[0053] Step S2: At each test monitoring time, based on the distribution characteristics of contact resistance at all historical acquisition times within its preset historical test period, obtain all suspected abnormal contact times within its preset historical test period, as well as the contact stability coefficient of the test probe. Based on the contact resistance at each suspected abnormal contact time and the change characteristics of the contact resistance within its preset time domain, determine all abnormal contact times under each abnormal contact type from all suspected abnormal contact times.

[0054] It should be noted that the analysis method for abnormal contact between the test probe and the device test point, as well as the method for adjusting the applied pressure of the test probe, are the same at each test monitoring moment during semiconductor device testing. Here, we will only take any test monitoring moment as an example for analysis and description, and will not go into detail again.

[0055] Considering that when the test probe and the device test point have good contact, the measured contact resistance should be relatively stable. Therefore, the more concentrated the distribution of contact resistance collected during the historical test period corresponding to the test monitoring time, the greater the possibility that the test probe and the device test point have stable contact during that historical test period. Conversely, it indicates that the contact is not stable enough and the possibility of poor contact is relatively greater.

[0056] Therefore, in each test monitoring moment, the present invention first obtains the contact stability coefficient of the test probe and all suspected abnormal contact moments within the preset historical test period based on the distribution characteristics of the contact resistance at all historical acquisition moments within the preset historical test period. This provides a certain reference for subsequent analysis of the pressure application anomaly coefficient of the test probe and adjustment of the applied pressure.

[0057] It should be noted that the relative fluctuation of the contact resistance at suspected abnormal contact times may be due to different sampling times under different contact failure conditions, or it may be random noise. Stable contact resistance does not necessarily indicate good contact performance; it may indicate stable contact failure. Therefore, the contact stability coefficient does not directly reflect a good contact effect between the test probe and the device test point or that the pressure applied by the test probe is reasonable. It only represents a relatively stable contact state.

[0058] Considering the large amount of contact resistance data, and the fact that poor contact can cause a sudden increase in resistance, analyzing the changes in the maximum value of contact resistance can also help assess the rationality of the pressure applied by the test probe. Therefore, all contact resistances to be analyzed can be determined based on the temporal fluctuation characteristics of the contact resistance. The contact resistances to be analyzed are those under the condition of a sudden increase, which may be the contact resistances collected under the condition of suspected poor contact, and may also contain abnormal data such as noise.

[0059] Furthermore, considering that density-based clustering can help analyze the similarity of data clusters, the contact resistance to be analyzed in the cluster with the highest density reflects its relatively stable and concentrated characteristics, while the contact resistance to be analyzed in the other clusters can be considered as abnormal contact resistance with certain fluctuation and deviation characteristics, thereby determining all suspected abnormal contact moments.

[0060] Based on this, in a preferred embodiment of the present invention, the method for obtaining the suspected abnormal contact time includes:

[0061] The preset historical test period is the corresponding time period between each test monitoring time and the previous adjacent test monitoring time; at each test monitoring time, based on the fluctuation characteristics of the contact resistance at all historical acquisition times within its preset historical test period, all contact resistances to be analyzed are determined, and all contact resistances to be analyzed are density clustered to obtain all clusters;

[0062] The cluster with the highest density is selected as the target cluster, and all contact resistances to be analyzed in the remaining clusters outside the target cluster are selected as suspected abnormal contact times corresponding to the historical acquisition times.

[0063] As an example, the time period between each test monitoring time and the previous adjacent test monitoring time is first used as a preset historical test period; then, all contact resistances to be analyzed within each preset historical test period are obtained.

[0064] In a preferred embodiment of the present invention, the method for obtaining the contact resistance to be analyzed includes:

[0065] Based on the contact resistance fitting curves at all historical acquisition times within the preset historical test period, the contact resistance corresponding to each maximum value in the resistance variation curve is taken as the contact resistance to be analyzed; specifically, the resistance variation curves of the contact resistance during the preset historical test period are fitted based on the least squares method, and the determination of the maximum value is a well-known technique, which will not be elaborated here.

[0066] Then, the DBSCAN clustering algorithm is used to cluster all the contact resistances to be analyzed within the preset historical test period to obtain all clusters. The contact resistances to be analyzed in each cluster have certain similarity and concentration characteristics, which can help evaluate the distribution of the contact resistances to be analyzed. The DBSCAN clustering algorithm and the determination of the density of each cluster are existing technologies. Implementers can also use other density clustering algorithms, which will not be elaborated here.

[0067] Finally, the cluster with the highest density was selected as the target cluster, and all contact resistances to be analyzed in the other clusters were selected as suspected abnormal contact times corresponding to the historical acquisition times.

[0068] Preferably, in one embodiment of the present invention, considering that the density of the obtained clusters reflects the concentration characteristics of the contact resistance to be analyzed, when there is no stable contact relationship between the test probe and the device test point, the fluctuation of the contact resistance to be analyzed is relatively large, exhibiting relatively discrete and random characteristics, and the density of the contact resistance to be analyzed in each cluster will be relatively small; while the target cluster, as the cluster with a relatively large degree of aggregation among all clusters, indicates that the density of the target cluster is also small when the density difference between the other clusters and the target cluster is small, and the contact between the test probe and the device test point is unstable; conversely, when the density of the target cluster is relatively large and the number of contact resistances to be analyzed within it is large, it indicates that the contact between the test probe and the device test point is stable; then the method for obtaining the contact stability coefficient includes:

[0069] The second densest cluster is used as the reference cluster. Based on the density difference between the target cluster and the reference cluster, as well as the difference in the number of contact resistances to be analyzed within the corresponding cluster, the contact stability coefficient of the test probe is obtained.

[0070] Specifically, the cluster with the second highest density is used as the reference cluster and compared with the target cluster. The difference between the density of the target cluster and the density of the reference cluster is mapped to the sigmoid function and the range is adjusted to obtain the density deviation parameter. Similarly, the difference between the total number of contact resistors to be analyzed in the target cluster and the total number of contact resistors to be analyzed in the reference cluster is mapped to the sigmoid function and the range is adjusted to obtain the quantity deviation parameter. Then, the density deviation parameter and the quantity deviation parameter are multiplied and fused to obtain the contact stability coefficient of the test probe.

[0071] In other embodiments, the implementer may also fuse the density deviation parameter and the quantity deviation parameter by means of weighted summation, or use other mapping functions to adjust the value range to avoid negative differences that would affect the calculation.

[0072] Considering that the suspected abnormal contact moments may also include random noise corresponding to the acquisition moments, the corresponding (to be analyzed) contact resistance may have similar fluctuation deviation characteristics as the (to be analyzed) contact resistance acquired under actual contact failure conditions, which will affect the subsequent contact failure analysis results, and may in turn affect the subsequent adjustment effect of the pressure applied to the test probe.

[0073] Furthermore, considering that the short-term changes in contact resistance are usually different under different degrees of poor contact, for example, the magnitude of the contact resistance may be different, and thus the change characteristics of the contact resistance in the preset time domain are not consistent; at the same time, the change characteristics of the contact resistance in the preset time domain corresponding to noise are also somewhat different from the change characteristics of the contact resistance in the preset time domain corresponding to poor contact.

[0074] Based on this, all abnormal contact moments under each abnormal contact type can be determined from all suspected abnormal contact moments according to the contact resistance at each suspected abnormal contact moment and the change characteristics of the contact resistance in the preset time domain; each abnormal contact type represents a poor contact situation of different degree or state, providing a basis for subsequent analysis of contact effect.

[0075] Preferably, please refer to Figure 2 It illustrates a flowchart of a method for obtaining all abnormal contact moments under each abnormal contact type according to an embodiment of the present invention, specifically including:

[0076] Step S201: For each suspected abnormal contact moment, the time period between the corresponding moment of the previous adjacent minimum contact resistance and the next adjacent minimum contact resistance is taken as a preset time domain. Within the preset time domain, the time period to the left of the suspected abnormal contact moment time sequence is taken as the abnormal start time period, and the time period to the right of the suspected abnormal contact moment time sequence is taken as the abnormal fallback time period.

[0077] Considering that each suspected abnormal contact moment can be regarded as the peak moment in a poor contact event or noise event, and by analyzing the change characteristics of contact resistance in each poor contact event or noise event, it can help determine the attribute of the suspected abnormal contact moment, that is, whether it is a real poor contact or noise. Therefore, the preset time domain of each suspected abnormal contact moment can be determined first, and the preset time domain corresponds to the time period of a poor contact event or noise event.

[0078] Specifically, in the resistance change curve corresponding to the contact resistance within the preset historical test period, all minimum values ​​(minimum contact resistance) are determined, and the time period between the time corresponding to the previous adjacent minimum contact resistance and the time corresponding to the next adjacent minimum contact resistance at each suspected abnormal contact moment is taken as the preset time domain.

[0079] Within the preset time domain, the time period between the acquisition time corresponding to the previous adjacent minimum contact resistance and the suspected abnormal contact time (the time period to the left of the suspected abnormal contact time sequence) is the abnormal start time period, used to analyze the sudden change trend of contact resistance; the time period between the suspected abnormal contact time and the acquisition time corresponding to the previous adjacent minimum contact resistance (the time period to the right of the suspected abnormal contact time sequence) is the abnormal fall-off time period, used to analyze the fall-off trend of contact resistance.

[0080] Step S202: Based on the contact resistance at each suspected abnormal contact moment, the changes in contact resistance during the abnormal start period and the abnormal fallback period, determine the resistance change feature vector at each suspected abnormal contact moment.

[0081] Considering that the higher the contact resistance, the greater the possibility and severity of poor contact;

[0082] Furthermore, considering the rate of change of contact resistance during the initial period of an anomaly, it can reflect the change in the contact state between the test probe and the test point of the device to a certain extent; the faster the rate of change, the more likely there is a sudden contact failure such as probe bouncing, and conversely, the more likely there is a gradual contact failure such as probe slippage; at the same time, the rate of change may also characterize the degree of abrupt change in noise.

[0083] Furthermore, considering that the change in contact resistance within the preset time domain corresponding to each suspected abnormal contact moment can be regarded as a pulse, the shorter the duration of the change in contact resistance during the abnormal fallback period, and the smaller the difference between the minimum contact resistance at the end of the abnormal fallback period and the minimum contact resistance at the beginning of the abnormal start period, the more obvious the short-term spike characteristics within the preset time domain, and the more likely it is to be pulse noise; at the same time, it can also reflect the relevant state characteristics of poor contact.

[0084] Based on this, in a preferred embodiment of the present invention, the method for obtaining the resistance change feature vector includes:

[0085] The rate of change of contact resistance during the initial period of an anomaly is used as the characteristic parameter of resistance abrupt change; the product of the difference between the previous and next adjacent minimum contact resistances within a preset time domain and the duration of the anomaly fallback period is used as the characteristic parameter of resistance fallback; the contact resistance, resistance abrupt change characteristic parameter, and resistance fallback characteristic parameter at the suspected abnormal contact moment are normalized and then used as vector elements to construct a resistance change characteristic vector.

[0086] It should be noted that the difference is represented by the absolute value of the difference, thereby calculating the resistance drop characteristic parameter. Since the value ranges of contact resistance, resistance change characteristic parameter, and resistance drop characteristic parameter may vary greatly, which may affect the subsequent similarity analysis and clustering results, it is also necessary to linearly normalize the vector elements in each dimension. For example, for contact resistance, linear normalization is performed on each contact resistance in the dimension of contact resistance at all suspected abnormal contact times within the preset historical test period. The calculation of this and the rate of change is a well-known technique and will not be elaborated further.

[0087] Step S203: Based on the similarity features of the resistance change feature vector, all abnormal contact moments are selected from all suspected abnormal contact moments, and all abnormal contact moments are clustered to obtain all moment clusters. All abnormal contact moments in the same moment cluster correspond to the same abnormal contact type.

[0088] The resistance change feature vector characterizes the poor contact state or noise change characteristics at the corresponding suspected abnormal contact time. However, poor contact may occur frequently or repeatedly in time, while noise is usually more random. Based on this, we can further screen all abnormal contact times from all suspected abnormal contact times based on the similarity characteristics of the resistance change feature vector, thereby eliminating some noise interference to a certain extent. Abnormal contact times can be regarded as the collection times corresponding to poor contact.

[0089] In a preferred embodiment of the present invention, considering that the LOF (Local Outlier Factor) algorithm is an unsupervised learning algorithm for anomaly detection, the larger the LOF value, the more likely it is to be an anomalous outlier; the method for obtaining the anomalous contact time includes: obtaining the LOF value of each suspected anomalous contact time based on the resistance change feature vector; and taking the suspected anomalous contact time with the LOF value greater than a preset threshold as the anomalous contact time.

[0090] It should be noted that the application of the LOF algorithm and the calculation of the LOF value are well-known technologies and will not be elaborated further; in this example, the preset threshold is set to 1.5, but implementers can adjust it themselves.

[0091] Furthermore, all abnormal contact moments can be clustered together, and abnormal contact moments with similar resistance change characteristics can be grouped into a cluster to characterize similar abnormal contact states or poor contact types.

[0092] Specifically, based on the resistance change feature vector at each abnormal contact moment, the DBSCAN clustering algorithm is used to cluster all abnormal contact moments to obtain several moment clusters, and each moment cluster corresponds to an abnormal contact type.

[0093] Step S3: Within the preset historical test period at each test monitoring time, based on the total number of abnormal contact types and the distribution of contact resistance variation characteristics within the preset time domain at each abnormal contact time under each abnormal contact type, and in conjunction with the contact stability coefficient, obtain the abnormal pressure application coefficient of the test probe at each test monitoring time.

[0094] Considering that within the preset historical test period during the test monitoring time, the more abnormal contact types there are, the more the contact failure situation is constantly changing, and the more abnormal the pressure applied by the test probe becomes, it indirectly indicates that the pressure applied by the test probe is too low and cannot guarantee a good contact state. At the same time, although the contact stability coefficient cannot directly characterize the degree of contact failure, it can provide the stability of the contact state between the test probe and the test point, and to a certain extent, it provides a certain reference for abnormal pressure application. Furthermore, the distribution of the contact resistance variation characteristics within the preset time domain at the time of abnormal contact under each abnormal contact type can provide relevant references for the contact failure state, and to a certain extent, it can help analyze and adjust the applied pressure of the test probe.

[0095] Based on this, the embodiments of the present invention can obtain the pressure application anomaly coefficient of the test probe at each test monitoring time. The larger the pressure application anomaly coefficient, the smaller the applied pressure of the test probe, which is insufficient to maintain a good contact state, thus providing a basis for subsequent adjustments.

[0096] Preferably, in one embodiment of the present invention, the method for obtaining the pressure application anomaly coefficient includes:

[0097] Please see Figure 3 The flowchart illustrates a method for obtaining a pressure application anomaly coefficient according to an embodiment of the present invention, specifically including:

[0098] Step S301: Combine the negative correlation mapping result of the contact stability coefficient with the total number of abnormal contact types to obtain the first abnormal parameter.

[0099] Based on the above logic, the first abnormal parameter of the pressure applied by the test probe can be determined firstly based on the contact stability coefficient and the number of abnormal contact types. Specifically, the contact stability coefficient is mapped to the exponential function exp(-x) with the natural constant e as the base to adjust the logic. Then, the mapping result is multiplied by the total number of abnormal contact types to obtain the first abnormal parameter.

[0100] In other embodiments of the present invention, the implementer may also use other negative correlation mapping methods, such as performing reciprocal operations.

[0101] Step S302: Based on the discreteness of the resistance drop characteristic parameters at all abnormal contact times under each abnormal contact type, and the similarity of the resistance distribution between the contact resistance to be analyzed at all abnormal contact times under each abnormal contact type and the contact resistance to be analyzed in the target cluster, obtain the second abnormal parameter.

[0102] Considering that the resistance drop characteristic parameter not only reflects the possibility of a real contact failure to a certain extent, but also characterizes the abnormal characteristics of the contact state, the more similar the distribution of the resistance drop characteristic parameter at all abnormal contact times in a certain abnormal contact type, the more it indicates that the contact state at all abnormal contact times has similar abnormal contact characteristics.

[0103] Furthermore, considering that the distribution characteristics of the contact resistance to be analyzed within the target cluster preliminarily characterize the distribution characteristics of the contact resistance to be analyzed under the most stable contact state, the more similar the distribution characteristics of the contact resistance to be analyzed under different abnormal contact types are to the corresponding distribution characteristics in the target cluster, the more likely there is a systematic poor contact between the test probe and the conductor test point, and the greater the possibility of abnormal pressure application; conversely, if the distribution characteristics are not similar, it may indicate that there may still be noise or interference, and the possibility of abnormal pressure application is relatively smaller.

[0104] Based on this, the second abnormal parameter can be obtained.

[0105] As an example, the discrete characteristics are first measured by the coefficient of variation. For each type of abnormal contact, the coefficient of variation of the resistance drop characteristic parameters at all abnormal contact times is negatively correlated and mapped. For example, the logic is adjusted in the exponential function exp(-x) with the natural constant e as the base. The smaller the coefficient of variation, the more concentrated the resistance drop characteristic parameters are, and the more likely the abnormal contact state is to be characterized as a systematic defect.

[0106] Then, the frequency distribution histograms of the contact resistance to be analyzed in the target cluster and the frequency distribution histograms of the contact resistance to be analyzed corresponding to the abnormal contact time under each abnormal contact type are obtained respectively. The frequency distribution change curves corresponding to each frequency distribution histogram are fitted (each data point in the frequency distribution change curve is the frequency of the contact resistance to be analyzed). The DTW similarity between the target cluster and the frequency distribution change curves corresponding to each abnormal contact type is further calculated. The larger the DTW similarity, the more similar the abnormal contact state under the abnormal contact type is to the contact state corresponding to the target cluster. The above operation methods are all well-known techniques and will not be described in detail.

[0107] For each type of abnormal contact, the negative correlation mapping result corresponding to the obtained coefficient of variation is multiplied and fused with the DTW similarity. The smaller the coefficient of variation and the larger the DTW similarity, the more likely the abnormal contact state is a systematic poor contact rather than an occasional one, and the more likely it is caused by abnormal pressure applied by the test probe. Finally, the products corresponding to all abnormal contact types are summed and the sum is used as the second abnormal parameter.

[0108] In another embodiment of the present invention, the implementer may also use the cluster density or clustering effect of the cluster at the time corresponding to each abnormal contact type to replace the negative correlation mapping result of the coefficient of variation, so as to characterize the similar stability characteristics of the contact state under the abnormal contact type; and may also use the normal distribution to replace the frequency distribution to evaluate the similarity characteristics between the contact state corresponding to the target cluster and the contact state under each abnormal contact type, thereby evaluating the systematic or repetitive nature of the contact anomaly rather than its randomness, and thus helping to assess the abnormal probability of applying pressure.

[0109] Step S303: Combine the first abnormal parameter and the second abnormal parameter to obtain the pressure application abnormality coefficient.

[0110] As an example, the first abnormal parameter is multiplied and fused with the second abnormal parameter to obtain the pressure application abnormality coefficient; in other examples, the implementer may also use weighted or summative methods to fuse them.

[0111] Step S4: At each test monitoring moment, adjust the applied pressure of the test probe according to the pressure application anomaly coefficient of the test probe.

[0112] Once the pressure application anomaly coefficient of the test probe is determined at each test monitoring time, the applied pressure can be further adjusted to ensure the contact effect and thus improve the accuracy of the test results.

[0113] Preferably, in one embodiment of the present invention, considering that a larger pressure application anomaly coefficient indicates that the applied pressure of the test probe is insufficient to guarantee a good contact state, thus leading to abnormal contact, it is necessary to further increase the applied pressure; based on this, adjusting the applied pressure of the test probe includes:

[0114] At each test monitoring time, the pressure adjustment weight is determined based on the pressure application anomaly coefficient, and the difference between the applied pressure at each test monitoring time and the corresponding applied pressure at the previous adjacent test monitoring time is taken as the pressure change amplitude.

[0115] The pressure change amplitude is weighted using pressure adjustment weights, and the weighted result is used as the pressure adjustment value. The pressure adjustment value is then added to the applied pressure at the test monitoring time to obtain the applied pressure at the next test monitoring time.

[0116] As an example, firstly, at each test monitoring moment, the pressure application anomaly coefficient is mapped to the adjustment range of the sigmoid function, and the constant 1 is added to the mapping result to obtain the pressure adjustment weight. Then, the difference between the applied pressure at each test monitoring moment and the corresponding applied pressure at the previous adjacent test monitoring moment is used as the pressure change amplitude. The pressure change amplitude is used as the adjustment benchmark and multiplied by the pressure adjustment weight to obtain the pressure adjustment value. The larger the pressure application anomaly coefficient, the greater the applied pressure is on the previous adjustment benchmark to ensure the contact pressure of the test probe to the device test point and ensure the contact effect. Finally, the pressure adjustment value is added to the applied pressure at the current test monitoring moment to obtain the applied pressure at the next test monitoring moment.

[0117] It should be noted that if the applied pressure at the next test monitoring time is greater than the critical applied pressure, the applied pressure at the next test monitoring time will be set to the critical applied pressure or a pressure value slightly less than the critical applied pressure to avoid damaging the semiconductor device.

[0118] In one embodiment of the present invention, after adjusting the applied pressure of the test probe at each test monitoring moment in real time, the changes in the pressure application anomaly coefficient of the test probe at recent test monitoring moments can be further analyzed. For example, when the pressure application anomaly coefficient at the last 5 test monitoring moments is less than a preset threshold such as 0.3, and the pressure application anomaly coefficient obtained after each adjustment of the applied pressure does not fluctuate much, for example, when the negative correlation normalization value of the coefficient of variation is greater than a set threshold such as 0.9, it indicates that the adjustment of the applied pressure or the relationship between the test probe and the test point has reached a stable state, and the adjustment can be stopped to stably perform semiconductor device testing.

[0119] The present invention also proposes a semiconductor device test probe, including a test probe body and a test control system. The test control system is used to control the test probe on the test device and can realize the semiconductor device test method of the semiconductor device test probe described in steps S1-S4 above, so as to automatically adjust the pressure applied by the test probe to the test point of the device, thereby performing semiconductor device testing.

[0120] In summary, this invention first acquires the contact stability coefficient of the test probe at each test monitoring moment during the semiconductor device testing process, along with all suspected abnormal contact moments within a preset historical test period. Based on the contact resistance at each suspected abnormal contact moment and its variation characteristics within a preset time domain, it identifies all abnormal contact moments under each abnormal contact type from all suspected abnormal contact moments. Then, based on the contact stability coefficient, the total number of abnormal contact types, and the distribution of contact resistance variation characteristics within a preset time domain for each abnormal contact type, it acquires the pressure application anomaly coefficient of the test probe and adjusts the applied pressure of the test probe. This invention assesses the contact state by analyzing the variation characteristics of contact resistance, and then evaluates the abnormality of the applied pressure of the test probe to adjust the applied pressure, thereby ensuring a good contact state between the test probe and the device test point and improving the testing accuracy of semiconductor devices.

[0121] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0122] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A semiconductor device testing method of a semiconductor device testing probe, characterized by, The method comprises: During the semiconductor device testing process, the contact resistance between the test probe and the device test point is collected in real time to obtain the applied pressure of the test probe at each test monitoring time; At each test monitoring time, the distribution central feature of the contact resistance at all historical collection times in the preset historical test period is obtained to obtain all suspected abnormal contact times in the preset historical test period and the contact stability coefficient of the test probe, and the contact resistance at each suspected abnormal contact time and the change feature of the contact resistance in the preset time domain of each suspected abnormal contact time are used to determine all abnormal contact times of each abnormal contact type from all suspected abnormal contact times; In the preset historical test period of each test monitoring time, the total number of abnormal contact types and the distribution of the change feature of the contact resistance in the preset time domain of the abnormal contact time of each abnormal contact type are obtained in combination with the contact stability coefficient to obtain the pressure application abnormality coefficient of the test probe at each test monitoring time; At each test monitoring time, the applied pressure of the test probe is adjusted according to the pressure application abnormality coefficient of the test probe; The method for obtaining the suspected abnormal contact time comprises: The preset historical test period is the corresponding period between each test monitoring time and the adjacent test monitoring time; at each test monitoring time, the fluctuation feature of the contact resistance at all historical collection times in the preset historical test period is used to determine all to-be-analyzed contact resistances, and the density clustering is performed on all to-be-analyzed contact resistances to obtain all clusters; The cluster with the largest density is taken as a target cluster, and the corresponding historical collection time of all to-be-analyzed contact resistances in the remaining clusters outside the target cluster is taken as a suspected abnormal contact time; The method for obtaining the contact stability coefficient comprises: The cluster with the second largest density is taken as a reference cluster, the difference between the density of the target cluster and the density of the reference cluster is mapped into a sigmoid function to obtain a density deviation parameter; the difference between the total number of to-be-analyzed contact resistances in the target cluster and the total number of to-be-analyzed contact resistances in the reference cluster is mapped into a sigmoid function to obtain a quantity deviation parameter; and the density deviation parameter and the quantity deviation parameter are multiplied to obtain the contact stability coefficient of the test probe; The method for obtaining the pressure application abnormality coefficient comprises: The contact stability coefficient is mapped into an exponential function exp(-x) with a natural constant e as the base number, the negative correlation mapping result of the contact stability coefficient is multiplied by the total number of abnormal contact types to obtain a first abnormal parameter; For each abnormal contact type, the variation coefficient of the resistance drop characteristic parameter of all abnormal contact time points corresponding thereto is mapped into an exponential function exp(-x) with a natural constant e as a base number; then a frequency distribution histogram of the contact resistance to be analyzed in the target cluster is obtained, and a frequency distribution histogram of the contact resistance to be analyzed corresponding to the abnormal contact time points under each abnormal contact type is obtained, and a frequency distribution change curve corresponding to each frequency distribution histogram is fitted, each data point in the frequency distribution change curve being the frequency of the contact resistance to be analyzed; further, the DTW similarity between the target cluster and the frequency distribution change curve corresponding to each abnormal contact type is calculated; for each abnormal contact type, the mapping result of the variation coefficient is multiplied by the DTW similarity, and the products corresponding to all abnormal contact types are accumulated, and the accumulation sum is taken as a second abnormal parameter; Fusion of the first abnormal parameter and the second abnormal parameter obtains a pressure application abnormality coefficient.

2. The semiconductor device testing method of claim 1, wherein The method for obtaining the contact resistance to be analyzed comprises: Based on the resistance change curve fitted based on the contact resistance at all historical acquisition time points in a preset historical test period, each maximum value of the resistance change curve is taken as the contact resistance to be analyzed.

3. The semiconductor device testing method of claim 1, wherein, The method for obtaining all abnormal contact time points under each abnormal contact type comprises: For each suspected abnormal contact time point, the time period between the time point of the previous adjacent minimum contact resistance and the time point of the next adjacent minimum contact resistance is taken as a preset time domain, and in the preset time domain, the time period on the left side of the suspected abnormal contact time point is taken as an abnormal start time period, and the time period on the right side of the suspected abnormal contact time point is taken as an abnormal drop time period; According to the contact resistance of each suspected abnormal contact time point, the change of the contact resistance in the abnormal start time period and the abnormal drop time period, a resistance change feature vector of each suspected abnormal contact time point is determined; Based on the similar features of the resistance change feature vectors, all abnormal contact time points are screened from all suspected abnormal contact time points, and all the abnormal contact time points are clustered to obtain all time clusters, and all the abnormal contact time points in the same time cluster correspond to the same abnormal contact type.

4. The semiconductor device testing method of claim 3, wherein, The method for obtaining the abnormal contact time point comprises: Based on the resistance change feature vector, the LOF value of each suspected abnormal contact time point is obtained, and the suspected abnormal contact time point with an LOF value greater than a preset threshold value is taken as an abnormal contact time point.

5. The semiconductor device testing method of claim 1, wherein, The adjustment of the application pressure of the test probe comprises: At each test monitoring time point, a pressure adjustment weight is determined according to the pressure application abnormality coefficient, and the difference between the application pressure corresponding to each test monitoring time point and the application pressure corresponding to the previous adjacent test monitoring time point is taken as a pressure change amplitude; The pressure change amplitude is weighted by using the pressure adjustment weight, and the weighted result is taken as a pressure adjustment value; the application pressure at the next test monitoring time point is obtained by adding the pressure adjustment value to the application pressure at the test monitoring time point.

6. A semiconductor device test probe comprising a test probe body, characterised by, Further comprising a test control system, the system implements the steps of the semiconductor device test method of the semiconductor device test probe according to any one of claims 1-5.

Images