Wafer chip detection system and method

Abstract

The invention relates to the technical field of wafer chip detection, in particular to a wafer chip detection system and method. The method comprises the following steps: fixing a wafer chip to be detected on a bearing tray by using a locking mechanism, and scanning the surface area of the wafer chip line by line at a preset illumination angle to generate a corresponding reflected light intensity distribution sequence; when a continuous fluctuation section appears in the reflected light intensity distribution sequence, extracting a boundary curve of the section, judging whether the extension length and curvature change of the boundary curve meet crack characteristic conditions or not, and after crack judgment is completed, controlling a probe to make contact with an electrode bonding pad of a wafer chip, applying constant current, recording voltages at the two ends and calculating a contact resistance value; the contact resistance value is compared with a preset threshold value, and if the contact resistance value exceeds the threshold value, the contact resistance value is marked as an electrode failure risk point; according to the invention, by detecting the wafer chip, the joint detection of the wafer crack and the electrode failure is realized, and the detection accuracy and reliability are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wafer chip detection, and in particular to a wafer chip detection system and method. BACKGROUND

[0002] As an important basis for integrated circuit manufacturing, the quality of wafer chips is directly related to the performance and reliability of subsequent devices. During wafer processing and packaging, the wafer surface is easily affected by stress, mechanical force and environmental factors, resulting in cracks. Once the micro cracks expand, it may cause chip functional failure or even whole piece scrap. In addition, the stability of the contact resistance of the electrode pad on the wafer as the key channel for external circuit connection is also an important indicator of chip reliability. In the existing detection technology, optical microscopes or laser interference are usually used to observe the wafer surface to determine whether there are crack defects, the detection range is limited and it is difficult to quantitatively identify micro cracks, which has the risk of missed detection and misjudgment. There is a lack of fine judgment means for abnormal distribution of electrode pad contact resistance, and it is difficult to realize unified evaluation of structural defects and electrical failure. SUMMARY

[0003] Therefore, it is necessary to provide a wafer chip detection system and method to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, a wafer chip detection method is applied to a wafer chip, the wafer chip is fixed on a bearing tray, the bearing tray includes a locking mechanism, and the method includes the following steps: Step S1: using the locking mechanism to fix the wafer chip to be tested on the bearing tray, and scanning the surface area of the wafer chip row by row under a preset illumination angle to generate a corresponding reflected light intensity distribution sequence; Step S2: when a continuous fluctuation section appears in the reflected light intensity distribution sequence, the boundary curve of the section is extracted, and it is determined whether the extension length and the curvature change of the boundary curve meet the crack characteristic condition. After the crack determination is completed, the probe contacts the electrode pad of the wafer chip, applies a constant current and records the voltage across the two ends, and calculates the contact resistance value; Step S3: comparing the contact resistance value with a preset threshold value, if the contact resistance exceeds the threshold value, marking it as an electrode failure risk point, obtaining the electrical determination result of the wafer chip, combining the crack determination result and the electrical determination result, and outputting the detection result of the wafer chip.

[0005] The present application also provides a wafer chip detection system applied to a wafer chip, the wafer chip is fixed on a bearing tray, the bearing tray includes a locking mechanism, and is used for executing the wafer chip detection method as described above. The wafer chip detection system includes: The receiving tray is equipped with a connecting component and a locking mechanism. The connecting component is equipped with a limit post and a limit hole. The limit post and the limit hole slide together. The locking mechanism is a buckle, a pull buckle or a magnetic buckle. The detection module is used to fix the wafer chip under test to the carrier tray using a locking mechanism, and scan the surface area of ​​the wafer chip line by line under a preset illumination angle to generate the corresponding reflected light intensity distribution sequence. The contact resistance value calculation module is used to extract the boundary curve of a continuous fluctuation segment when a continuous fluctuation segment appears in the reflected light intensity distribution sequence, determine whether the extension length and curvature change of the boundary curve meet the crack characteristic conditions, and after the crack determination is completed, control the probe to contact the electrode pad of the wafer chip, apply a constant current and record the voltage at both ends, and calculate the contact resistance value. The wafer chip threshold determination module compares the contact resistance value with a preset threshold. If the contact resistance exceeds the threshold, it is marked as a risk point for electrode failure, and the electrical determination result of the wafer chip is obtained. Combining the crack determination result and the electrical determination result, the detection result of the wafer chip is output.

[0006] The beneficial effects of this invention are as follows: By scanning the surface of a wafer chip line by line under a preset illumination angle, a complete sequence of reflected light intensity distribution can be obtained, and crack features can be determined based on the boundary curves extracted from the fluctuating sections. Compared with traditional methods that rely on single-point detection or manual microscopic observation, this method can achieve large-area, automated scanning and analysis, avoiding missed detections and false judgments, ensuring high-precision identification of micro-cracks, and thus significantly improving the comprehensiveness and reliability of wafer structure defect detection.

[0007] By introducing an electrical testing step after crack detection, a probe is used to contact the wafer electrode pads and apply a constant current. The voltage across the probes is collected in real time to calculate the contact resistance, thereby quantitatively determining whether there is a risk of electrode failure. This method can not only identify potential failure points caused by material stress or process defects, but also complement the results of structural inspection, effectively avoiding the problem of missing electrode contact anomalies when relying solely on optical inspection, and achieving multi-dimensional quality assurance for wafer chips.

[0008] By combining crack detection results with electrical detection results, comprehensive testing of wafer chips, from structural integrity to electrical performance stability, can be achieved. This dual-detection mechanism effectively improves the accuracy and reliability of the detection results, enabling early detection of latent cracks and warnings of electrode contact failure risks, thus meeting the quality control requirements of wafer chips in high-reliability applications. Attached Figure Description

[0009] Fig. 1 This is a schematic diagram of the steps in a wafer chip inspection method; Fig. 2 This is a schematic diagram of the modules of a wafer chip inspection system; Fig. 3 Photo of a tray supporting wafer chips; Fig. 4 Photograph of a wafer chip structure; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0010] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0011] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0012] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0013] To achieve the above objectives, please refer to Figs. 1 to 4 A method for inspecting wafer chips, applied to wafer chips, wherein the wafer chips are fixed to a carrier tray, the carrier tray including a locking mechanism, the method comprising the following steps: Step S1: Fix the wafer chip to be tested onto the carrier tray using the locking mechanism, and scan the surface area of ​​the wafer chip line by line under the preset illumination angle to generate the corresponding reflected light intensity distribution sequence; Step S2: When a continuous fluctuation segment appears in the reflected light intensity distribution sequence, extract the boundary curve of the segment, determine whether the extension length and curvature change of the boundary curve meet the crack characteristic conditions, and after the crack determination is completed, control the probe to contact the electrode pad of the wafer chip, apply a constant current and record the voltage at both ends, and calculate the contact resistance value. Step S3: Compare the contact resistance value with a preset threshold. If the contact resistance exceeds the threshold, mark it as an electrode failure risk point and obtain the electrical judgment result of the wafer chip. Combine the crack judgment result with the electrical judgment result to output the detection result of the wafer chip.

[0014] In one embodiment, an adsorption locking mechanism in a batch flipping device is used to adjust the diameter... The wafer chips are fixed on the surface of the quartz tray, which has Flatness accuracy. Using wavelength. The laser light source, with the incident angle set as After being expanded by a flat-top lens, a line spot is formed, with a scanning step size of [missing information]. The scanner moves line by line along the X direction. The reflected signal is generated by a linear CCD (4096 pixels, sampling rate...). The system receives light and forms a one-dimensional light intensity sequence, which is then sequentially spliced ​​to obtain the reflected light intensity distribution sequence of the entire wafer surface. The difference between adjacent points in the light intensity distribution sequence is calculated; when the difference remains in the same direction for 15 consecutive sampling points, it is marked as a candidate fluctuation segment. This segment is then subjected to Bessel interpolation and weighted moving average to generate a smoothed boundary curve. The curvature sequence is obtained by calculating the tangent angle difference and the segment length ratio point by point, and compared with a preset crack threshold (curvature). And the extension length The comparison is performed, and if the conditions are met, the crack is confirmed. Then, the XYZ three-axis platform is controlled to move, driving the probe to press against the target electrode pad. Maintain contact under force and apply constant current, sampling frequency Record the voltage across both ends and take the average value, then calculate the contact resistance.

[0015] Query the current electrode's process standard threshold table: when contact resistance Determined to be normal. It was determined to be a warning. The system is identified as having a failure risk. After calculating the resistance, it is automatically categorized, and the pad status is marked in real time on the detection interface. If both a crack detection anomaly and a resistance exceeding the threshold are present, an alarm is immediately triggered on the host computer, and a non-conformance report is generated.

[0016] In another embodiment, a mechanical snap-locking mechanism is used, in conjunction with dual light sources, the first light source being... Ultraviolet LED, the second light source is Infrared laser, respectively with and Alternating incident angles are used to enhance the contrast between shallow cracks and deep defects. The reflected signals are received by a high-sensitivity CMOS camera at a frame rate of... The reflected light intensity distribution sequence is then stitched together in real time using an FPGA. Wavelet packet decomposition filtering is applied to the light intensity sequence to remove high-frequency noise before segment detection. The crack detection threshold is set to the curvature. And the extension length The probe uses a tungsten needle with a tip radius of... Before crimping, the position error is precisely calibrated by optical alignment to be less than [value missing]. Constant current source output current The voltage sampling window is The calculated contact resistance value is stored in the database and bound to the pad number.

[0017] The resistance threshold is dynamically adjusted based on the pad size and location, for example... × Small pad threshold set ,and × Large pad threshold set The test results are visualized in wafer coordinates, with cracked areas marked in red and electrical failure risk points marked in yellow. A test report is then generated and archived for production line quality traceability.

[0018] Of particular importance, step S1 includes: The locking mechanism is used to firmly press the wafer chip under test onto the positioning surface of the carrier tray; After the wafer chip is fixed, the incident angle of the light source is adjusted to a preset angle range; The driving scanning device moves the sampling points along the surface of the wafer chip row by row according to a predetermined path to obtain the reflected light intensity value of each sampling point; They are arranged sequentially to form a sequence of reflected light intensity distributions covering the surface area of ​​the wafer chip.

[0019] In one embodiment, the wafer chip to be tested is placed on the central positioning surface of the carrier tray, and a vacuum adsorption locking mechanism is activated to fix the wafer chip. The locking mechanism includes mechanical grippers at four corners and a central vacuum adsorption area. The mechanical grippers gently hold the edges of the wafer chip with a preset pressure to avoid damage to the chip surface. The central vacuum adsorption generates negative pressure to tightly adhere the wafer chip to the tray positioning surface, ensuring that the chip remains stable and does not shift during scanning. After the wafer chip is fixed, the incident angle of the laser light source is adjusted. The angle of the light source is set within a preset oblique incident angle range by a precision angle adjustment device. This angle range can effectively highlight the optical characteristics of surface defects. The two-dimensional scanning device is activated and moves row by row along the surface of the wafer chip from left to right and from top to bottom according to the grid path. The scanning device carries a photoelectric sensor that pauses briefly at each sampling point to collect the reflected light intensity signal at that point and convert it into a digital quantity. The entire wafer surface is divided into a dense sampling grid, with each row containing hundreds of sampling points, resulting in tens of thousands of reflected light intensity data points. The collected light intensity data are arranged sequentially according to the scanning path to form a one-dimensional reflected light intensity distribution sequence, which fully records the optical reflection characteristics of each region on the surface of the wafer chip.

[0020] In another embodiment, for large-size wafer chips, a partitioned locking strategy is adopted, dividing the locking mechanism into two independent systems: a central region and an edge region. The central region uses a flexible airbag clamping method, while the edge region uses multi-point mechanical grippers to ensure the flatness of the entire wafer chip surface meets inspection requirements. The light source adjustment employs automatic angle control, automatically optimizing the incident angle based on the wafer chip's material properties and surface roughness to achieve optimal defect detection. The scanning device is driven by a high-precision linear motor, and the scanning speed and stepping accuracy can be dynamically adjusted according to inspection requirements. During scanning, the dwell time at each sampling point is optimized to ensure both stable acquisition of the light intensity signal and overall inspection efficiency. The photoelectric sensor has a wide dynamic range, accurately capturing subtle changes in light intensity under different surface conditions. After scanning, the obtained reflected light intensity distribution sequence undergoes preliminary data processing to remove outliers and noise interference, forming high-quality raw data for subsequent defect identification and analysis.

[0021] Of particular importance is that the driving scanning device moves the sampling points row by row along the surface of the wafer chip according to a predetermined path, and obtains the reflected light intensity value of each sampling point, including: The drive scanning device moves the sampling point position line by line along the surface of the wafer chip according to the set travel path under the external control command; During the scanning process, the optical acquisition unit is triggered at each sampling point to acquire the light intensity signal reflected from the surface of the wafer chip; The collected light intensity signals are recorded sequentially as the light intensity values ​​of the corresponding sampling points according to the scanning order.

[0022] In one embodiment, a scan start command is sent to the motion controller. Upon receiving the command, the motion controller immediately drives the dual-axis precision linear motor assembly to begin executing a preset grid scanning path. The scanning device moves to the starting position at the upper left corner of the wafer chip surface, and then moves at a constant speed from left to right along the first row. Each time it reaches a preset sampling point, the motion controller sends a trigger signal to the optical acquisition unit. Upon receiving the trigger signal, the optical acquisition unit immediately activates the photoelectric converter to acquire the laser intensity signal reflected from the wafer chip surface at that moment. The photoelectric converter converts the light signal into a voltage signal, which is then converted into a digital quantity by an analog-to-digital converter. After the first row of scanning is completed, the scanning device automatically moves to the starting position of the next row and continues the scanning acquisition process from left to right, repeating this row-by-row until the entire wafer chip surface is covered. Throughout the scanning process, the digital light intensity value acquired at each sampling point is transmitted in real time to the data storage buffer and stored sequentially according to the scanning time order. The data storage automatically assigns a unique sequence number to each light intensity value to ensure accurate reconstruction of the spatial position information of each sampling point during subsequent data processing. After completing the acquisition of the last row, the scanning device automatically returns to the standby position.

[0023] In another embodiment, a closed-loop servo control strategy is employed to ensure high-precision positioning of sampling points, addressing the high-density sampling requirements. The dwell time of the scanning device at each sampling point is precisely controlled, ensuring sufficient signal integration time for the optical acquisition unit while avoiding excessive dwell time that could negatively impact overall scanning efficiency. The optical acquisition unit is equipped with a high-sensitivity photodiode and a low-noise signal amplifier, enabling accurate capture of subtle changes in surface-reflected light signals. During signal acquisition, the real-time position information and acquisition timestamp of the scanning device are recorded synchronously, establishing complete coordinates and time signatures for each light intensity value. The acquired light intensity signals are then processed by hardware filtering and software denoising before being stored in a high-speed data storage array in a fixed-precision format. The scanning path employs a serpentine reciprocating pattern to reduce idle travel time and improve scanning efficiency. The entire optical scanning acquisition process features real-time monitoring, enabling the detection of abnormal signals and timely adjustment of acquisition parameters. Upon completion of the scan, a data integrity report is automatically generated, confirming that the light intensity data of all sampling points has been correctly acquired and stored, providing reliable raw data support for subsequent image reconstruction and defect detection.

[0024] Preferably, step S2 includes: Retrieve continuous fluctuation segments in the reflected light intensity distribution sequence and mark the start and end points of the segments; Using the starting and ending points as boundaries, the light intensity curves of the segments are extracted, and the boundary curves are formed after smoothing out the noise. Measure the extension length of the boundary curve and establish a combined judgment condition for length and curvature; When the extension length exceeds the preset threshold and the curvature change meets the crack characteristic conditions, it is confirmed that there is a crack in the section. After crack detection is completed, control probes are pressed against electrode pads, a constant current is applied and the voltage at both ends is recorded, and the contact resistance value is calculated.

[0025] In one embodiment, after completing the wafer chip fixation and obtaining the reflected light intensity distribution sequence, a sliding window search is performed on the reflected light intensity distribution sequence. The window width is set to 20 sampling points, and the step size is 5 sampling points. If the light intensity fluctuation amplitude of more than 15 consecutive sampling points within a certain window is greater than the mean ± 2σ (σ is the global light intensity standard deviation), then the starting point of the window is recorded as the starting point of the fluctuation segment; when the fluctuation amplitude of consecutive sampling points falls back to the mean ± σ range and remains for more than 10 sampling points, it is recorded as the ending point. This yields the start and end positions of a fluctuation segment.

[0026] After the segment is marked, the light intensity curve corresponding to that segment is extracted, and the original curve is subjected to Savitzky-Golay smoothing filtering with a filter order of 3 and a window width of 11 points to remove random noise and preserve the curve trend. The smoothed boundary curve is obtained and stored for subsequent crack determination.

[0027] After the boundary curve is extracted, the extension length of the boundary curve is calculated using the arc length accumulation algorithm. Specifically, this involves calculating the coordinates of adjacent sampling points. and Calculate the increment ; Then sum over the entire curve segment to obtain the total extension length. Simultaneously, discrete curvature calculations are performed on the boundary curves, tangent directions are constructed at each sampling point, and the angle difference between adjacent tangents is calculated. , and then The ratio to the length of the corresponding arc segment As local curvature values, a curvature change sequence is obtained. The curvature change sequence is input into a preset combination judgment condition, for example: when And the range of curvature changes When the boundary curve is determined to conform to the characteristics of a crack, it is determined that the boundary curve conforms to the characteristics of a crack.

[0028] If the above conditions are met, the section is marked as a crack section, and the coordinates of the crack's initiation and termination points and its extension length are recorded. After crack identification is completed, the probe control module is activated, and the precision micro probe is used to... The pressure is pressed onto the surface of the target electrode pad. Then, a constant current source is applied. A constant current is applied, and the voltage value across the probe is acquired using a high-precision digital multimeter. Calculate electrode contact resistance and will The values ​​are saved to the detection database to support subsequent electrical determinations.

[0029] In another embodiment, it is assumed that the extension length of a certain fluctuation segment is... The curvature variation range is The condition was determined to be a crack after comparison with other conditions. Subsequently, a probe was pressed onto the corresponding electrode pad, and the voltage value was measured. ,exist The contact resistance is calculated under constant current. Higher than the preset threshold The result was marked as "crack + high-resistance risk pad" in the detection database.

[0030] Preferably, retrieving continuous fluctuation segments in the reflected light intensity distribution sequence and marking the start and end points of these segments includes: The reflected light intensity distribution sequence is arranged into a continuous data stream according to the sampling order, and the light intensity difference between adjacent sampling points is calculated in the continuous data stream; When the difference maintains the same direction of change across multiple consecutive sampling points, it is marked as a candidate fluctuation segment; Identify the first point at the front end of the fluctuation candidate segment where the difference deviates from the baseline threshold, and set it as the starting point; Identify the point at the end of the fluctuation candidate segment where the last difference falls back to the baseline threshold, and set it as the termination point; The continuous fluctuation section is determined by using the starting point and the ending point as boundaries.

[0031] In one embodiment, the reflected light intensity distribution sequence obtained by scanning the surface of a wafer chip is processed. The light intensity values ​​obtained by scanning line by line are arranged into a continuous data stream according to the sampling order to ensure that each data point corresponds to a specific location on the wafer surface. In the continuous data stream, the light intensity difference between adjacent sampling points is calculated sequentially, and the trend of light intensity change is determined based on the sign and magnitude of the difference. When the difference between multiple adjacent sampling points continuously changes in the same direction, such as continuously increasing or continuously decreasing, it is determined as a fluctuation candidate segment. At the beginning of the fluctuation candidate segment, the first point that significantly deviates from the background noise range is detected as the starting point. The background noise range here can be determined by statistically analyzing the normal fluctuation amplitude of most differences in the entire light intensity distribution sequence, generally taking its average fluctuation range plus a safety margin. When the light intensity difference of a point exceeds this range, it is considered that a valid fluctuation has occurred, and thus it is marked as the starting point. At the end of the fluctuation candidate segment, the change of the difference continues to be tracked until the light intensity difference gradually falls back into the normal fluctuation range. This point is marked as the termination point, indicating the end position of the fluctuation segment.

[0032] The marked start and end points are used as boundaries to define the range of the continuous fluctuation section. In subsequent processing, this section will be extracted to generate boundary curves, serving as candidate regions for determining the presence of cracks.

[0033] In another embodiment, when scanning a diameter When scanning wafer chips, press Light intensity data was collected at intervals. During processing, a continuous light intensity variation was observed in a certain area from point 300 to point 340. The starting point was marked at point 302, and the ending point was marked at point 338, corresponding to a physical length of approximately [missing information]. This section was subsequently recorded and used as a key area for inspection in subsequent crack detection steps.

[0034] Preferably, the process of extracting the light intensity curve of a segment using the starting and ending points as boundaries, and smoothing out curve noise to form the boundary curve includes: The light intensity data of the continuous fluctuation segment is extracted by using the starting and ending points of the segment as the index range, and the extracted data is arranged in the sampling order to form the original light intensity curve. The original light intensity curve is processed by moving average, and a low-pass filter is added after moving average to suppress local spike disturbances. The filtered curve is interpolated and resampled, and the resampled result is used as the boundary curve and stored.

[0035] In one embodiment, the data processing unit receives the start and end point information of a determined continuous fluctuation segment, and uses the start and end points of the segment as an index range to extract data for the corresponding segment from the reflected light intensity distribution sequence. The extracted data is arranged in the sampling order to form the original light intensity curve of the segment. This original curve can fully reflect the light intensity variation characteristics within the segment, but it usually contains spike disturbances introduced by optical noise, detector electrical interference, and random environmental factors.

[0036] A moving average process is performed on the original light intensity curve. In this process, the processing unit traverses the original curve point by point using a fixed-length sliding window and calculates the average light intensity of all sampling points within the window. This average value replaces the original value at the center point of the window. This method effectively reduces high-frequency noise in the curve while maintaining its overall trend.

[0037] After the moving average processing is completed, a low-pass filter operation is further applied to the smoothed curve. Specifically, the processing unit calls the built-in digital filtering module to perform secondary processing on the moving average result to suppress residual high-frequency spikes and local interference signals. The cutoff frequency of the low-pass filter is preset according to the sampling rate and the characteristic frequency band of the curve, ensuring that high-frequency disturbances are removed while preserving the main trend as much as possible.

[0038] Next, to avoid segmental discontinuities or excessive sparsity in the curve due to the limited number of sampling points, the data processing unit performs interpolation resampling on the filtered curve. The interpolation method uses piecewise cubic spline interpolation or linear interpolation, selected based on the actual smoothness of the curve. The interpolation resampling process generates more interpolation points between the original sampling points, resulting in a curve that is superior to the initial sampling result in both continuity and smoothness.

[0039] The resampled curve is used as the boundary curve, and the boundary curve data is stored in the result buffer for subsequent structural feature extraction or defect region boundary fitting. This step enables a stable, smooth, and high-resolution representation of light intensity changes in continuously fluctuating regions.

[0040] In another embodiment, assuming the detected fluctuation segment starts at point 156 and ends at point 298, 143 light intensity data points are captured. The light intensity values ​​of the original light intensity curve range from 2800 to 3200 digital values, with approximately 15-20 local peak disturbances. Setting the moving average window to 5 points, the fluctuation amplitude of the filtered light intensity curve ranges from... Reduce to After processing with a low-pass filter with a cutoff frequency of 12%, local spike disturbances were basically eliminated, and the curve continuity was good. Resampling to 256 points via cubic spline interpolation resulted in boundary curve data occupying approximately 2KB of storage space, with a processing time of approximately 3.5 milliseconds.

[0041] Preferably, measuring the extension length of the boundary curve and establishing a combined judgment condition for length and curvature includes: The boundary curve is segmented sequentially along the sampling points, and the distance between adjacent sampling points is calculated for each segment. The distances of each segment are summed up, and the extension length of the boundary curve is recorded. Establish the tangent direction at each sampling point, calculate the angle difference between adjacent tangents, and use the ratio of the angle difference to the corresponding segment length as the local curvature index. The extension length and curvature index are stored to form the criteria for crack combination judgment.

[0042] In one embodiment, boundary curve data containing 256 sampling points is acquired and sequentially labeled as points 1 to 256. Adjacent sampling points are segmented, and the... Point and the The straight-line distance between points is calculated using the formula: Piece distance = This yields 255 segmented distance values. Summing all segmented distances gives a total extension length of 18.6 mm for the boundary curve. Then, a tangent direction is established at each sampling point, and the tangent angle is determined using the slope calculation method based on two points before and after the sampling point. Tangent angle at point = [(No. Point ordinate - the (point y-coordinate) / (the first point) x-coordinate of point - number (x-coordinate of the point). Calculate the angle difference between adjacent tangents and take the absolute value as the curvature index. Divide the angle difference of each segment by the corresponding segment length to obtain the local curvature index, in degrees / mm. Establish the data correspondence, combining the total extension length of 18.6 mm with the average local curvature index of 2.3 degrees / mm to form the geometric characteristic parameters of the boundary curve. According to the preset crack judgment condition: when the extension length is greater than 15 mm and the average curvature index is greater than 2.0 degrees / mm, it is judged as a suspected crack. This boundary curve meets the crack characteristic condition.

[0043] In another embodiment, the boundary curve containing 180 sampling points is processed, and 179 segment distances are calculated, ranging from 0.08 to 0.12 mm, with a total cumulative extension length of 16.8 mm. In the tangent direction calculation, the three-point method is used to calculate the tangent angle, i.e., using the slope of the straight line formed by the current point and one point before and after it. The obtained tangent angle ranges from -15 degrees to +25 degrees, with an average difference of 1.8 degrees between adjacent tangent angles. When calculating the local curvature index, the 179 angle differences are divided by their corresponding segment lengths to obtain a sequence of local curvature indexes, with a maximum value of 8.5 degrees / mm, a minimum value of 0.2 degrees / mm, and an average value of 1.9 degrees / mm. A combined judgment condition is established: an extension length of 16.8 mm combined with an average curvature of 1.9 degrees / mm. Since the average curvature is less than the threshold of 2.0 degrees / mm, the boundary curve is judged to be a surface scratch rather than a crack. The extension length, curvature distribution data, and judgment results are stored together in the detection database.

[0044] Preferably, a tangent direction is established at each sampling point, the angle difference between adjacent tangents is calculated, and the ratio of the angle difference to the corresponding segment length is used as a local curvature index, including: At each sampling point on the boundary curve, the direction of the line connecting that point and the adjacent point is taken as the tangent direction and the direction angle is recorded. Pair adjacent tangent directions, calculate the angle difference, and form an angle difference sequence; In the angle difference sequence, each angle difference is bound to its corresponding sampling point interval, and the start and end positions of the interval are marked; For each sampling interval of a marker, the curve length of that interval is measured, and the curve length is used as the denominator parameter to perform a ratio calculation with the corresponding angle difference; The result of the ratio calculation is recorded as the local curvature index of the sampling interval, and arranged sequentially according to the sampling point position to form a sequence of local curvature index data.

[0045] In one embodiment, a boundary curve containing 200 sampling points is obtained and labeled as points 0 to 199. A tangent direction is established at each sampling point, using the line connecting that point to its next adjacent point as the tangent direction; that is, the tangent direction at point i is the line connecting the i-th point to the next adjacent point. Time The vector direction. The formula for calculating the tangent direction angle is: Direction angle = (Vertical axis increment, horizontal axis increment) yields 199 directional angle values, ranging from -180 degrees to +180 degrees. Adjacent tangent directions are then paired, i.e., the... The tangent and the first Tangent pairings were performed, resulting in 198 pairs of tangent directions. The angle difference between each pair of tangents was calculated, forming a sequence of data containing 198 angle difference values. Each angle difference was then bound to its corresponding sampling point interval. The interval corresponding to the angular difference is from the ... Point to number Point, and record the start and end positions of the interval as [ , ]. Measure the curve length for each marked interval, i.e., calculate the length from the first marked interval. Point of the Classic Point to number Total length of the broken line at a point: Interval length = the first point Point to number Point distance + the first Point to number Point distance. Using the angle difference as the numerator and the corresponding interval length as the denominator, a ratio is calculated to obtain the local curvature index: angle difference / interval length, in degrees / millimeters. These are arranged in order of sampling point location, forming a sequence of 198 local curvature index values, and stored in the data buffer.

[0046] In another embodiment, the boundary curve containing 150 sampling points is processed to establish 149 tangent direction angles, with an angle distribution range of -165 degrees to +170 degrees. 148 angle differences are obtained by pairing adjacent tangents, where the maximum angle difference is 25.6 degrees, the minimum angle difference is 0.3 degrees, and the average angle difference is 4.2 degrees. During the binding of angle differences with sampling intervals, the interval corresponding to the 50th angle difference is... The curve length of this interval is calculated as follows: the distance from point 50 to point 51 (0.095 mm) plus the distance from point 51 to point 52 (0.087 mm), for a total interval length of 0.182 mm. The local curvature index of this interval is calculated as follows: After calculating all 148 intervals, the numerical range of the local curvature index sequence was between 2.1 and 85.3 degrees / mm. Twelve intervals had curvature indices exceeding 50 degrees / mm, mainly concentrated at sharp bends in the curves. The local curvature index sequence was stored sequentially, with each value occupying 4 bytes of storage space, for a total sequence data size of 592 bytes. The processing time was approximately 1.2 milliseconds, providing accurate geometric analysis data for subsequent crack feature determination.

[0047] Preferably, after crack detection is completed, the probe is pressed against the electrode pad, a constant current is applied and the voltage across the two ends is recorded. The contact resistance value is calculated, including: After the crack detection is completed, the drive probe is moved to the target position of the electrode pad, and the probe tip is aligned with the electrode pad; After alignment, a pressing action is performed to make the probe contact the electrode pad surface at a constant speed and maintain the set pressure; After the contact is stabilized, a constant current source is applied to the probe, with the current amplitude preset to the milliampere level, and the current output is kept constant. During the application of constant current, the voltage signal across the probe is acquired, and the voltage timing data is recorded using a high-frequency sampling method. The average voltage value is calculated from the collected voltage time-series data, and the contact resistance value is calculated by dividing the average voltage value by the applied current.

[0048] In one embodiment, after crack detection is completed, a precision motor is activated to drive the probe assembly, moving the probe from its current position to directly above the target electrode pad according to preset coordinate parameters. The relative position of the probe tip and the electrode pad is monitored in real time using microscopic vision, and fine-tuning alignment is performed to ensure that the deviation between the probe central axis and the center point of the pad is controlled within a certain range. Within. After alignment, activate the pressing mechanism and set the probe descent speed to [value] per second. At a constant speed, when the probe tip contacts the pad surface, the sensor detects a contact signal and stops the descent. Vertical pressure is continued to be applied up to a set value of 15 grams, and this pressure value is maintained stable in real time via a pressure sensor. Contact stabilization is then achieved. Then, the constant current source is activated to apply current to the probe. A constant current, with current accuracy controlled within positive and negative ranges. Within this range, the continuous output time is During the current application period, voltage signals across the probe were continuously acquired at a high sampling frequency of 10,000 times, resulting in 10,000 voltage data points. The first and last 200 data points were removed to eliminate transient effects, and the arithmetic mean of the remaining 9,600 voltage data points was calculated to obtain the average voltage value. Average voltage value Divided by the applied current The calculated contact resistance value is .

[0049] In another embodiment, for the microelectrode pad, after the probe moves to the target position, it is precisely aligned using a high-magnification optical lens, with the alignment accuracy requirement reaching [a certain level]. Within. The downward pressing motion uses a slower speed, set to within one second. The contact pressure was adjusted to 8 g / L to avoid damaging the tiny pads. The applied constant current was adjusted to... To minimize impact on sensitive circuits, the sampling frequency is set to 5000 times per second, and the continuous sampling time... 4000 voltage data samples were obtained. After removing the first and last 100 data points, the average voltage value was calculated. The contact resistance value is obtained through division. The test parameters, including current, voltage, contact pressure, and test time, along with the resistance calculation results, are recorded in the test database. After the test, the probe automatically lifts and detaches from the pad surface. The entire test process takes approximately [time missing]. This resistance value will be compared with a preset threshold to provide accurate measurement data for subsequent electrical performance assessment.

[0050] Preferably, step S3 includes: Query and load the preset threshold level corresponding to the current electrode pad; The measured contact resistance values ​​are compared item by item according to the threshold level to determine the threshold range in which they fall. If the contact resistance value is higher than the highest level threshold, mark the pad as a failure risk point and record the real-time parameters of the pad. If the contact resistance value is within the warning range, mark it as needing retesting, record the retesting priority, and add the pad to the retesting queue; The crack judgment result and the electrical judgment result are compared according to the preset merging rules. If both exceed the limit, it is recorded as a serious non-compliance and an alarm is triggered. If only one is abnormal, it is recorded as a marginal non-compliance and a review instruction is issued.

[0051] In one embodiment, based on the number of the electrode pad under test, the corresponding preset threshold level parameter is queried from the threshold configuration database and loaded. This pad belongs to a critical signal line pad, and the threshold level is set to: Excellent range. - Qualified range - Warning range - Failure range is greater than The measured contact resistance value Each item was compared according to its threshold level, and then compared with the upper limit of the excellent / good range. Comparison, discovery Greater than This does not fall under the excellent category. It remains within the upper limit of the acceptable range. Compare, Less than The contact resistance of the pad is determined to be within the acceptable range. The determination result is recorded as "Electrical performance qualified," and the measurement timestamp, resistance value, test conditions, and other parameters are stored in the pad data record. Since no crack defects were found in the pad, the crack determination result is "Physical structure qualified." According to the preset merging rules, when both the crack determination and the electrical determination are qualified, the overall determination result is "Pad qualified," and no alarm or retest instruction needs to be triggered. The final test result is written to the wafer chip quality report, and pad P0156 is marked as green and qualified.

[0052] In another embodiment, electrode pad P0089 is tested. This pad is a power supply pad, and the threshold level is relatively strict: excellent range. - Qualified range - Warning range - Failure range is greater than The measured contact resistance value is After comparing each item, it was found that the value exceeded the failure range threshold. Immediately mark the pad as a potential failure point. Record the pad's real-time parameters, including: test current. Test voltage The contact pressure was 8 grams, the test temperature was 23.5 degrees Celsius, and the humidity was 45%. Meanwhile, optical inspection revealed that the length of the pad was... The mean curvature is A suspected crack with a crack density of 1.5 millimeters is identified as having an "abnormal physical structure." Based on preset merging rules, if both the crack identification and electrical identification fail, the overall assessment is "critically non-compliant." An automatic alarm is triggered, displaying information including the pad number, defect type, and risk level, along with a red warning window popping up on the user interface. The pad is marked as a non-compliant product and highlighted in red. The entire assessment process takes approximately [time missing]. This ensures the efficiency requirements of real-time detection.

[0053] Preferably, if the contact resistance value is higher than the highest level threshold, the pad is marked as a failure risk point, and the real-time parameters of the pad are recorded, including: Determine whether the contact resistance value is above the highest level threshold; When the contact resistance is higher than the highest level threshold in multiple consecutive samples, a failure risk mark for the pad is inserted into the current wafer chip data table; Read the precise position parameters and real-time measurement parameters of the pad in the wafer coordinate system, and write them into the judgment record; The resistance measurement results are stored with fixed precision, data validity is verified, and the stored results are bound to the corresponding pad numbers.

[0054] In one embodiment, the contact resistance value of electrode pad P0243 was measured to be... Immediately determine whether the value exceeds the highest level threshold for this type of pad. The comparison results show Greater than The threshold condition is met. To ensure measurement accuracy, a continuous multiple sampling verification mechanism is activated, performing five consecutive measurements on the pad under the same test conditions, with measurement intervals of [missing information]. The resistance values ​​obtained from five consecutive measurements are as follows: , , , , All measurement results exceeded Threshold. If the pad is confirmed to have a failure risk, immediately insert a failure risk marker "RISK_FAIL" in row 243 of the current wafer chip datasheet. Read the precise position parameters of the pad in the wafer coordinate system: X coordinate is... The Y-coordinate is The rotation angle was 0.25 degrees. Real-time measurement parameters recorded included: ambient temperature 24.2 degrees Celsius, relative humidity 48%, test current 5mA, and average test voltage. Contact pressure 12 grams, test duration Write all parameter information to the judgment log file, recording timestamps accurate to the microsecond level. Record the resistance measurement results. The data is stored in a fixed-precision format with one decimal place. The data range is checked to confirm that the value is within a reasonable range, and then uniquely bound to the pad number P0243 and stored in the failure risk database.

[0055] In another embodiment, the critical power pad P0067 is inspected, and the contact resistance value is initially measured. This far exceeds the highest threshold value for this type of pad. Three consecutive verification measurements were initiated, with a measurement interval set to 300ms. The resulting resistance value sequence is as follows: , , The pad was identified as a failure risk pad after three consecutive measurements exceeding the threshold. A high-risk flag "HIGH_RISK" was inserted into the wafer datasheet, and the priority was set to emergency handling. The precise X-axis coordinates of the pad were read. Y-axis Z-axis height The recorded real-time parameters include: atmospheric pressure 101.2 kPa, and test current. Peak voltage steady-state voltage Voltage fluctuation range probe contact resistance Total test duration Measure the values It is stored with a fixed two decimal places, and data integrity is verified to ensure no data loss. Finally, it is stored in a one-to-one binding relationship with pad number P0067.

[0056] The present invention also provides a wafer chip inspection system, applied to a wafer chip, wherein the wafer chip is fixed on a carrier tray, the carrier tray includes a locking mechanism for performing the wafer chip inspection method described above, and the wafer chip inspection system includes: The receiving tray is equipped with a connecting component and a locking mechanism. The connecting component is equipped with a limit post and a limit hole. The limit post and the limit hole slide together. The locking mechanism is a buckle, a pull buckle or a magnetic buckle. The detection module 101 is used to fix the wafer chip to be tested to the carrier tray using a locking mechanism, and scan the surface area of ​​the wafer chip line by line under a preset illumination angle to generate a corresponding reflected light intensity distribution sequence. The contact resistance value calculation module 102 is used to extract the boundary curve of a continuous fluctuation segment when a continuous fluctuation segment appears in the reflected light intensity distribution sequence, determine whether the extension length and curvature change of the boundary curve meet the crack characteristic conditions, and after the crack determination is completed, control the probe to contact the electrode pad of the wafer chip, apply a constant current and record the voltage at both ends, and calculate the contact resistance value. The wafer chip threshold determination module 103 is used to compare the contact resistance value with a preset threshold. If the contact resistance exceeds the threshold, it is marked as an electrode failure risk point, and the electrical determination result of the wafer chip is obtained. Combining the crack determination result and the electrical determination result, the detection result of the wafer chip is output.

[0057] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for detecting wafer chips, characterized in that, Applied to wafer chips, the wafer chips are fixed to a carrier tray, the carrier tray includes a locking mechanism, and the method includes the following steps: Step S1: Fix the wafer chip to be tested onto the carrier tray using the locking mechanism, and scan the surface area of ​​the wafer chip line by line under the preset illumination angle to generate the corresponding reflected light intensity distribution sequence; Step S2: When a continuous fluctuation segment appears in the reflected light intensity distribution sequence, extract the boundary curve of the segment, determine whether the extension length and curvature change of the boundary curve meet the crack characteristic conditions, and after the crack determination is completed, control the probe to contact the electrode pad of the wafer chip, apply a constant current and record the voltage at both ends, and calculate the contact resistance value. Step S3: Compare the contact resistance value with a preset threshold. If the contact resistance exceeds the threshold, mark it as an electrode failure risk point and obtain the electrical judgment result of the wafer chip. Combine the crack judgment result with the electrical judgment result to output the detection result of the wafer chip.

2. The wafer chip testing method according to claim 1, characterized in that, Step S2 includes: Retrieve continuous fluctuation segments in the reflected light intensity distribution sequence and mark the start and end points of the segments; Using the starting and ending points as boundaries, the light intensity curves of the segments are extracted, and the boundary curves are formed after smoothing out the noise. Measure the extension length of the boundary curve and establish a combined judgment condition for length and curvature; When the extension length exceeds the preset threshold and the curvature change meets the crack characteristic conditions, it is confirmed that there is a crack in the section. After crack detection is completed, control probes are pressed against electrode pads, a constant current is applied and the voltage at both ends is recorded, and the contact resistance value is calculated.

3. The wafer chip testing method according to claim 2, characterized in that, Retrieve continuous fluctuation segments in the reflected light intensity distribution sequence and mark the start and end points of these segments, including: The reflected light intensity distribution sequence is arranged into a continuous data stream according to the sampling order, and the light intensity difference between adjacent sampling points is calculated in the continuous data stream; When the difference maintains the same direction of change across multiple consecutive sampling points, it is marked as a candidate fluctuation segment; Identify the first point at the front end of the fluctuation candidate segment where the difference deviates from the baseline threshold, and set it as the starting point; Identify the point at the end of the fluctuation candidate segment where the last difference falls back to the baseline threshold, and set it as the termination point; The continuous fluctuation section is determined by using the starting point and the ending point as boundaries.

4. The wafer chip testing method according to claim 2, characterized in that, Using the starting and ending points as boundaries, the light intensity curves of the segments are extracted, and the boundary curves are smoothed after noise removal, including: The light intensity data of the continuous fluctuation segment is extracted by using the starting and ending points of the segment as the index range, and the extracted data is arranged in the sampling order to form the original light intensity curve. The original light intensity curve is processed by moving average, and a low-pass filter is added after moving average to suppress local spike disturbances. The filtered curve is interpolated and resampled, and the resampled result is used as the boundary curve and stored.

5. The wafer chip testing method according to claim 2, characterized in that, Measuring the extension length of the boundary curve and establishing combined criteria for length and curvature include: The boundary curve is segmented sequentially along the sampling points, and the distance between adjacent sampling points is calculated for each segment. The distances of each segment are summed up, and the extension length of the boundary curve is recorded. Establish the tangent direction at each sampling point, calculate the angle difference between adjacent tangents, and use the ratio of the angle difference to the corresponding segment length as the local curvature index. The extension length and curvature index are stored to form the criteria for crack combination judgment.

6. The wafer chip testing method according to claim 5, characterized in that, At each sampling point, a tangent direction is established, the angle difference between adjacent tangents is calculated, and the ratio of the angle difference to the corresponding segment length is used as a local curvature index, including: At each sampling point on the boundary curve, the direction of the line connecting that point and the adjacent point is taken as the tangent direction and the direction angle is recorded. Pair adjacent tangent directions, calculate the angle difference, and form an angle difference sequence; In the angle difference sequence, each angle difference is bound to its corresponding sampling point interval, and the start and end positions of the interval are marked; For each sampling interval of a marker, the curve length of that interval is measured, and the curve length is used as the denominator parameter to perform a ratio calculation with the corresponding angle difference; The result of the ratio calculation is recorded as the local curvature index of the sampling interval, and arranged sequentially according to the sampling point position to form a sequence of local curvature index data.

7. The wafer chip testing method according to claim 2, characterized in that, After crack detection is completed, the control probe is pressed against the electrode pad, a constant current is applied and the voltage across the two ends is recorded. The contact resistance value is calculated, including: After the crack detection is completed, the drive probe is moved to the target position of the electrode pad, and the probe tip is aligned with the electrode pad; After alignment, a pressing action is performed to make the probe contact the electrode pad surface at a constant speed and maintain the set pressure; After the contact is stabilized, a constant current source is applied to the probe, with the current amplitude preset to the milliampere level, and the current output is kept constant. During the application of constant current, the voltage signal across the probe is acquired, and the voltage timing data is recorded using a high-frequency sampling method. The average voltage value is calculated from the collected voltage time-series data, and the contact resistance value is calculated by dividing the average voltage value by the applied current.

8. The wafer chip testing method according to claim 1, characterized in that, Step S3 includes: Query and load the preset threshold level corresponding to the current electrode pad; The measured contact resistance values ​​are compared item by item according to the threshold level to determine the threshold range in which they fall. If the contact resistance value is higher than the highest level threshold, mark the pad as a failure risk point and record the real-time parameters of the pad. If the contact resistance value is within the warning range, mark it as needing retesting, record the retesting priority, and add the pad to the retesting queue; The crack judgment result and the electrical judgment result are compared according to the preset merging rules. If both exceed the limit, it is recorded as a serious non-compliance and an alarm is triggered. If only one is abnormal, it is recorded as an edge non-compliance and a review instruction is issued.

9. The wafer chip testing method according to claim 8, characterized in that, If the contact resistance value is higher than the highest level threshold, the pad is marked as a failure risk point, and the real-time parameters of the pad are recorded, including: Determine whether the contact resistance value is above the highest level threshold; When the contact resistance is higher than the highest level threshold in multiple consecutive samples, a failure risk mark for the pad is inserted into the current wafer chip data table; Read the precise position parameters and real-time measurement parameters of the pad in the wafer coordinate system, and write them into the judgment record; The resistance measurement results are stored with fixed precision, data validity is verified, and the stored results are bound to the corresponding pad numbers.

10. A wafer chip inspection system, characterized in that, This system is applied to wafer chips, where the wafer chips are fixed to a carrier tray, and the carrier tray includes a locking mechanism for performing the wafer chip inspection method as described in claim 1. include: The receiving tray is equipped with a connecting component and a locking mechanism. The connecting component is equipped with a limit post and a limit hole. The limit post and the limit hole slide together. The locking mechanism is a buckle, a pull buckle or a magnetic buckle. The detection module is used to fix the wafer chip under test to the carrier tray using a locking mechanism, and scan the surface area of ​​the wafer chip line by line under a preset illumination angle to generate the corresponding reflected light intensity distribution sequence. The contact resistance value calculation module is used to extract the boundary curve of a continuous fluctuation segment when a continuous fluctuation segment appears in the reflected light intensity distribution sequence, determine whether the extension length and curvature change of the boundary curve meet the crack characteristic conditions, and after the crack determination is completed, control the probe to contact the electrode pad of the wafer chip, apply a constant current and record the voltage at both ends, and calculate the contact resistance value. The wafer chip threshold determination module compares the contact resistance value with a preset threshold. If the contact resistance exceeds the threshold, it is marked as a risk point for electrode failure, and the electrical determination result of the wafer chip is obtained. Combining the crack determination result and the electrical determination result, the detection result of the wafer chip is output.

Images