A wafer inspection method and system based on multi-channel C-scan

Abstract

This invention provides a wafer inspection method and system based on multi-channel C-scanning, aiming to address issues such as multi-channel sensitivity differences, Z-axis height deviation, and stitching artifacts in traditional systems. Before inspection, the maximum defect point and defect-free point on a standard wafer are selected, the gain of each channel is adjusted to a uniform amplitude, and the echo amplitude is recorded to achieve sensitivity calibration. During inspection, a laser displacement sensor is used to collect wafer surface height data in real time. Modeling is combined with ARIMA+Kalman filtering or Gaussian process regression algorithms, and the Z-axis focus distance is dynamically adjusted through PID feedback to compensate for signal attenuation caused by surface warping. For image stitching, three grayscale mapping methods are used to optimize channel response: linear adaptive, gamma adaptive (automatically aligning channel means after a reference γ is set), and manual setting. Coordinate alignment is achieved by calculating sub-pixel offsets based on a two-dimensional cross-correlation function. A 5% overlap area is set at the channel junction, and grayscale transitions are achieved through linear gradient weighted averaging.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor detection technology, and in particular to a wafer detection method and system based on multi-channel C scanning. Background Art

[0002] The application of multi-channel ultrasonic C-scan systems presents a series of pressing technical challenges. Traditional multi-channel ultrasonic C-scan systems, due to hardware differences across channels, such as inconsistent probe sensitivity and varying signal amplifier gain, make it difficult to achieve consistent grayscale representation in the final inspection results. This grayscale inconsistency significantly interferes with the accurate interpretation of defects in the inspection object, leading to operators making misjudgments or missed detections when analyzing and judging defects based on the test results, severely impacting the accuracy and reliability of inspections.

[0003] As described in the prior art CN110531344B, its circuit has the advantages of automatically adjusting the resonance parameters and gain deviation, so that the unit of each ultrasonic channel can be in the best working state, achieving the highest sensitivity and the most stable gain. However, when faced with a warped or uneven wafer surface, the distance between the probes of each channel and the sample will be different. This difference will cause different degrees of signal attenuation, and the reason is that the sound wave propagation path has changed. This difference in signal attenuation further amplifies the grayscale difference in the test results, making the already complex test data even more difficult to analyze and interpret, adding great difficulties to the accurate assessment of wafer quality.

[0004] Furthermore, when processing multi-channel scan results, directly stitching the scans from each channel can produce noticeable stitching artifacts at channel boundaries due to issues like grayscale jumps and coordinate misalignment. These stitching artifacts not only affect the integrity and aesthetics of the inspection image, but more importantly, they can obscure true defect information near the channel boundaries, severely impacting overall inspection accuracy and hindering the application and development of multi-channel ultrasonic C-scan systems in high-precision inspection scenarios. Summary of the Invention

[0005] The pixel-level stitching module of this invention calculates the stitching offset between channel images using a two-dimensional cross-correlation function, achieving sub-pixel coordinate alignment. A 5% overlap region is set at the channel junction, and a weighted average algorithm with linear gradient weights is used for grayscale fusion. This eliminates stitching artifacts, resulting in a continuous, smooth image after stitching multi-channel scan results. This improves overall image quality and provides high-quality image data for subsequent defect analysis and processing.

[0006] The present invention provides a wafer inspection method based on multi-channel C-scanning, comprising: sequentially adjusting the gain of each channel to a specified amplitude; after starting scanning, dynamically adjusting the Z-axis height and obtaining the scanning results of each channel; optimizing the grayscale consistency of the scanning results of each channel; and performing stitching coordinate alignment and executing stitching on the scanning results of each channel.

[0007] In one embodiment of the present invention, sequentially adjusting the gain of each channel to a specified amplitude includes: selecting the position P1 of the center point of the largest defect on the wafer standard; moving channel k to P1, focusing on the Z axis to find the maximum echo amplitude; and adjusting the gain of channel k to a specified amplitude Amax, where k is a positive integer greater than or equal to 1.

[0008] In one embodiment of the present invention, the further step includes: selecting a defect-free position P2 on the standard wafer; and moving the channel k to P2, and recording the interface echo amplitude Amink of the defect-free position of the channel k at this time.

[0009] In one embodiment of the present invention, dynamically adjusting the Z-axis height includes: measuring the wafer surface distance of the wafer, fitting the wafer surface coordinates, and obtaining a pre-gradient curve for each scanning channel; recording the wafer distance of each pixel of each scanning channel during the scanning process to obtain an actual curve for each scanning channel; averaging the actual curve and the pre-gradient curve of the next scanning channel to obtain a Z-axis adjustment compensation amount; and adjusting the height of the next scanning channel according to the Z-axis adjustment compensation amount.

[0010] In one embodiment of the present invention, obtaining the Z-axis adjustment compensation amount further includes: treating the historical height data as a time series, using the ARIMA model to predict the future position trend, and then fusing the real-time sensor data through Kalman filtering to correct the error to obtain the Z-axis adjustment compensation amount; or treating the wafer surface as a spatial random process, using Gaussian process regression to establish a probability model of height distribution, and obtaining the Z-axis adjustment compensation amount, wherein the modeling formula is: , k(x, x') is the covariance function, μ(x) is the mean function, which represents the prior mean of the Gaussian process at the input x.

[0011] In one embodiment of the present invention, performing stitching coordinate alignment includes:

[0012] Select the minimum defect area Z1 on the standard wafer;

[0013] Move channel 1 to the center of Z1 and record the coordinates of channel 1 at this time (x1, y1);

[0014] Scan the minimum defect area Z1 with channel 1 to obtain the defect image FLAW1;

[0015] When k is greater than 1, move channel k to the coordinate (x1, y1), scan the minimum defect area Z1 and record the corresponding defect image FLAWk;

[0016] Create the following function:

[0017] ;

[0018] Get the coordinate position [i1k, j1k] where the maximum value is obtained, which is the offset of the image stitching between channel 1 and channel k;

[0019] Align channel scan results according to corresponding offsets;

[0020] Where x′, y′ are the pixel coordinates of the defect area;

[0021] x and y are offset parameters;

[0022] T(x′,y′) is the pixel value of the defect image of the reference channel, i.e., FLAW1;

[0023] I(x+x′,y+y′) is the pixel value of the defect image of the channel to be registered, that is, FLAWk;

[0024] x+x′, y+y′ are the pixel coordinates of the image to be registered after shifting.

[0025] In one embodiment of the present invention, grayscale consistency optimization includes: linear adaptation, setting the grayscale mapping curve of the signal amplitude of channel k to Grayk=(Ak-Amink) / (Amax-Amink); or gamma adaptation, setting the grayscale mapping curve of the signal amplitude of channel k to Grayk=((Ak-Amink) / (Amax-Amink))^gammak, where when k=1, gammak is set by the customer, and the rest are calculated based on the average value of each channel; or manually setting all gammas; where gamma is a nonlinear parameter used to adjust image contrast and brightness distribution.

[0026] In one embodiment of the present invention, it further includes:

[0027] A 5% overlap area is set at the channel junction;

[0028] Define the image weight: w_top(y) = 1 - (y / H_overlap);

[0029] Define the image weight: w_bottom(y) = y / H_overlap;

[0030] Grayscale transition is performed through the formula,

[0031] I_fused(y) = w_top(y) × I_top(y) + w_bottom(y) × I_bottom(y);

[0032] Where y is the vertical coordinate in the overlapping area;

[0033] H_overlap is the height of the overlapping area;

[0034] I_top(y) is the grayscale value of the upper image at coordinate y;

[0035] I_bottom(y) is the grayscale value of the bottom image at coordinate y.

[0036] The present invention also provides a wafer inspection system based on multi-channel C scanning, including: a multi-probe array module, which includes multiple scanning channels and is used to divide the wafer into multiple rectangular areas for parallel scanning; a sensitivity calibration module, which is used to adjust the sensitivity of each channel to be uniform through standard samples; a dynamic Z-axis adjustment module, which is used to realize dynamic Z-axis adjustment based on real-time height measurement; and a pixel-level stitching module, which is used to stitch the scanning results of each channel and optimize the grayscale consistency.

[0037] In one embodiment of the present invention, the dynamic Z-axis adjustment module includes: a laser displacement sensor for measuring the distance of the wafer surface; and a PID feedback controller for adjusting the scanning channel height according to the Z-axis adjustment compensation amount.

[0038] The present invention has the following beneficial effects:

[0039] (1) By adjusting the gain of each channel to the specified amplitude Amax at the maximum defect point P1 of the standard film, and recording the echo amplitude Amin at the defect-free point P2, the sensitivity between channels is unified. This effectively solves the grayscale inconsistency problem caused by hardware differences such as probe sensitivity and amplifier gain in traditional systems, and provides a unified grayscale benchmark for subsequent defect interpretation.

[0040] (2) Using a laser displacement sensor to measure the wafer surface height, combined with ARIMA+Kalman filtering or Gaussian process regression (GPR) modeling, the Z-axis focus distance is adjusted in real time through a PID feedback controller to compensate for the signal attenuation differences caused by wafer surface warpage. This system adapts to the unevenness of the wafer surface in real time, eliminates the signal attenuation differences caused by the different distances between the probe and the sample, further optimizes the grayscale consistency, and improves the adaptability of the detection system to wafers with complex surfaces, ensuring the reliability of the detection results.

[0041] (3) Through three methods: linear adaptation, gamma adaptation (user sets the base gamma and automatically aligns the mean of other channels), and manual setting, grayscale mapping is performed on each channel signal to compensate for the response differences between channels. The stitching offset between channel images is calculated based on the two-dimensional cross-correlation function to achieve sub-pixel coordinate correction and solve the coordinate misalignment problem during stitching. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A flow chart of a wafer inspection method based on multi-channel C-scanning according to an embodiment of the present invention is shown;

[0043] Figure 2 The figure shows the composition of a wafer inspection system based on multi-channel C-scan in one embodiment of the present invention. DETAILED DESCRIPTION

[0044] In the following description, the present invention is described with reference to various embodiments. However, those skilled in the art will recognize that the embodiments may be implemented without one or more of the specific details or with other alternative and / or additional methods, materials, or components. In other cases, well-known structures, materials, or operations are not shown or described in detail to avoid obscuring the inventive aspects of the present invention. Similarly, for purposes of explanation, specific numbers, materials, and configurations are set forth to provide a comprehensive understanding of the embodiments of the present invention. However, the present invention is not limited to these specific details.

[0045] In the present invention, each embodiment is only intended to illustrate the aspects of the present invention and should not be construed as limiting.

[0046] In this specification, reference to "one embodiment" or "the embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. The appearances of the phrase "in one embodiment" in various places in this specification are not necessarily all referring to the same embodiment.

[0047] In addition, the numbering of the steps of the methods of the present invention does not limit the order in which the steps are to be performed. Unless otherwise specified, the steps of the methods may be performed in different orders.

[0048] The present invention will be further described below with reference to the accompanying drawings in conjunction with specific embodiments.

[0049] Figure 1 A flow chart of a wafer inspection method based on multi-channel C-scanning in one embodiment of the present invention is shown.

[0050] In one embodiment of the present invention, a wafer inspection method based on multi-channel C-scan includes the following steps:

[0051] The channel sensitivity before detection is unified as 100. In one embodiment of the present invention, the specific process is as follows:

[0052] a. Select the center point P1 of the largest defect on the standard piece.

[0053] b. Move channel 1 to P1 and automatically focus on the Z axis to find the maximum echo amplitude.

[0054] c. Automatically adjust the gain of channel 1 to the specified amplitude Amax, such as 80% of the full screen.

[0055] d. For channels 2 to 4, perform steps b and c in sequence.

[0056] e. Select the defect-free position P2 on the standard wafer.

[0057] f. Move channel 1 to P2 and record the interface echo amplitude Amin1 at the defect-free position of channel 1 at this time.

[0058] g. Perform step f on channels 2 to 4 in sequence and record the corresponding echoes Amin2 to Amin4.

[0059] During the detection process, the dynamic Z-axis height adjustment 200 is performed. In one embodiment of the present invention, the specific process is as follows:

[0060] The wafer surface distances H0, H90, H180, and H270 at 0°, 90°, 180°, and 270° of the wafer are measured, and the wafer surface coordinates are obtained by fitting to obtain the pre-gradient curve of each scanning channel.

[0061] The wafer distance of each pixel in each scanning channel is recorded to obtain the actual curve of each scanning channel, and the Z-axis adjustment compensation amount of the next scanning channel is obtained by averaging the pre-gradient curve of the next scanning channel.

[0062] In this embodiment, linear weighting is used. Optional algorithms are as follows. The following algorithms will achieve more accurate results with a slight calculation delay.

[0063] Algorithm based on time series prediction (ARIMA+Kalman filtering): Historical height data is regarded as a time series, and the ARIMA model is used to predict future position trends. The real-time sensor data is then integrated through Kalman filtering to correct errors.

[0064] Surface modeling-based algorithm (Gaussian process regression, GPR): The wafer surface is regarded as a spatial random process, and Gaussian process regression is used to establish a probability model of height distribution.

[0065] The modeling formula is:

[0066] ,

[0067] Where k(x, x') is the covariance function and μ(x) is the mean function, which represents the prior mean of the Gaussian process at input x.

[0068] In one embodiment of the present invention, the inter-channel pixel-level stitching and grayscale correction 300 includes the following steps:

[0069] Optimization of grayscale consistency between channels:

[0070] Linear Adaptive:

[0071] Channel 1 signal amplitude (A1) grayscale mapping curve (taking linear as an example):

[0072] Gray1=(A1-Amin1) / (Amax-Amin1).

[0073] Channel 2 signal amplitude (A2) grayscale mapping curve (taking linear as an example):

[0074] Gray2=(A2-Amin2) / (Amax-Amin2).

[0075] Channel 3 signal amplitude (A3) grayscale mapping curve (taking linear as an example):

[0076] Gray3=(A3-Amin3) / (Amax-Amin3).

[0077] Channel 4 signal amplitude (A4) grayscale mapping curve (taking linear as an example):

[0078] Gray4=(A4-Amin4) / (Amax-Amin4).

[0079] Gamma Adaptive:

[0080] Grayk =((Ak-Amink) / (Amaxk-Amink))^gammak.

[0081] Channel 1 signal amplitude (A1) grayscale mapping curve:

[0082] Gray1=((A1-Amin1) / (Amax-Amin1))^gamma1.

[0083] Channel 2 signal amplitude (A2) grayscale mapping curve:

[0084] Gray2=((A2-Amin2) / (Amax-Amin2))^gamma2.

[0085] Channel 3 signal amplitude (A3) grayscale mapping curve:

[0086] Gray3=((A3-Amin3) / (Amax-Amin3))^gamma3.

[0087] Channel 4 signal amplitude (A4) grayscale mapping curve:

[0088] Gray4=((A4-Amin4) / (Amax-Amin4))^gamma4.

[0089] Gamma1 is set by the user, and gamma2, gamma3, and gamma4 are calculated based on the average value of each channel.

[0090] Gamma manual setting:

[0091] The mapping principle is consistent with Gamma Adaptive, where gamma1-gamm4 are all set by the customer.

[0092] Stitching coordinate alignment:

[0093] Select the minimum defect area Z1 on the standard chip.

[0094] Channel 1 moves to the center of Z1 and records the coordinates of channel 1 at this time (x1, y1).

[0095] The minimum defect area Z1 is scanned to obtain a defect image FLAW1 (eg, a pixel area of ​​1024*1024).

[0096] Channels 2 to 4 move to the coordinates (x1, y1) in sequence, scan the minimum defect area Z1, and record the corresponding defect images FLAW2 to FLAW4.

[0097] Let function R[i,j] be the two-dimensional cross-correlation function of images [FLAW1, FLAW2].

[0098] ;

[0099] The coordinate position [i12, j12] of the maximum value is the offset of the image stitching between channel 1 and channel 2.

[0100] Repeat to obtain the splicing offsets [i13, j13] and [i14, j14] of channels 3 and 4.

[0101] The splicing of channels 2 to 4 is corrected according to the corresponding offset;

[0102] Where x′, y′ are the pixel coordinates of the defect area;

[0103] x and y are offset parameters;

[0104] T(x′,y′) is the pixel value of the defect image of the reference channel, i.e., FLAW1;

[0105] I(x+x′,y+y′) is the pixel value of the defect image of the channel to be registered, that is, FLAWk;

[0106] x+x′, y+y′ are the pixel coordinates of the image to be registered after shifting.

[0107] Grayscale fusion of overlapping areas:

[0108] A 5% overlap area is set at the channel junction;

[0109] Define the image weight: w_top(y) = 1 - (y / H_overlap);

[0110] Define the image weight: w_bottom(y) = y / H_overlap;

[0111] Grayscale transition is performed through the fusion formula.

[0112] I_fused(y) = w_top(y) × I_top(y) + w_bottom(y) × I_bottom(y);

[0113] Where y is the vertical coordinate in the overlapping area;

[0114] H_overlap is the height of the overlapping area;

[0115] I_top(y) is the grayscale value of the upper image at coordinate y;

[0116] I_bottom(y) is the grayscale value of the bottom image at coordinate y.

[0117] Figure 2 The figure shows the composition of a wafer inspection system based on multi-channel C-scan in one embodiment of the present invention.

[0118] In one embodiment of the present invention, a wafer inspection system based on multi-channel C-scan includes:

[0119] The multi-probe array module 10 includes multiple scanning channels for dividing the wafer into multiple rectangular areas for parallel scanning.

[0120] The sensitivity calibration module 20 dynamically calibrates the detection sensitivity of each channel using standard samples, eliminating grayscale inconsistencies caused by hardware variations such as probe sensitivity and signal amplifier gain. The module selects the largest defect point and a non-defective point on the standard wafer, adjusts the gain of each channel to a uniform value (e.g., 80% of the full screen), and records the echo amplitude of the non-defective point to achieve inter-channel signal amplitude consistency. Ultimately, inter-channel grayscale variations are reduced to less than 5%, providing a unified grayscale benchmark for subsequent defect interpretation.

[0121] The dynamic Z-axis adjustment module 30 measures the wafer surface height in real time based on a laser displacement sensor, and automatically adjusts the Z-axis focusing distance through a PID feedback controller to adapt to the warping or unevenness of the wafer surface and compensate for the signal attenuation caused by the difference in the distance between the probe and the sample.

[0122] The pixel-level stitching module 40 achieves pixel-level precise stitching of multi-channel scanned images through grayscale consistency optimization, coordinate alignment and grayscale fusion of overlapping areas, eliminates stitching artifacts, and generates continuous, smooth, high-quality detection images.

[0123] Grayscale consistency optimization: Linear adaptation, gamma adaptation (user-set baseline gamma and automatic alignment of other channel means), and manual setting are used to perform grayscale mapping on each channel signal to compensate for response differences between channels.

[0124] Stitching coordinate alignment: Calculate the stitching offset of images between channels based on the two-dimensional cross-correlation function, achieve sub-pixel coordinate correction, and solve the coordinate misalignment problem during stitching.

[0125] Grayscale fusion in overlapping areas: A 5% overlapping area is set at the junction of channels, and a weighted average algorithm with linear gradient weights is used to smooth grayscale transitions and eliminate grayscale jumps at the stitching boundaries.

[0126] In one embodiment of the present invention, the dynamic Z-axis adjustment module includes a laser displacement sensor and a PID feedback controller. The laser displacement sensor is used to measure the distance between the scanning channel and the wafer surface, and the PID feedback controller adjusts the height of the scanning channel by feedback of the dynamic Z-axis compensation value.

[0127] Although various embodiments of the present invention are described above, it should be understood that they are presented as examples only and not as limitations. It will be apparent to those skilled in the relevant art that various combinations, modifications and variations may be made thereto without departing from the spirit and scope of the present invention.

Claims

1. A wafer detection method based on multi-channel C-scan, characterized in that: include: By determining the maximum echo amplitude of each channel, adjust the gain of each channel to the specified amplitude in turn; After starting the scan, the Z-axis adjustment compensation amount is obtained by establishing the pre-gradient curve of each channel and combining it with the actual curve of each channel. The Z-axis height is dynamically adjusted according to the Z-axis adjustment compensation amount, and the scanning results of each channel are obtained; Optimize the grayscale consistency of the scanning results of each channel; as well as The stitching offset between channels is obtained through the defect image of the minimum defect area in the scanning result of each channel, and the stitching coordinates are aligned according to the stitching offset.

2. The wafer detection method based on multi-channel C-scan according to claim 1, characterized in that: By determining the maximum echo amplitude of each channel, adjusting the gain of each channel to the specified amplitude in turn includes: Select the center point position P1 of the largest defect on the standard wafer; Move channel k to P1 and focus on the Z axis to find the maximum echo amplitude; and Adjusts the gain of channel k to the specified amplitude Amax, where k is a positive integer greater than or equal to 1.

3. The wafer detection method based on multi-channel C-scan according to claim 2, characterized in that: Also includes: Selecting a defect-free position P2 on the wafer standard; and Move channel k to P2 and record the interface echo amplitude Amink at the defect-free position of channel k at this time.

4. The wafer detection method based on multi-channel C-scan according to claim 1, characterized in that: By establishing the pre-gradient curve of each channel and combining it with the actual curve of each channel, the Z-axis adjustment compensation amount is obtained. The Z-axis height is dynamically adjusted according to the Z-axis adjustment compensation amount, including: Measure the wafer surface distance of the wafer, obtain the wafer surface coordinates by fitting, and obtain the pre-gradient curve of each scanning channel; Record the wafer distance of each pixel in each scanning channel during the scanning process to obtain the actual curve of each scanning channel; Averaging the actual curve and the pre-gradient curve of the next scanning channel to obtain the Z-axis adjustment compensation; and Adjust the height of the next scanning channel according to the Z-axis adjustment compensation amount.

5. The wafer detection method based on multi-channel C-scan according to claim 4, characterized in that: Obtaining the Z-axis adjustment compensation also includes: Treat historical height data as a time series, use the ARIMA model to predict future position trends, and then use Kalman filtering to fuse real-time sensor data to correct errors and obtain the Z-axis adjustment compensation; or The wafer surface is regarded as a spatial random process, and a probability model of height distribution is established using Gaussian process regression to obtain the Z-axis adjustment compensation. The modeling formula is: H(x)~GP(μ(x), k(x, x′)), k(x, x′) is the covariance function, and μ(x) is the mean function, which represents the prior mean of the Gaussian process at the input x.

6. The wafer detection method based on multi-channel C-scan according to claim 2, characterized in that: The stitching offset between channels is obtained through the defect image of the minimum defect area of ​​the scanning result of each channel, and the stitching coordinate alignment is performed according to the stitching offset, including: Select the minimum defect area Z1 on the standard wafer; Move channel 1 to the center of Z1 and record the coordinates of channel 1 at this time (x1, y1); Scan the minimum defect area Z1 with channel 1 to obtain the defect image FLAW1; When k is greater than 1, move channel k to the coordinate (x1, y1), scan the minimum defect area Z1 and record the corresponding defect image FLAWk; Create the following function: Get the coordinate position [i1k, j1k] where the maximum value is obtained, which is the offset of the image stitching between channel 1 and channel k; Align channel scan results according to corresponding offsets; Where x′, y′ are the pixel coordinates of the defect area; x and y are offset parameters; T(x′,y′) is the pixel value of the defect image of the reference channel, i.e., FLAW1; I(x+x′,y+y′) is the pixel value of the defect image of the channel to be registered, that is, FLAWk; x+x′, y+y′ are the pixel coordinates of the image to be registered after shifting.

7. The wafer detection method based on multi-channel C-scan according to claim 2, characterized in that: Grayscale consistency optimization includes: Linear adaptation, set the channel k signal amplitude grayscale mapping curve to Grayk = (Ak-Amink) / (Amax-Amink); or Gamma adaptation: Set the grayscale mapping curve of the signal amplitude of channel k to Grayk = ((Ak - Amink) / (Amax - Amink))^gammak. When k = 1, gammak is set by the user, and the rest are calculated by aligning the average values ​​of each channel; or Manually set all Gammas; Gamma is a nonlinear parameter used to adjust the image contrast and brightness distribution.

8. The wafer detection method based on multi-channel C-scan according to claim 7, characterized in that: Also includes: A 5% overlap area is set at the channel junction; Define the top image weight: w_top(y)=1-(y / H_overlap); Define the image weight: w_bottom(y)=y / H_overlap; Grayscale transition is performed through the formula, I_fused(y)=w_top(y)×I_top(y)+w_bottom(y)×I_bottom(y); Where y is the vertical coordinate in the overlapping area; H_overlap is the height of the overlapping area; I_top(y) is the grayscale value of the upper image at coordinate y; I_bottom(y) is the grayscale value of the bottom image at coordinate y.

9. A wafer inspection system based on multi-channel C-scan, characterized in that: include: A multi-probe array module, comprising multiple scanning channels, for dividing the wafer into multiple rectangular areas for parallel scanning; The sensitivity calibration module is used to adjust the gain of each channel to the specified amplitude by determining the maximum echo amplitude of each channel; Dynamic Z-axis adjustment module, used to establish pre-gradient curves for each channel and obtain Z-axis adjustment compensation based on the actual curves of each channel; as well as The pixel-level stitching module is used to obtain the stitching offset between channels through the defect image of the minimum defect area in the scanning results of each channel, and align the stitching coordinates and optimize the grayscale consistency based on the stitching offset.

10. The wafer inspection system based on multi-channel C-scan according to claim 9, characterized in that: The dynamic Z-axis adjustment module includes: Laser displacement sensing is used to measure the distance between each channel and the wafer surface to establish a pre-gradient curve; and a PID feedback controller is used to adjust the scanning height of each channel according to the Z-axis adjustment compensation amount.

Images