JPH08274136A - Hybrid ic inspection device - Google Patents

Abstract

PURPOSE: To enable a hybrid IC inspection device to detect defects with high accuracy at a high speed by detecting defects having slight concentration varia tion by dividing an object to be detected into blocks by rotating a camera and, at the same time, executing subtraction and selecting thresholds by using multilevel picture images. CONSTITUTION: A camera is rotated around the midpoint of a diagonal line in a designated direction by a rotating shaft correcting angle. A hybrid IC substrate is inspected by dividing the substrate into several blocks. In order to detect a defect, a picture 33 to be inspected is subtracted from a standard picture 34 having no defect. Both the pictures 33 and 34 are multilevel pictures. As the result of subtraction, a differential picture 35 containing an actual defect and the contour noise 36 of a pattern is obtained. The contour noise 36 is removed by multiplying the differential picture 35 by a contour picture 37. As a result, all contour noise is removed and a defect picture 38 containing only the defect is obtained. Therefore, a hybrid IC inspection device can be made compact and the inspection speed of the device can be improved. At the same time, the inspection device can detect defects with high accuracy.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention uses a hybrid IC (hereinafter
The present invention relates to an apparatus and method for automatically and highly accurately detecting a defect occurring in a circuit pattern (referred to as HIC).

[0002]

2. Description of the Related Art A transmission illumination system in which illumination is performed from below the substrate is often used to detect defects such as defects, protrusions, and deviations that are inherent defects in a circuit pattern of a HIC substrate or the like. Further, since the inspection method is performed by comparing and comparing (matching) a reference pattern and a pattern to be inspected, strict alignment (position correction) is required.

For position correction, correction in the XY axis directions (hereinafter,
XY axis correction) and rotation axis correction (hereinafter, θ axis correction) are required. The XY-axis correction is performed by moving the XY table on which the HIC board is placed, while the θ-axis correction is performed by using the transmitted light. Therefore, the XY table and the light source are rotated on the θ-table. ing.

Further, as a conventional inspection method, a method of performing matching by a reference defect-free circuit pattern and an inspection target circuit pattern having a defect, or a rule based on a design rule (design rule) of the inspection target circuit pattern There are check methods etc. The present invention specifically describes a solution to the problem in the former inspection method.

[0005]

In the prior art, especially H
When an IC substrate is inspected using transmitted light, the θ-axis correction has the following drawbacks. It is X on the θ table
Since the Y table and the light source are rotated while being mounted, a rotary table having a large power is required, and as a result, the apparatus itself becomes large and the operation speed becomes slow, which leads to deterioration of the performance of the inspection apparatus. Further, the conventional inspection method has the following problems.

Defects that are particularly difficult to detect in the hybrid IC pattern are defects such as pinholes and beard-like protrusions that occur in the circuit pattern and have a small change in density.
For these detection, the pattern of the conventional method is "0" for the substrate.
It was difficult to perform an operation between images using a binary image represented by 1 ". The reason is that in order to generate this binary image, a constant density value is set for a multi-valued image captured by a camera. If they are distinguished by (threshold value), only the regular pattern width can be extracted, and defects such as pinholes and mustache-like projections having a minute density change may not be detected because they do not reach the threshold value.

Further, in order to detect a defect, when a logical operation (exclusive OR) between a defect-free standard image and an inspection image having a defect is performed, contour noise occurs due to the limit of the positioning accuracy of the substrate, Sometimes it is impossible to distinguish between true defects and this contour noise.

[0008]

In order to solve the problem concerning position correction for HIC board inspection, the θ-axis correction for board position adjustment is not performed on the inspection table on which the HIC board is placed, but the camera is rotated. This is done by a highly accurate positioning method. This can be realized by mounting a small-scale rotating mechanism for rotating a lightweight camera on the upper part of the camera, and as a result, the inspection apparatus can be made compact and the operating speed can be improved.

On the other hand, in order to solve the problem relating to the inspection method, unlike the conventional binary image, both the inspection image and the standard image use multi-valued images, and the arithmetic operation (subtraction) is executed without performing the logical operation. We propose a method to detect defects with minute changes in density by selecting the optimum threshold. Further, in order to reduce the contour noise, a contour image of a circuit pattern is created together with a defect-free standard image, and contour noise generated by arithmetic operation during matching is erased using this contour image. As a result, the actual defect and the contour noise can be distinguished from each other, and the defect can be detected with high accuracy.

[0010]

In the inspection, since the pattern of the HIC substrate is enlarged and the inspection is conducted, the inspection field of view becomes small.
The pattern of the substrate is divided into a plurality of blocks. For this reason, for the positioning of the substrate, first the θ-axis correction by the rotation of the camera is performed once at the first inspection, and then the XY axis is adjusted for each block.
It is realized by the method of axis correction.

With regard to the inspection method, it is possible to detect a defect having a minute density value by using a multivalued image. A defect, which is a pattern mismatch portion, is actually detected by arithmetic operation of the standard image and the inspection image. Although these defects have various density values, even a small defect can be detected by setting a threshold value slightly higher than the noise level. Further, the protrusion and the defect of the defective portion have different density values in terms of characteristics, and therefore are detected by setting two optimum threshold values.

Further, using the defect-free standard image, the contour image of the circuit pattern is created by the enlarging / reducing process of the image processing, and the standard image and the contour image are stored in the memory as a pair of reference images. Next, the contour noise generated by the arithmetic operation at the time of matching is erased using this contour image, the actual defect and the contour noise are distinguished, and only the defect is detected.

[0013]

EXAMPLE First, as shown in FIG. 1, a HIC substrate 1 (hereinafter referred to as a reference substrate), which serves as a reference for inspection, is placed on an XY table 2 and the position of the substrate is measured. This reference board is
It is a defect-free substrate having no defects, and its image obtained from the camera 3 is called a standard image. In addition, H having a defect
The IC board is hereinafter referred to as an inspection board, and the image obtained from the camera 3 is used as the inspection image. First, in order to measure the position of the reference substrate, the defect-free HIC substrate 1 is mounted on the XY table 2.
The HIC substrate 1 is illuminated from below using the transillumination device 5, and an image of a diagonally located portion of the HIC substrate 1 is captured by the camera 3. At this time, the image from the camera 3 is stored in the memory of the image processing device 6, and the state of the image is observed by the monitor 7. Next, in the captured image, as shown in FIG. 2, the upper right and lower left diagonal windows 11,
12 is set to obtain a pair of upper right and lower left diagonal eigenpoints 13 and 14 which are reference positions for inspection. The processing here is to detect horizontal and vertical straight lines 17 and 18 indicating the contours of the outermost circuit pattern in the horizontal and vertical windows 15 and 16, respectively.
The diagonal eigenpoint 13 is obtained as the intersection of these two straight lines. The lower left diagonal eigenpoint 14 is also obtained in the same manner.

Next, the standard image 19 of the positioned HIC reference board 1 is captured. As shown in FIG.
The C board is divided into several blocks, and a standard image 19 and a contour image 20 in which only the contour of the circuit pattern is left are obtained for each block and stored in a memory in the image processing apparatus 6. The standard image 19 and the contour image 20 created in this way are called up at any time during inspection and used for comparison and contrast with inspection images having defects. The actual inspection procedure will be described below. First, the HIC board 1 to be inspected is mounted on the XY table 2 of FIG. 1, and then the position and orientation are measured to see how much the positions of the inspection board and the reference board are deviated. This is similar to the method used when measuring the position of the reference board, and the upper right and lower left diagonal windows 11 and 12 are used to obtain the upper right and lower left diagonal eigenpoints 13 and 14 of the inspection board. then,
From the difference in the positions of the diagonal eigenpoints of the reference board and the inspection board, XY
The correction amount in the axial direction and the θ-axis correction amount, which is the rotation, are calculated.

First, the correction amount in the XY axis directions is obtained using FIG. The shape shown by the solid line in FIG. 4 is the reference substrate 25,
The portion indicated by the dotted line indicates the position of the inspection board 26 on the XY table 2. Here, the coordinates of the right upper diagonal eigenpoints 22 and 21 of the reference substrate 25 and the inspection substrate 26 are respectively (XR0, YR
0), (XR1, YR1), the correction amount ΔX in the X direction and the correction amount ΔY in the Y direction are as follows.

[Equation 1]

[Equation 2]

Next, the θ-axis correction amount of rotation is determined as follows using FIG. If the coordinates of the lower left diagonal eigenpoint 24 of the reference substrate 25 are (XL0, YL0) and the coordinates of the lower left diagonal proper point 23 of the inspection substrate 26 are (XL1, YL1), the X direction of the reference substrate 25 and the Y direction are Y. LX0, LY0 corresponding to the direction length
Is

(Equation 3)

[Equation 4] Becomes Further, if the angle formed by the straight line connecting the upper right corner and the lower left diagonal eigenpoints 22 and 24 on the reference substrate 25 and the X axis is θ1,

(Equation 5) Becomes Similarly, on the inspection board 26, LX1, L
Set Y1 as follows.

(Equation 6)

(Equation 7)

Here, when the angle formed by the straight line connecting the upper right corner and the lower left diagonal eigenpoints 21 and 23 on the inspection board 26 and the X axis is θ2,

(Equation 8) Becomes Therefore, from the equations (5) and (6), the difference Δθ in the rotation angle between the reference substrate 25 and the inspection substrate 26 is given by the following equation.

[Equation 9] A signal of this rotation correction amount is sent from the control computer 10 to the rotary table drive device 8, and the camera rotation device 4 connected to the camera 3 moves the camera 3 in the designated direction by the rotation axis correction amount Δθ degrees to the midpoint of the diagonal line. Rotate to the center. This rotation axis correction Δθ is performed only once the first time the inspection is started for one HIC substrate.

Next, as shown in FIG. 3, the HIC substrate 1 is divided into several blocks 29 and the inspection is advanced. However, although the camera 3 is aligned with the XY directions of the tilted inspection board 26, the XY table 2 that actually moves is different from the XY directions, so it is necessary to correct the X direction and the Y direction for each block. Below, the correction amount of the XY axes to be performed for each block is obtained using FIG.

FIG. 5 is a diagram showing the XY-axis correction amount necessary for each block, and the X- and Y-axis directions 27 and 28 of the coordinates.
Are the XY directions of the XY table 2 which is actually moved.
Here, the nth block in the X direction is Xb, and the nth block in the Y direction is n.
Let the second block be Yb. The block size in the X direction is Bx, and the block size in the Y direction is By. In FIG. 5, the substrate composed of several inclined blocks 29 is the inspection substrate 26 to be inspected, and Δθ in the equation (9)
FIG. 7 illustrates a case where the value of is positive. in this case,
The upper right position coordinate (Xn, Yn) of the nth block is
It looks like this:

[Equation 10]

[Equation 11]

Further, when the value of Δθ becomes negative, that is, when the substrate position in FIG. 5 rises to the right, the upper right position coordinates (Xn, Yn) of the nth block are as follows.

(Equation 12)

(Equation 13) As described above, the control computer 10 calculates the position coordinate 30 of the upper right corner of each block, and the XY table 2 moves in accordance with this position coordinate by the signal from the XY table driving device 9. Then, the defect is detected by matching the standard image and the inspection image for each block.

Next, an inspection method for highly accurate defect detection will be described. The pinhole-like defect 31 and the whisker-like protrusion defect 32 in FIG. 6 are defects that occur in the circuit pattern of the HIC substrate, and are defects that are small in shape and have small changes in color density. These defects are caused by the density change waveform 3 of the inspection image.
Shown in 3.

In the inspection using the transillumination method of illuminating from below the HIC substrate 1, a circuit pattern coated with a metal conductive material is dark (a low concentration value) and a thin ceramic substrate is shown. It is bright (high density value). Therefore, in order to distinguish the pattern from the substrate, the conventional method uses a binary image in which the pattern is represented by "0" and the substrate is represented by "1". However, in order to generate this binary image, with respect to the density change waveform 33 of the inspection image,
When distinguishing by the density value (threshold value) indicated by BB ′, only the regular pattern width can be extracted, and it becomes impossible to detect the pinhole-shaped defect 31 and the beard-shaped protrusion defect 32 that do not reach the threshold value. Therefore, in order to detect these pinhole-like defects 31 and whisker-like protrusion defects 32, it is impossible to detect both defects at the same time even if the threshold value is set to AA ′ or CC ′.

Therefore, the present invention proposes a method capable of accurately detecting even a defect having a small density difference. Images such as standard images and contour images used in the following description are the density change waveforms of the images in FIG. First, in order to detect a defect, an arithmetic operation (subtraction) is performed between the defect-free standard image 34 and the inspection image 33 having a defect. In this case, unlike the conventional binary image, both the standard image 34 and the inspection image 33 are multivalued images. As a result of the subtraction, the difference image 35 including the actual defect and the contour noise 36 of the pattern is obtained. The wavy noise in the difference image 35 is contour noise 36, and the protruding portion is a defect. These defects are pattern non-matching portions and have various density values, but even a small defect can be detected by setting a threshold slightly higher than the noise level. Further, the contour noise 36 is always generated due to the positioning limit, and can be removed by the method described below.

In order to eliminate the contour noise 36 of the circuit pattern caused by the execution of the subtraction, the contour image 20 stored in the memory of the image processing device 6 is used to create an image in which the inside and outside of the contour are expanded by several pixels. Then, using this image, the contour noise 36 is masked and erased. This is executed by the following process. First, the difference image 35 and the contour image 37
And calculation (multiplication) with. At this time, the density values of the contour portion and the substrate portion of the circuit pattern in the contour image 37 are set to “0” and “1”, respectively, and the contour noise 36 is eliminated by multiplication with the difference image 35. As a result, the contour noise is completely erased, and the defect image 38 in which only the defect remains is obtained.

In the defect image 38, the negative data is a protruding defect (beard-like protrusion defect 32), and the positive data is a defective defect (pinhole-like defect 31). Also,
Since the defect of the protrusion and the defect have different density values in characteristics due to the difference of the surrounding light environment, DD ′ and E in FIG.
By setting two optimum threshold values as indicated by E ′, the detection is performed with high accuracy. In order to set the optimum threshold value, it is necessary to measure the maximum noise level by subtracting the standard image 34 and the inspection image 33 in advance.
Finally, the pinhole-like defect 31 and the beard-like protrusion defect 32 in the binarized defect image 29 are binarized by the threshold value, and the area is counted as "1" data to obtain the size of the defect.

[0026]

As described above, the θ-axis correction for aligning the HIC board is not performed on the inspection table on which the HIC board is placed.
This can be easily done by rotating the camera side.
Therefore, the inspection device can be made compact and the inspection speed can be increased by simply performing the θ-axis correction on the camera side and using a highly accurate positioning method as in the present invention.

On the other hand, with respect to the inspection method, unlike the conventional binary image, both the inspection image and the standard image are multivalued images, and the arithmetic operation (subtraction) is executed without performing the logical operation and the optimum threshold value is set. By selection, it becomes possible to detect a defect having a slight change in density. Also, to reduce contour noise,
By creating a contour image of a circuit pattern together with a defect-free standard image and erasing contour noise caused by arithmetic operations during matching using this contour image, it is possible to distinguish between an actual defect and the contour noise, and a highly accurate defect It becomes possible to detect.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an example of a hybrid IC inspection apparatus of the present invention.

FIG. 2 is a diagram showing a method of obtaining diagonal eigenpoints for alignment of a HIC substrate.

FIG. 3 is a conceptual diagram of a standard image and a contour image of a defect-free HIC substrate used at the time of matching, which are obtained by imaging from a camera.

FIG. 4 is a diagram showing a method of obtaining an XY-axis correction amount and a θ-axis correction amount for aligning the HIC substrate.

FIG. 5 is a diagram showing a method for obtaining an XY axis correction amount for each block necessary for block inspection of a HIC board.

FIG. 6 is a diagram showing an image density change waveform in each processing step for performing high-accuracy defect detection.

[Explanation of Codes] 1 HIC substrate 2 XY table 3 Camera 4 Camera rotating device 5 Transmitted illumination device 6 Image processing device 7 Monitor 8 Rotating table driving device 9 XY table driving device 10 Control computer 11 Upper right diagonal window 12 Lower left diagonal Window 13 Top right diagonal eigenpoint 14 Bottom left diagonal eigenpoint 15 Horizontal window 16 Vertical window 17 Horizontal straight line 18 Vertical straight line 19 Standard image 20 Contour image 21 Coordinates of the right upper diagonal eigenpoint of the inspection board 22 Right upper diagonal eigenpoint of the reference board Coordinates of points 23 Coordinates of diagonal eigenpoints of lower left of inspection board 24 Coordinates of diagonal eigenpoints of lower left of reference board 25 Reference board 26 Inspection board 27 X direction of XY table for moving HIC board 28 Y of XY table for moving HIC board Direction 29 Each block for HIC board division inspection 30 Each block Upper right corner coordinates 31 Pinhole defects 32 Whisker defects 33 Density change waveform of inspection image 34 Standard image density change waveform 35 Difference image density change waveform 36 Contour noise 37 Contour image density change waveform 38 Defect image density Variation waveform 39 Density variation waveform of binarized defect image

Claims (5)

[Claims] 1. A device for detecting a defect in a circuit pattern of a hybrid IC substrate, wherein an inspection table on which the substrate is mounted is fixed, and a camera is rotatably installed to correct a rotation axis for aligning the substrate. A hybrid IC inspecting device, characterized in that the substrate is divided into several blocks and inspected. 2. A means for positioning a substrate for each block uses the rotation amount of the substrate, the x- and y-direction block sizes, and the x and y-direction block numbers, and formulas (12) and (13) are used. The hybrid IC according to claim 1, which is determined by
Inspection device. 3. A camera rotation means for correcting a rotation axis, an xy table for moving the substrate in x and y directions, and a transillumination device for illuminating the substrate from below. Alternatively, the hybrid IC inspection device according to item 2. 4. A defect having a minute density change by performing subtraction between a multivalued image obtained by capturing the substrate with a camera and a multivalued image having the same type as the substrate and having no defect, and selecting two optimal threshold values. A hybrid IC inspection method, which is characterized by detecting 5. As a means for erasing the contour noise of the circuit pattern caused by execution of the subtraction, an image in which the inside and outside of the contour of the substrate circuit pattern are expanded by several pixels is created, and the contour noise is masked using this image to erase the contour noise. The hybrid IC inspection method according to claim 4, which is performed.