JP3447538B2 - Pattern inspection method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an inspection method for inspecting a pattern formed on a green sheet, a film carrier or the like.

[0002]

2. Description of the Related Art Conventionally, PGA (Pin Grid Array) has been known as a mounting technique suitable for a demand for increasing the number of pins of ICs and LSIs. In PGA, a ceramic substrate is used as a base of a package to which a chip is attached, and wiring is performed up to a lead wire extraction position. To make this ceramic substrate, a so-called green sheet is used in which alumina powder is kneaded with a liquid binder to form a sheet, and a paste containing a metal having a high melting point is screen-printed on the green sheet. Then, by stacking a required number of such sheets and firing them, so-called simultaneous firing is performed, in which the green sheets are sintered and the paste is metallized.

As another mounting technique, TAB is used.
(Tape Automated Bonding) is known. The TAB method is a polyimide film carrier (TAB tape).
The copper foil pattern formed above is bonded to the electrodes of the IC chip to form external leads. The copper foil pattern is formed by attaching a copper foil to a film with an adhesive and etching this.

In such a green sheet or film carrier, the pattern is visually inspected by a person using a microscope after the pattern is formed. However, in order to visually inspect a fine pattern, there is a problem that skill is required and eyes are seriously used. Therefore, as an alternative to visual inspection, a technique has been proposed in which a pattern formed on a film carrier or the like is imaged by a TV camera and automatically inspected (for example, Japanese Patent Laid-Open No. 7-11086).
3 gazette).

FIG. 8 is a diagram for explaining a conventional inspection method for detecting a protrusion or a defect disclosed in Japanese Patent Laid-Open No. 7-110863. In this inspection method, when the error amount a at the tip of the protrusion of the measured pattern Pa extracted from the image captured by the TV camera is equal to or greater than a predetermined threshold value, or the error amount at the deepest part of the defect of the measured pattern Pc. When b was equal to or larger than the predetermined threshold value, it was determined that there was a defect.

[0006]

As described above, in the conventional inspection method, the protrusion or the defect of the measured pattern is detected based on the error amount between the master pattern which is the reference of the inspection and the measured pattern. However, in such an inspection method, the adjacent measured pattern Pb is compared with the error amount a of the protrusion.
There is a problem in that even if the gap amount A between and is sufficiently large and there is no electrical problem, it is detected as a defect. Similarly, the pattern remaining amount B is sufficiently larger than the error amount b of the defect, and even if there is no electrical problem, it is detected as a defect. Therefore, due to such over-detection, the conventional inspection method has a problem that the yield of products is reduced. The present invention has been made to solve the above problems, and an object of the present invention is to provide a pattern inspection method capable of preventing overdetection of pattern defects.

[0007]

According to the present invention, as described in claim 1, edge data indicating a pattern edge is extracted from a grayscale image of a pattern to be measured, and a master pattern and edge data as a reference for inspection are extracted. The amount of error is calculated, the protrusions that are the defect candidates of the pattern to be measured are detected based on this amount of error, and the edge data of the pattern to be measured whose error amount is the maximum value a in a predetermined area centered on the defect candidate is obtained. Find the edge data of the adjacent measured pattern that is the tip of the protrusion and has the smallest distance from the tip of the protrusion, and set the distance at this time as the gap A with the adjacent measured pattern. Based on the rate a / A, it is determined whether or not the defect candidate is a defect. When creating the master pattern, the center lines of two opposing straight lines indicating both edges of the pattern are found in the master pattern registered as a set of straight lines indicating the pattern edges. The bisector that divides the angle formed by two continuous straight lines of the master pattern into two equal parts is obtained, and the circumference of the straight line of the master pattern is divided into regions that belong to each straight line by these bisectors. The edge data of the measured pattern existing in 1) is associated with the straight line of the master pattern to which the area belongs. The distance from the intersection of the perpendicular line drawn from the edge data of the measured pattern and the straight line of the master pattern to the edge data is the error amount between the master pattern and the edge data.

Further, as described in claim 2, edge data indicating a pattern edge is extracted from the grayscale image of the pattern to be measured, the error amount between the master pattern and the edge data as a reference of the inspection is obtained, and this error amount A defect which is a defect candidate of the measured pattern is detected based on the above, and the edge data of the measured pattern whose error amount has the maximum value b in a predetermined area centered on the defect candidate is set as the deepest part of the defect, and this defect is detected. The distance between the deepest portion and the edge data that constitutes both edges of the measured pattern together with the deepest portion of the pattern and the deepest portion is the remaining amount B of the pattern, and based on the error amount and the pattern remaining rate b / B, It is designed to determine whether or not the defect candidate is a defect.

[0009]

BEST MODE FOR CARRYING OUT THE INVENTION Next, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a flow chart showing a pattern inspection method according to an embodiment of the present invention, and FIG. 2 is a block diagram of a pattern inspection apparatus used in this inspection method. In FIG. 2, 1 is a green sheet,
2 is an XY table on which the green sheet 1 is placed, 3 is a line sensor camera that images the green sheet 1, 4 is edge data indicating the edges of the pattern to be measured from the grayscale image obtained by the camera 3, and is a master. An image processing apparatus for inspecting the measured pattern by obtaining an error between the pattern and the measured pattern, 5 is a host computer for controlling the entire apparatus, and 6 is a display device for displaying the inspection result.

First, a master pattern created in advance before the inspection will be described. Host computer 5
Reads design value data (hereinafter referred to as CAD data) of a green sheet created by a CAD (Computer Aided Design) system and written on a magnetic disk by a magnetic disk device (not shown) (step 101 in FIG. 1). Then, the edge data of the pattern is extracted from the read CAD data, and this is used as a master pattern that serves as a reference for inspection (step 102). The edge data of the extracted master pattern is a set of straight lines indicating pattern edges.

Needless to say, the green sheet 1 is produced based on the CAD data and the pattern is screen-printed on the sheet 1. Then, the host computer 5 obtains the center line L1 of the two opposing straight lines A1 and A2 indicating both edges of the master pattern as shown in FIG. When a perpendicular line H1 perpendicular to the center line L1 is drawn, the length between the intersections of the perpendicular line H1 and the straight lines A1 and A2 is the width W0 of the master pattern. In this way, the host computer 5 obtains the center line and the width of the master pattern for each straight line of the master pattern and stores them together with the edge data.

Next, the inspection of the measured pattern will be described. First, the green sheet 1 is imaged by the camera 3. Then, the image processing device 4 digitizes the grayscale image output from the camera 3 and temporarily stores it in an internal image memory (not shown) (step 103). Since the camera 3 is a line sensor in which pixels are arranged in the X direction, two-dimensional image data is stored in the image memory by moving the XY table 2 or the camera 3 in the Y direction.

Then, the image processing device 4 aligns the positioning mark of the measured pattern stored in the image memory with the positioning mark of the master pattern sent from the host computer 5 to align the measured pattern with the master pattern. Is performed (step 104).

After alignment, the image processing device 4
Binarizes the grayscale image of the measured pattern (step 105). The grayscale image data of the measured pattern includes the pattern and the other background (base material), but there is a density difference between the pattern and the background, so there is a difference between the density value of the pattern and the density value of the background. If you set the value of as a threshold,
The pattern is converted to "1" and the background is converted to "0". Thus, the pattern edge and the inside are pixel "1".
It is possible to obtain a pattern to be measured filled with.

Subsequently, the image processing device 4 extracts edge data indicating the edge coordinates of the pattern to be measured by a labeling process for giving the same label (name) to the connected pixels in the binary image (step 106). FIG. 4 is a diagram for explaining the labeling process. Here, the pattern is represented by a white circle and the background (base material) is represented by a black circle.

For example, when edge data is extracted from a binary image as shown in FIG. 4, the binary image is sequentially scanned in the TV raster direction (from left to right in FIG. 4), and boundary tracking is still performed. An unbounded boundary point is found, this is set as a new tracking start point n1, and its X and Y coordinates are stored.
Then, for example, a boundary point connected clockwise from this tracking start point n1 is searched, and the X and Y coordinates of this boundary point are stored until the tracking start point n1 is returned.

Boundary points such as n1, n2, n3 ... Are extracted, and when the boundary tracing of one pattern is completed, a boundary point which has not been subjected to boundary tracing again is searched for, and the boundary of the next pattern is traced. In this way, the measured patterns are labeled one after another. In the present embodiment, when scanning in the raster direction, the white pixel is the boundary point when rising from black to white, and the black pixel is the boundary point when falling from white to black.

Next, the image processing apparatus 4 extracts the edge data n1, n2, n3 ...
.. is associated with the straight line of the master pattern (step 107). In order to make this correspondence, as shown in FIG. 5, the continuous straight lines A1 and A2, A of the master pattern are formed.
.., which bisects the angles formed by 2 and A3 ..

The circumference of the straight lines A1, A2, A3, ... Of the master pattern is divided by the bisectors A2 ', A3' ,. As a result, the edge data n1, n2, n3 ... Of the measured pattern existing in each area are associated with the straight lines A1, A2, A3 ... Of the master pattern to which the area belongs. . For example, in FIG. 5, the edge data n1 to n3 are associated with the straight line A1, and the edge data n4 to n6 are associated with the straight line A2.

Next, the image processing apparatus 4 obtains an error amount Δd between the master pattern and the measured pattern, and this error amount Δd
The defect candidates of the pattern to be measured are detected based on (step 108). FIG. 6 is a diagram for explaining a method of calculating the error amount Δd.

First, each edge data n of the measured pattern
1, n2, n3, ...
Draw a perpendicular line to. The distance from the intersection of the perpendicular line drawn from the edge data n1 of the measured pattern and the straight line A1 of the master pattern to the edge data n1 is the error amount Δd between the master pattern and the edge data n1 (Δd in FIG. 6).
1), and the distance from the intersection of the perpendicular line drawn from the edge data n2 and the straight line A1 to the edge data n2 is the error amount Δd (Δd2 in FIG. 6) between the master pattern and the edge data n2. The error amount Δd can be similarly obtained for the other edge data.

The image processing device 4 determines that there is no defect when the error amount Δd is small. Further, the image processing device 4 is
When the error amount Δd of certain edge data is equal to or larger than the predetermined threshold value E, this edge data is determined as a defect candidate, and the position recognized as the defect candidate is stored.

After performing the above inspection on the entire pattern to be measured, the image processing apparatus 4 searches for the tip of the protrusion or the deepest part of the defect in a region of a predetermined size centered on the defect candidate ( Step 109). FIG. 7 is a diagram for explaining a method for searching the tip end portion of the protrusion and the deepest portion of the defect, and a method for calculating a gap residual rate and a pattern residual rate, which will be described later. In FIG. 7A, two opposing straight lines A1 and A2 and A5 and A6 are straight lines indicating both edges of the master pattern. Here, the master pattern in which both edges are formed by the straight lines A1 and A2 and the master pattern in which both edges are formed by the straight lines A5 and A6 are parallel to each other. Similarly, in FIG. 7B, two opposing straight lines A3 and A4 are straight lines indicating both edges of the master pattern.

As shown in FIG. 7A, in a master pattern in which two opposing straight lines A1 and A2 indicate both edges,
The width of the master pattern is previously calculated as W0. Here, the distance between the facing edge data of the measured pattern is obtained by drawing a perpendicular line from the edge data n of the measured pattern to the center line L1 of the master pattern. This is the width W of the measured pattern. When the width W of the measured pattern is larger than the width W0 of the master pattern, it can be determined that a protrusion exists at the position of this defect candidate.

Then, the image processing device 4 makes the area of a predetermined size centered on the defect candidate recognized as a protrusion,
The edge data for which the error amount Δd has the maximum value a is defined as the tip end na of the protrusion. On the other hand, as shown in FIG. 7B, when the width W of the measured pattern is smaller than the width W0 of the master pattern, it can be determined that there is a defect at the position of this defect candidate. Therefore, the image processing device 4 determines the error amount Δ in an area of a predetermined size centered on the defect candidate recognized as a defect.
The edge data for which d has the maximum value b is the deepest part nb of the defect.

Next, the image processing device 4 uses the tip n of the protrusion.
The minimum gap amount A between a and the adjacent pattern to be measured is calculated to calculate the gap residual ratio a / A, and the minimum residual pattern amount B of the defective portion is calculated to calculate the pattern residual ratio b / B (step 110). First, the image processing device 4
As shown in FIG. 7A, after a vertical line is drawn from the tip end portion na of the protrusion to the center line L1 of the master pattern, the perpendicular line is moved in the direction opposite to the center line L1 (that is, in the adjacent measured pattern). Direction).

Subsequently, the image processing apparatus 4 is adjacent to each other in an arc shape having a radius R with the distance between the tip end na of the perpendicular line extending in the direction opposite to the center line L1 with the tip end na of the protrusion as the center. Find the edge data of the measured pattern. If no edge data is found, the radius R is increased by a fixed amount and the same scanning is repeated. The scanning angle at this time is as shown in FIG.
As shown in (a), the range is ± θ with the perpendicular as the center.

When the image processing apparatus 4 repeats the same scanning while increasing the radius R and finds the edge data nc of the adjacent measured pattern, the radius R at this time is set to the minimum gap amount A with the adjacent measured pattern. And In this way, the gap residual rate a / A can be calculated. The reason why the adjacent measured pattern is searched in an arc shape is that the measured pattern having a protrusion and the measured pattern adjacent to this pattern are not always parallel to each other.

Further, as shown in FIG. 7B, the image processing device 4 draws a perpendicular from the deepest part nb of the defect to the center line L2 of the master pattern, and puts this perpendicular in the direction opposite to the deepest part nb. By extending, the edge data on the opposite side forming both edges of the measured pattern together with the deepest part nb are searched. When the image processing device 4 finds the edge data nd facing the deepest portion nb, the image processing device 4 sets the distance between the edge data nd and the deepest portion nb as the minimum pattern remaining amount B in the defective portion. In this way, the pattern residual rate b / B can be calculated.

Next, the image processing device 4 calculates the error amount Δd,
Based on the gap residual rate a / A and the pattern residual rate b / B, it is determined whether the measured pattern is normal (step 111). That is, the image processing device 4 determines that the error amount Δd is normal when the error amount Δd is smaller than the predetermined threshold value E.

On the other hand, the image processing device 4 determines that the error amount Δd due to the protrusion is normal when the gap residual rate a / A is smaller than the predetermined threshold value F even if the error amount Δd is not less than the predetermined threshold value E. To do. Similarly, the image processing device 4 determines that the pattern residual rate b / B is normal when the pattern residual rate b / B is smaller than the predetermined threshold value G even if the error amount Δd due to the loss is equal to or larger than the predetermined threshold value E.

Further, the image processing apparatus 4 has an error amount Δd of a threshold value E or more and a gap residual rate a / A of a threshold value F or more, or a pattern residual rate b / B of a threshold value G or more. If so, it is determined that there is a defect.

As described above, in the present invention, even if the error amount of the protrusion is equal to or more than the threshold value E, the gap amount A between the adjacent measured patterns is sufficiently large with respect to the error amount a of the tip portion (that is, , The gap residual ratio a / A is smaller than the threshold value F), it is regarded as normal. Similarly, even if the error amount of the defect is equal to or larger than the threshold value E, the residual pattern amount B of the defect portion is sufficiently larger than the error amount b of the deepest portion of the defect (the pattern residual rate b / B is the threshold value). If it is smaller than G), it is regarded as normal. In this way, over-detection can be prevented.

In an etched circuit pattern such as a TAB tape, a fine short S as shown in FIG. 8 (a) occurs. The part of fine short S is 0.
Since the width is only 5 to several μm, the resolution is insufficient to obtain an accurate image with a camera. Therefore, when the camera captures the area of FIG. 8A, an image as shown in FIG. 8B is obtained. However, in the image of FIG. 8B, since the protrusions are present in two adjacent patterns, the gap amount A is smaller than that in the case of a single protrusion as shown in FIG. 8C. Therefore, by examining the gap residual rate as in the present invention, it is possible to detect a fine short circuit that could not be detected conventionally.

In this embodiment, the master pattern is created from the CAD data. However, a pattern to be measured which is determined to be a non-defective product is imaged, edge data is extracted from the pattern to be measured, and the straight line is extracted. The master pattern may be created by converting the master pattern into a master pattern.

[0036]

According to the present invention, as described in claim 1, the pattern to be measured is inspected based on the error amount and the gap residual rate, so that even if the error amount of the protrusion is a defect. When the gap A with the adjacent measured pattern is sufficiently large with respect to the error amount a at the tip of the protrusion, the measured pattern is regarded as normal, so overdetection can be prevented and the product yield is improved. Can be made. Also,
It is possible to detect a fine short circuit that could not be detected conventionally.

Further, as described in claim 2, by inspecting the pattern to be measured based on the error amount and the pattern residual rate, even if the error amount of the defect is a defect, the deepest part of the defect is detected. The pattern remaining amount B of the defective portion with respect to the error amount b
When is sufficiently large, the pattern to be measured is regarded as normal, so overdetection can be prevented, and the yield of products can be improved.

[Brief description of drawings]

FIG. 1 is a flow chart showing a pattern inspection method according to an embodiment of the present invention.

FIG. 2 is a block diagram of a pattern inspection device.

FIG. 3 is an explanatory diagram for obtaining a width of a master pattern.

FIG. 4 is a diagram for explaining a labeling process of a measured pattern.

FIG. 5 is a diagram for explaining correspondence between edge data of a measured pattern and a straight line of a master pattern.

FIG. 6 is a diagram illustrating a method of obtaining an error amount between a master pattern and a measured pattern.

FIG. 7 is a diagram for explaining a method of searching for a tip end portion of a protrusion and a deepest portion of a defect, and a method of calculating a gap residual rate and a pattern residual rate.

FIG. 8 is a diagram showing a state where a fine short exists between adjacent patterns and a diagram showing an image when the camera captures the fine short.

FIG. 9 is a diagram for explaining a conventional inspection method.

[Explanation of symbols]

1 ... Green sheet, 2 ... XY table, 3 ... Line sensor camera, 4 ... Image processing device, 5 ... Host computer, 6 ... Display device.

Claims (2)

(57) [Claims] 1. An edge data indicating a pattern edge is extracted from a grayscale image of a pattern to be measured, an error amount between a master pattern as a reference for inspection and the edge data is obtained, and a defect of the pattern to be measured is based on the error amount. The edge data of the measured pattern having the maximum error value a in the predetermined area centered on the defect candidate is detected as the tip of the projection, and the distance from the tip of the projection is the minimum. Then, the edge data of the adjacent measured pattern is obtained, and the distance at this time is set as the gap A with the adjacent measured pattern, and whether or not the defect candidate is a defect is determined based on the error amount and the gap residual rate a / A. A pattern inspection method characterized by determining whether or not. 2. An edge data indicating a pattern edge is extracted from a grayscale image of a measured pattern, an error amount between a master pattern as a reference for inspection and the edge data is obtained, and a defect of the measured pattern is detected based on the error amount. A candidate defect is detected, and the edge data of the measured pattern in which the error amount has the maximum value b in a predetermined region centered on the defect candidate is set as the deepest part of the defect and the deepest part of the defect is opposed to the deepest part. The distance between the edge data that constitutes both edges of the measured pattern together with the portion and the deepest portion is the remaining amount B of the pattern, and whether the defect candidate is a defect or not based on the error amount and the pattern remaining rate b / B. A pattern inspection method characterized by determining whether or not.