JPH0228888A - Pattern inspection method - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for inspecting patterns such as printed circuit patterns. The present invention relates to a pattern inspection method that can accurately detect not only the presence or absence of defects such as wire breaks, but also their positions.

[Prior Art] For example, a circuit pattern drawn on a ceramic green sheet has many uneven edges a as shown in FIG. This unevenness a on the pattern edge is not a defect in the circuit pattern, so it is an original defect (broken wire, half-broken wire C,
When detecting short d, half-shi i-1 to e), unevenness a
It is necessary to avoid erroneously detecting defects b - e.

Further, since the positional deviation f is not a defect, it is necessary to prevent it from being erroneously detected as a defect.

When inspecting the above-mentioned circuit pattern, conventionally, as described in Japanese Patent Laid-Open No. 59-192945, the detected pattern is binarized, and this binarized pattern is expanded by a certain amount.
After shrinkage, the patterned connections between specific positions are extracted, and the extracted connections are compared with non-defective product data or design data to detect defects.

[Problems to be Solved by the Invention] The above-mentioned conventional technology can accurately detect the presence or absence of a defect even when the pattern edge has irregularities. However, since the presence or absence of a defect is detected using one continuous pattern as a unit, it is not possible to specify where in the pattern the defect exists, and no consideration is given to this.

However, in recent years, circuit pattern sheets have become larger and the length of one pattern has tended to become longer. Therefore,
Even if it is possible to detect only the presence or absence of a defect, if it is not possible to identify where in the circuit pattern the defect is located, it will take time to visually confirm the defective location of the pattern that is determined to have a defect, and it will be necessary to correct the defect. The problem is that a lot of time is wasted until the process is completed.

An object of the present invention is to provide a pattern inspection method that can accurately detect not only the presence or absence of a defect but also its location.

[Means for Solving the Problem] In order to achieve the above object, the invention according to claim 1 binarizes the detected pattern, expands the binarized pattern,
The connection relationships of the expanded patterns are extracted, the extracted connection relationships are compared with the connection relationships of normal patterns, and errors in patterns that should have been separated but have incorrect connection relationships are detected. Divide the edge coordinates into pads that exist on the pattern, extract a series of edge coordinates that incorrectly connect pads from the segmented edge coordinates, and find the minimum distance between the series of edge coordinates. Let the pair of edge coordinates be the position coordinates of a short or half-short defect. Moreover, in the invention according to claim 2,
Binarize the detected pattern, shrink the binarized pattern, extract the connection relationship of the contracted pattern, compare the extracted connection relationship with the connection relationship of the normal pattern, and check whether the pattern should originally be connected as one. If the patterns are separated, the edges of those patterns are detected, and the pair of edge coordinates with the minimum distance between the series of edge coordinate series is determined as the position coordinates of the wire breakage or semi-breakage defect.

[Action Co. In the invention described in claim 1, the binarized pattern is expanded.

As a result, not only short circuits but also half short circuits among defects are represented as connected states. The fact that originally separate locations are connected indicates that there is a short or half-short defect, which is detected by comparing the connection relationship between the expansion pattern and the normal pattern. Then, an edge coordinate series of a connection state location where pads are illegally connected is extracted. Since the part represented by the connected state due to a defect is usually thinner than the width of the expanded pattern of the original pattern, the pair of coordinates with the minimum interval between the edge coordinate series indicates the defect location. .

In the invention according to claim 2, the binarized pattern is contracted.

As a result, among defects, not only disconnections but also half-disconnections are represented in a separated state. The fact that a place where the wires are originally connected is in a separated state indicates that there is a defect such as a wire breakage or a half wire breakage, and this is detected by comparing the connection relationship between the contracted pattern and the normal pattern. Then, the edge coordinate series of this separated state location is extracted, and the coordinate pair with the minimum interval between the edge coordinate series indicates the defective location.

[Example] Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

FIGS. 1(a) and 1(b) are diagrams explaining the present invention in detail. The detected binarization pattern ■ illustrated in FIG.
The configuration is such that the pads ■-■, ■-○, and ■-■ are connected (+- is attached to the center of each pad ■-■), and in the example shown, There are defects of half open wire C and half short e. First, by expanding the detected binarized pattern, an expanded pattern (2) is obtained. Further, by contracting the detected binarized pattern, a contracted pattern (2) is obtained. As a result, a half short circuit is converted to a short circuit (a short circuit remains a short circuit, of course), and a half open circuit is converted to a disconnection (a disconnection remains a circuit disconnection). From the binarized expansion or contraction patterns ■, ■ converted in this way, the connection relationships based on the pattern between the pad center positions whose positions are determined in advance are extracted, and as shown in the figure, the directed tree structure data ■, Expressed as ■. This will be called connection data. The connection data is a directed tree with a depth of 1, and the arrows indicate the connection relationships between pads. For example, the connection data shown in FIG.
Indicates that ■, ■, and ■ are connected. Of these,
The pad (pad ■ in the example of FIG. 4) that is the root of a directed tree and has an arrow pointing toward itself is called a parent pad. There is only one parent pad in the connection data for one connected pattern.

On the other hand, the connection relationship between normal pads is represented by a -fold cyclic list structure, as illustrated in FIG.

This is called design data. Follow the arrow to cycle through the list 1
By going around, you can get all the pad numbers to be connected. In the case of FIG. 5, it is shown that pads ■, ■, ■ should be connected.

Now, suppose that the connection data (2) and (2) as shown in FIG. 1 are extracted for the expanded binary pattern (2) and the contracted binary pattern (2), respectively.

When we examine the connection relationship corresponding to each arrow on the design data for the connection data ■ for the expansion pattern ■, we find ■→
The connection relationships of ■ and ■→■ are not included in the design data. From this result, pads ■, ■, ■.

It can be seen that the pattern containing ■ is short (or semi-circumcised). On the other hand, if we examine the parent pads ■, ■, ■, ■ of the connection data V for the contraction pattern ■ on the design data, we find that the parent pads ■, ■ are the only parent pads in one circular list of the design data. For parent pad ■,
(2) coexists in one cyclic list. This indicates that this pattern is separated into two parts, so the pad ■
.

It can be seen that the pattern containing ■ is a disconnection (or semi-disconnection).

With the above steps, patterns with defects are detected,
The location of the defect cannot yet be determined.

First, the edge of the pattern where the defect has occurred is traced, and a series of edge coordinates is determined.

The edge coordinate series for the expansion pattern is (xo.

yo)-(x□, yl)--(Xn+yn), find the center position and adjacent edge coordinates of pads ■, ■, ■, ■ in the pattern, and calculate these by (Xixryi
tL (Xi4+ yiJ+ (Xis+ y, 5
Let L (X; +; +y, 6) (Figure 6). By these, the edge coordinate series (Xo+yJ-...-(X
When dividing n+ yn), (Xix+ yil)
”' (Xj6+ ytGL(Xis+ 3'iG)
”' (Xis+ yisL (X+s+ yi
s)...' (Xi4+ Xi4)+ (Xi4+
Xi4)"' (Xix+y, □) is divided into four parts. There are connections to root, pad ■→■, and ■→■ that refer to the design data, but there are no connections between ■→■ and ■→■. .Therefore, (XiG.

Xi6) ””(Xis+ yis)+ (Xi
<+ Xi4) "'(X, □+yix) is an invalid edge coordinate series. Next, calculate the distance between each point of these two edge coordinate series, and find the minimum edge coordinate pair, (x++yix). Yl□).

Let (X 121 y +t) be the short (or half-short) position (FIG. 1). On the other hand, the edge coordinate series for the contracted pattern with defects is (X
O' g yo' )-...-(xp lypo
L (Xo1+y n')-...-(Xqo, yq'
) (Figure 1), calculate the distance between each point in these two edge coordinate series, and find the minimum edge coordinate pair, (
Let X j 0+ yj 0) + (xj1+ yjl) be the disconnected (or half-broken) position.

FIG. 3 is a block diagram of an apparatus for carrying out the above-described pattern inspection method of the present invention.

The pattern detected by the pattern detector 1 is converted into a binary pattern by the binarization circuit 2, and the pattern is converted into a binary pattern by the dilation circuit 3 and the contraction circuit 4.
Then, they are respectively expanded and contracted and then stored in the memories 5 and 6. As the pattern detector 1, a TV camera, a CCD line sensor, a constant or fast moving stage, etc. can be used, and any device that photoelectrically converts a two-dimensional optical image of a pattern to be inspected can be used. In order to faithfully detect the optical image of the circuit pattern of the object to be inspected, depending on the object,
Any known detection method such as bright field illumination detection, transmitted illumination detection, or trend detection may be used. Furthermore, as the binarization circuit 2, a publicly known fixed threshold binarization method, floating threshold binarization method, etc. can be used. An optimal binarization method may be used depending on the situation. FIG. 7 shows a specific example of the expansion and contraction circuit. The input binary pattern signal is sequentially input from an input terminal 20 to a shift register 21 having four serial inputs and serial outputs. The length of the shift register 21 is determined by the length of the pattern detector 1 (FIG. 3) in the main scanning direction (in the case of a TV camera,
It is assumed to be equal to the number of pixels in the horizontal scanning direction (in the case of a CCD line sensor, its self-scanning direction). The 5×5 D flip-flops 22 and the shift register 21 operate synchronously according to the same clock. Select the output of D flip-flop 22,
If it is led to the ND gate 23, a contracted image signal 25 is obtained, and if it is led to the OR gate 24, an expanded image signal 26 is obtained. 7th
The example shown in the figure is an example of a circuit that contracts and expands two pixels, respectively, and the contraction and expansion circuits can be realized with a similar configuration. By scanning the pattern detector or the object to be inspected, the expansion and contraction patterns of the entire pattern to be inspected are stored in the memory 5,
6 (FIG. 3).

Memories 5 and 6 are connected to bus 7 of microcomputer 9. After the entire pattern is stored in the memories 5 and 6, the contents of the memories 5 and 6 are sequentially read out by the connection relation extraction device 8, and the pad coordinates and numbers input in advance from the external storage device 10 of the microcomputer 9 are read out in sequence. Connection data is generated according to the information. Specifically, the connection data has a data structure as shown in FIG.
The connection data (■) of the expansion pattern (■) shown in the figure is as shown in FIG. That is, the pad number indicating the starting point of the arrow of the directed water in the connection data is set to be the address, and the parad number indicating the ending point is set to be the data. The connection relationship extraction device 8 may be a device that extracts connection relationships through software processing by a normal microcomputer, a device that processes signals using hardware to extract connection relationships, or a device that extracts connection relationships by processing signals using hardware, or a device that extracts connection relationships by processing signals using hardware. Various types of equipment can be used. The extracted connection data is transferred to the microcomputer 9 and compared with design data input from the external storage device 10 to detect circuit patterns in which defects have occurred.

Next, the microcomputer 9 references the memory to trace the edges of the pattern in which the defect has occurred,
Generate edge coordinate series. The edge tracking method will be described using FIG. Here, the connectivity of the pattern is defined as 4 connections, that is, connections to the top, bottom, left and right of the pixel of interest. The same applies to the case of 8 connections. First, pixels are examined in one direction from one point in the pattern, that is, one of the pad center positions (X mark) in the pattern, in the case of Fig. 10, to the right, and then the pixels are examined from one point outside the pattern. The point in front of you is the starting point (Xoyyo). Next, according to Table 1, (
The pixels around Xo+yo) are examined in order, and the direction in which the first pixel of the pattern is hit is defined as the next edge point (X, yl). In the case of Fig. 10, since the starting point was in the direction of E (see Fig. 11), turn N.

I checked E and S in that order and found the first pixel of the pattern in the S direction, so I moved in that direction and found the next edge point (xl.

yx). Next, check E, S, and W in that order and proceed in the W direction. Return to (Xo+Mo) and repeat the above processing until the next point becomes (X-work, yl).

Table 1 Through the above processing, the clockwise edge coordinate series of the pattern (X o + '/ o ) (X 1 * V
1) It is possible to obtain...

Next, division points as shown in FIG. 6 are found from the edge coordinate series obtained from the expansion pattern and division is performed. For example, in the case shown in FIG. 13(a), the distance d of the edge coordinates is calculated with respect to the pad center position (x mark) in the pattern. The relationship between this distance d and the coordinate series is the 12th
The curve shown in Figure (a) is formed, and the center of the point where the distance d is less than a certain threshold Th is defined as the dividing point (X+s+yts) of the edge coordinate series based on the pad center position. Figure 12 (
b), as shown in FIG. 13(b), if the pad center position is not at the end point of the pattern, multiple division points (X1S
O+3'is. ) (X1S1s yes). The above process for determining division points is performed for all pad center positions within the pattern. Thereafter, design data is referred to from the divided edge coordinate series to extract an incorrect edge coordinate series, and the coordinate pair with the minimum distance between mutual edges is determined as the position of the short or half-short defect.

Furthermore, the edge coordinate series of the pattern in which the wire breakage or half-wire breakage occurs, obtained from the shrinkage pattern, is at least 2
Since the wire consists of two series, the coordinate pair with the minimum distance between mutual edges is determined as the position of the wire breakage or half wire breakage defect.

Note that the calculation of the distance between the pad center position and the edge point or between the edges is as follows: 1Δx1+1Δy1 or (Δx)'+(Δ
y)′ or ~α1−εF]]11Y, but there is an advantage that calculating 1Δx1+1Δy requires the least calculation time.

[Effects of the Invention] According to the present invention, by expanding and contracting the detected binary pattern and comparing the connection relationship of the pattern with design data,
Since defects are detected, only the critical defects can be faithfully detected even in patterns with uneven edges, misalignment, and local deformation. Also,
Since it is possible to focus on the edge of the pattern where the defect occurs and even pinpoint the defect location, it has the effect of significantly reducing the number of man-hours required for manual defect confirmation and correction work following automatic inspection. be.

[Brief explanation of the drawing]

1(a) and (b) are explanatory diagrams of a pattern inspection method according to an embodiment of the present invention, FIG. 2 is an explanatory diagram of defects in a circuit pattern, and FIG. 3 is a diagram of a pattern according to an embodiment of the present invention. A configuration diagram of a pattern inspection device to which the inspection method is applied, Figures 4 and 5 are explanatory diagrams of connection data and design data, Figure 6 is an explanatory diagram of dividing points of edge coordinate series, and Figure 7 is an expansion/contraction circuit. 8 is a data configuration diagram of connection data, FIG. 9 is a data configuration diagram of design data, FIG. 10 is an explanatory diagram of edge coordinate series generation, FIG. 11 is an edge tracking direction definition diagram,
Figure 12 (a), (b) and Figure 13 (a), (
b) is an explanatory diagram of edge division point calculation. 1... Pattern detector, 2... Binarization circuit, 3...
- Expansion circuit, 4... Contraction circuit, 5, 6... Memory,
7... Bus, 8... Connection relationship extraction device, 9... Microcomputer, 10... External storage device, 11.
...Defect determination output. Agent Patent Attorney Tadashi Akimoto Figure 10

Claims (2)

[Claims] 1. Binarize the detected pattern, expand the binarized pattern, extract the connection relationship between specific coordinates on the pattern, compare it with the connection relationship of a normal pattern, and identify incorrect connections in patterns that should have been separated. Detect the edge coordinate series of the pattern where there is, divide the edge coordinate series at points close to the specific coordinates, and create two edge coordinate series that connect the patterns to be separated among the divided edge coordinate series. A pattern inspection method characterized in that a pair of edge coordinates having a minimum distance between each point of the edge coordinate series is determined as the position coordinates of a short or half-short defect. 2. Binarize the detected pattern, shrink the binarized pattern, extract the connection relationship between specific coordinates on the pattern, compare it with the connection relationship of a normal pattern, and find out which patterns should originally be connected into one. If they are separated, detect the edge coordinate series of those patterns, and find the edge coordinate pair that has the minimum distance between the points in the edge coordinate series,
A pattern inspection method characterized in that the position coordinates of a wire breakage or a half wire breakage defect are determined.