JPH06138165A - Pattern inspection method and device - Google Patents

Abstract

PURPOSE:To prevent excess detection of defects in wiring pattern by adding one pixel number at the center and the number of pixels in both oblique directions from the center after converting to the vertical can horizontal directions and multiplying a trigonometric function correction value to the added value for calculating the total length of line width in the oblique directions. CONSTITUTION:In image input part 32, wiring pattern is photographed, the electric signal according to the wiring pattern is binarized in binarizing circuit 33 and stored as an original pattern. The original pattern is processed by filtering in pre-processing part 34, measured with a length sensor in a radial length measurement part 35 in the radial direction from the center position in the line width direction, and the center position corresponding to the pixel is detected at center detection part 36 and sent to a coding circuit 37. The line width is measured basing on the pixel to be in the center position, coded to convert to radial code, which is referred to a dictionary in a defect judgment part 28 and the defect in the wiring pattern is detected from the goodness of the judgment.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pattern inspection method for automatically inspecting printed circuit board wiring patterns for defects.

In recent years, the wiring pattern formed on a printed board used has been densified with high complexity and miniaturization of the device. In the defect inspection of this wiring pattern, an automatic visual inspection has come to be performed.

In the automatic visual inspection, a wiring pattern is optically read, converted into an electric signal to obtain a binarized image, and a defect is detected from the image. The defect is detected from the binarized image. There is a demand for high accuracy with a small error.

[0004]

2. Description of the Related Art A conventional pattern inspection apparatus captures an image of a wiring pattern from an optical system such as an optical sensor, converts it into a binary image, and stores it in a binary image. The center of the pattern is detected by measuring the length of the binarized image, the pattern lengths in a plurality of directions are measured at the center, and the measured length value is encoded into a radial code. Then, the pass / fail judgment is made by associating the detected radial code with the dictionary in which the pass / fail of the radial code is stored in advance. This method has the versatility that dictionary data can be exchanged and can be applied to various patterns.

Here, the coding will be described. Figure 1
FIG. 0 shows a conventional coding circuit diagram. As shown in FIG. 10 (A), in the coding, after the center of the pattern is detected, the coding process is performed in four vertical, horizontal, and diagonal directions. , Length code (2
bit), balance code (1bit), direction code (1bi
t)).

Direction different by 180 ° (for example, 0 °
Direction (180 ° direction) and the sum (diameter) of the length measurement values of the radius are calculated. This value is divided into four values. Ie threshold
(1) When it is smaller than the above, it is regarded as “S” (short). First
If the threshold value is equal to or more than the threshold value and is equal to or less than the second threshold value (2), it is regarded as “C” (collect). Further, when it is larger than the second threshold value, it is set to “L” (long), and when the measured value on one side is “OV threshold value” or more, it is set to “OV” (overflow) and one of two directions. When the above is "OV",
The diameter is also "OV". A length code (2 bits) is created by this processing. In this case, the correspondence between each length and the code is OV = 1, L = 1, C = 2, S = 3.

Here, L is the diameter, and OV means that the measured length of the sensor is not less than the specified maximum length.

Next, the difference between the above two length measurement values is obtained.
When the value is equal to or less than the "balance margin" (that is, the balance is almost centered), the balance code is "1", and when not, it is "0". If one of them is OV, the balance is made when both are OV, and the one is unbalanced if they are OV.

When the balance is "0", the direction code (1 bit) indicates which is longer. It is "1" when the top is long, and "0" when the bottom is long. When the balance is "0", the direction code is also "0".

By the above processing, a 4-bit code is obtained for one set of directions. Find this in four directions,
Use 16-bit code. This is a radial code, and Table 1 shows the correspondence between each state and the radial code. It should be noted that the OVs are paired when both are OVs, and the OVs are longer when one is OV.

[0011]

[Table 1] Such processing is realized by the coding circuit shown in FIG. In FIG. 10B, first, assuming that the lateral outputs from the length measuring sensor are, for example, r1 and r9,
The r1 and r9 are input to the balance coding circuit 11.
In this balance coding circuit 11, the difference calculation circuit 12
Then, the difference between the outputs r1 and r9 is calculated, and the comparator 13 determines whether or not the difference is larger than a predetermined threshold margin (balance margin).

If not balanced, the comparator 14
And direction bit that indicates which is longer depending on gate 15
Is output. That is, the balance coding circuit 11
Output 2bit code of balance and direction.

On the other hand, the outputs r1 and r9 are simultaneously input to the length coding circuit 16. Here, first, the adder circuit 17 calculates the sum of both outputs, and then each comparator 1
8 and 19, the magnitude comparison with the predetermined first and second threshold values is performed, and a comparison output is output. These results are 2
It is encoded in the value of bit. That is, it indicates which range the length is for two thresholds. Further, the other r1 and r9 are compared with the OV threshold value by the comparators 20 and 21 to determine whether or not they are OV.

All these data are input to the data conversion circuit 22 to obtain a 4-bit code. Then, the above processing is performed for each of the length measurement values of the outputs r2 and r9, ... R8 and r16, and as a result, a 16-bit value (radial code) is obtained.

Next, FIG. 11 shows a diagram for explaining the measurement of the line width. FIG. 11A shows a wiring pattern 23 in which straight lines of a pattern (width) in the 90 ° direction and a pattern (width) in the 45 ° direction are combined. Figure 11
10B and 10C show an example of line width calculation, and FIG.
In (B), the line width is calculated by A + B in the directions of 0 ° and −180 °, and in FIG. 11 (C), I in the directions of 135 ° and −135 °.
The line width is calculated by NT (C × √2) + INT (D × √2). Note that INT means rounding into an integer.

A small square indicates one pixel, and A,
B, C, and D indicate the number of pixels, and the black squares indicate the central pixel. The length does not include the central one pixel, and the length is the sum of the measured values. In the case of FIG. 10 (B), the length of the line width of the straight line in the 90 ° direction is smaller than the actual length by 1 because the center pixel is not inserted, but the length is not exactly measured. The threshold value of 1 does not include the central pixel, so no error occurs.

[0017]

However, the length of the line width in the direction of 45 ° is the sum of the corrected length measurement values that have been corrected by multiplying the length measurement values by √2, and the diagonal of one pixel at the center. Since it is a value obtained by subtracting √2 pixels in the direction, it is smaller than the actual length minus one pixel at the center. For this reason, there is a problem that thin defects are excessively detected on the straight line in the 45 ° direction.

For example, as shown in FIG. 11 (D), when the pixels of only the target line width portion are shown, ◯ mark, ×
Considering the marked pixel as the center of the line width, the threshold value C (correction) of the line width is 6. Further, although the pixel marked with a circle and the pixel marked with a x have the same center of line width, the length of the line width calculated centering on the circle is 6,
The length calculated centering on the X mark is 5. Therefore, the pixel marked with a circle is C (correct) and is judged to be normal, while the pixel with a cross is S (short) and judged as a thin defect.

Therefore, the present invention has been made in view of the above problems, and an object thereof is to provide a pattern inspection method for preventing defect detection and performing defect inspection with high accuracy.

[0020]

FIG. 1 shows an explanatory view of the principle of the method of the present invention. In FIG. 1, a wiring pattern formed on a sample to be inspected is imaged to obtain a binarized wiring pattern corresponding to a pixel, and a plurality of radial lengths are measured from each central position of the binarized wiring pattern. Then, the process of the pattern inspection method for inspecting the defect of the wiring pattern by calculating the entire line width including the center position from the measured value and coding is shown. A plurality of radial lengths are measured from the center position. In step 2, thereafter, when calculating the diagonal line width total length, one pixel at the center position and the number of pixels in both diagonal directions from the center position are converted in the vertical direction or the horizontal direction and added. In step 3, the added value is multiplied by the trigonometric function correction value for the diagonal direction to calculate the total line width.

[0021]

As shown in FIG. 1, a wiring pattern is imaged to obtain a binary wiring pattern corresponding to a pixel, and a plurality of radial lengths are measured from each center position. This measurement is detected corresponding to the pixel of the length measuring sensor.

When calculating the total line width in the diagonal direction by this measurement, one pixel at the center position and the number of pixels in both diagonal directions from the center position are converted in the vertical or horizontal direction and added, and a trigonometric function is added to the added value. The total line width is calculated by multiplying the correction value. This trigonometric function correction value is converted into an oblique angle with respect to the vertical or horizontal direction, and the actually measured length is calculated.

That is, since the actually measured length is calculated after conversion in the vertical or horizontal direction and addition, the length measurement accuracy can be improved and excessive detection can be prevented.

[0024]

FIG. 2 shows a block diagram of an embodiment of the present invention.
FIG. 2 shows an overall configuration diagram of the pattern inspection apparatus.

In the pattern inspection apparatus 31 shown in FIG. 2, the wiring pattern formed on the sample to be inspected, which is the inspection object, is arranged, and the image input unit 32 has an optical system such as an illumination, a lens, and an optical sensor (CCD sensor). The wiring pattern is imaged. In the image input unit 32, an electric signal corresponding to the imaged wiring pattern is sent to the binarization circuit 33.

In the binarization circuit 33, this electric signal is binarized and stored in the storage area as an original pattern (binarized wiring pattern). The original pattern is subjected to filtering processing such as smoothing and edge extraction in the preprocessing unit 34, and is sent to the radial length measuring unit 35. In the radial length measuring unit 35, a length measuring sensor (described later) is arranged radially from the center position in the line width direction.
The center position corresponding to the pixel is detected by the center detector 36 and is sent to the encoding circuit 37.

Then, in the coding circuit 37, the line width is measured based on the pixel at the center position, coded and converted into a radial code. This radial code is used by the defect determination section 38 to refer to the dictionary and detect a defect in the wiring pattern based on its quality.

Each part of the pattern inspection device 31 will be described in more detail.

FIG. 3 is a diagram for explaining the operation principle of the radial length measuring unit in FIG. 2, and FIG. 4 is a diagram for explaining the arrangement of the length measuring sensors.

FIG. 3A shows a radial length measuring sensor circuit in the radial length measuring unit 35, which is provided with, for example, one shift register constituted by continuously connecting flip-flops for 7 lines. The pattern binary image is operated on this radial length measuring circuit. The radial measurement circuit has a 7-bit x 7-bit radial measurement sensor area 4
1 is provided. As shown in FIG. 3B, the radial length measuring sensor area 41 is provided with radial length measuring sensors 42 _{1 to} 42 ₈ having pixels radially extending in eight directions from a bit at the center thereof. . The radial length measurement sensors 42 _{1 to} 42 ₈ are for detecting the number of pixels having “1” in each of the eight directions, that is, the length of the wiring pattern portion. In FIG. 3B, numerals 0 to 3 indicating positions from the center are shown along each direction.

Although FIGS. 3 (A) and 3 (B) have a simple structure for the sake of explanation, in practice, as shown in FIG. 3 (C), for example, the length measuring arm of the radial length measuring sensor has A1 to A1. A16
Are arranged in 16 directions, and the corresponding r1 to r16
Indicates the output value (measured value).

Here, FIG. 4 shows a diagram for explaining the arrangement of the length measuring sensors. In FIGS. 4A to 4D, the radial length measurement sensor area is actually 63 bits × 63.
The measuring arm A1 which is composed of bit areas and extends in the direction of the line sensor 16 from the center as shown in FIG.
It has an area of A16 and each measuring arm Ai (i = 1 to 1)
6) has a length of 32 bits (including the central bit). With this radial length measurement sensor, 0 ° in each direction from the center pixel of the radial length measurement sensor (Fig. 4 (A)), 22.5 °
(Fig. 4 (B)), 45 ° (Fig. 4 (C)), 67.5 ° (Fig. 4)
(D)), 337.5 ° wiring pattern length measurement values r1 to t
16 is obtained.

Next, FIG. 5 shows a circuit diagram of the center detector of FIG.

In FIG. 5, the output r from the length measuring sensor
1 and r9 are input to the balance calculation circuit 51 for opposite dimension measurement. Here, first, the difference calculation circuit 52 calculates the difference between the two outputs, and then the comparator 53 determines whether or not the difference value is larger than a predetermined threshold margin. Then, the output of the balance detection signal is input to the balance condition determination circuit (ROM) 54. Similarly, the sets r2 and r10, r3 and r1 of the outputs from the length measurement sensor
.., r8 and r16 are input to the corresponding counter measurement value balance calculation circuits 51, and balance detection signals are input to the balance condition determination circuit 54, respectively. The output is used as a balance condition signal (center signal), and the center is when the balance is equal to or more than a specified number.

Next, FIG. 6 shows a circuit diagram (2) of the coding circuit of FIG. 2, and FIG. 7 shows a circuit diagram (2) of the coding circuit of FIG. FIG. 6 shows the case of encoding into one direction code (4 bits), and FIG. 7 shows the case of encoding into each of the other directions, and the circuit diagram of FIG. 6 is provided for each direction. The same parts as those in FIG. 10B are designated by the same reference numerals.

In FIG. 6A, the coding circuit 37
Is a length calculation circuit (ROM) 62 and two comparators 6 instead of the length coding circuit 16 in FIG.
A length coding circuit 61 composed of 3, 64 is provided. In this case, FIG. 10B shown in FIG.
The length coding circuit 16 may be provided in parallel, or only the length coding circuit 16 may be provided depending on the pattern inspection method.

Further, in FIG. 7, the corresponding one-way coding circuit 71 shown in FIG. 6A is applied to the corresponding straight line sets r2 and r10, ... R8 and r16 shown in FIG.
And a gate circuit 72 generates a radial code (16 bits) based on the center signal from these outputs.

Here, FIG. 8 shows a diagram for explaining the principle of line width measurement in FIG. 8A and 8B show pixels in which the line width is detected in the radial length measuring unit 35. FIG. 7A shows a case of 0 ° and FIG. 8B shows a case of −45 °. 11 (B) and 11 (C).

First, in the first length measurement method, in FIG. 8A, one pixel at the center position (black-painted portion) and the number of pixels in both diagonal directions from the center position are added (A + B +).
1), there is no difference from the conventional method. +1 in this case
Is the difference in the method of setting the threshold value at the time of encoding, and the result is the same.

On the other hand, in FIG. 8B, the length of the line width is one pixel in the center and the number of pixels of the measured values on both sides thereof is the number of pixels in the horizontal direction (or in the vertical direction) (C, D). Convert to and add. In this case, the line width is INT {(C + D + 1) × √
2}. Note that √2 is 45 ° (=
It is a trigonometric function correction value (1 / cos θ) when converting the horizontal conversion value in the case of θ) into the measured length. Further, INT is a symbol for rounding a numerical value to make it an integer. This process can reduce the over-detection of thin defects.

Specifically, using FIG. 11D, the line width lengths at both the line width centers O and X are 7. In this case, since the threshold value C (correction) of the line width is 7, both O and × are C (correction),
It is judged to be normal.

The second length measuring method is as shown in FIG.
One pixel at the center position is excluded as A + B.
The length of the one pixel is set in consideration by the first and second coding thresholds in
It is similar to the length measurement method of. The same applies to the third length measurement method described later.

In FIG. 8B, the line width is INT {(C +
D) × √2}, which also can reduce the excessive detection of narrow defects in the wiring pattern. Specifically, as shown in FIG. 11D, the line width lengths at the line width centers ◯ and X are both 6. In this case, since the threshold value C (collect) of the line width is 6,
Both ○ and × are C (correct), and are judged to be normal.

Next, in the third length measuring method, the line width is INT {C × √2 + (√2-1) / 2} + INT in FIG. 8B.
It is shown by {D × √2 + (√2-1) / 2}. Here, (√2-1) / 2 is the correction value of the actual measurement error in the diagonal direction in one pixel. That is, the length of the line width is calculated by adding the value obtained by subtracting 1 pixel from the diagonal length of 1 pixel at the center to the length of the actual measured value and dividing by 2 and rounding off. Is the sum of This process can reduce the over-detection of thin defects.

Specifically, as shown in FIG. 11D, the line width lengths at the line width centers ◯ and X are both 6. In this case, since the threshold value C (correction) of the line width is 6, both O and × are C (correction), and it is determined that the line width is normal.

Then, returning to FIG. 6 and FIG. 7, an example of practical length calculation will be shown using the length measurement results r1 to r16 of FIG. In the first length measurement method described above, the length coding circuit 61 of FIG. 6A is used, and Table 2 shows an example of length calculation in the 45 ° direction in which the error occurs most. Table 3 shows an example of the calculation. r3 and r11 are 45 ° and 22 respectively
The corrected length measurement value (number of pixels) in the 5 ° direction is shown. r3
Input the value of r11 and output the numerical value written in the table as the length.

[0047]

[Table 2]

[0048]

[Table 3] Note that the blank circles in Tables 2 and 3 indicate skips (see FIG. 4C).

Then, Table 4 shows the error with respect to the actual length of the length calculation example of the present invention, and Table 5 shows the error with respect to the conventional length calculation example as a comparison.

[0050]

[Table 4]

[0051]

[Table 5] The error in the calculation example of the present invention in Table 4 is in the range of -0.5 to 0.3 pixels, while in the calculation example of Table 5, the threshold value is smaller by one pixel at the center, so the value obtained by adding one pixel at the center is set. . The shaded values indicate that the error with respect to the actual length is larger than the error of the first length measurement method. That is, in the conventional calculation, the error is −0.1 to −1.2 pixels, and there are many places where the error is large in the shaded portion as compared with Table 4. Therefore, it is possible to prevent the excessive detection of the thinness by using the new length calculation.

Subsequently, in the above-described second length measuring method, horizontal and vertical directions r1 and r9, r5 and r13 are shown in FIG.
The length coding circuit 16 shown in FIG. 6B is used, and the other lengthwise diagonal directions are calculated using the length coding circuit 61 shown in FIG.

As a result, a conventional circuit can be used as a circuit, and the present invention can be easily implemented.

Then, in the third length measuring method described above, FIG.
Only the length coding circuit 16 of (B) is used. An error with respect to the actual length according to the present invention is shown in Table 6 by taking 45 ° as an example of the calculation in this case, and compared with Table 5.

[0055]

[Table 6] As shown in Table 6, the error is in the range of −0.9 to 0.8 pixels, and when compared in the shaded area, there are few places where the error is larger than in Table 5, and the conventional length code as shown in FIG. The circuit including only the digitizing circuit 16 can reduce excessive detection of slender thin defects.

Next, FIG. 9 shows a diagram for explaining the defect determining section of FIG. FIG. 9 shows the gate circuit 7 in FIG.
The radial code (16 bits) output from No. 2 is input as an address to the defect determination dictionary (ROM) 73, and data on whether the thin defect of the wiring pattern is good or bad is output. It should be noted that the quality judgment result is recorded in advance.

[0057]

As described above, according to the present invention, one pixel at the center position and the number of pixels in both diagonal directions from the center position are converted in the vertical direction or the horizontal direction and added, and the added value for the diagonal direction is added. By calculating the total diagonal line width by multiplying the trigonometric correction value, it is possible to prevent excessive detection of defects in the wiring pattern and perform defect inspection with high accuracy.

[Brief description of drawings]

FIG. 1 is a diagram illustrating the principle of the method of the present invention.

FIG. 2 is a configuration diagram of an embodiment of the present invention.

FIG. 3 is a diagram for explaining the operation principle of the radial length measuring unit in FIG.

FIG. 4 is a diagram for explaining the arrangement of length measuring sensors.

5 is a circuit diagram of a center detector of FIG.

6 is a circuit diagram (1) of the encoding circuit of FIG.

7 is a circuit diagram (2) of the encoding circuit of FIG.

8 is a diagram for explaining the principle of line width measurement in FIG.

FIG. 9 is a diagram for explaining the defect determination unit in FIG.

FIG. 10 is a conventional coding circuit diagram.

FIG. 11 is a diagram for explaining measurement of line width.

[Explanation of symbols]

 31 pattern inspection device 32 image input unit 33 binarization circuit 34 preprocessing unit 35 radial length measurement unit 36 center detection unit 37 coding circuit 38 defect determination unit 16,61 length code circuit 62 length calculation circuit 63, 64 comparison vessel

Claims (4)

[Claims] 1. A wiring pattern formed on a sample to be inspected is imaged to obtain a binarized wiring pattern corresponding to a pixel, and a plurality of radial lengths from each central position of the binarized wiring pattern are obtained. A pattern inspecting method for inspecting a defect of the wiring pattern by measuring the total line width including the center position from the measured value and encoding the measured line length, and measuring a plurality of radial lengths from the center position. A step of converting one pixel at the center position and the number of pixels in both diagonal directions from the center position in the vertical direction or the horizontal direction and adding when calculating the total line width in the diagonal direction after measurement; A step of multiplying a value by a trigonometric function correction value in the diagonal direction to calculate the line width total length, and a pattern inspection method comprising: 2. The pattern according to claim 1, wherein when converting the number of pixels in the vertical direction or the horizontal direction, one pixel at the center position is excluded and converted, and correction is performed in the encoding. Inspection method. 3. A wiring pattern formed on a sample to be inspected is imaged to obtain a binarized wiring pattern corresponding to a pixel, and a plurality of radial lengths from each central position of the binarized wiring pattern are obtained. A pattern inspecting method for inspecting a defect of the wiring pattern by measuring the total line width including the center position from the measured value and encoding the measured line length, and measuring a plurality of radial lengths from the center position. When calculating the total line width in the diagonal direction after measurement, the number of pixels in both diagonal directions from the center position is converted into the vertical direction or the horizontal direction, and each is multiplied by a trigonometric function correction value for the diagonal direction. And a step of adding a correction value of a diagonal measurement error in one pixel of the center value to each of the multiplied values, and a step of adding each of the added values to calculate the total line width And a pattern inspection method comprising: 4. A wiring pattern formed on a sample to be inspected is imaged to obtain a binarized wiring pattern corresponding to a pixel, and a plurality of radial lengths from each central position of the binarized wiring pattern are obtained. In a pattern inspection apparatus that performs defect inspection of the wiring pattern by measuring and coding the total line width including the center position from the measured value, a plurality of radial lengths from the center position are measured. When calculating the total line width in the diagonal direction after measurement, one pixel at the center position and the number of pixels in both diagonal directions from the center position are converted in the vertical direction or the horizontal direction and added to obtain the added value. A pattern inspection apparatus including a length encoding circuit (61) for multiplying and encoding the trigonometric function correction value in the diagonal direction.