JPS59119245A - Automatic positioning device - Google Patents

Abstract

PURPOSE:To position a pattern to be detected in precise relation to an optical detection system easily by using the optical detection system which has a pattern defect and detecting the mutual position relation of the pattern to be detected. CONSTITUTION:An image of the pattern 1 to be detected is formed on a linear image sensor 2a through an objective 2b and an X-Y feed mechanism 56 operates for every line scan on the sensor 2a. A video signal (a) is inputted to a flip-flop group 8d through a binary-coding circuit 8a and shift registers 8b and 8c, and when the number of boundary point signal (f) from a gate 8e exceeds a specific value, a flip-flop 8g is set by a preset counter 8f. Then, only when the output of an exclusive OR gate 10a is (1), Y-axial measuring table feed pulses (g) are counted by a counter 10d and an offset direction is detected by a flip-flop 10b. A processor 11 operates a turntable 5a on the basis of said data to adjust the inclination of the pattern to be detected to zero.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to a device for automatically aligning the installation position of a pattern to be inspected to an accurate position in a device for inspecting defects in printed wiring patterns. be.

[Prior art]

In a printed wiring pattern defect inspection device that compares and checks two identical patterns to be inspected side by side, the system shown in Fig. 1 (
As shown in α), there are multiple identical wiring patterns (A, B, C
, D...) Corresponding points of adjacent identical patterns (for example, A and B) on the arranged substrate? By performing a comparative check, pattern defects can be inspected.

As shown in FIG. 1 (Al), this device uses one-dimensional image sensors 2α and 3α of a plurality of optical detection units 2 and 3 to focus on corresponding points in minute areas on the left and right patterns and detect each. Detects the “presence” of both patterns,
゛If the two are equal, it is normal; if they are different, the At
It is considered a defect.

In the above, the scanning of the point of interest in the lateral direction c (hereinafter referred to as the X direction) is the -dimensional image sensor 2α.

3b, and the vertical direction C (hereinafter referred to as the y direction) is carried out by moving the substrate 1 to be tested in the y direction. In such a comparison check method, it is assumed that the test patterns A, B, .

3α needs to be spaced at intervals equal to the horizontal arrangement pitch of the pattern to be tested. Furthermore, image sensor 2
Regarding the arrangement of α and 5α, both are placed exactly on the X-axis (or an axis parallel to the X-axis), and the position of the test pattern relative to this is set at an inclination angle θ (two test patterns A).

It is a necessary condition that the angle formed by the straight line connecting the corresponding points of B and the X-axis is set to zero.

Conventionally, reference line markers M and A are provided in the x-movement direction on the left and right sides of the test board as shown in FIG. The rotary table 5 on the X-Y feed mechanism 5h constituting the measurement table 5 detects the substrate 1 to be tested so that the positions in the y-axis direction are equal on the left and right sides according to a command from the control device 4.
The inclination was adjusted by rotating it by α. However, in such a method, markers 1i and N are required on the left and right sides of the original inspection board 1, and the image sensor 2αI3α described above is required.
and optical detection units 2 and 3 equipped with objective lenses 2h and 5b.
Two reference position detectors are required on the side, and if the marker position i is changed due to an individual change of the pattern, it is necessary to reset this and the position of 7 for that city, etc. Therefore, the mechanism tends to be complicated and handling is inconvenient. Here, in particular, the position setting Fi of the reference position detector 6.7 has to be performed repeatedly, which requires high precision, which is extremely inconvenient. Therefore, it has been desired to develop a method that uses a simple mechanism to easily and accurately align the test pattern.

[Purpose of the invention]

SUMMARY OF THE INVENTION An object of the present invention is to provide an automatic printed circuit board alignment device that can accurately and easily align a test pattern with respect to a detection optical system.

[Summary of the invention]

If the inclination of the test pattern could be detected by the image sensor 2α and Riz itself, it would not be necessary to set the reference line marker described above, but also to set the reference position detector to detect this, and the above-mentioned drawbacks would be eliminated. It should be possible.

The main point of the present invention is to automatically align the substrate to be tested by detecting the mutual positional relationship between the patterns A and B using the detection optical sections 2 and 3 that detect original pattern defects. shall be.

The test pattern is composed of various pattern elements.
Since the dimensions and lengths of each element are various, we have prepared a pattern element that is advantageous for detecting the inclination angle in advance by using a corresponding rectangular specimen whose longitudinal direction is parallel to the X-axis as shown in Fig. 2. Select pattern elements A and Bf. In this method, coordinate data on the substrate of elements selected by visual inspection or the like is stored in advance, and pattern elements α and b are moved to the vicinity of the center of the field of view of the optical detection section using this coordinate data. Now, as shown in FIG. 2, the lower edges Eα and E of each of the pattern elements AB
Assume that b is shifted by i in the y direction. Here the first
When the measurement table 5 shown in the figure is sent down in the y-axis direction by minute steps (image sensor field of view 2 towns, unit length equal to the length of each picture element 3α1), the image sensors 2α, 3
The image of α changes as shown in FIG. 2 (α) to rC+. That is, the image sensor field of view 2α.

The "brightness" or "darkness" (corresponding to "pattern" 1 or "absence" of pattern) of each picture element in 3 towns changes as shown by "1#" or "0#" in the figure.Second In the figure (α) is edge E
The state before α and Eh are detected, (hl is the state when the left image sensor 2α detects the edge Eα, (C1 is the state when the right image sensor 3a detects the edge Eh. As a judgment condition for edge detection , there is no pattern (“
It is necessary for the point that was 0 to become the pattern "C"(1#) this time, but when this is simultaneously established for multiple picture elements "n or more 1" within the field of view of the image sensor, it is considered edge detection. do. The y-axis direction feed t from when one image sensor detects an etch until the other image sensor detects an etch is equal to the y-axis direction offset value ΔY of the left and right test patterns. Once the value of d is determined, the tilt angle θ can be easily determined using the following formula, and the substrate plate may be rotated so as to make the value of θ zero.

θ=5. n, where t is the horizontal arrangement interval between the test patterns A and B. Regarding the above-mentioned "nine or more", in principle this may be "one or more", but if it is one, the unevenness of the pattern If a defect like this exists on an etch, there is a risk of false detection, so it is better to set 'n' to an appropriate number.

Also, the running angle θ is originally small, about 1710 for TA mil.
Since the number of edge detection pixels is approximately the same as that of the following, a large number of edge detection picture elements are generated at the same time.

The value of offset i detected here is ±1 pit (±1
Step) Contains the following errors.

In order to suppress this error below an allowable level, it is necessary to make the pixel detection resolution of the optical detectors 2 and 3 sufficiently higher than the resolution for the final pattern defect inspection performance. Even in the tube, the pixel detection resolution is sufficiently high and there is no difference in time.

[Embodiments of the invention]

A block diagram of an automatic printed circuit board positioning apparatus according to an embodiment of the present invention is shown in FIG. The image of the test pattern is captured by the one-dimensional image sensor 2a (
3α) Image is formed on the -dimensional image sensor 2a (3
The video information in the X-axis direction-line entered α) is sequentially read out from left to right as brightness data for each picture element, and a video signal (A) is obtained as an output. On the other hand, the X-Y feed mechanism 5
b is linearly fed in the y-axis direction in small steps in synchronization with each line scan of the -dimensional image sensor 2α (3α) according to a command from the processing device 12. As a result, the video signal (a) undergoes X-Y scanning on the test pattern, similar to normal TV screen scanning. This is converted into a binarization circuit 8α (9
By comparing the gray level (intermediate level between light and dark) (b) with α), the output is “1# (bright: pattern 1 present)”
), “0” (dark: pattern “no”) binary video signal (
C). Shift register 8h, Bc (9h, 9
c) is to delay this binarized video signal (c) by the time of one line scanning of the -dimensional image sensor 2α (3α), and as a result, the shift registers 8b, Bc (9h,
9c) output signal) and (b) The data is one step and two steps before the table feed in the y-axis direction for the binary video signal (c), respectively.In reality, the image sensor 2α (3α) is Since each line consists of 1024-bit picture elements, shift registers 8h, Bc (9h,
The number of bits in 9c) is also 1024 bits each. Signal (c), (=l.

&19 are respectively placed in the flip-flop group M (9d), and a signal is obtained through the gate (G)8e (9g) with the connection shown in the figure. Here, if the aforementioned straight edge portion Eα (Eb) parallel to the X-axis is selected as the pattern element to be tested, data indicating a boundary point to the signal will be obtained. (When it is "1", it is a boundary point; when it is "0", it is not a boundary point.) The boundary point is originally one pit in the y-axis direction, and is "0'" this time.
1". However, as mentioned above, since the pixel resolution of the one-dimensional image sensor is high, the edge Eα (Eh) is actually a straight line that can be seen with the microscopic eye. As shown in Figure 4, the line is uneven.For this reason, the boundary point detection method is y l1ill + direction 1.
Compare this with the data several bits before the pit, not just before the pit, and in the X-axis direction, compare not only one pit of interest, but multiple pits including the left and right of this, all "0" or "1'". It is added on the condition that the following is true.

For example, if i
-m, when the point of interest is the center flip-flop (mouth in the figure), the state 1 step before the y-direction feed υ corresponds to the 7 flip-flops at the bottom of the figure) for 3 bits including the left and right. When all “0” is established for the 3 bits including the left and right in the state after one bit of y-direction feed (corresponding to the flip-flop on the upper side of the diagram), the point of interest is set. If it is defined as a boundary point, the signal (to) is “1#”
When , the boundary point is detected. In this example, a point where all 3 bits in the X-axis direction are "0" or 1# is compared with "1" and "0" with a point 2 bits away in the Y-axis direction. In reality, the circuit design is such that appropriate values can be arbitrarily selected for one bit in the X-axis direction and two bits in the Y-axis direction. The number of “11” of this boundary point signal is
Preset counter B within the range of one line scan in the axial direction
f (9f) is counted, and when this value exceeds the specified value (n1 pieces), a preset counter 8f (9f) fc overflow is generated, which causes the flip-flop By (9f) to overflow.
y) is set to 11". This becomes a signal that detects the boundary line.

With the above, boundary lines can be detected stably and with high resolution. The blocks described above are provided one set each, corresponding to the left and right optical detection units 2 and 3, and the latter part number is indicated within 0 in the text.

By the way, in the flip-flops 8g and 9g, the y-axis simultaneous feed υ amount of the measurement table 5 from when one becomes "1#" until the other becomes "1" is the y-axis direction offset value K between the test patterns. In order to find this offset value d, the output of flip-flops 8! If? Counter 10d through pulse count 1oc with table feed pulse
If the offset value d is counted, this counted value becomes equal to the offset value d. Also, regarding the offset direction, flip-flop 8! If the output of I is connected to the data terminal (C) of the D-type flip-flop 10A, and the output of the exclusive OR gate 10α is connected to the trigger terminal (T+), the output of the D-type flip-flop IOA indicates the offset direction. That is,
When flip-flop 8f becomes 1 bit first, this becomes 11.
#, otherwise this is "0", this is 1",'o
1 corresponds to the upper left and upper right of the test pattern. The data on the offset value and the offset direction are sent to the processing device 11, and the amount and direction of self-correction of the pattern inclination are determined from the data, and the data is passed through the pulse motor control/drive circuit 12α to drive the pulse motor 5c. , the rotary table 5cL is operated to adjust the inclination of the pattern to be tested to zero. Preset counter Bf (9f) and 7 lip flops 8! I(9!1) is reset every line scan of the -dimensional image sensor 2α13α, and the counter 1od and the D-type flip-flop 10A are reset before the y-axis scan of the measurement chifur. The pattern elements of interest in the pattern to be tested are selected in advance for each type of substrate, and the coordinate values on this substrate are stored in the processing device, and before the tilt is detected, the processing device automatically processes the
It is assumed that the measurement table is moved in the XY direction to the vicinity of the required target element.

The above is an example of the present invention, but since the quality of the video signal of the image sensor is not very good including noise defects etc.
In actual circuits, analog signals are converted into binary values. 57N
Measures have been taken to improve the S/N ratio by spatial filtering in the digital processing section, but a detailed description of these measures will be omitted.

In addition, most of the parts required for the present invention have a function that is used for original pattern defect inspection.

Reference numbers of parts (D 8f, Bg, 9f, 91.
10cz, 10A.

Only 10c, 10g, 12a, 5a and 5c, making it economical.

〔Effect of the invention〕

As described above, according to the present invention, automatic alignment of the test pattern can be performed accurately and stably, and there is no need to provide a reference position marker on the test board. It has the effect of greatly improving operability since there is no need to align the body position.

[Brief explanation of drawings]

Figure 1 (α) is a layout diagram of the test pattern on the test board, Figure 1 (bl is a block diagram of the alignment device in the conventional method, and Figure 2 is the offset between the left and right test patterns). Detection method explanatory diagram, FIG. 3 is a block diagram of an embodiment of a printed circuit board automatic positioning device which is an embodiment of the present invention, and FIG. 4 is an enlarged view of edge Eα in FIG. 2. 1.・Test pattern 2, 3...Optical detection section 4
...Control device 5...Measurement table 8α,
9α...Binarization circuit 8h, 8c, 9h, 9c...Shift register No. 1 Atsushi Kuchi 2 Illustration 4 Procedural amendment (voluntary) Indication of case 1982 Patent application No. 226886 Title of the invention
Immediate meeting with person JM who corrects the automatic alignment device Patent applicant name '51Qll: 1 meeting 1 11
Tate Seisakusho (and 1 other person) Agent #! Book Aa, page 12, fg, line 9, ``8f'' is corrected as ``89''. 2. Figure 3 in the drawings is corrected as shown in the attached Figure 6. List of the above attached documents

Claims (1)

[Claims] A plurality of optical detection units that detect images of the pattern, a circuit unit that detects the inclination of the left and right patterns through the detection units, a test board XY feeding table for scanning the image, and a system that reduces the detected inclination angle to zero. An automatic positioning device characterized by having a rotary table that performs correction operations.