External appearance inspection apparatus

Abstract

The present invention provides accurate external appearance inspection technology that has superior reproducibility. An external appearance inspection apparatus (10) is characterized by comprising: a camera (32) that is composed of an imaging optical system and an image pickup element and acquires images (18) to be used in an external appearance inspection; a mechanism (31) that, when the field of view of the camera (32) has been divided into M × N (2 ≤ M × N) equal parts, changes the positional relationship between the camera and an inspection subject (30) by a movement pitch equal to the division pitch; and an image processing inspection unit (23) that processes the images (18) obtained from the camera (32) and performs an inspection, wherein at least one image (18) is captured by the camera (32) each time the mechanism (31) implements a movement at the movement pitch, M × N images of the entire inspection area of the inspection subject (30) are acquired, and the acquired images (18) are subjected to inspection processing by the image processing inspection unit (23).

Description

Visual inspection equipment

[0001] The present invention relates to a visual inspection technique for inspecting various products using image processing technology.

[0002] Visual inspection equipment that uses a camera, a device that captures images, to inspect the appearance of various products is becoming widespread.A camera consists of an imaging optical system that forms an optical image of the object under test, and an image sensor (generally a CMOS image sensor) that photoelectrically converts the light intensity of that image into an electronic image, and the size of the image obtained in one capture is called the field of view.

[0003] The field of view can be changed by changing the magnification of the lens. However, since pixel resolution, which is the size on the test object equivalent to one pixel of the imaging device, is a key factor in determining inspection performance, the magnification of the imaging optical system must be determined to obtain the desired pixel resolution, and the field of view cannot necessarily be freely set. Therefore, when inspecting a test object larger than the field of view, the camera or the test object must be moved and multiple fields of view must be used to inspect the entire test object. For example, if the camera's field of view size is 11 x 11 mm and the test object is 20 x 20 mm, four fields of view can be used to inspect the entire test object. This type of inspection is hereinafter referred to as multi-field inspection.

[0004] As a specific example, bump inspection in the semiconductor packaging process (post-processing) is described below. Integrated circuit (IC) chips must be packaged in resin to protect against deterioration and heat. Packaging involves mounting the chip on a multilayer wiring board, connecting the electrodes on the board to those on the chip, and then sealing the chip in resin. While wire mounting, which connects electrodes via wires, has traditionally been the norm, flip-chip mounting, which connects ball-shaped microelectrodes (hereafter referred to as "bumps") in a two-dimensional array, has become increasingly popular in recent years.

[0005] In flip-chip mounting, electrodes are directly superimposed on each other, so it is important that the bumps are uniform in size. This is because any variation in bump size can cause poor connection. Therefore, it is necessary to accurately inspect the shape of the bumps before connection (see, for example, Patent Document 1).

[0006] Visual inspection of bumps is performed as part of in-line inspection during the manufacturing process, by photographing the arrayed bumps from above with a camera. In recent years, full inspection has become common, so inspection speed is of great importance. Current bumps range in diameter from 10 to 80 μm, are arranged at pitches ranging from 20 to 130 μm per multilayer wiring board, and hundreds to tens of thousands of them are mounted. There is a demand for high-speed inspection of all of these bumps per multilayer wiring board, taking from one to several tens of seconds.

[0007] In recent years, packaging using the chiplet concept has been promoted, and as more chips are mounted on multilayer wiring boards, the size of the multilayer wiring boards has increased. On the other hand, as circuits have become more miniaturized, the bump diameter has become smaller, and fine pixel resolution is required for appearance inspection, which has resulted in a smaller field of view. As a result, multi-field inspection using a large number of fields of view has become necessary.

[0008] Patent No. 4984332

[0009] One problem with image-based inspections is the issue of aberration in the imaging optical system. Due to aberration in the imaging optical system, the image quality is not uniform within the field of view; for example, the state of aberration differs between the center and periphery of the field of view, resulting in different image quality. If the image quality differs, the inspection results may not be the same. For example, when inspecting to check the height of a bump, there may be a difference between the height value at the center of the field of view and the height value at the periphery of the field of view, even for the same bump. (First issue)

[0010] Furthermore, as test objects become smaller and require higher-precision inspections, the relationship between the digital pixel layout and the position of the test object can affect the inspection results. For example, because a ball-shaped bump has a relatively mirror-like surface, when illuminated from directly above, the reflected light is limited to the apex. Therefore, as the diameter of the bump becomes smaller, the area from which the reflected light returns also becomes smaller, and the size of the captured apex becomes smaller than the pixel size of the camera. This results in the possibility that the reflected light from the apex of the bump is captured by a single pixel or by multiple adjacent pixels. Therefore, depending on the case, the morphological information of the bump differs, resulting in a problem of reduced accuracy and reproducibility of the inspection results (the second problem).

[0011] The present invention has been made in consideration of the above circumstances, and has as its object to provide an appearance inspection technique that is highly reproducible and accurate.

[0012] The visual inspection device of the present invention comprises a camera composed of an imaging optical system and an image sensor for acquiring images to be used for visual inspection, a mechanism for changing the positional relationship between the camera and the test object at a movement pitch approximately equal to the division pitch when the field of view of the camera is divided into M x N (2 ≦ M x N) equal parts, and an image processing inspection unit for processing the images acquired from the camera and performing inspection, characterized in that the camera acquires at least one image each time it moves by the movement pitch, acquiring M x N images of all the test area of ​​the test object, and the acquired images are inspected and processed by the image processing inspection unit.

[0013] The present invention provides a highly accurate appearance inspection technique.

[0014] 1 is a block diagram of a visual inspection device according to a first embodiment of the present invention. (A) A top perspective view of a multilayer wiring board on which bumps are arranged in an array, (B) A side perspective view of a multilayer wiring board on which chips are mounted via bumps. (A) to (F) Diagrams explaining the operation of the visual inspection device according to the embodiment. (G) to (L) Diagrams explaining the operation of the visual inspection device according to the embodiment. (A) A side view showing reflected light from the vertices of bumps of different sizes entering the camera, (B) A top view of the same. (A) An image of a bump showing the light reflected from the vertex detected by one pixel, (B) An image of a bump showing the light reflected from the vertex detected by two pixels, (C) An image of a bump showing the light reflected from the vertex detected by four pixels. Arithmetic expressions related to the visual inspection algorithm.

[0015] (First Embodiment) Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings. For the sake of specificity, bump inspection will be taken as an example. FIG. 1 is a block diagram of a visual inspection apparatus 10 according to a first embodiment of the present invention. FIG. 2(A) is a top perspective view of a multilayer wiring substrate (subject to test) 30 on which bumps 14 are arranged in an array. FIG. 2(B) is a side perspective view of an IC package 37 in which a chip 36 is mounted on the multilayer wiring substrate 30 via the bumps 14.

[0016] As shown in FIG. 2 , in an IC package 37, electrodes of a multilayer wiring substrate 30 (hereinafter simply referred to as "substrate 30") and electrodes of an IC chip 36 are electrically connected by bumps 14. This method of electrical connection using bumps 14 is called flip-chip mounting. In this embodiment, the bumps 14 arranged on the substrate 30 are inspected. Thereafter, a chip 36 is mounted, and the chip is packaged 35 with resin to protect against deterioration and heat, completing the IC package 37. Note that, although a substrate 30 with bumps 14 arranged thereon is shown as an example of an object to be tested by the visual inspection apparatus 10 in this embodiment, this is not particularly limited.

[0017] As shown in Fig. 1, a plurality of substrates 30 (3 x 5 = 15 in the figure) on which bumps 14 (Fig. 2) are arranged are accommodated on a tray 38. The tray 38 thus accommodating a plurality of substrates 30 is placed on an inspection table 39 and inspected by the visual inspection device 10. After inspection, the tray 38 is transferred to the next process, and a new tray 38 containing substrates 30 yet to be inspected is placed on the inspection table 39. Note that the multilayer wiring substrate is one example of the substrate 30 on which bumps 14 are arranged, and there is no particular limitation.

[0018] 3A to 3F and 4G to 4L are explanatory diagrams illustrating the operation of the appearance inspection device 10 for bumps 14 according to the embodiment. The explanation will continue with reference to FIG. 1. The field of view 12 of the camera 32 is divided into four equal 2x2 sections, each of which is referred to as a segment 11. The device includes a mechanism 31 that sequentially moves the field of view 12 stepwise toward a test area 14a where a group of bumps 14 exists at a movement pitch equal to the division pitch of the segments 11, and an image processing and inspection unit 23 that processes and inspects an image 18 of the field of view 12 acquired from the camera 32.

[0019] In the above description, the segments 11 are assumed to be four equal parts of the field of view 12 of the camera 32, but this does not need to be exact. When performing multi-field inspection, it is common to set a slight overlapping area between adjacent fields of view. For example, in the case of a bump, if the bump is divided and imaged in adjacent fields of view, the measurement process becomes complicated. Therefore, to ensure that the bump is imaged without being divided into either field of view, the images are overlapped so that an area equal to or greater than the bump pitch is imaged in both fields of view. It is sufficient to exclude such areas and divide the field of view into four equal parts; it does not need to be exactly one-fourth of the field of view. This is indicated by the fact that the sizes of the field of view 12 and the four segments 11 in Figures 3 and 4 do not match.

[0020] Also, dividing into four equal parts is just one example, and in general, it is divided into M×N equal parts (2≦M×N). As a result, by sequentially moving the field of view 12 in steps at a movement pitch equal to the division pitch of the segments 11, all bumps are imaged M×N times. This allows the morphological information 24 of the bumps 14 to be determined using image information from M×N times, thereby improving the determination accuracy.

[0021] The mechanism 31 sequentially moves the camera 32 in a step-like manner at a moving pitch equal to the division pitch of the segments 11. Note that, as a modified example of the mechanism 31, although not shown, it is also possible to adopt a form in which an inspection table 39 on which a tray 38 is placed is sequentially moved in a step-like manner.

[0022] The operation of the bump appearance inspection device 10 is as follows: first, as shown in FIG. 3(A), the field of view 12 of the camera 32 is positioned so that it includes one segment 11 located at the corner of the test area 14a where a group of bumps 14 exists, and an image 18 is acquired in the image processing inspection unit 23 (FIG. 1).

[0023] Next, as shown in Figure 3(B), the field of view 12 is advanced by one step along the row of bumps 14 at a movement pitch equal to the division pitch, and images 18 are acquired by the image processing inspection unit 23. Furthermore, the sequential advancement of the field of view 12 along the row of bumps 14 and the acquisition of images 18 are repeated up to the opposite end, as shown in Figure 3(C). Thereafter, the field of view 12 is advanced by one step in the perpendicular direction, and as shown in Figure 3(D), the sequential advancement of the field of view 12 in the opposite direction and the acquisition of images 18 are repeated up to the opposite end, as shown in Figure 3(E). Similarly, the sequential advancement in steps at a movement pitch equal to the division pitch of the segments 11 and the acquisition of images 18 are repeated from Figure 3(F) to Figures 4(G) to (L).

[0024] In this way, the image processing inspection unit 23 acquires M×N (four in the figure) images 18 for each of the plurality of segments 11. Each of the same segment 11 captured in the plurality of images 18 is captured in a different region of the field of view 12 divided into M×N regions.

[0025] Generally, images 18 are affected by aberrations in the imaging optical system. Generally, aberrations increase from the center to the periphery of the field of view, resulting in a tendency for image quality to deteriorate. Furthermore, the effects of this aberration are not necessarily point-symmetric due to manufacturing errors such as decentering. On the other hand, bumps 14 captured in different regions of the field of view 12 are located in different locations with different aberration conditions in the images 18, and it is thought that each image has different accuracy. Therefore, the M×N (four images shown) images 18 may be detected in the center or periphery of the field of view 12. Then, by statistically processing these detection values, the accuracy of the morphological information 24 used to determine the bumps 14 can be made uniform.

[0026] 5A is a side view showing the reflected light 26 (26x, 26y) from vertices 27 of bumps 14 (14x, 14y) of different sizes that enters the camera 32. FIG. 5B is a top view of the same. Because the ball-shaped bump 14 is relatively close to a mirror surface, when illuminated from directly above, the reflected light 26 (26x, 26y) that enters the camera 32 is limited to the vertices 27 (27x, 27y). Therefore, as the diameter of the bump 14 decreases, the area of ​​the vertex 27 from which the reflected light 26 returns also becomes smaller.

[0027] 6A is an image 18 of a bump 14 showing the reflected light 26 from a vertex 27 detected by one pixel 22, (B) is an image 18 of the bump 14 showing the reflected light 26 detected by two pixels 22, and (C) is an image 18 of the bump 14 showing the reflected light 26 detected by four pixels 22. As the bump 14 becomes smaller and the size of the captured vertex 27 becomes equal to or smaller than the size of the camera's pixel 22, the vertex 27 may be captured by one pixel or by multiple adjacent pixels.

[0028] The movement pitch of the field of view 12 is the size of the segment 11 (= division pitch), but by making the movement pitch an integer multiple of the pixel resolution and adding 1 / M (in the case of vertical movement pitch) or 1 / N (in the case of horizontal movement) of the pixel resolution, if the amount of movement is accurate, it is possible to obtain M x N images 18 in which the positional relationship of the bump to the pixel 22 is shifted by 1 / M vertically and 1 / N horizontally.

[0029] In other words, the division pitch of the segment 11 is multiplied by the magnification of the imaging optical system, and the result is divided by the size of the pixel 22 of the camera 32 (pixel size), and the result is rounded to the nearest integer. Furthermore, 1 / M times the pixel size is added vertically and 1 / N times the pixel size is added horizontally, and the values ​​obtained by dividing these values ​​by the magnification of the imaging optical system become the vertical and horizontal sizes of the movement pitch.

[0030] Such pixel shifting has the effect of improving pixel resolution, making it possible to reduce the effect on measurement values ​​caused by differences in the positional relationship between pixels and the object, as shown in Figure 6, and improving the accuracy and reproducibility of inspection results.

[0031] The image processing inspection unit 23 (FIG. 1) judges the morphological information 24 of the target bump 14 based on the four (M×N) vertex pixels 22a acquired by sequentially feeding the image. The morphological information 24 specifically includes the position, height, and outer diameter of the bump 14, as well as other shape abnormalities of the bump 14, and further includes the height of the substrate 30 and surface foreign matter.

[0032] The measurement values ​​of these morphological information 24 are obtained one by one from each of the four (M x N) images 18 showing the shape of each bump 14 based on the brightness values ​​of the vertex pixels 22a and / or their surrounding pixels 22, for a total of four (M x N) values, and these multiple measurement values ​​are then statistically processed to output one representative value for each bump 14. Note that the measurement values ​​and their representative values ​​representing the morphological information 24 of the bump 14 are not limited to one item, but are determined for multiple items.

[0033] 7 shows the formula for the algorithm of the visual inspection. The representative value V of the shape information 24 is calculated from the plurality of measured values ​​v according to the following three algorithms. i However, the method for calculating the representative value V is not limited to these algorithms.

[0034] The first algorithm is "simple averaging." This "simple averaging" is performed by multiplying M × N measured values ​​v i (i = 1 to M × N) according to the formula (1) in Figure 7, and calculate the representative value V. Due to the effect of averaging (law of number of units, central limit theorem), one measurement value v iThis improves reliability compared to selecting the representative value V as the representative value V.

[0035] The second algorithm is a "weighted average by reliability." This "weighted average by reliability" is calculated by multiplying M × N measured values ​​v i (i=1 to M×N) i is used as a weighting coefficient to determine the corresponding measurement value v i Then, a weighted average is performed to calculate the representative value V. This "weighted average based on reliability" can suppress the effects of pixel cracking shown in FIGS. 6(B) and 6(C), improving the reliability of the representative value V.

[0036] Here, the reliability value is a quantity that represents the degree of reliability of the inspection value or measurement value of each pixel, and differs depending on the visual inspection method. For example, the brightness value of a pixel represents the amount of light, which is the signal, so it can be said to represent reliability in the sense of whether the signal is being obtained reliably, and is an example of a reliability value.

[0037] The third algorithm is "weighted averaging by distance from the center." This "weighted averaging by distance from the center" is performed by taking M×N measured values ​​v i The distance L from the center position of the field of view 12 of the pixel 22 from which each of (i = 1 to M × N) was derived i (i = 1 to M × N) is used as a weighting coefficient to calculate the corresponding measurement value v i Then, a weighted average is performed to calculate the representative value V. This "weighted average based on distance from the center" places emphasis on values ​​measured near the center of the field of view, thereby suppressing the effects of aberration and improving the reliability of the representative value V.

[0038] The present invention can reduce the effects of aberrations in the imaging optical system (problem 1) and pixel cracks (problem 2) in visual inspections using images, thereby improving the reliability of the inspection. However, there is a clear negative factor of increased inspection time. The larger M×N, the more reliable the inspection, but the longer the inspection time. However, by incorporating the functions of the present invention into a visual inspection device, the user can select optimal reliability and speed, which is considered to be a significant advantage.

[0039] 10...appearance inspection device, 11...segment, 12...field of view, 14...bump, 14a...group of bumps (test area), 15...adjustment unit, 18...image, 22...pixel, 22a...vertex pixel, 23...image processing inspection unit, 24...morphological information, 26...reflected light, 27...vertex, 30...multilayer wiring board (substrate, test object), 31...mechanism, 32...camera, 35...packaging, 36...IC chip, 37...IC package, 38...tray, 39...inspection table.

Claims

1. A camera composed of an imaging optical system and an imaging device for acquiring an image used for appearance inspection, a mechanism unit for changing the positional relationship between the camera and the object to be inspected at a movement pitch equal to the division pitch when the field of view of the camera is divided into M×N (2≤M×N) equal parts, and an image processing inspection unit for processing the image obtained from the camera and performing an inspection. The mechanism unit acquires at least one image with the camera each time it moves by the movement pitch, acquires M×N images of all the inspection areas of the object to be inspected, and the acquired images are subjected to inspection processing by the image processing inspection unit. An appearance inspection apparatus characterized by the above.

2. In the appearance inspection apparatus according to claim 1, a value obtained by dividing the value obtained by multiplying the division pitch by the imaging optical system magnification by the pixel size of the camera is rounded to an approximate integer value, and further, 1 / M times the pixel size is added in the vertical direction and 1 / N times in the horizontal direction, and the values obtained by dividing these values by the imaging optical system magnification are set to be the vertical and horizontal sizes of the movement pitch. An appearance inspection apparatus thus configured.

Images