JPH03178145A - Inspecting equipment for bonding ball - Google Patents

Abstract

PURPOSE:To precisely measure the diameter of a bonding ball with as little influence as possible of quantization errors and image density gradient, by extracting pixels with zero-cross, from pixels and subpixels subjected to secondary differentiation, by using a zero-cross detecting means. CONSTITUTION:An image sensing means 10 picks up the image of a bonding ball 5 at the tip of bonding wire 4. A pixel forming means 11 executes digital processing of the output video signal and forms pixels. A pixel forming means 12 executes gradation interpolation between elements in an inspection region having a specified size out of pixels formed by the pixel forming means 11, and forms subpixels. Each pixel and each subpixel formed in the above process are stored in a monochromatic gradation image memory 13. The picture elements and pixels read from the memory 13 are subjected to secondary differentiation, and the zero-cross as the result of the secondary differentiation is detected by a zero cross detecting means 14. On the basis of the detected results, the diameter of the bonding ball 5 is measured. Thereby quantization error can be reduced, and dimension measurement precision can be improved.

Description

[Detailed Description of the Invention] [Summary] This invention relates to a bonding ball inspection device that measures and inspects the diameter of the bonding ball, which is an important element in managing the bonding state in wire bonding of integrated circuits, etc. The purpose of this invention is to measure the diameter of the bonding ball with high precision without being affected by the concentration gradient as much as possible. A pixel generating means for digitally processing a video signal to generate pixels, and generating sub-pixels by performing gradation interpolation between each pixel in an inspection area of a predetermined size among the pixels generated by the pixel generating means. a sub-pixel generating means, each pixel generated by the pixel generating means in the inspection area, and a II light gradation image memory storing each sub-pixel generated by the sub-pixel generating means, and the gray gradation. pyrocross detection means for performing second-order differentiation on pixels and sub-pixels read from the image memory and detecting pyrocrosses as a result of the second-order differentiation, and detecting the bonding ball based on the detection result by the zero-cross detection means It is configured to visually measure the diameter of.

[Industrial application field]

The present invention relates to a bonding ball inspection device, and more particularly to bonding ball inspection equipment for measuring and inspecting the diameter of the bonding ball, which is an important element in managing the bonding state in wire bonding of integrated circuits and the like.

In recent years, integrated circuits (ICs) and large-scale integrated circuits (LSIs) have become widely used and in large quantities, so the visual inspection of the wire bonding status of ICs and LSIs has been automated. It has become. In this visual inspection of the wire bonding state, it is necessary to accurately measure the diameter of the bonding ball.

(Prior Art) FIG. 7 shows a configuration diagram of an example of a conventional bonding ball inspection device 5A. In the figure, 1 is a lead frame, on which an IC chip 2 is placed.

A pad 3 is formed on the IC chip 2.

4 is a wire, one end of which is fixed to the pad 3 to form a bonding ball 5.

The lead frame 1 is connected to a frame feeder 6 which is a conveying device.
As a result, it is transported directly below the illumination device N7 and the imaging device 8. Thereby, the bonding ball 5 is illuminated by the illumination device 7 and imaged by the Il imaging device @8. The video signal taken out from the imaging device 8 is supplied to the information processing device 9, where the shading is corrected by analog processing or digital processing to make the image into a uniform gray state, and then the image is automatically converted to a binary value with the same value. The wire bonding state of the captured image is inspected.

Here, in the conventional bonding ball inspection apparatus, a 2ifI conversion threshold is obtained as a process in the information processing Kff9, the size of the bonding ball diameter is measured from the grayscale image converted to 2fa using this binarization threshold, and the size of the bonding ball diameter is measured from this diameter. The shape of 5 is being inspected.

[Problem to be solved by the invention]

However, in the conventional bonding ball inspection apparatus described above, a quantization error occurs because the video signal is digitized. This quantization error can be reduced by increasing the resolution of the imaging optical system, but since the field of view becomes narrower as the resolution increases, there are restrictions on increasing the resolution. For this reason, conventionally there is a relatively large quantization error, which causes a dimensional error in the diameter of the bonding ball 5.

Further, conventionally, a grayscale image converted into 2III by binarization equivalent value is affected by the density gradient of the image, and the shape of the bonding ball 5 changes depending on the binarization threshold value.

The present invention has been made in view of the above points, and an object of the present invention is to provide a bonding ball inspection device capable of determining the open side of the bonding ball diameter with high accuracy without being affected by quantization errors or image density gradients as much as possible. With the goal.

[Means to solve the problem]

FIG. 1 shows a block diagram of the principle of the present invention. In the same figure, 10
is an imaging means that images the bonding ball at the tip of the wire bonded to the pad. 11 is a pixel generation means that digitally processes the video signal to generate pixels.

Reference numeral 12 denotes a sub-pixel generating means, which performs i-tone conversion between each pixel in an inspection area of a predetermined size to generate sub-pixels.

Further, 13 is a gray scale image memory which stores each pixel and sub-pixel within the above-mentioned inspection area. Furthermore, 14 is a pyrocross detection means that performs second-order differentiation of the pixels and sub-pixels from the gradation image memory 13, detects the pyrocross as a result of the second-order differentiation, and thereby measures the bonding ball diameter dimension #I. .

[Effect]

In the present invention, the subpixel generation means 12 performs tone interpolation between each pixel (pixel) in the inspection area to generate subpixels, and stores them in the grayscale image memory 13.
It is stored together with each pixel in the inspection area. Therefore,
Since the number of pixels taken out from the gray scale image memory 13 is increased by the number of sub-pixels compared to the conventional case, the digital quantization error of the image is reduced.

By the way, human vision involves lateral inhibition.
There is a function that emphasizes and detects object contours using hibition>. On the other hand, the zero-crossing method is known as a method for accurately detecting the outline of an object in a digital image (Marr and H1dreth).

1980). This method uses a Gaussian filter (Gaussian F 1l) to detect intensity changes in the image.
ter ), the blurred image is subjected to the Laplacian (2), the zero crossings (bi0 crossings) of the image are extracted, and the intensity change displacement M (
outline).

The strength of the image! Mathematically, the task of blurring 1 function 1 (x, y) with a Gaussian number G is expressed as G*l, and this Laplacian is written as 2 (G*I), but in reality, the above Gaussian - Pyrocross is detected by extracting the address of a pixel in which a pixel with a positive value and a pixel with a negative value are adjacent to each other in a 2G*I image that has been passed through a filter and applied a Laplacian.

The present invention detects the outline of an image based on this pyrocrossing method. However, in the present invention, since the sub-pixel generation means 12 performs tone interpolation from the original pixels by linear approximation, the image after the tone interpolation has an increased pixel density and blur. . Therefore, in the present invention, the zero-cross detection means 1 uses pixels and sub-pixels that have been subjected to Laplacian <2), that is, subjected to second-order differentiation, without applying a Gaussian filter to the image.
4, the pixels in which the pyrocross occurs are extracted, thereby making it possible to accurately detect the outline of the bonding ball. This ZERO detection is less affected by concentration gradients than a binary image generated by 2IA conversion.

In this way, in the present invention, by increasing the number of pixels by subpixels, digital quantization errors can be reduced, and the same effect as blurring the image can be obtained, and the outline of the bonding ball can be detected more accurately by zero-cross detection. can.

(Embodiment) FIG. 2 shows a configuration diagram of an embodiment of the present invention. In the figure, the same components as those in FIG. 7 are denoted by the same reference numerals, and their explanations will be omitted. In FIG. 2, the illumination device 7 and the 'ta image 5A
Place B constitutes the imaging means 10 shown in FIG. In FIG. 2, 21 is an image input circuit, and the pixel generating means 11
It is configured to digitally process input video signals.

22 is an image memory, and 23 is an image processing memory, which constitute the gray scale image memory 13.

Reference numeral 24 denotes an image processing processor, which constitutes the sub-pixel generating means 12 and the zero cross detecting means 14, and performs operations according to a flowchart shown in FIG. 3, which will be described later. 25 is an image display circuit that causes a display 26 to display an image. Pixel data transfer between the image input circuit 212, image memory 22.degree. image processing memory 231, image processing processor 24, and image display circuit 25 is performed via the image bus 27.

Further, 2B is a program memory in which a program for controlling the entire bonding ball inspection apparatus is stored. Reference numeral 29 denotes a system processor, which centrally controls the circuits of the entire bonding ball inspection apparatus according to the program from the program memory 28.

A camera controller 30 controls the operation of the Iall device 8. 31 is a lighting controller that adjusts the amount of illumination light for the IC chip 2 of the lighting device 7 and controls the S of the video signal.
/N is optimized. Reference numeral 32 denotes a frame feeder controller, which controls the transport of the frame feeder 6, which is the transporter 5A. 33 is the I10 controller, keyboard and joystick 34. Provides an interface between printer 35 and system processor 29. The keyboard and joystick 34 are used to set test conditions, and the printer 35 is used to print out test results.

36 is a floppy disk controller (FCC) and a hard disk controller (+-IDC), which are connected to a floppy disk 37 and a hard disk 38. The floppy disk 37 and hard disk 38 store test results, test standards, and the like. Furthermore, 39 is a system bus, which is a bus for data transfer between the above-mentioned circuits.

Next, the operation of this embodiment will be explained. The imaging device 8 is I
The bonding ball 5 formed at the end of the wire 4 bonded to the C chip 2 on the IC chip 2 side is imaged;
The video signal obtained thereby is supplied to the image input circuit 21. The image input circuit 21 digitally processes the input video signal to generate image data in which pixel groups are synthesized in time series, and stores the image data in the image memory 22 via the image bus 27. The system processor 29 cuts out pixels (image data) of an area f*D at a predetermined position and size from the image memory 22 and stores them in the image processing memory 23.

Using all the pixels of the area wtD thus stored in the image processing memory 23, the image processing processor 24 realizes the sub-pixel generation means 12 and the zero-cross detection means 14 according to the flowchart shown in FIG.

First, as shown in FIG. 3, the image processing processor 24 reads out (cuts out) all pixels in the area (step 4).
1). Next, the image processing processor 24 uses 1! between each read pixel! ! Performs interpolation (sub-pixel generation) (third
In the figure, step 42). As shown in FIG. 4, this inter-pixel gradation interpolation is performed by selecting any four adjacent pixels from the black circle P1 to
Shown in P4, mat: - solera f) picture M floor mla'E:P
+ = f' (i.

j>, P2 = f' (i+l, j), P3 = f'
(i, j+1), P4 = f' (i+1.j+1)
Then, this is a method of generating a sub-pixel having a gradation value of f (x, y) expressed by the following formula (%) at the position indicated by a white circle SP. However, in the above formula, α represents the horizontal distance from Pl, and β represents the vertical distance from Pl.

By this inter-pixel gradation interpolation, the above four pixels P1 to P4
In between, sub-pixels are generated in the same manner as above at each position indicated by white circles in FIG. As a result, the image of the bonding ball 5 before the above interpixel gradation interpolation is changed to FIG. 5 (A>
As shown in Figure (B), the jaggedness in the diagonal image part was large, but by performing the above interpixel interpolation, the image of the bonding ball 5 has a large jaggedness in the diagonal image part as shown in Figure (B). is significantly reduced, and an image with significantly reduced quantization errors caused by digitization can be obtained.

The sub-pixels generated in this manner are stored in an area of the image processing memory 23 that is different from the original pixel storage area in the inspection area.

Next, the image processing processor 24 reads out each pixel and sub-pixel within the inspection area from the image processing memory 23 and performs quadratic differential processing on them (step 43 in FIG. 3). A zero cross of the second-order differential result is detected (step 44 in the same example). That is, each process of steps 43 and 44 is performed by the zero cross detection means 1.
This is the process for realizing 4.

Here, in step 43 above, the image 111! l Inspection area vt by image and sub-pixels read out from memory 23
As the image of D is shown in FIG. 6(A), pad 3. Assuming that the image represents the shapes of the wire 4 and the bonding ball 5, the relationship between the pixel position of the part indicated by the solid line a in FIG. 10A and its pixel IM tone value is as shown in FIG. The outline of the bonding ball 5 shows a density gradient.

Pixel g of this image! 1ill value is expressed as f (x, y), and when second-order differentiation is applied to this image in step 43 above, the second-order differentiation result D (X.

y) is D(x,y)=2*f(x,y) It is expressed as shown in FIG. 6(C).

The zero cross detection in step 44 is performed in the right direction and the left direction from the center O shown in FIG. 6(C),
Detect the first pyrocross positions I C+ and C2, respectively.

Here, the detection of pirocros from the center O to the right is D (
x+1. 10 (x.

y), and zero cross detection from center 0 to the left is done by extracting pixels with D (X-1,1>O,D
A value D that satisfies both conditions (x+1.y)<O (x,y
) by extracting pixels with

Furthermore, the second-order differential processing in step 43 and the zero cross detection processing in step 44 are performed not only in one direction with respect to the solid line a in FIG. 6(A), but also in each direction indicated by the broken line in FIG. Do the same for each. However, the detection direction is set so as not to cover the wire 4. This is because the wire 4 is not the bonding ball 5.

Next, the pixel processing processor 24 measures the diameter of the bonding ball 5 from the difference between the two pillow cross detection values in each detection direction obtained in this manner (step 45 in FIG. 3). In this case, the center of the bonding ball 5 is determined from the position of the maximum value first detected by searching the plane of each pixel of the gradation-converted image from the direction opposite to the image of the wire 4, and By carrying out the above steps 43 to 45 on a straight line passing through , the diameter of the bonding ball 5 can be accurately measured four times.

Further, by measuring the diameter of the bonding ball 5 in each of a plurality of directions, distortion in the shape of the bonding ball 5 can be inspected.

〔Effect of the invention〕

As described above, according to the present invention, the number of pixels in the inspection area can be increased by the amount of sub-pixels, so quantization errors can be reduced compared to the conventional method, and all pixels in the inspection area including sub-pixels can be By simply applying Laplacian (2) to , the contour of the bonding ball can be accurately detected by zero-cross detection, which improves the accuracy of bonding ball dimension measurement compared to the case of using a binarized gray image using binarized ill. This method has the advantage of improving the accuracy of bonding ball inspection compared to the conventional method.

[Brief explanation of drawings]

Fig. 1 is a block diagram of the principle of the present invention, Fig. 2 is a block diagram of an embodiment of the invention, Fig. 3 is a flowchart for explaining the operation of an embodiment of the invention, and Fig. 4 is an explanation of 1m1lill pixel interpolation. 5 is a diagram showing an example of images before and after interpixel WA tone interpolation, FIG. 6 is an explanatory diagram of ball diameter measurement by zero-cross detection, and FIG. 7 is a configuration diagram of an example of a conventional device. In the figure, 2 is an IC chip, 3 is a pad, 4 is a wire, 5 is a bonding ball, 10 is an 1111 means, 11 is an image generation means, 12 is a subpixel generation means, 13 is a gray scale image memory, 1 and 4 are The Pirokuros detection means is shown.

Claims (1)

[Claims] An imaging means (10) for imaging a bonding ball at the tip of a wire bonded to a pad;
pixel generating means (11) for digitally processing the output video signal of step 10) to generate pixels; Sub-pixel generation means (1) that generates sub-pixels by tonal interpolation
2), and a grayscale image memory (13) for storing each pixel generated by the pixel generation means (11) and each subpixel generated by the subpixel generation means (12) in the inspection area, respectively. and zero-cross detection means (14) for performing second-order differentiation on the pixels and sub-pixels read out from the gray scale image memory (13) and detecting zero-crossings of the second-order differentiation results. A bonding ball inspection device characterized in that the dimension of the diameter of the bonding ball is measured based on a detection result by a zero cross detection means (14).