JPH09311014A - Inspection apparatus for protruding part of semiconductor integrated circuit device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain an inspection apparatus by which the dislocation of a solder protruding part is inspected with good accuracy and at high speed and by which its shape can be inspected. SOLUTION: An inspection apparatus is constituted so as to contain an approximate position judgment means 10 in which the approximate coordinates of a protruding part as an object to be inspected are compared with coordinates comprising the protruding part originally, which outputs a dislocation defect signal when a protruding part which is judged to be deviated sharply from an inspection standard exists, which ends the inspection of the protruding part so as to be shifted to the position inspection of a next protruding part and which requests a high-accuracy judgment only regarding a protruding part which is judged to be not deviated sharply from the inspection standard and a high-accuracy position judgment means 15 by which subpixel-protruding-part coordinate data is compared with coordinates comprising the protruding part originally and which judges a dislocation defect when the protruding part does not exist in a preset dislocation tolerance.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a projection inspection apparatus for a semiconductor integrated circuit device, and more particularly to a semiconductor integrated circuit called an area bump or a ball grid array (BGA) in which a large number of solder bumps are regularly arranged. The present invention relates to a projection inspection apparatus for inspecting an individual bump displacement and a shape defect of an apparatus.

[0002]

2. Description of the Related Art FIG. 5 is a block diagram showing a conventional example. The inspection object 41 is illuminated by the illumination 42 and the camera 4
3 captures an image of the inspection target 41. The analog signal output from the camera 43 is AD-converted by the AD converter 44 and output as multi-valued image data. The multi-valued image signal is binarized by the binarizing means 45 and converted into a binarized image signal. The labeling unit 46 receives the binarized image signal and performs a labeling process, and the determining unit 47 detects a missing / misaligned / defective size of the projection based on the labeling data. In the position shift inspection, a position shift allowable range is set from the coordinates of the adjacent protrusions, and if a label exists within the range, it is determined that the position shift is not defective. (See, for example, JP-A-7-10263)

In the prior art, since the coordinates of the protrusion are calculated from the binarized image, the edge coordinates of the protrusion are quantized by the image resolution, and the detected coordinates have a quantization error. Since it is included, there is a problem in that it is not possible to accurately inspect the displacement of the solder protrusion.

[0003]

According to a first aspect of the present invention, there is provided a projection inspection apparatus for a semiconductor integrated circuit device, which comprises a plurality of projections for electrically connecting to a circuit board or the like. Illumination that irradiates the part from diagonally above, a camera that is installed above the projection part to be inspected and captures an image of the projection part to be inspected, and an analog signal output from the camera is converted to multi-valued (gray) image data AD to
The conversion means, the multi-valued image data storage means for storing the multi-valued image data output from the AD conversion means, the multi-valued image data input, and data having a preset binarization level or higher 1 ”, data smaller than the binarization level is converted to“ 0 ”and binarized image data is output, and binarized image data storage for inputting and storing the binarized image data. Means, a labeling means for reading out the binarized image data from the binarized image data storage means, performing labeling processing and outputting the label data, and inputting the label data to calculate the barycentric coordinates for each label to be inspected In a projection inspection device for a semiconductor integrated circuit device, which comprises a rough coordinate detection means for making rough coordinates of a projection, (A) comparing the rough coordinates of the projection to be inspected with the coordinates where the projection should originally exist. And inspection standard If there is a protrusion that is judged to be significantly out of alignment, a misalignment defect signal is output, the inspection for that protrusion is discontinued, and the position of the next protrusion is inspected. A rough position determination means for making a high-precision determination request only for a protrusion that is determined not to exist, and (B) a search range for calculating a search range of a size slightly larger than the preset protrusion size centered on the coordinates of the inspection target protrusion Calculation means, (C) template image storage means in which multi-valued image data for one protrusion is previously registered, (D) search range calculation means, search range, template image storage means, template image and multi-valued Multi-valued image data is input from the image storage means, and the multi-valued image data and template image data at each coordinate in the search range are used to obtain the multi-valued pattern. The multi-valued pattern matching calculation means for performing the correlation processing to calculate the correlation value, and (E) the two-dimensional curve interpolation of the coordinate correlation data output from the multi-valued pattern matching calculation means to calculate the peak coordinates in the two-dimensional curve. Then, the sub-pixel calculating means for setting the sub-pixel projection coordinates is compared with (F) the sub-pixel projection coordinate data and the coordinates where the projection should originally exist. And a high-accuracy position determining means for determining a defect.

According to a second aspect of the present invention, in the projection inspecting device for a semiconductor integrated circuit device according to the first aspect, the maximum value of each coordinate correlation data output from the multi-level pattern matching calculating means is searched for and the maximum correlation is obtained. When the maximum correlation value is smaller than a value set in advance, a shape determining unit that determines that the shape is defective is provided.

The projection inspection apparatus for a semiconductor integrated circuit device according to the third aspect of the present invention can be inspected even when the projection is a solder bump in the first aspect.

According to the present invention, a multi-valued image for one projection is registered in advance as a template image, and a multi-valued pattern matching operation is performed with a captured image, and then the coordinates of the projection are calculated by sub-pixel operation. Since the calculation is performed, it is possible to accurately calculate the positional deviation amount of the protrusion. A labeling process is performed after the binarization, the approximate coordinates of the solder protrusions are detected, and then a narrow range only in the vicinity thereof is inspected in detail by multi-level pattern matching, so that high-speed inspection can be performed. Further, since the correlation value of each solder projection is calculated based on the image of the normal solder projection registered in advance, the correlation value of the solder projection having a poor shape becomes small, and it can be detected as a defective shape.

[0007]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing a first embodiment of the present invention. The illumination 1 irradiates the solder ball of the BGA 2 to be inspected from obliquely above. Camera 3 is the BGA to be inspected
2 is mounted above the device 2 and captures an image of the inspection target BGA2. If the illumination 1 is, for example, illumination capable of irradiating the inspection target portion from a ring-shaped direction, the image of the solder protrusion portion taken into the camera 3 is imaged with the specularly reflected light portion in a ring shape. Illumination that can be applied to the inspection target from the ring-shaped direction includes, for example, a ring-shaped fluorescent lamp, an LED illumination arranged in a ring shape, and a fiber illumination arranged in a ring-like shape with an emission outlet for illumination.

The A / D converter 4 receives the analog signal a output from the camera 3 and outputs multi-valued image data b which has been A / D converted to a digital signal. The multivalued image data b output from the AD conversion means 4 is input to and stored in the multivalued image data storage means 5. The binarizing means 6 receives the multi-valued image data b1 output from the multi-valued image data storing means 5, and sets data equal to or higher than a preset binarization level to "1" and data from the binarization level to "1". Small data
The data is converted to 0 "and the binarized image data c is output. Irregularly reflected light from the surface on which the solder protrusion is attached also enters the camera 3, but regular reflection from the surface of the solder protrusion enters the camera 3. Since the amount of incident light is smaller than that of light, the binarization level set in advance by the binarization means 6 is set to "1" for the region of the regular reflection light and "1" for the region of the irregular reflection light.
By setting to be "0", only the area of the regular reflection light from the solder protrusion can be set to "1" without being affected by the irregular reflection light from the mounting surface of the solder protrusion.

The multi-valued image data input to the binarization means 6 is not the multi-valued image data b1 output from the multi-valued image storage means 5 as shown in FIG. It is also possible to input the multi-valued image signal b output from 4.

The binarized image data c output from the binarizing unit 6 is input to and stored in the binarized image data storage unit 7. The labeling means 8 reads the binarized image data c1 from the binarized image data storage means 7, performs a labeling process, and outputs label data d. The protruding portion approximate coordinate detection means 9 inputs the label data d, calculates the barycentric coordinates for each label, and sets the protruding portion approximate coordinates, and outputs the protruding portion approximate coordinate data e. It is also possible to calculate the center coordinates from the coordinates of the circumscribing rectangle for each label and use them as the approximate coordinates of the protrusion.

The approximate position judging means 10 compares the approximate coordinates of the projection output from the approximate coordinates detecting section 9 with the coordinates at which the projection should originally be located. Only the approximate coordinates of the protrusions between the lower limit of the search range are searched, and the approximate coordinate data signal f of the solder protrusion to be inspected with high accuracy is output. In order to calculate the coordinates that should have a projection, the product to be inspected should be set with the coordinates relative to the camera set to predetermined coordinates, and the absolute coordinates of the projection should be registered in advance. A method of calculating projection coordinates after recognizing the alignment mark by providing an alignment mark on the upper surface of the mounted substrate, or detecting the coordinates of the adjacent projections or the projection coordinates existing at the corners of the projection array. Then, a method of calculating the projection coordinates from the relative coordinates may be considered.

The purpose of the rough position judging means 10 is not to carry out a high-precision inspection described below for all the protrusions, but to project a protrusion having a small amount of positional deviation from the rough coordinates obtained so far. The projections with a large amount of misalignment are regarded as rejection of the misalignment inspection. In other words, the position deviation inspection for all the protrusions is performed at a higher speed than when the inspection is performed.

The search range calculating means 11 inputs the approximate coordinate data signal f, calculates a search range from a preset search size such that the approximate coordinate data is at the center of the search range, and outputs search range data g. . The search size is set slightly larger than the protrusion size. The multivalued image data for one projection is registered in the template image storage unit 12 in advance. The multi-valued pattern matching calculation means 13 includes multi-valued image data b1 output from the multi-valued image storage means 5, search range data g output from the calculation means 11, and template image storage means 1.
2, the template image data h is input, and the multi-valued pattern matching processing is performed from the multi-valued image data within the search range of the multi-valued image data and the template image data to calculate correlation value data k at each coordinate.

The sub-pixel operation means 14 interpolates the correlation value data k at each coordinate output from the multi-valued pattern matching operation means 13 into a two-dimensional curve, calculates the maximum value coordinate in the two-dimensional curve, and As the projection coordinates, sub-pixel projection section coordinate data 1 is output. As the interpolation, spline curve interpolation, multidimensional curve interpolation, etc. can be considered in addition to two-dimensional curve interpolation.

The high-precision position determination means 15 compares the sub-pixel projection coordinate data 1 with the coordinates where the projection should originally be located, and determines that the positional deviation is defective if the positional deviation does not fall within a preset allowable range of positional deviation. If the maximum correlation value retrieved by the sub-pixel operation unit 14 is smaller than a preset value, the high-accuracy position determination unit 15 determines that the shape is defective.

Next, FIG.
This will be described in detail with reference to (a) and (b). Figure 2 (a)
Is a template image. The image size is equivalent to one normal solder protrusion. FIG. 2B shows multi-valued image data within the search range. The search range is larger than the template image size.

The multi-valued pattern matching calculating means 13 calculates a correlation value from the template image data and image data having the same size as the template image data among the image data within the search range by using equation (1). Equation (2) can also be used for calculating the correlation value.

[0019]

[0020]

In the calculation of the correlation value, the multivalued image data used for the correlation calculation is shifted one pixel at a time within the search range, and the correlation value is obtained in the entire region within the search range. For example, when the search range is larger than the template size by 3 pixels both vertically and horizontally, n to (n + 3) (n: X coordinate) in the horizontal (X) direction, and m to (m + 3) (m: Y coordinate) in the vertical (Y) direction. ) Since the calculation is shifted, the correlation calculation is performed 16 times in a total of 4 × 4. The multi-level pattern matching calculating means 13 outputs the correlation value for the number of times the multi-level image coordinates within the search range in which the correlation value was calculated and the correlation value were calculated.

FIG. 3 is a graph for explaining the sub-pixel calculation. In the sub-pixel operation, first, the largest value among the correlation values obtained by the multi-value pattern matching operation means 13 is searched, and the XY coordinates (Xma
x, Ymax). The graph of FIG. 3 is obtained by plotting the correlation value in the X direction at Ymax and obtaining a quadratic curve function passing through three points centered on the maximum value of the correlation value. From the calculated quadratic curve function, the X coordinate at which the maximum point of the quadratic curve is obtained is defined as the sub-pixel X coordinate of the solder protrusion. This is because the correlation value decreases as the actual projection coordinates and the template image become farther apart, so that by performing curve interpolation on the correlation values around the maximum point of the correlation value, it is possible to calculate the solder projection coordinates at the sub-pixel. It can be understood. Further, the points for obtaining the graph are not limited to three points. The graph is not limited to the quadratic curve.

Similarly, the sub-pixel coordinates in the Y coordinate can be calculated from the correlation value in the Y direction at Xmax. Also, the solder protrusions are partially missing,
When the size is different, the maximum value of the correlation value is smaller than that of the normal solder protrusion, so the maximum value of the correlation value for the good solder protrusion and the maximum value of the correlation value for the poorly shaped solder protrusion are determined in advance. It can be seen that by setting the determination value between the values, the shape defect can be detected.

With the method using multi-level pattern matching according to the present invention, the coordinates of the solder protrusions can be detected with high precision. When the correlation calculation is performed, the number of calculations becomes extremely large and it takes a long time. However, in this method, the correlation calculation is performed only in a small range after detecting the approximate coordinates of the protrusion from the binarized image in advance, so that high speed operation is possible. Can be achieved.

FIG. 4 is a block diagram showing a second embodiment of the present invention. The multivalued image data b output from the AD conversion means 4 is input to and stored in the multivalued image data storage means 5b. The binarization means 6 receives the multi-valued image data b output from the AD conversion means 4, and inputs data having a binarization level equal to or higher than a predetermined binarization level to "1", and data having a binarization level smaller than "1". The data is converted to 0 "and the binarized image data c is output. Other than this is the same as FIG.

[0026]

According to the projection inspection apparatus for a semiconductor integrated circuit device of the present invention, a multi-valued image for one projection is registered in advance as a template image, and a multi-valued pattern matching operation is performed with a captured image. After that, since the coordinates of the protrusion are calculated by the sub-pixel calculation, there is an effect that the displacement of the solder protrusion can be performed with high accuracy.

Further, since the labeling process is performed after binarization, the approximate coordinates of the solder protrusions are detected, and then a narrow range only in the vicinity thereof is inspected by multi-value pattern matching. Can be done. Since the correlation value of each solder protrusion is calculated based on the image of the normal solder protrusion registered in advance, the shape of the defective solder protrusion is smaller than the correlation value calculated from the solder protrusion. Defective shape of the solder protrusion can be detected.

[Brief description of drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIGS. 2A and 2B are pattern diagrams for explaining a multi-valued pattern matching calculation unit 13 in FIG.

FIG. 3 is a graph for explaining a sub-pixel operation unit 14 in FIG. 1;

FIG. 4 is a block diagram showing a second embodiment of the present invention.

FIG. 5 is a block diagram showing an example of the related art.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Illumination 2 BGA 3 Camera 4 A / D conversion means 5 Multivalued image data storage means 6 Binarization means 7 Binarized image data storage means 8 Labeling means 9 Projection approximate coordinate detection means 10 General position determination means 11 Search range Calculation means 12 Template image storage means 13 Multi-valued pattern matching calculation means 14 Sub-pixel calculation means 15 High-precision position determination means 41 Inspection object 42 Lighting 43 Camera 44 AD conversion means 45 Binarization means 46 Labeling means 47 Determination means a Analog signal b, b1 Multivalued image data c, c1 Binary image data d Label data e Projection approximate coordinate data ef Outline coordinate data signal g Search range data h Template image data k Correlation value data l Subpixel projection Coordinate data

Continuation of front page (51) Int.Cl. ⁶ Identification code Office reference number FI technical display location // H01L 21/321 H01L 21/92 604T 604Z

Claims (3)

[Claims] 1. An illumination for irradiating a projection portion to be inspected of a semiconductor integrated circuit device having a plurality of projection portions for electrically connecting to a circuit board or the like from diagonally above, and an inspection mounted above the projection portion to be inspected. A camera that captures an image of the target protrusion, an AD conversion unit that AD-converts an analog signal output from the camera into multi-valued (gray) image data, and AD
The multi-valued image data storage means for storing the multi-valued image data output from the conversion means and the data having a preset binarization level or higher by inputting the multi-valued image data are
1 ”, data smaller than the binarization level is converted to“ 0 ”and binarized image data is output, and binarized image data storage for inputting and storing the binarized image data. Means, a labeling means for reading out the binarized image data from the binarized image data storage means, performing labeling processing and outputting the label data, and inputting the label data to calculate the barycentric coordinates for each label to be inspected In a projection inspection apparatus for a semiconductor integrated circuit device, comprising: a rough coordinate detection unit that is a rough coordinate of a projection, (A) comparing the rough coordinates of the projection to be inspected with the coordinates where the projection should originally exist. However, if there is a protrusion that is determined to be significantly out of the inspection standard, a misregistration defect signal is output, the inspection for that protrusion is aborted, and the position of the next protrusion is inspected. Far off (B) a search range for calculating a search range having a size slightly larger than a preset projection size centered on the coordinates of the projection to be inspected, Calculation means, (C) template image storage means in which multi-valued image data for one protrusion is registered in advance, (D) search range calculation means from search range and template image storage means from template image to multi-valued Multi-valued pattern matching calculation means for inputting multi-valued image data from the image storage means, performing multi-valued pattern matching processing from the multi-valued image data at each coordinate in the search range and template image data, and calculating a correlation value And (E) two-dimensional curve interpolation is performed for each coordinate correlation data output from the multi-valued pattern matching calculation means. (F) Sub-pixel calculation means for calculating the peak coordinates on the curve to obtain the sub-pixel protrusion coordinates, and (F) comparing the sub-pixel protrusion coordinate data with the coordinates where the protrusion should originally be, to obtain a preset positional deviation allowable range. A projection inspection device for a semiconductor integrated circuit device, comprising: a high-precision position determination means for determining a position misalignment defect if it is not included. 2. A maximum value of each coordinate correlation data output from the multi-level pattern matching calculation means is searched and set as a maximum correlation value. If the maximum correlation value is smaller than a preset value, it is determined that the shape is defective. 2. The projection inspection apparatus for a semiconductor integrated circuit device according to claim 1, further comprising a shape determination unit that performs the shape determination. 3. The projection inspection apparatus for a semiconductor integrated circuit device according to claim 1, wherein the projection can be inspected even when the projection is a solder bump.