Ball grid array defect analysis method and equipment

Abstract

The present invention provides an analysis method for a ball grid array on a chip, comprising acquiring detection data of a defective ball grid in a ball grid array on a chip, wherein the detection data describes a reference point physical position and a physical coverage range of each defective ball grid in the chip, translating the reference point physical position and the physical coverage range into reference pixel positions and pixel coverage range represented by pixels based on a predetermined pixel resolution, and generating an indication of a defect type characterizing the defective ball grid based on the reference pixel positions and the pixel coverage range.

Description

Technical Field

[0001] The present invention relates to ball grid array technology, and in particular to the analysis of defective ball grids. Background Art

[0002] Ball grid array (BGA) is a new chip packaging form. It uses solder balls (or columns) arranged in a matrix on the bare bottom of the chip as the lead-out terminal. It has superior characteristics than pin packaging, so it has been widely used to adapt to the development of the semiconductor industry. However, due to the limitations of packaging technology, problems such as missing solder balls and bridging of two or more solder balls may occur during chip implantation. Therefore, it is necessary to detect these possible problems to avoid the reflow of the problematic chip. Here, the ball grid with such problems is called a "defective ball grid". At the same time, it is expected to analyze the detected defective ball grid so as to improve the implantation process or equipment to avoid or reduce the occurrence of defective ball grids.

[0003] At present, the traditional commonality analysis of defective areas of ball grid array (BGA) is to manually define the defective area by using a grid and then calculate the number of defects in the area. For example, the ball grid array area of ​​the entire chip is divided into 16 areas, and then based on the defective balls identified in advance, the ratio of defective balls in each area is counted, so as to achieve the distribution estimation of defective balls on the entire ball grid array. However, it can be seen that this statistical method inevitably has the problems of long time consumption and low classification accuracy, and only this numerical statistics cannot accurately reflect the exact area where the defect occurs, so it is not conducive to the improvement of ball planting process or equipment. Summary of the invention

[0004] The present invention proposes a scheme for counting defective ball grids based on a graphical method, which can not only visually present the defect types of the defective ball grids, but also accurately determine the locations where the defects occur.

[0005] According to one aspect of the present invention, there is provided an analysis method for a ball grid array on a chip, comprising: acquiring detection data of a defective ball grid in the ball grid array on a chip, the detection data describing a reference point physical position and a physical coverage range of each defective ball grid in the chip; translating the reference point physical position and the physical coverage range into a reference pixel position and a pixel coverage range represented by pixels based on a predetermined pixel resolution; and generating an indication of a defect type characterizing the defective ball grid based on the reference pixel position and the pixel coverage range.

[0006] According to another aspect of the present invention, a statistical platform for automatically visualizing defective ball grids based on big data is provided, which integrates valuable information by counting data of defective ball grids on a large number of chips, thereby providing a more valuable reference for process improvement. Therefore, the method according to the present invention further includes: for each chip among a plurality of chips, obtaining a plurality of first ball grid array graphics and / or second ball grid array graphics showing defective ball grids of the first and second defect types; superimposing the first ball grid array graphics and / or the second ball grid array graphics to generate a heat map indicating the problem defects on the chip, wherein different areas on the heat map indicate the frequency and position of the defective ball grid. In a preferred embodiment, the plurality of chips come from the same chip position in a tray used in the ball grid implantation process, wherein the frequency represents the statistical distribution of the defective ball grid occurring at the chip position. Thus, the present invention can count the statistical distribution of the defective ball grid occurring at different chip positions in the tray, thereby improving the ball grid implantation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1A and 1B A defective ball grid is schematically shown;

[0008] Figure 2 A flow chart showing a ball grid defect analysis method;

[0009] Figure 3 A schematic diagram showing a defective ball grid on a standard template is shown;

[0010] Figure 4 A flow chart showing the generation of a defect spherical grid representation on a standard template;

[0011] Figure 5 A schematic diagram for determining the position of a defective ball grid is shown;

[0012] Fig. 6A and 6B A thermal map of a defective ball grid based on big data is shown. DETAILED DESCRIPTION

[0013] The method and device provided by the embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. Although the preferred embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0014] Figure 1A and 1B The embodiment of defective ball grid in the ball grid array on the chip is shown as an example, wherein Figure 1AIn the partial view of the ball grid array shown in FIG. 1 , a missing ball grid appears, that is, no solder ball is implanted at position A; Figure 1B In the partial view of the ball grid array shown in FIG, a bridged ball grid appears, wherein the solder balls implanted at positions B and C are connected together. In the defective ball grid detection using the prior art, the detected defective ball grid is defined by a reference point R and its coverage range, such as Figure 1A and 1B As shown in , the coverage is represented by a rectangular box SQ in the figure, which covers the missing solder balls, and the reference point R is a corner point of the rectangular box. Therefore, the missing ball grid can be represented as {(x0, y0), (L, W)}, where (x0, y0) represents the coordinates of the reference point R, and L, W represent the side lengths of the rectangular range box. In one example, the detection data is recorded in actual physical positions, for example, the coordinates of the reference point and the size of the rectangular box are recorded in microns, and the coordinates of each defective ball grid are usually recorded with the upper left corner of the bare bottom view of the chip SoC as the coordinate origin, for example Figure 1A The coordinate system XY is shown as an example. Figure 1A The missing solder ball in , since only one solder ball is covered, its covering frame is a square, that is, L = W, and for example Figure 1B The coverage of the two solder balls bridged in the figure is a rectangle, that is, L≠W. Usually, the reference point R has a predetermined relative position relationship with the related physical coverage SQ. For example, in this example, the reference point R is a corner point of SQ. Therefore, the coordinates of any point in the coverage SQ can be determined by using the coordinates of the reference point R. That is, for any point (x i ,y i ), whose coordinates are (x0+Δx i ,y0+Δy i ), where Δx i ,Δy i They represent the distance from the reference point R to any point i within the coverage range.

[0015] Figure 2 The analysis method for ball grid array on chip according to one embodiment of the present invention is shown. In step 201, an original detection data set ORG_DataSet about defective ball grid for a certain type of chip SoC is obtained, wherein each data in the data set includes the reference point and its coverage data {(x0, y0), (L, W)} for identifying the defective ball grid as described above.

[0016] According to this example, in order to graphically show each defective spherical grid and its defect type, it is necessary to characterize the position of each defective spherical grid in a pixel manner. To this end, each data represented by physical coordinates in the detection data set DataSet is converted into a pixel representation. Specifically, in step 203, the original reference point physical position (x0, y0) of each defective spherical grid in the detection data set ORG_DataSet is first converted into a physical position representation (x′0, y′0) in the second coordinate system. As an example, Figure 1A The second coordinate system X′-Y′ is shown, where the origin (0, 0) is located at the lower left corner of the chip. The conversion to the coordinate system X′-Y′ here is to facilitate normal viewing habits and operations. The conversion between the two coordinate systems can be achieved through a conversion matrix. Figure 1A The coordinate system XY and X′-Y′ shown are actually flipped 180 degrees, so the transformation matrix T can be defined as follows:

[0017]

[0018] Therefore, the position of each reference point after conversion can be expressed as:

[0019]

[0020] Thus, the dataset TRS_DataSet under the coordinates X′-Y′ can be generated. Since the coverage does not change during the conversion process, for the convenience of explanation, the coverage of each defect spherical grid in the dataset TRS_DataSet is still represented by (L, W). Therefore, in the dataset TRS_DataSet, each defect can be represented as {(x′0, y′0), (L, W)}.

[0021] In step 205, based on the desired image resolution dpi, the physical position of each reference point in the data set TRS_DataSet and the physical coverage {(x′0, y′0), (L, W)} are translated into reference pixel positions (Px0 and Py0) and pixel coverage (Lp, Wp) expressed in pixels, where Lp and Wp represent the horizontal and vertical dimensions of the coverage expressed in pixels, thereby generating a pixel data set Pixel_DataSet of the defective spherical grid.

[0022] For example, when the image resolution dpi is conventionally expressed in inches, and when the physical position of each reference point R and its coverage SQ are expressed in micrometers in this example, the mapping relationship between the physical position and the pixel position is as follows:

[0023] Px0=x′0·dpi / 25400

[0024] Py0=y′0·dpi / 25400

[0025] Lp=L·dpi / 25400

[0026] Wp=W·dpi / 25400

[0027] The above is only the conversion relationship between micrometers and inches to represent pixels. It is not difficult to understand that for the situation where physical positions and pixels are represented by other units, it can be achieved by adjusting the scale factor (25400 in this example).

[0028] In step 207, an indication of the defect type of the defective ball grid may be generated based on the reference pixel positions (Px0 and Py0) and their corresponding pixel coverage (Lp, Wp) generated by the translation in step 205. In this example, a graphic symbol is used to indicate the defect type of the defective ball grid for easy visual inspection.

[0029] According to the example of the present invention, in order to clearly and comparatively distinguish the defective ball grid from the normal ball grid, the present invention adopts the standard ball grid array template TP as a reference benchmark, as an example, Figure 3 As shown, for a normal ball grid, its position is defined by the center of the ball grid, i.e. (Px, Py), and each normal ball grid is represented by a standard solid circle, as shown by arrow 1, while for a bridged ball grid, as shown by arrow 2, a special graphic representation is used to seamlessly connect the two defective ball grids. Similarly, for a missing ball grid, other graphics such as a ring can be used to represent it. In this way, the state of the entire ball grid array can be standardized on the standard template TP. It should be pointed out here that Figure 3 A schematic diagram of a portion of a standard ball grid array template TP represented by solid circles is shown, on which ideally all normal ball grids are marked as standard solid circles, and each ball grid is located at a normal design position, also referred to herein as a standard pixel position. According to the present invention, the standard template TP to be displayed graphically has a resolution of dpi, which is consistent with the pixel resolution when the data is translated in step 205.

[0030] It should be noted here that in some implementations, it is not necessary to perform coordinate transformation of the physical location data. For example, it is obviously not necessary to perform transformation on a data set collected directly in the second coordinate system X′-Y′, so step 203 can be skipped directly in the analysis.

[0031] Figure 4Shows a flowchart of marking defective ball grids on a standard template TP according to an example of the present invention. In step 401, pixel data of the first defective ball grid is extracted from the data set Pixel_DataSet, including {((Px0 and Py0))(Lp, Wp)}. In step 403, the shape of its coverage profile SQ is determined based on the coverage data (Lp, Wp). When Lp is substantially equal to Wp and close to a predetermined minimum value Vmin, it can be determined that the current pixel data represents a missing ball grid; when Lp >> Wp, for example, the ratio of Lp to Wp is close to 2, it can be determined that the current pixel data represents a bridged ball grid, and the bridging occurs between two solder balls. In step 403, if it is determined that Lp is substantially equal to Wp, proceed to step 405.

[0032] In step 405, based on the reference pixel positions ((Px0 and Py0)), the exact position of the missing ball grid within the coverage SQ is determined, in this example, its center position is determined. For this purpose, as Figure 5 shown, with the reference pixel point position ((Px0 and Py0)) as the reference, each point within the pixel coverage defined by (Lp, Wp) is traversed to determine a point C0, and the sum of the Euclidean (or Gaussian) distances between this point C0 and the standard ball grids whose pixel positions in the standard template TP are outside the pixel coverage SQ is the shortest, denoted as D0, that is:

[0033]

[0034] In the formula, Px C0 = Px0 + ΔPx C0 , Py C0 = Py0 + ΔPy C0 , and ΔPx C0 , ΔPy C0 represent the lateral and longitudinal distances of point C0 from the reference point R, where 0 < ΔPx C0 < Lp, 0 < ΔPy C0 < Wp. In addition, in the formula, (Px i , Py i ) represents the position of the standard ball grid outside the coverage SQ, and N represents the number of standard ball grids outside the coverage SQ. Here, the point C0 with the shortest Gaussian distance is defined as the center of the missing ball grid, and its position data (Px C0 , Py C0 ) is stored in the memory.

[0035] According to another example of the present invention, in order to reduce the amount of calculation, a standard spherical grid within a certain range outside the pixel coverage range SQ can also be specified to calculate the center C0 of the defective spherical grid under the SQ coverage range, such as specifying a spherical grid within a quarter area of ​​the spherical grid array in the standard template, etc.

[0036] If it is determined in step 403 that Lp>>Wp, proceed to step 407. As mentioned above, if Lp>>Wp, it indicates that there are at least two bridged ball grids under the current coverage range. Based on the value size and ratio of Lp and Wp, the number of defective ball grids under the coverage range can be determined. For example, in this example, the ratio of Lp to Wp is close to 2, and Wp is close to the minimum value Vmin, then it can be determined that there are two bridged ball grids under the current coverage range. Therefore, in step 407, based on the reference pixel position (Px0 and Py0) and its coverage range (Lp, Wp), the center position of the two bridged ball grids is determined. To this end, still taking the reference pixel position (Px0 and Py0) as the reference, traverse each point in the rectangular pixel coverage range defined by (Lp, Wp) to determine two points C1 and C2, which have the shortest and second shortest sum of the Euclidean distances with the standard spherical grid whose pixel position is outside the pixel coverage range (Lp, Wp) in the standard template TP, namely D1 and D2. Correspondingly, the pixel positions of the defective spherical grids C1 and C2 are stored in the memory in association, namely C1: (Px0 and Py0). C1 , Py C1 ), C2:(Px C2 , Py C2 ).

[0037] In step 409, it is determined whether all defective spherical grid data in the data set Pixel_DataSet have been obtained. If not, the process returns to step 401 to continue reading the next data, and repeats steps 403-407 until all data are processed.

[0038] In step 411, the position data (Px) of the defective ball grid determined in step 409 is used to determine the exact position of the defective ball grid. C , Py C ), the corresponding defective ball grid is marked on the standard template TP, and presented in a different graphic from the standard ball grid. For example, for each missing ball grid position data (Px C , Py C ), on the standard template TP by (Px C , Py C ) mark, create a hollow circle graphic to indicate that at position (Px C , Py C), the hollow circle pattern is significantly different from the solid circle used to mark the standard ball grid, so as to highlight the defective ball grid. In this way, the situation where the ball grid is missing can be marked on the standard template TP representing the ball grid array of the entire chip, so that the ball grid missing situation can be visually checked.

[0039] For the bridged ball grid case, for example, for the two ball grids C1 and C2 determined in step 407, on the standard template TP, (Px C1 , Py C1 ) and (Px C2 , Py C2 ) marked positions, the bridge ball grids at these two positions can be marked with the same or different graphics as the standard graphics. In this example, they are still represented by the same solid circle form as the normal ball grid, such as Figure 3 However, in order to illustrate the bridging defect, according to an example of the present invention, the two position coordinates (Px C1 , Py C1 ) and (Px C2 , Py C2 ) performs interpolation calculations to calculate the middle point between the two positions. For example, for the middle point C12 between C1 and C2, its coordinates can be determined as follows:

[0040]

[0041] At the same time, at the generated middle point C12, the same solid circle graphic is created to fill the space between C1 and C2. In this way, the middle points between C1 and C12 and between C12 and C2 are calculated, and solid circles are inserted at each calculated middle point to further fill the space between C1 and C12 and between C12 and C2. The interpolation process of the middle point is repeated until the space between the two solid circles representing C1 and C2 is seamlessly filled. Thus, the standard template is presented as follows Figure 3 The ball grid pattern shown hereby allows intuitive viewing of the two ball grids and their positions where bridging occurs. In this way, all the bridging ball grids occurring on the chip ball grid array can be marked on the standard template TP, thereby forming a graphic display TP'.

[0042] It should be noted here that different types of defective ball grids can be marked on the same template TP or marked separately, so as to facilitate viewing of each defect situation.

[0043] In the above manner, each defective ball grid and its type in a data set of defective ball grids detected on a chip can be marked on the template TP, thereby forming a graphic display TP' reflecting the defects of the ball grid array on the chip.

[0044] In order to analyze the causes of defects caused by processes or equipment on the production line, it is necessary to analyze the defects based on big data. On the production line, each chip is usually placed on a tray for batch implantation, where the tray is pre-set with multiple positions to hold the chips. For example, the tray may have 4 rows and 9 columns of holding positions, which can hold 4*9 chips. Therefore, in another implementation of the present invention, for each holding position in the tray, for example, for the 1st row and 1st column position, a detection big data set DS of M chips that have been implanted at the holding position is obtained. 1,1 , DS 1,1 ={DataSet 1 1,1 ,DataSet 2 1,1 ,...DataSet M 1,1 ,}, for the position of row 1 and column 2, obtain the detection data set DS of M chips that have been ball-planted at this placement position 1,2 , DS 1,2 ={DataSet 1 1,2 ,DataSet 2 1,2 ,...DataSet M 1,2 In this way, a large detection data set DS of M chips that have been ball-planted at the 4th row and 9th column placement position is obtained. 4,9 , DS 4,9 ={DataSet 1 4,9 ,DataSet 2 4,9 ,...DataSet M 4,9 It should be noted here that the number of samples M in the data set obtained at each placement position may be the same or different.

[0045] According to an example of the present invention, for each large data set DS obtained at each placement position, a corresponding ball grid array graphic display TP′ can be generated for each data set DataSet in the large data set DS based on the specific defect type. For example, for a missing ball grid, an array display graphic TP1′-TP M′, and then superimpose these M TP′ graphics to form a specific graphic reflecting the missing ball grids on the M chips at the placement position, such as a heat map, and the superposition method is as follows: If the same position on the two superimposed graphics is a standard graphic (such as a solid circle), that is, a normal ball grid, then the solid circle of the same graphic remains unchanged. If the same position on the two superimposed graphics is a graphic representing a missing ball grid (such as a hollow circle), then the graphic remains unchanged, but the grayscale of the graphic is increased or the color of the graphic is changed to indicate that the number of missing ball grids at this position has increased. If the graphics at the same position on the two superimposed graphics are different, for example, the previous graphic has a solid circle at this position, and the next graphic has a hollow circle at this position, then in the superimposed graphic, the graphic at this position is changed to a graphic representing a missing ball grid, that is, a hollow circle, to indicate that a missing ball grid has appeared at this position. In this way, the superimposed graphics are superimposed with TP3′, TP4′… to form a comprehensive big data graph superimposed with M TP′ graphics, such as Fig. 6A shown. Fig. 6A A schematic diagram showing the distribution and degree of missing ball grids produced on a complete standard template based on the detection data of a large number of chips is shown, wherein different grayscales represent the degree of ball grid missing at that location; or in another example, different gradient colors can be used to indicate the degree of ball grid missing. Figure 6B The figure shows the distribution and degree of the bridged ball grid produced on a complete standard template based on the detection data of a large number of chips. Fig. 6A , 6B , the graphic on the right represents the gray level indicator.

[0046] It should be pointed out here that Fig. 6A The complete template shown is made according to the predetermined resolution rpi based on the bare bottom image of the actual chip, wherein the bare bottom is distributed with a ball grid array, a capacitor area in the middle area, and a calibration area at the edge. In the process of making the template based on the bare bottom image, the areas such as the capacitor area and the calibration area that do not contain the ball grid are removed through image recognition processing, thereby obtaining a complete ball grid array image for this model of chip. On this basis, the complete ball grid array image is subjected to noise filtering and binarization processing to obtain a complete standard template TP. The binarization here means: for the background, a 0 value is used, and for each ball grid in the background, a standard graphic representation represented by a 1 value is used, or vice versa.

[0047] In the above manner, the ball grid array implantation quality at a fixed placement position on the tray can be statistically analyzed, and the implantation process at the position can be improved according to the statistical results. The ball implantation process of a batch of chips on the tray can be visually viewed, so that the ball implantation process of the entire tray can be viewed and improved.

[0048] The analysis method according to the present invention can be implemented by a central processing unit CPU or a processor executing a computer-readable program or instructions stored in a memory, wherein the processor implements the method of the present invention by executing the instructions in the memory, and the analysis result is graphically presented on a display.

[0049] The present invention is shown and described in detail above through the accompanying drawings and preferred embodiments. However, the present invention is not limited to these disclosed embodiments. Based on the above multiple embodiments, those skilled in the art can know that the code review methods in the above different embodiments can be combined to obtain more embodiments of the present invention, and these embodiments are also within the protection scope of the present invention.

Claims

1. A method for analyzing a ball grid array on a chip, comprising: Acquire detection data of defective ball grids in a ball grid array on a chip, wherein the detection data describes a reference point physical position and a physical coverage range of each defective ball grid in the chip; Translating the reference point physical position and physical coverage into reference pixel position and pixel coverage represented by pixels based on a predetermined pixel resolution; Based on the reference pixel position and pixel coverage, generating a graphic symbol indication representing the defect type of the defective ball grid; The physical coverage has a first characteristic type and a second characteristic type, The analysis method further comprises: Acquire a standard template of a graphic symbol of a standard ball grid array of the chip, wherein the standard template includes standard pixel position data and a standard graphic symbol of each standard ball grid, wherein the graphic symbol of the standard ball grid array has the pixel resolution; Taking the reference pixel position as a reference, traverse each point within the pixel coverage range to determine at least two points therein, wherein each of the at least two points has the shortest Gaussian distance to a standard spherical grid in the standard template whose standard pixel position is outside the pixel coverage range; mark the at least two points as first confirmed positions of a defective spherical grid of a first defect type and store pixel position data of the first confirmed position; or Taking the reference pixel position as a reference, traversing each point within the pixel coverage range to determine a point therein, wherein the point has the shortest Gaussian distance to a standard spherical grid in the standard template whose standard pixel position is outside the pixel coverage range; The one point is marked as a second confirmed position of a defective sphere grid of a second defect type and pixel position data of the second confirmed position is stored.

2. The analysis method according to claim 1, further comprising: Acquire detection raw data of a defective ball grid in the ball grid array on the chip, wherein the raw data includes the physical position of the original reference point of the defective ball grid represented in a first coordinate system and the physical coverage range; The original reference point physical position of the defective spherical grid is converted into the reference point physical position in the second coordinate system.

3. The analysis method according to claim 2, further comprising: Based on the characteristic type of the physical coverage, the defect type of the defective ball grid is determined.

4. The method according to claim 1, further comprising: Interpolation calculations of intermediate values ​​of two points are iteratively performed between the first confirmed positions of the at least two points, wherein each interpolation represents an inserted point between the first confirmed positions, until the space between the two standard graphic symbols representing the at least two points is completely filled with the standard graphic symbols representing the inserted points to form an indication of the first defect type.

5. The analysis method according to claim 4, further comprising: The defect ball grid of the first defect type filled with the standard graphic symbol is presented on the standard template to cover the standard ball grid corresponding to the pixel position of the first confirmed position in the standard template to obtain a first ball grid array pattern.

6. The analysis method according to claim 1 further includes presenting the defect ball grid of the second defect type with another graphic symbol on the standard template to cover the standard ball grid corresponding to the pixel position of the second confirmed position in the standard template to obtain a second ball grid array graphic, wherein the other graphic symbol is different from the standard graphic symbol and is used to indicate the second defect type.

7. The analysis method according to claim 5 or 6, further comprising: For each chip of the plurality of chips, acquiring a plurality of first ball grid array patterns and / or second ball grid array patterns showing defective ball grids of the first and second defect types; The first BGA pattern and / or the second BGA pattern are superimposed to generate a heat map indicating problematic defects on the chip, wherein different areas on the heat map indicate the frequency and location of defective BGAs.

8. The method of claim 7, wherein the plurality of chips are from a same chip position in a tray used in a ball grid implantation process, wherein the frequency represents a statistical distribution of occurrences of the defective ball grid at the chip position. 9 . The method according to claim 8 , further comprising calculating the statistical distribution of the defective ball grid at different chip positions in the tray.

10. The method of claim 9, wherein the statistical distribution can be used to improve the ball grid implant process.

11. The analysis method according to claim 2, further comprising generating the standard template by: acquiring an image of the chip including the ball grid array; Preprocessing the image includes: identifying and eliminating areas of the image that are not relevant to the ball grid array to form a ball grid matrix; filtering the ball grid array to eliminate noise; Binarizing the ball grid array region in the image to obtain a binary image of the ball grid array; The second coordinate system is established for the image and the center point position coordinates of the binary image of each spherical grid are determined in the second coordinate system, and the standard graphic symbol is used to represent the standard spherical grid located at the center point position coordinates.

12. An apparatus for analyzing defective ball grids, comprising: monitor; a memory having a computer readable program stored therein, A processor, configured to execute the readable program of the right to implement the method of any one of claims 1-11.

Images