JP5537801B2 - Positioning apparatus, positioning method, and bonding apparatus - Google Patents

Description

  The present invention relates to a positioning device, a positioning method, and a bonding device.

  Conventionally, a bonding apparatus is used when bonding a chip formed by dicing a wafer. That is, a wafer ring is set on the wafer table of the bonding apparatus, and a plurality of chips are attached to a wafer sheet stretched on the wafer ring. Then, the chip is picked up by a pick-up device and mounted on a substrate such as a lead frame.

By the way, when picking up, it is necessary to align the chip. That is, this type of pickup device generally picks up a chip by sucking it with a suction collet. For this reason, if the chip is misaligned, it may not be possible to mount the chip at the proper position.

  Therefore, conventionally, it has been proposed to perform chip alignment using pattern matching. Pattern matching is a process of associating the same pattern that exists in a certain image with which part in another image, and a typical example is template matching. . That is, as shown in FIG. 5, a method in which a pattern called a template 51 expressed in units of pixels is prepared in advance, and this is overlapped with the image 52 while moving to detect a highly similar pattern from the search target image. It is. There are various position detection devices using pattern matching as described in Patent Document 1, Patent Document 2, and the like.

  By the way, in order to align the chip using pattern matching, the following steps are performed. First, an image of a chip conveyed to a pickup position is picked up by an image pickup means such as a camera, and this picked-up image is written in a memory such as a personal computer (personal computer) that performs image processing calculation. Then, a pattern by a software operation for comparing similarities such as SAD (sum of absolute difference of luminance), SSD (sum of square of absolute difference of luminance), NCC (normalized cross-correlation function), etc. with respect to the image in the memory Using the matching algorithm, the position information of the search object is obtained.

  If the position of the chip (semiconductor chip) is shifted from the normal position based on the position information, the position is corrected (corrected), and the corrected chip is picked up by the pickup device. It will be mounted on a mounting part such as a lead frame. Thereafter, the next chip to be picked up is transferred to the pick-up position, and the above steps are sequentially performed.

The SAD can be obtained by the following equation 1, the SSD can be obtained by the following equation 2, and the NCC can be obtained by the following equation 3.

JP 2008-283006 A JP 2006-41089 A

  However, in such a case, when information is written to a personal computer, exposure time and image transfer time are required. In addition, a calculation time for obtaining the position information of the search object using the pattern matching algorithm is also required. That is, image processing is started only after exposure and image transfer are completed and image information is stored in the memory. For this reason, if the image is large, it takes time to transfer and the timing for starting the image processing is delayed. In addition, the larger the image, the longer the image processing time.

Therefore, it can be proposed to reduce the processing time by performing a pattern matching algorithm by hardware calculation. However, with this hardware calculation, it is difficult to detect the position of an object that has rotated more than a certain amount within the image field of view. Even if it can be done, it must be an expensive configuration.

  In view of the above problems, an object of the present invention is to provide a positioning device and a positioning method capable of reducing working time, and another object is to perform chip bonding work in a short time and accurately. An object of the present invention is to provide a bonding apparatus that can be used.

The positioning device of the present invention is a positioning device that specifies the positions of search objects arranged in a matrix, and includes an imaging unit that captures an image of the search object, and image information acquired by the imaging unit. Image capturing means for capturing, and information acquisition means for acquiring two-dimensional information on the XY plane as position information of the search object on the image by a pattern matching algorithm using FPGA for the image information transferred from the image capturing means And the θ direction as more detailed position information by using a pattern matching algorithm using SAD, SSD, or NCC similarity comparison in the area of the position information of the search object obtained by the information acquisition means And sub-information acquisition means for obtaining information. As another positioning device, a dynamic reconfigurable processor may be used for a pattern matching algorithm with the information acquisition unit.

This FPGA can be constructed on an image capture board constituting the image capture means, or a dynamic reconfigurable processor can be constructed on an image capture board constituting the image capture means. Here, FPGA (Field Programmable Gate Array) is a kind of gate array in which a user can write a unique logic circuit. A programmable logic device that is particularly rewritable is called as follows. As an example of the internal structure, a logic block consisting of a combination circuit of about 4 inputs and a sequential circuit that can configure any logic is arranged in a grid form of several tens of tens of tens, and wiring between them is a simple crossbar. One having a structure connected by a switch is mentioned. A dynamic reconfigurable processor is a reconfigurable processor that can change the contents of a configuration register during execution of a program. As an LSI (Large Scale Integrated Circuit) that simultaneously realizes high computing performance and flexibility, there is a reconfigurable processor (Re-configurable Processor). The reconfigurable processor includes an arithmetic circuit such as an ALU (Arimatic Logic Unit) arranged in an array and a switch that connects the arithmetic circuits. The function of the arithmetic circuit and the wiring between the arithmetic circuits can be reconfigured according to the contents of a register called a configuration register, and the program is executed by changing the configuration contents according to the purpose. A dynamic reconfigurable processor is equipped with a large number of processing units consisting of computing units, shifts, registers, distributed memories, etc., and by switching the configuration and connection of these processing units in a short time, the area has excellent flexibility and high speed. It can be realized with efficiency.

By the way, in this specification, a pattern matching algorithm using FPGA or a pattern matching algorithm using a dynamic reconfigurable processor is called a pattern matching algorithm by hardware operation, and pattern matching using similarity comparison of SAD, SSD, or NCC The algorithm is called a pattern matching algorithm by software operation. According to the positioning apparatus of the present invention, since a pattern matching algorithm based on hardware calculation can be used, image processing can be started during image transfer. In addition, if the orientation of the search object in the θ direction is unknown at the start of work, the sub information acquisition means performs software calculation on the position information obtained by the pattern matching algorithm by hardware calculation at the start. More detailed position information, that is, the θ-direction orientation and sub-pixel position information can be obtained by the pattern matching algorithm. Thus, in total, detailed position information in the XY direction and the θ direction can be obtained.

  The bonding apparatus of the present invention includes the positioning apparatus described above. In such a bonding apparatus, it is possible to perform image processing of the chip position in a short time.

The positioning method of the present invention is a positioning method for specifying the position of a search object arranged in a matrix, and captures an image of the search object and captures this image information by an image capturing means. With respect to this image information, a pattern matching algorithm using FPGA obtains two-dimensional information on the XY plane as position information of the search object on the image, and further obtains two-dimensional information on the XY plane. In the position information area, the θ direction information as more detailed position information is obtained by using a pattern matching algorithm using similarity comparison of SAD, SSD, or NCC . As another positioning method, a dynamic reconfigurable processor may be used as a method of obtaining two-dimensional information on the XY plane as position information of a search object on an image.

  In the present invention, image processing can be started during image transfer. For this reason, the time required for the positioning operation can be shortened, and the efficiency of the positioning operation can be improved. Further, the sub-information acquisition means can obtain more detailed position information, that is, the θ-direction posture, etc., by a pattern matching algorithm based on software calculation, so that the overall positioning operation can be shortened and the accuracy can be improved. it can. Moreover, since the software calculation of the sub information acquisition means can be limited to the peripheral portion of the position information obtained by the hardware calculation, this software calculation processing does not require much time, and such detailed information Even if you get the total work time does not increase. And it becomes possible to detect the attitude | position of the search target which failed in the detection by the restrictions on hardware calculation using the pattern matching algorithm by software calculation as a sub-information acquisition means.

  The information acquisition means can be constituted by an FPGA or a dynamic reconfigurable processor, and stable hardware calculation can be performed. In addition, an FPGA or the like can be incorporated into an image capturing board, and the entire apparatus can be made compact.

  The bonding apparatus of the present invention can perform image processing of a chip position in a short time, can perform chip positioning in a short time, and can improve the efficiency of bonding work.

  Hereinafter, an embodiment of the present invention will be described with reference to FIGS.

  FIG. 1 shows a positioning apparatus, and FIG. 2 shows a bonding apparatus using this positioning apparatus. In the bonding apparatus, a chip 1 cut out from a wafer 9 (see FIG. 3) is pipped up at a pickup position P and transferred (mounted) to a bonding position Q of a substrate 2 such as a lead frame. As shown in FIG. 1, the wafer 9 is adhered onto a wafer sheet (adhesive sheet) 11 stretched on a metal ring (wafer ring) 10 and divided into a large number of chips 1 by a dicing process ( Divided). Therefore, the chips 1 are arranged in a matrix as shown in FIG. 3, and are the search object W of the present invention.

  As shown in FIG. 2, this bonding apparatus includes a collet (adsorption collet) 3. The collet 3 is moved by an unillustrated moving mechanism in the direction of arrow A and lowered in the direction of arrow B on the pickup position P, and in the direction of arrow C and lowered in the direction of arrow D on the bonding position Q. The reciprocation between the pickup position P and the bonding position Q in the directions of arrows E and F is possible. In the moving mechanism, the movement of the arrows A, B, C, D, E, and F is controlled by a control unit configured by, for example, a microcomputer. In addition, as a moving mechanism, it can comprise with various mechanisms, such as a cylinder mechanism, a ball screw mechanism, and a motor linear mechanism.

  The suction collet 3 is provided with a head (suction nozzle) 4 having an suction hole 8 opened on the lower surface thereof, and the chip 1 is vacuum-sucked through the suction hole 8, and the chip 1 is placed on the lower end surface (tip surface) of the head 4. Adsorbs. When this vacuum suction (evacuation) is released, the chip 1 is detached from the head 4.

  Further, the wafer 9 divided (divided) into a large number of chips 1 is arranged on, for example, an XYθ table 5 (see FIG. 3), and the XYθ table 5 is provided with push-up means having push-up pins. That is, the push-up means pushes up the chip 1 to be picked up from below and makes it easy to peel from the adhesive sheet. In this state, the chip 1 is adsorbed to the adsorbing collet 3 that has been lowered.

  By the way, when the chip 1 is picked up at the pickup position P, the chip 1 needs to be in a normal posture. For this reason, the positioning apparatus according to the present invention is used in this bonding apparatus.

  As shown in FIG. 1, the positioning device includes an imaging unit 20 such as a CCD camera that captures an image of a search target, an image capturing unit 21 that captures image information acquired by the imaging unit 20, and the image. Information acquisition means 22 for obtaining position information of a search object on the image by a pattern matching algorithm by hardware calculation for the image information transferred from the capturing means 21, and the information acquisition by a pattern matching algorithm by software calculation Sub-information acquisition means 23 for obtaining more detailed position information from the position information of the search object obtained by means 22 is provided.

  The image capturing unit 21 captures image data captured by the image capturing unit 20, and an FPGA 25 for configuring the information acquisition unit 22 is disposed in the image capturing unit 21. An FPGA (Field Programmable Gate Array) is a kind of gate array in which a user can write a unique logic circuit. A programmable logic device that is particularly rewritable is called as follows. As an example of the internal structure, a logic block consisting of a combination circuit of about 4 inputs and a sequential circuit that can configure any logic is arranged in a grid form of several tens of tens of tens, and wiring between them is a simple crossbar. One having a structure connected by a switch is mentioned.

  In this case, the positioning of the chip 1 (search object) is performed using pattern matching. Pattern matching is a process of associating the same pattern that exists in a certain image with which part in another image, and a typical example is template matching. . In other words, this is a technique in which a pattern called a template is expressed in pixel units in advance, and this is superimposed on the image while moving, and a pattern with high similarity is detected from the search target image.

  By the way, data is written from the image capturing means 21 to the memory 27 of the personal computer (personal computer) 26, and the CPU 28 performs image processing calculation based on the data in the memory 27. For this reason, here, an exposure time and an image transfer time are required.

That is, when an input image, a template (pattern image), a threshold value for matching or the like are input from the personal computer 26, a circuit that returns matched coordinates may be designed, and this circuit may be mounted on the FPGA 25.

  The algorithm implemented in the circuit is almost the same as the calculation method of the correlation coefficient in software. However, by parallelizing the points that need to be calculated for each pixel and pipelining a series of calculations, an FPGA with a low operating frequency can be calculated faster than that of software.

  If necessary, the sub information acquisition unit 23 performs SAD (sum of absolute differences in luminance), SSD (absolute difference in luminance) on image data in the memory 29 within the area narrowed down by the information acquisition unit. The position information of the search object on the image is obtained with high accuracy by using a pattern matching algorithm based on a software operation that performs similarity comparison such as the sum of squares of) and NCC (normalized cross correlation function). That is, the sub information acquisition unit 23 is configured by the CPU 28 or the like.

  The SAD (total sum of absolute differences in luminance) can be obtained by the above-mentioned formula 1 in the background art column, and the SSD (sum of squares of absolute differences in brightness) can be obtained by the formula 2 in the background art column. NCC (normalized cross-correlation function) can be obtained by the above equation 3 in the background art column.

  In this bonding apparatus, for example, the pattern matching by the software calculation by the sub information acquisition unit 23 is performed immediately after the bonding process for the next wafer 9 is started (wafer change) after the bonding process of all the chips 1 of one wafer 9 is completed. (Just after).

  Immediately after the wafer change, the θ direction posture of the wafer 9 may be unknown (fluctuated). For this reason, it is necessary to perform positioning in the θ direction. However, it is difficult to match with the hardware calculation. For this reason, in this embodiment, pattern matching by software calculation is performed to correct the position in the θ direction.

  Next, a bonding method using the above bonding apparatus will be described. First, the chip 1 at the pickup position is picked up by the image pickup means 20, the image information is taken in by the image take-in means 21, and pattern matching is performed on the image information from the image take-in means 21 by hardware calculation by the FPGA 25. The position information of the search object on the image is obtained by an algorithm. This position information is two-dimensional information on the XY plane.

  Further, in the CPU 28 of the personal computer 26, SAD (total luminance absolute difference), SSD (luminance absolute difference) in the area of the detected position information obtained by hardware calculation based on the information in the memory 27 of the personal computer 26. The position information of the search object on the image, that is, the θ direction information is obtained by using a pattern matching algorithm by software calculation represented by NCC (normalized cross correlation function), NCC (normalized cross correlation function), and the like.

  Thereby, the position information of the chip 1 at the pickup position P can be obtained, and based on this position information, the XYθ table 5 is displaced and the position of the chip 1 is aligned. Thereafter, the collet 3 disposed at the pickup position P is lowered as indicated by the arrow B, and the chip 1 is adsorbed by the collet 3. At this time, as described above, the chip 1 is pushed up through the adhesive sheet 11 by the push-up pin of the push-up means, and the chip 1 is easily peeled off from the adhesive sheet 11. Thereafter, the collet 3 adsorbing the chip 1 is raised as indicated by an arrow A. Next, after moving the collet 3 in the direction of arrow E, the collet 3 is lowered along the direction of arrow D, and the chip 1 adsorbed on the collet 3 is mounted on a substrate 2 such as a lead frame. Thereafter, the collet 3 moves up in the direction of arrow C and then moves in the direction of arrow F to return to the initial state located above the pickup position P. If the chip 1 is attracted at the pickup position P, the push-up pin of the push-up means is lowered.

  Thereafter, the chip 1 to be picked up next is positioned at the pickup position P, and this chip 1 is also positioned by the positioning device. In this case, the alignment in the θ direction is performed in the previous time (initial stage), and the alignment of the wafer 9 in the θ direction is performed. For this reason, the alignment in the θ direction is also performed on the chip 1.

  Therefore, this time, only pattern matching of hardware calculation is performed. By this pattern matching, positioning in the XY directions is performed, position correction is performed, and pick-up work is started. Thereby, the mounting operation of the chip 1 on the substrate 2 is completed. Thereafter, the mounting operation after performing only the pattern matching of the same hardware operation is performed. After the wafer 9 is mounted, the wafer 9 is replaced and the above process is repeated.

  For this reason, the operation time of the pattern matching by the software calculation and the pattern matching by the hardware calculation becomes a graph as shown in FIG. That is, if the software calculation is performed, the pattern matching time is long, and if the hardware calculation is performed, the pattern matching time is short. Therefore, it can be seen that in the bonding operation using this apparatus, there are few software operations that require a long pattern matching time, and most hardware operations that use a short pattern matching time are sufficient.

  By the way, normally, the first imaging object immediately after the wafer exchange has a different θ orientation from that before the exchange, and detection failure often occurs in pattern matching by hardware calculation. For this reason, it is preferable to perform pattern matching by software calculation only when pattern matching by hardware calculation fails immediately after wafer replacement. By doing so, since the θ posture is constant until the next wafer exchange, if the θ of the template model is matched, detection failure in pattern matching by hardware calculation is eliminated. That is, it becomes possible to detect the attitude of the search object that has failed to be detected due to restrictions in hardware computation using the pattern matching algorithm based on software computation as the sub information acquisition means.

  In the present invention, since a pattern matching algorithm based on hardware computation can be used, image processing can be started during image transfer. For this reason, the time required for the positioning operation can be shortened, and the efficiency of the positioning operation can be improved. In addition, if the orientation of the search object in the θ direction is unknown at the start of work, the sub information acquisition means performs software calculation on the position information obtained by the pattern matching algorithm by hardware calculation at the start. More detailed position information, that is, the θ-direction orientation and sub-pixel position information can be obtained by the pattern matching algorithm. For this reason, the accuracy of the positioning operation as a whole can be improved. Moreover, since the software calculation of the sub information acquisition means is within the position information obtained by the hardware calculation, this software calculation processing does not take much time, and even if such details are obtained, the entire information is obtained. Work time does not increase.

  By the way, in the said embodiment, although FPGA was used for the information acquisition means 22, it may replace with this and may use a dynamic reconfigurable processor. Among the reconfigurable processors, the contents of the configuration register can be changed during program execution. As an LSI (Large Scale Integrated Circuit) that simultaneously realizes high computing performance and flexibility, there is a reconfigurable processor (Re-configurable Processor). The reconfigurable processor includes an arithmetic circuit such as an ALU (Arithmic Logic Unit) arranged in an array and a switch for connecting the arithmetic circuits. The function of the arithmetic circuit and the wiring between the arithmetic circuits can be reconfigured according to the contents of a register called a configuration register, and the program is executed by changing the configuration contents according to the purpose. A dynamic reconfigurable processor is equipped with a large number of processing units consisting of computing units, shifts, registers, distributed memories, etc., and by switching the configuration and connection of these processing units in a short time, the area has excellent flexibility and high speed. It can be realized with efficiency.

  For this reason, even if a dynamic reconfigurable processor is used for the information acquisition unit 22, pattern matching of hardware calculation can be performed, and the same operational effects as when the FPGA is used can be obtained.

  As described above, the embodiment of the present invention has been described. However, the present invention is not limited to the above-described embodiment, and various modifications are possible. For example, in the above-described embodiment, the sub-position for positioning the θ-direction posture is used. Although the information acquisition means 23 is provided, the information acquisition means 23 may be omitted. Further, in the case where the sub information acquisition unit 23 is provided, the sub information acquisition unit 23 is used for positioning in the θ-direction posture immediately after the wafer replacement. However, the sub information acquisition is performed after a predetermined number of chips 1 are picked up in one wafer 9. Pattern matching by means 23 may be performed. That is, in the above-described embodiment, pattern matching by software calculation is performed on a single wafer 9 only immediately after wafer replacement. However, pattern matching is performed on a single wafer 9 by multiple software calculations. You may do it.

  In the above-described embodiment, positioning is performed using the positioning device according to the present invention only at the time of pip-up, but this positioning device is used for positioning the substrate (lead frame) 2 when the chip 1 is mounted on the substrate 2 (lead frame). You may make it perform positioning using.

  In the embodiment, the positioning using the positioning device according to the present invention is performed only at the time of pip-up. However, the positioning using the positioning device is performed for positioning the substrate 2 when the chip 1 is mounted on the substrate 2. May be.

  Moreover, you may make it perform positioning using the positioning device concerning this invention not only in a bonding apparatus but in the other apparatus which needs positioning. In this case, the search object W is arranged in a matrix like the chip 1 of the wafer 9.

It is a simplified diagram showing a positioning device of an embodiment of the present invention. 1 is a simplified diagram illustrating a bonding apparatus according to an embodiment of the present invention. It is a simplified perspective view which shows a wafer. It is a graph which shows the relationship between the processing time of a software calculation and a hardware calculation. It is explanatory drawing of pattern matching.

Explanation of symbols

20 Image pickup means 21 Image capture means 22 Information acquisition means 23 Sub information acquisition means

Claims (5)

A positioning device for specifying the positions of search objects arranged in a matrix,
The image capturing means for capturing the image of the search object, the image capturing means for capturing the image information acquired by the image capturing means, and the pattern matching algorithm using FPGA for the image information transferred from the image capturing means. Information acquisition means for acquiring two-dimensional information on the XY plane as position information of the search object on the image, and the position information area of the search object obtained by the information acquisition means, SAD, SSD, or NCC A positioning apparatus comprising: a sub-information acquisition unit that obtains θ direction information as more detailed position information by using a pattern matching algorithm that uses similarity comparison. A positioning device for specifying the positions of search objects arranged in a matrix,
An image capturing unit that captures an image of a search target, an image capturing unit that captures image information acquired by the image capturing unit, and pattern matching using a dynamic reconfigurable processor for image information transferred from the image capturing unit Information acquisition means for acquiring two-dimensional information on the XY plane as position information of the search object on the image by an algorithm, and SAD, SSD in the area of the position information of the search object obtained by the information acquisition means or by using a pattern matching algorithm using a similarity comparison of the NCC, position-decided Me device you characterized in that a sub-information obtaining means for obtaining θ direction information as more detailed position information. A bonding apparatus comprising the positioning device according to claim 1 . A positioning method for specifying the positions of search objects arranged in a matrix,
An image of the search object is picked up, the image information is acquired by an image acquisition means, and an XY plane is obtained as position information of the search object on the image by a pattern matching algorithm using FPGA for the image information. Using a pattern matching algorithm that uses a SAD, SSD, or NCC similarity comparison in the region of the position information of the search object from which the two-dimensional information of the XY plane was obtained, A positioning method characterized by obtaining θ direction information as more detailed position information . A positioning method for specifying the positions of search objects arranged in a matrix,
An image of the search target is captured, the image information is captured by an image capturing unit, and the position of the search target on the image is detected by a pattern matching algorithm using a dynamic reconfigurable processor for the image information. A pattern matching algorithm using SAD, SSD, or NCC similarity comparison is obtained in the area of the position information of the search object from which the two-dimensional information of the XY plane is obtained as information, and the two-dimensional information of the XY plane is obtained. A positioning method characterized in that the θ direction information is obtained as more detailed position information.

Images