JPS6258659B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention is an automatic die bonder that aligns and adsorbs semiconductor pellets placed on an XY table and separated from a semiconductor wafer into a plurality of pieces to the adsorption point of a collect for die bonding. The present invention relates to a pellet positioning device.

In the die bonding process, semiconductor pellets (hereinafter referred to as ``pellets''), such as ICs, are bonded to the die pad of a lead frame, but since the pellets are extremely small, they must be precisely aligned. For this reason, it is necessary to position the pellets separated from the wafer into a large number of pieces using an automatic positioning device so that the pellets are attracted to the collect of the automatic die bonder at an accurate position.
The pellets adsorbed by the collet are moved onto the die pad of the lead frame and die-bonded.

A conventional pellet positioning device of this type has a first
The configuration was as shown in the diagram shown in the figure. Pellets 1 are separated into a large number of pellets and are adhered onto a transparent synthetic resin sheet 2 made of a transparent vinyl sheet or the like. These pellets 1 are attached to a transparent synthetic resin sheet 2 by making score lines on the semiconductor wafer, and are separated by stretching this synthetic resin sheet all around. The pellets 1 in this state are shown in FIG. 2, and each pellet 1 is arranged on a synthetic resin sheet 2 with regular separation widths.

Returning to FIG. 1, the synthetic resin sheet to which the separated pellets 1 are attached is placed on a support plate 4 on an XY table 3 that is moved in the X-axis and Y-axis directions. An illumination device 5 is disposed diagonally above the pellet group 1. This illumination device 5 includes a lamp 6, a diffusion plate 7 that diffuses the light from the lamp, and a condensing lens 8 that condenses the light that has passed through the diffusion plate and irradiates it onto the pellets 1 within a predetermined small area. It is designed to uniformly irradiate a large amount of light.

The irradiated light beam from the illumination device 5 is specularly or specularly reflected on the surface of the pellet 1, and on the optical axis of the light beam,
A planar imaging device 9 is disposed at an obliquely upper position. This imaging device 9 is a charge coupled device (CCD)
Or an industrial television camera 10 using an image pickup tube
and an imaging lens 11 that forms a predetermined image on the imaging surface of the television camera.

Illumination from the illumination device 5 illuminates the synthetic resin sheet 2, and reflected light enters the imaging lens 11.
Since the reflected light is specularly reflected, most of the reflected light from one surface of the pellet is incident.

When the video signal from the television camera 10 is processed and quantized by the A-D converter 12 using a threshold value that can separate the sheet 2 image and the pellet 1 image, the defective marks on the sheet 2 and pellet 1 sides become "0", i.e. Black, pellet side can be converted to "1", that is, white.

When the signal from this AD converter 12 is received on a monitor television, an image as shown in FIG. 3 is obtained. An image 2a is formed on the synthetic resin sheet 2, an image 1a is formed on the pellet 1, and an image 1b is formed if there is a defective mark attached to the surface of the pellet 1 during the electrical inspection.

13 is a digital control device, which is configured as follows. Control circuit 1 that sends control signals to each device
7, a graphic memory 14 that stores the digital signal from the A-D converter 12 as planar graphic information,
A position detection device 15 that performs arithmetic processing on the digital figure stored in this figure memory, and a shape determination device 16
It is made up of

The processing method using the above-mentioned conventional apparatus is as follows. The video signal from the imaging device 9 is input to an A/D converter 12, processed by an appropriate threshold value according to a control signal from a control circuit 17, and stored in a graphic memory 14 as a digital signal. Thereafter, the whole or part of the surface information in the figure memory 14 is processed by the position detection device 15, and the shape determination device 1
6, the shape of the pellet 1 is inspected, and if it is a good pellet 1, its center position is calculated, and the control circuit 17 instructs the XY table driving device 18 to move the amount, and the automatic die bonder adsorbs the collect 19. Move the XY table 3 so that the center of pellet 1 is at the point.

The operation of the control circuit 17 is as follows. Supports write signals and read signals to the graphic memory 14, provides various reference information to the position detection device 15 and the shape determination device 16, and
Performs input/output control, further obtains information from these position detection device 15 and shape determination device 16,
Instructs the XY table drive device 18 to move the table.

As a result, the XY table drive device 18 operates, the XY table 3 is moved, and the center of the pellet 1 reaches the collection point of the collect 19. Subsequently, the control circuit 17 provides a die bonding start request signal to a die bonder (not shown). Due to the operation of the die bonder, the collet 19 descends and pellets 1
It adsorbs, rises, moves to the die bonding point, and performs die bonding.

When the die bonding of one pellet 1 is completed, the control circuit 17 commands to move the XY table 3 to the predicted adsorption point position for the next pellet 1, read the video signal from the television camera 10 for this position, and first determine the shape. If a defect mark is attached to the pellet 1 or the surrounding area is chipped, the pellet is determined to be defective and the XY table is moved to the next predicted position. When it is determined that the pellet is a good pellet, the position detecting device 15 determines the center position of the pellet 1, and the XY table 3 is moved so as to reach the collection point.

In the conventional device described above, in order to position the pellet 1 on the XY table 3 at the suction point, each time before die-bonding one pellet 1, the video signal is read from the television camera 10 into the figure memory 14, and the shape is determined. In particular, since the position is calculated and this pattern recognition process is repeated for each pellet 1 on the synthetic resin sheet 2,
When there are a large number of defective pellets, many recognitions are performed in vain, and even parts of the pellets 1 to which no pellets are attached are recognized for confirmation. For this reason, not only wasted time was wasted, but the die bonder itself was forced to wait, reducing productivity.

This invention was made in order to eliminate the drawbacks of the conventional device as described above, in which graphic processing is performed each time while moving the XY table. The center position of only the good pellets is calculated at high speed, and a good product map is created using the data converted to the coordinate values of the XY table, and this good product map is referred to and read out before each die bonding, and the collection is adsorbed. The XY table is moved sequentially so that the center of the pellet is at the point, so that pellet 1 is picked up at the correct position, eliminating wasteful recognition of defective pellets or areas where no pellets exist, and increasing efficiency. To provide a pellet positioning device for an automatic die bonder which can accurately position pellets, prevent wasteful waiting time on the die bonder body, and improve productivity.

FIG. 4 is a block diagram showing a pellet positioning device according to an embodiment of the present invention.
Reference numerals 12, 14, and 18 are the same as those in the conventional device. Reference numeral 20 denotes an illumination device that illuminates a group of pellets on the synthetic resin sheet 2 obliquely in a band-like manner within a predetermined range. As shown in FIG. 5, this illumination device 20 has a fluorescent lamp 21 placed at the focal point of a reflective concave mirror 22. The light from the fluorescent lamp 21 is reflected by the reflective concave mirror 22 to become a parallel beam of light, which mainly illuminates a band-shaped area.

Returning to FIG. 4, a planar imaging device 23 is disposed obliquely above the synthetic resin sheet 2 on the optical axis of the reflected light. Imaging lens 11 of this imaging device 23
A right angle prism 2 is located between the camera 10 and the TV camera 10.
4, the incident optical path is divided into two, and the light traveling straight through the prism is directed to the TV camera 10.
A predetermined image is formed on the imaging plane. On the other hand, the light branched at right angles by the prism 24 enters the linear imaging device 25. This linear imaging device 25 includes an imaging lens 27 through which the branched light passes, and a large number of images that form an image of a linear area of a wide field of view on the synthetic resin sheet 2 on a photosensitive surface. It consists of a line sensor 26 and an aperture 28 provided in front of an imaging lens 27 in order to make the sensitivity on the photosensitive surface of this image line sensor the same as that of the television camera 10. The magnification of the image formed on the image line sensor 26 is changed by the combination of lens systems.

The illumination device 20 is arranged at an angle where the light specularly reflected on the surface of the pellet 1 enters the imaging lens 11 so that the image line sensor 26 can capture the band-shaped area on the synthetic resin sheet 2, and
In the band-shaped area on the synthetic resin sheet 2, the sheet surface between the pellets 1 is arranged at an angle such that the illumination of the illumination device 20 is blocked. The reflectance due to illumination is high on the surface of the pellet 1 and is low on the defective mark surface and the synthetic resin sheet 2 surface in that order. Therefore, the surface of the pellet 1 is specularly reflected and illuminated at the incident angle, and has a high reflectance, so it is more sensitive to the optical system than the defective mark surface or the synthetic resin sheet 2 surface, which has a low reflectance and has a small amount of reflected light. The image is captured with increased output.

The TV camera 10 images the first group of pellets on the synthetic resin sheet 2 in the area shown in FIG. It is sent to a detection circuit means consisting of a counter 32 and an image memory control circuit 33, and is stored in the image memory 14 in this detection circuit means. At this time, A-
The optimal threshold for digitization by the D converter 12 is set as follows. First, the image memory control circuit 33 controls the image memory 14 and the area counter 32 while the control circuit 31 sets the threshold value of the A-D converter 12 to the lowest value, and the digital planar figure at the lowest threshold value is generated. is read in the image memory 14, and the number of "1"s on the entire screen stored in the image memory 14 is counted by the area counter 3.
2, and the counted number is stored in the arithmetic processing unit, that is, the microcomputer 35. Next, the control circuit 31 converts the A-D converter 12
With the threshold value set slightly higher, the number of "1"s on the entire screen is counted as described above, and the counted number is stored in the microcomputer 35. By repeating this operation, the microcomputer 35 generates a map of the number of "1"s on the entire screen at each level value.
Create with. Here, the number of counts of "1" repeatedly increases and decreases depending on the change of the threshold value, and the threshold value at which the number of counts becomes the largest, that is, the threshold value that becomes the peak when the number of counts increases. It is known that multiple values exist. It is also experimentally known that if you select the two peaks with the highest number of counts from among these multiple peaks and find the average value from the threshold values of these two peaks, the average value becomes the optimal threshold value. It is being Therefore, the microcomputer 35 calculates an average value from the threshold values of the two peaks, and sends this average value as the optimum threshold value to the A-D converter 12 via the control circuit 31. Finally, set the threshold for Monitoring a video signal quantized at this optimum level value results in an image similar to that shown in FIG.

Next, for the digital planar figure read into the image memory 14 at the optimum level value, 1 good pellet is
The horizontal (x) and vertical (y) dimensions of image memory 1
4. Area counter 32 and image memory control circuit 3
The bit length is calculated by the microcomputer 35 using the detection circuit means consisting of 3 and stored.

After setting the level and measuring the size of each pellet using the television camera 10, the image line sensor 26 is used to calculate the position of only the non-defective pellets among all the pellets on the synthetic resin sheet 2. The configuration and operation of this device are as follows.

The sensitivity of the photosensitive surface of the image line sensor 26 is set to be the same as that of the television camera 10, and the optimal threshold value for converting the video signal from the television camera from A to D and the video signal from the image line sensor 26 are - You can use the same level for the D converter. Therefore, using the threshold value obtained as described above, the line video signal from the image line sensor 26 is transferred to the A-D converter 29.
The signal is converted into a digital signal and sent to the non-defective pellet position detection device 30.

As shown in FIG. 7A, the digital signal from the A-D converter 29 detects a line L including one row of pellets in the X direction of the image of the upper surface of the synthetic resin sheet 2 entering the image line sensor 26. When read, it is sent out as an output signal S, as shown in Figure B. Sheet image 2a and defective mark image 1b
The part is "0", and the part of pellet image 1a is "1"
becomes. W is the horizontal length of the pellet image 1a, and P is the extraction point. The detected length of one line L is determined by the combination of the imaging lens 11 and the imaging lens 27. In this embodiment, in order to make the magnification per bit of the television camera 10 and the image line sensor 26 the same, a uniform length of approximately several tens of millimeters is detected. Furthermore, in order to detect all the pellets 1 on the synthetic resin sheet 2 by the image line sensor 26, since the image line sensor 26 is fixed, the XY table 3 is moved by a table drive device using a drive signal from the control circuit 31. 18, every time one line is read in the Y (vertical) direction, it is synchronized and moved by a fixed pitch to scan the entire Y direction, and then the XY table 3 is driven by a drive signal from the control circuit 31 to drive the table drive device 18. Accordingly, it is shifted by one run length in the X-axis direction, and similarly to the above, it is moved by a fixed pitch in the Y direction every time one line is read to scan the entire Y direction and process the entire screen. The fixed pitch feed amount of the XY table 3 is set to, for example, about 50 to 10 μm at one time so that the size of the defective mark image 1b can be detected.

As shown in the functional block diagram of FIG. 6, the good pellet position detection device 30 consists of a counter 37 (electrical hardware) with a shift register 36 for calculating the good pellet position. A-D converter 2
The digital signal sent out from 9 is first held in the line shift register 36 including one row of pellets in the X direction. In this embodiment, since a 2048-bit image line sensor 26 is used, the capacity of the shift register 36 is also 1 bit×2048. Next, one line of information held in this shift register is sent bit by bit to the control circuit 3.
It is sent to the transfer counter 37 by the command signal D of 1.
The counter 37 is preset with the horizontal (X direction) length of the pellet image 1a obtained using the television camera 10 in bit length from the control circuit 31 via the data bus E. Next, the signal transferred from the shift register 36 is converted to 1 by the counter 37.
Count each pellet, and each time when this counted number matches the preset number above,
A coincidence display signal F is sent to the control circuit 31. When this signal F is input to the arithmetic processing unit, that is, the microcomputer 35 via the control circuit 31, the microcomputer 35 determines the number of lines (y) at which the match occurs, and the number of lines (y) at which the match occurs. The location (x) on one line of the end of the pellet 1 (P in FIG. 7) at that time is stored in the memory of the microcomputer 35. This operation is performed over the entire surface of the synthetic resin sheet 2 by moving the XY table 3, and a map of coordinates where the length of the good pellet 1 in the X direction exists is created at high speed.

After moving the XY table 3 for several lines (at least the vertical length of the pellet 1), in the map created by the microcomputer 35,
Same coordinates in the X direction and length 1 in the Y direction
The part where the coincidence display signal F corresponding to the pellet is stored is detected, and the pellet 1 corresponding to this part is stored.
is considered to be a good product. Then, the center value of this good pellet is determined using screen coordinates (line and bit). The center value at this time is obtained by adding half the length of pellet 1 in the X direction to the X coordinate of point P in the horizontal direction (X), and by adding half the length of pellet 1 in the X direction to the It is obtained by adding half the length of the pellet 1 in the Y direction to the coordinates. The screen coordinates thus obtained are finally converted into XY table coordinates and registered in the non-defective product map memory 34. Such operations are performed for all lines by the microcomputer 35, and the table coordinate map of all good pellets is completed.

Next, when die bonding actually begins, the microcomputer 35 sequentially reads out the good product map memory 34, operates the table driving device 18 according to instructions from the control circuit 31, and moves the center point of the good pellet 1 on the XY table 3. By aligning with the suction point of the collet 19, the good pellets 1 are suctioned, transferred, and die-bonded.

Furthermore, just before the positioning of the XY table 3 is completed and the pellet 1 is attracted by the collect 19, the video signal from the television camera 10 is read into the image memory 14, and the area is counted for a flat portion of the size of the pellet 1. As a result, the presence or absence of the pellet 1 can be confirmed more precisely, and malfunctions can be prevented.

In the above embodiment, in order to detect the position of the good pellet 1, hardware processing is performed using the shift register 36 to extract the horizontal length point, a search is performed over the vertical length, and the calculation of the center point is performed by software. Although this calculation is performed using hardware, it is also possible to use dedicated hardware to calculate the center point, resulting in faster processing.

As described above, according to the present invention, before the start of die bonding, the center position of only good pellets among all the pellets on the synthetic resin sheet is searched and calculated at high speed, and the data is converted into coordinate values of an XY table. A good product map has been created, this good product map is read out, and the XY table is sequentially moved so that the center of the good pellet is aligned with the collection point, and each good pellet is adsorbed to the correct position on the collection. This has the effect of shortening the alignment time, almost eliminating the waiting time of the automatic die bonder body, increasing the speed, improving productivity, and reducing manufacturing costs.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing the pellet positioning device of a conventional automatic die bonder, Fig. 2 is a plan view showing the state of the separated pellets on the synthetic resin sheet of Fig. 1, and Fig. 3 is - An image diagram of the signal from the D converter received on a monitor television; Figure 4 is a configuration diagram showing a pellet positioning device for an automatic Daihonda according to an embodiment of the present invention; and Figure 5 is a perspective view of the illumination device shown in Figure 4. , Fig. 6 is a block diagram showing the function of the good pellet position detection device shown in Fig. 4, Fig. 7A
is an image formation diagram of a band-shaped range entering the image line sensor section in Figure 4, and Figure 7B is a digital signal output from the second AD converter in Figure 4 after detecting the image in Figure A on line L. It is a diagram. 1... Semiconductor pellet, 2... Synthetic resin sheet, 3... XY table, 10... Television camera, 11... Imaging lens, 12... First A-D
Converter, 14...Graphic memory, 18...Table drive device, 19...Collection, 20...Illumination device, 23...Planar imaging device, 25...Linear imaging device, 26...Image line sensor, 27...
Imaging lens, 29... Second A-D converter, 30
... Good pellet position detection device, 31 ... Control circuit, 32 ... Area counter, 33 ... Image memory control circuit, 34 ... Good product map memory, 36 ...
Shift register, 37...counter, and the same reference numerals in the drawings indicate the same or corresponding parts.

Claims (1)

[Claims] 1. A synthetic resin sheet on which a plurality of mutually separated semiconductor pellets are attached in the X and Y directions, an XY table on which this synthetic resin sheet is placed,
There is little reflected light on the table driving device that moves this XY table in the X-axis and Y-axis directions, on the defect mark part attached to the surface of the semiconductor pellet, and on the synthetic resin sheet between the semiconductor pellets, An illumination device is arranged to illuminate the surface of the semiconductor pellet at an angle that increases reflected light, and the reflected light from the semiconductor pellet caused by the illumination of this illumination device is imaged through a first lens system and converted into a video signal. a linear imaging device that splits the incident light of the imaging device in the middle, forms an image in a linear range via a second lens system, converts it into a video signal, and outputs the video signal; a first A-D converter that converts a video signal from the planar imaging device into a digital signal and outputs the digital signal;
detection circuit means for processing digital signals from the A-D converter of the semiconductor pellet and calculating the dimensions of the semiconductor pellet in the X direction and the Y direction using bit information; A video signal of a line including one row is sent from the linear imaging device, and a second A-D converter converts it into a digital signal and outputs it. Each line is sequentially held in a shift register, the bit information in the X direction of the semiconductor pellet calculated by the detection circuit means is preset in a counter, and the digital signal held in the shift register is counted by the counter. A good pellet position detection device that outputs a match display signal when the count matches the preset bit information, and a position on the above line in the X direction when the match display signal is output from this good pellet position detection device. The position is set as the X coordinate, the number of lines in the Y direction of the above line is stored as the Y coordinate, and when the area of the semiconductor pellet on the synthetic resin sheet is finished, the stored coincidence display signal is output. Among the coordinates, the X direction is the same coordinate, and the Y
If the coordinate values are continuous and correspond to the length of the semiconductor pellet in the Y direction, it is considered to be a good pellet, and the X and Y coordinate values at this time are set as the center position of the good semiconductor pellet. an arithmetic processing device for converting into table coordinates; a good product map memory for storing the table coordinates converted by this arithmetic processing device; In addition to transmitting signals,
It is equipped with a control circuit that outputs a drive signal to the table drive device, and before the start of die bonding,
All the pellets on the synthetic resin sheet are searched and the center position of only good pellets is stored in the good map memory.The information in the good map memory is read out sequentially, and the table driving device is driven by a drive signal from the control circuit. The automatic apparatus is characterized in that the above-mentioned XY table is moved to align the center of the good pellet with the collection suction position of the die bonder so that the pellet is adsorbed to the collet and die-bonded, and the positioning and adsorption of the good pellets is repeated one after another. Die bonder pellet positioning device.