JPH04306848A - Die bonding device - Google Patents

Abstract

PURPOSE:To enable a die-bonding process to be carried out in a short time so as to produce a required number of semiconductor elements separated according to characteristic ranks excepting defectives due to imperfect bonding. CONSTITUTION:The characteristic data and the position data of all chips of a wafer outputted from a wafer test device are compressed by a few types of data compression methods by a data processor, and the shortest one out of the compressed data are inputted into a die-bonding mechanism as test data. By this setup, a large amount of data outputted from the wafer tester can be transmitted in a short time. A control section outputs the position data of chips provided with specified characteristics data to a pickup section decoding the test data, and a die bonding operation is carried out separately according to characteristic rank. The control section discriminates defective bonding from non-defective bonding basing on the image data of the bonded chip, the number of chips judged non-defective is counted, and a bonding process made to continue until the count number reaches to the previously set required number.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a die bonding apparatus used for die bonding a wafer chip to a frame during the manufacturing process of semiconductor devices.

0002

2. Description of the Related Art A conventional die bonding apparatus is configured to sequentially pick up only those chips determined to be non-defective in the previous chip test process out of the diced chips using a suction collet and die bond them to a lead frame. ing. In the chip test process, chips are classified into non-defective and defective chips, and as shown in FIG. 6, a defective mark M is marked with red ink or the like on the defective products. The die bonding apparatus recognizes this defective mark M by image recognition, picks up only the chips on which the defective mark M is not marked with a suction collet, and die-bonds them to a lead frame. By the way, there are several ranks among chips that have been determined to be good, depending on their characteristics, but just because a completed semiconductor device is known to be good does not indicate which characteristic rank it belongs to among the good. I don't know if it is. Therefore, completed semiconductor devices are classified according to their characteristics in a final test. Therefore, it has been difficult to formulate a production plan that corresponds to the characteristics of semiconductor elements and the number of shipments required by users. In addition, in order to determine whether the chips are good or bad, it is necessary to use image recognition to check whether there is a defective mark M on all chips, so the trajectory L of the suction collet relative to the wafer becomes long, as shown in Figure 6. Therefore, the die bonding process took a relatively long time.

[0003] Therefore, the applicant of the present invention has developed a system that uses data compression to read a huge amount of data including wafer identifier data, characteristic data and position data of all the chips on the wafer from a wafer test equipment, and also The present application is for a die bonding apparatus that can decode the position data of each chip having specified characteristic data from the compressed data and select and die bond only the chips having specific characteristic data according to this position data. (Patent Application No. 3-23219)

0004

[Problems to be Solved by the Invention] This die bonding apparatus can produce semiconductor elements according to their characteristic ranks, and has the remarkable effect of shortening the time required for the die bonding process. There are still problems that need to be resolved. That is, since the data compression means is fixed to a certain specific method, the amount of compressed data may not necessarily be the minimum depending on the content of the data to be compressed, and a loss may occur in data transmission time, resulting in poor die bonding efficiency. There is a fear that this will happen.

[0005] Furthermore, since the number of chips that have been die-bonded can only be known from the monitor screen that displays the counting data, the operator can check whether the necessary number of chips have been bonded while looking at the monitor screen, etc. You have to decide whether. Moreover, this counting data includes
Since defective products as a result of the bonding process judgment are not subtracted, it is necessary to always perform the bonding process on more chips.As a result, there are always surplus bonded chips, which increases the complexity of inventory management, etc. Additional work will be required.

Therefore, the present invention makes it possible to accurately and automatically produce only the required number of semiconductor devices for each characteristic rank, and to shorten the time required to transmit the characteristic data and position data of each chip as much as possible. The technical problem is to provide a die bonding device that can perform the following steps.

[0007]

[Means for Solving the Problems] In the present invention, as a technical means for achieving the above-mentioned problems, a die bonding apparatus is constructed as follows. That is, there is a wafer test device that outputs characteristic data and position data of all chips on a wafer, and the data inputted from this wafer test device is compressed using multiple types of data compression methods, and the data size therein is It consists of a data processing section that outputs the smallest one as test data, and a die bonding mechanism section that performs die bonding processing based on the test data. The position data is output while decoding the test data, and the number of chips determined to be good by determining whether the bonding process is good or bad is determined based on the image data of each chip after bonding, and this count number is set in advance. a control unit that controls the die bonding process to continue until the required number of processes is reached, and a pickup unit that picks up chips according to position data output from the control unit and outputs the image data by an image recognition unit. It is constructed with the following features.

[0008]

[Operation] When a wafer that has been tested in a wafer test device is set in a die bonding device, the data processing unit compresses the characteristic data and position data of the set wafer from the wafer test device using multiple methods. At the same time, the data with the smallest data size among them is inputted as test data to the control section of the die bonding mechanism section at high speed. In this way, the huge amount of data from the wafer test equipment is compressed using the optimal data compression method and input to the bonding mechanism, so conditions are prepared for high-speed die bonding, and the bonding process can be completed in a relatively short time. be able to.

Next, when the characteristic rank of the chip to be die-bonded is specified, the control section transfers the position data of each chip belonging to the specified characteristic rank to the pickup section while decoding the compressed test data. output sequentially. Thereby, the pickup section performs the die bonding process only on chips belonging to the specified characteristic rank, and after each die bonding process, the image recognition section outputs image data of the external appearance of the chip to the control section. The control unit determines whether the bonding process is good or bad from each image data, counts each time the bonding process is judged to be good, and continues the bonding process until this counted number reaches a preset required number of processes. Control. Therefore, it is possible to accurately produce only the required number of semiconductor elements for each characteristic rank, and also to automatically produce them without requiring supervision by an operator.

0010

DESCRIPTION OF THE PREFERRED EMBODIMENTS A preferred embodiment of the present invention will be described in detail below with reference to the drawings. "FIG. 1" is a block configuration diagram showing the electrical configuration of one embodiment of the present invention. To explain the outline of this die bonding apparatus, the wafer test apparatus 2 outputs the identifier data 23 related to the wafer and all of the data on this wafer. Chip characteristic data 24 and position data 25
is output, this data is compressed by the data processing unit 4, and this compressed test data 41 is input to the bonding mechanism unit 1, and only the chips having the specified characteristic data 24 are die bonded, and image recognition is performed. The control unit 11 determines whether the bonding process is good or bad based on the image data 144 from the unit 143, and continues the die bonding process until the number of chips for which the bonding process has been determined to be good reaches the set required number.

Further, as shown in FIG. 5, a wafer sheet 32 slightly larger than the wafer 3 is placed on the back surface of the wafer 3 in order to prevent the diced chips 31 from scattering.
is pasted. At the edge of this wafer sheet 32,
A wafer identifier 33 such as a bar code is marked on the wafer. The wafer test apparatus 2 reads the wafer identifier 33 of the wafer 3 to be tested by the wafer identification section 21, and then checks the electrical characteristics of all the chips 31 of this wafer 3 using the test section 22.
The characteristic data 24 and position data 25 of the measurement results and the identifier data 2 from the wafer identification section 21 are
3 is input to the data processing section 4 directly online or via a floppy disk.

The data processing section 4 converts the characteristic data 24 into [
The characteristic data 24 is classified into four types of characteristic ranks: excellent, good, fair, and poor, and the characteristic data 24 is combined with the position data 25 and identifier data 23 of each chip 31, and then the data shown in FIG. After compressing data using four different methods, which will be described later with reference to FIG.

This compressed test data 26 is directly transmitted online to the control section 11 of the die bonding mechanism section 1.
will be transmitted in a short time. In addition, in "Figure 4", chips 31 of [Excellent] rank are dotted, and chips 3 of [Good] rank are dotted.
1 is indicated by a diagonal line, a chip 31 of rank [Acceptable] is indicated by two parallel lines, and a chip 31 of rank [Not acceptable] is indicated by an x mark.

Next, the four types of data compression methods mentioned above will be explained with reference to FIG. 3. (a) in the figure shows data before compression, and this data is expressed by the following equation. That is, data (X, Y) = {(X1 , Y1 ), (X1 ,
Y2 ), (X1 , Y3 ), ..., (Xn, Yn),
...}. “Figure 3” (b) shows the data in (a) as the first
The compressed data compressed using the method is shown, and len in the figure is
This is a number indicating how many pieces of Y data are consecutive from Ys, and this compressed data is based on the X coordinate of consecutive chips 31, the number of consecutive chips (len), and the start value of the consecutive Y coordinate (Ys ) is one block of data. When this compressed data is decoded, it becomes as follows. That is, data (X, Y) = {(X1 , Ys1 ), (X1
, Ys1 +1), (X1 , Ys1 +2), ...
...(X1, Ys1 +len-1), (X2, Ys
2),...}.

"Figure 3" (c) shows the data in (a) as a second
The compressed data compressed using the method is shown, and len in the figure is
This is a number that indicates how many pieces of Y data exist in the same X, and the X position coordinate of one row of X coordinates, the number of chips in one row (len), and the Y coordinate of each chip are considered one block of data. . After decoding this compressed data,
It becomes as follows. That is, data (X, Y) = {(X1 , Y1 ), (X1 ,
Y2 ), (X1 , Y3 ), ......(X1 , Yl
en1 ), (X2 , Y4 ), (X2 , Y5 )
...}.

"Figure 3" (d) shows the data in (a) as a third
This shows the compressed data compressed using the method shown in Figure 1. Ys in the figure is the start value of continuous Y data, Ye is the end value of continuous Y data, and the chips are continuous with the X coordinate where the chips are continuous. The start Y coordinate (Ys) and end Y coordinate of the chip are used as one block of data. When this compressed data is decoded, it becomes as follows. That is, data (X, Y) = {(X1 , Ys1 ), (X1 , Ys
1 +1),..., (X1 , Ye1 -1), (X1
, Ye1 ), (X2 , Ys2 ), (X2 , Ye
2),...(Xn, Ysn), (Xn, Yen),...}
becomes.

"Figure 3" (e) shows the data in (a) as the fourth
The compressed data compressed using the method is shown, and len in the figure is
It shows the number of consecutive data blocks in the same X x 2. That is, it indicates the number of bytes of Ys and Ye that follow on the compressed data. This compressed data is a number (le
n), the start Y coordinate (Ys), and the end Y coordinate (Ye) of a continuous chip unit are set as one block of data. When this compressed data is decoded, it becomes as follows. That is, data (X, Y) = {(X1, Ys1), (X1
, Ys1 +1), ..., (X1 , Ye1 -1), (
X1 , Ye1 ), (X1 , Ys2 ), (X1
, Ys2 +1), ... (X1 , Ye2 -1), (X
1 , Ye2 )...(X1 , Yslen), (X1
,Yslen+1),...,(X1,Yelen-1)
, (X1, Yelen)...}. Data processing section 4
outputs data with the smallest data size among the four types of compressed data described above as test data 41.

Next, regarding the operation of the above embodiment, see "Fig. 2".
This will be explained with reference to the flowchart, and the configuration of the bonding mechanism section 1 will also be explained accordingly. When the wafer 3 that has been tested by the wafer test apparatus 2 is set in the die bonding mechanism section 1, the wafer identification section 12 of the die bonding mechanism section 1 reads the wafer identifier 33 (step S1). The identifier data 121 read here is compared with the identifier data 23 included in the test data 41 from the wafer test section 2 to confirm that the wafer 3 matches. Then, the characteristic data 24 and position data 25 of the set wafer 3 are compressed using the data compression methods shown in FIG. 3 (b) to (e), and the data with the smallest data size is The selected test data 41 is written into the control unit 11 (step S2).

Here, the chip 3 to be die bonded
1, for example, [Excellent], and the required number of processes are instructed using the operation unit 13 and input into the control unit 11 (step S
3), and the control unit 12 stores the position data 112 of all the chips 31 belonging to the characteristic rank of [Excellent] (the position data 25 described above is within the wafer test apparatus 2, and the position data 112 is within the die bonding mechanism section). 1 (with a different sign from 12) from the compressed test data 41 (step S4), decodes it as described above (step S5), and picks up the required position data 112. 14 XY tables 142 (step S6).

The pickup section 14 includes an X-Y table 142 on which the wafer 3 is placed, and an X-Y table 142 on which the wafer 3 is placed.
2, a drive unit 141 for the suction collet, and an image recognition unit 143 that captures appearance information during and after wafer chip bonding and outputs it as image data 144. are doing. The above-mentioned position data 112 is stored in the X-Y table 142.
A pick-up signal 111 for picking up the chip 31 with the [excellent] characteristic rank is input to the suction collet drive section 141. As a result, the X-Y table 142 is moved so that the suction collet is located directly above the chip 31 to be die-bonded first (step S7), and the image data 144 is transferred from the image recognition section 143 to the control section 11. input (step S8)
, based on this image data 144, the control unit 11
After precisely positioning the table 142 (step S9), the first die bonding process is performed (step S10).

Next, the control unit 11 takes in image data 144 of the external appearance of the chip that has undergone the above-mentioned die bonding process from the image recognition unit 143 (step S11), and determines whether the bonding process is good or bad (step S12).
)do. If the bonding is defective, the process jumps to step S5, the compressed data is decoded, the position data of the next chip with the characteristic rank of [excellent] is output, and the bonding process is performed in the same processing operation as described above. On the other hand, if the bonding process is judged to be good, the production quantity counter is set to "1".
After counting up (step S13), it is determined whether this counted number has reached the required processing number set in step S3 (step S14), and the number of bonded chips for which the bonding process has been determined to be good is determined. The process jumps to step S5 and repeats the process up to step S14 until the number of processes reaches the required number of processes.

[0022] As a result, the chips belonging to the [excellent] characteristic rank are sequentially die-bonded to the lead frame (not shown) as many times as necessary. At this time, the pickup unit 13 has confirmed all the positions of the chips 31 belonging to the characteristic rank of [Excellent] based on the position data 112, and as shown by the trajectory l in FIG. It is possible to die-bond chips 31 with a characteristic rank of [Excellent] with minimal movement, and since the required number of production is known from the required processing number data, it is possible to die-bond only the desired number. can.

After die bonding, the semiconductor device is completed through normal steps. Since the characteristic data 24 is carried over to the steps after the die bonding step, it is possible to classify each completed semiconductor device by characteristic rank. Therefore, only the necessary number of semiconductor devices can be produced and shipped for each characteristic rank of "excellent", "good" and "fair", and inventory management etc. are extremely easy.

It should be noted that the present invention is not limited to the above-mentioned embodiments, but may include various modifications without departing from the scope of the claims. For example, in the embodiment described above, the required production quantity data is inputted as an instruction from the operation section 13 to the control section 11, but it may also be set and inputted to the data processing section 4. Furthermore, in the embodiment described above, the data processing section 4 classifies the characteristic data of each chip into characteristic ranks, but the control section 11 may also perform this.

[0025]

As described above, according to the die bonding apparatus of the present invention, a huge amount of characteristic data and position data of all chips included in a set wafer can be read at high speed by optimal data compression. It is possible to prepare the wafer for bonding, and the bonding process can be performed in a relatively short time even with the huge amount of data output from the wafer test equipment.

In addition, it is possible to selectively die-bond only chips belonging to a specified characteristic rank to produce each characteristic rank, and to process only the necessary number of chips while counting only the chips that have been determined to be good in the bonding process.
It is possible to grasp the production status of the die bonding process in real time without the need for operator monitoring, and it is possible to formulate a production plan that meets the user's requirements while minimizing the amount of inventory. Therefore, production adjustment and inventory adjustment are extremely easy, and delivery times can be shortened.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of a die bonding apparatus of the present invention.

FIG. 2 is a flowchart of processing operations of the device.

FIG. 3 is an explanatory diagram of a data compression method when reading data in the same device.

FIG. 4 is an explanatory diagram showing a trajectory of picking up a chip by the same device.

FIG. 5 is an explanatory diagram of a wafer set in the same apparatus.

FIG. 6 is an explanatory diagram showing the trajectory of a suction collet with respect to a wafer in a conventional apparatus.

[Explanation of symbols]

1 Die bonding mechanism section 2 Wafer test device 3 Wafer 4 Data processing section 11 Control section 14 Pick-up section 24 Characteristic data in the wafer test device 25 Position data in the wafer test device 31 Chip 41 Test data 112 Position data in the die bonding mechanism section 14
3 Image recognition unit 144 Image data

Claims (1)

[Claims] 1. A wafer test device that outputs characteristic data and position data of all chips on a wafer, and compresses each of the data inputted from the wafer test device using multiple types of data compression methods. It consists of a data processing section that outputs the smallest data size as test data, and a die bonding mechanism section that performs die bonding processing based on the test data. While decoding the test data, it outputs the position data of each of the above, and also determines whether the bonding process is good or bad based on the image data of each chip after bonding, and counts the number of chips that are judged to be good. A control unit that controls the die bonding process to continue until a preset required number of processes is reached, and a pickup that picks up chips according to position data output from the control unit and outputs the image data using an image recognition unit. A die bonding device comprising: