JPS634348B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for determining and recognizing the quality of a plurality of pellets formed on a semiconductor wafer, and is used, for example, when selecting only good pellets in a pellet bonding process. The present invention relates to a method for recognizing pellets.

Multiple pellets formed on a semiconductor wafer are passed through a wafer test to determine whether each pellet is good or bad, and then divided into individual pellets, and only the good pellets are bonded to lead frames, etc. There is. In other words, the conventional general method is to test the wafer in a testing device (such as a wafer prober) to determine the quality of each pellet.
Defective products are marked with a defective mark using ink approximately in the center of the pellet, and in the subsequent pellet bonding process, this defective mark is detected and only pellets without a defective mark are selected for bonding. -ing

For this reason, in the pellet bonding process, it is necessary to recognize whether the pellets are good or bad, and conventionally, a method of detecting defective marks using phototransistors has been adopted. This method detects defective marks on each pellet by scanning each pellet one by one, so the recognition results are highly reliable. However, since all pellets formed on the wafer are inspected, There is a problem that the operation is wasteful and the recognition time is long.

In other words, since a plurality of pellets are normally formed in a grid pattern on a circular wafer, one of the reasons why the pellets are considered defective is that the pellets formed at the periphery of the wafer have arcuate outer edges. However, there are some pellets that cannot obtain a predetermined outer shape and are considered defective. Therefore, defective products due to such causes can be judged and recognized simply by detecting whether the outer shape of the pellet is a predetermined shape (square). This is because it is not necessary to detect defective marks.

Therefore, for the purpose of shortening the recognition time, an attempt was made to first detect whether or not the pellet has a predetermined external shape, thereby omitting the trouble of detecting defective marks on defective pellets due to this. is being done. In this method, a wafer pellet is projected onto a TV monitor, and the entire projected pellet is scanned with a binary signal (for example, the presence of a pellet is set as "1" and the absence of a pellet is set as "0"). If the scanning result is above a predetermined value, the outer shape of the pellet is considered to be satisfactory, and the process depends on the detection of defective marks that occur almost simultaneously, and if the scanning result is below a predetermined value, the outer shape of the pellet is considered to be defective. The product is recognized as defective without waiting for the detection of a defective mark.

According to this, it is possible to shorten the time since it is not necessary to detect defective marks on all pellets, but
However, the fact remains that all pellets must be detected, and the reality is that sufficient effects are not achieved.

Therefore, an object of the present invention is to provide a pellet recognition method that can shorten the pellet recognition time for the entire wafer by omitting the detection of pellets that are naturally presumed to be defective due to defects in the pellet external shape. It's about doing.

In order to achieve this goal, the method of the present invention is such that when recognizing the quality of a plurality of semiconductor pellets formed on a semiconductor wafer, the formation state of the pellets on the wafer is memorized in advance, and the pellets are stored in a predetermined manner. Based on the memorized state of pellet formation on the wafer, it is determined where on the wafer the pellet being recognized is formed, and then detected To recognize whether a pellet is good or bad by estimating the shape of a pellet to be pelletized and detecting the shape of other pellets and the presence or absence of a defective mark without detecting pellets whose shape is presumed to be defective. It is characterized by The method of the present invention will be explained in detail based on the drawings.

FIG. 1 is a plan view of a circular semiconductor wafer 1. Generally, a plurality of rectangular pellets 2 are formed on this wafer 1 in a grid pattern as shown. The arrangement of the pellets 2 is substantially the same when they are formed using wafers and masks of the same shape. Usually, pellets formed around the wafer are formed with an incomplete outer shape in which at least one corner is missing due to the circumferential shape of the wafer. This wafer 1
After the pellets 2 are formed, they are subjected to a wafer testing device, and the defective pellets are marked with a defective mark approximately in the center, and are then diced along the outer contour of the pellets to make them ready for division. This wafer is then set in a pellet bonding device, and only good pellets are separated and taken out and bonded to a lead frame or the like.

FIG. 2 shows a device for recognizing the quality of each pellet of a wafer 1 set in a pellet bonding device, and a TV camera 3 for photographing the wafer 1.
, a TV monitor 4 for enlarging and projecting the image, and a control circuit 5. Furthermore, the wafer 1 is, for example,
It is set on the XY table 6, and the plane position is
You can set the direction to move. Further, 7 is a pickup for transferring good pellets to the bonding device 8, and 9 is a lead frame.

Conventionally, in such an apparatus, all the pellets 2 are scanned one by one from one end of the wafer 1 to the other, as shown by the broken line in FIG. It is performing detection.

On the other hand, in the method of the present invention, as shown in FIG. 3 showing the image on the TV monitor 4, each projected pellet 2 is moved one by one to a predetermined position on the screen by the action of the XY table 6 and the control circuit 7. These pellets are projected at a predetermined size in the four corners surrounded by parallel lines l ₁ and l ₂ , l ₃ and l ₄ , and perpendicular lines l ₅ and l ₆ , and l ₇ and l ₈ . S ₁ , S ₂ , S ₃ , S ₄
and a central portion _S5 located between these four corners, and each is individually scanned and detected using a binary signal. The binary signal is, for example, a "1" signal indicating the presence of a pellet, and a "0" signal indicating its absence, and this signal is sent to each of the four corners _S1 ,
By scanning and detecting in S ₂ , S ₃ , and S ₄ , it is possible to determine whether or not a corresponding corner of the pellet at each four corner exists by comparing the output signal with a reference value at each four corner. becomes clear. For example, in the example of FIG. 4, if there is _no pellet at corner S ₃ but there are pellets at other corners S ₁ , S ₂ , and S ₄ , this pellet is It turns out that it is missing.

Therefore, in this way, the four corners S ₁ , S ₂ , S ₃ , S ₄
By dividing and detecting the pellet, if the pellet has a defective external shape, it becomes possible to determine what shape the pellet has. Furthermore, since it is possible to determine the external shape of the pellet, it is also possible to determine where on the wafer the current pellet is formed. As a result, if the formation state of pellets on wafers is stored in advance in the control circuit 5, the shape of the next pellet to be detected can be easily estimated if the wafers are from the same lot. If it is estimated that the next pellet has an even more defective appearance, the control circuit fast-forwards the operation of the XY table 6, or moves the solid lines A ₁ , A ₂ , A ₃ in FIG.
As shown in the answer, the pellets 2a to 2e with poor external shapes are jumped over and sent to the next pellet with a good external shape. Therefore, pellets that are presumed to have a defective outer shape are not detected.
Accordingly, the detection time, that is, the recognition time for the entire wafer can be shortened.

On the other hand, the defective mark 1 is detected almost simultaneously with the detection of the outer shape.
0 is detected (recognized), and since defective marks can be detected by simply scanning the inside of the central part _S5 , the time required for detection is reduced compared to scanning the entire surface of the pellet. It becomes possible.

The electrical operation of dividing the pellet displayed on the TV monitor into the four corners and the center and scanning them as described above can be easily performed by an electrical engineer, so a detailed explanation will be given below. is omitted here.

As explained above, in the pellet recognition method of the present invention, when it is estimated that the pellet to be recognized has a defective shape, the pellet is jumped over and other pellets are recognized. The recognition operation time can be shortened compared to the case where recognition is performed on pellets.

[Brief explanation of the drawing]

Fig. 1 is a plan view of the wafer, Fig. 2 is a schematic diagram of the recognition device, Fig. 3 is a schematic diagram of the TV monitor surface, and Fig. 4 is an example of a pellet with a defective external shape.
FIG. 1...Wafer, 2...Pellet, 2a to 2e
...Bad pellets, 3...TV camera, 4...
TV monitor, 5... Control circuit, 6... XY table, 10... Defective mark, l ₁ to l ₈ ... Line,
_S1 to _S4 ...Four corners, _S5 ...Center.

Claims (1)

[Claims] 1. In recognizing the quality of a plurality of semiconductor pellets formed on a semiconductor wafer, the formation state of the pellets on the wafer is memorized in advance, and it is detected whether the pellets have a predetermined shape or not, and the Based on the formation state of the pellet, it is determined where on the wafer the pellet being recognized is formed, and the shape of the next pellet to be detected is estimated, and the shape is assumed to be defective. 1. A method for recognizing pellets, characterized in that the shape of other pellets and the presence or absence of defective marks are detected without detecting the pellets being pelleted.