JPS6131401B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a position detection device, and particularly to a device for bonding a lead wire to an electrode of a pellet die-bonded onto a lead frame.
This invention relates to a device for automatically detecting the position of electrodes on pellets.

The wire bonding process for electrodes such as transistor pellets is performed by die bonding pellets 2 onto lead frame 1, as shown in Figure 1.
Emitter electrode E and base electrode B on pellet
and lead frames 3 and 4 are connected by gold wires 5, respectively. In this step, when die bonding the pellet onto the lead frame, an error occurs in the bonding position, so it is necessary to detect the electrode positions E and B when wire bonding is performed.

Conventionally, this detection means relied on visual inspection by the operator. However, since the pellet size is very small, the work must be done under a microscope, which causes worker fatigue and requires improvement.

SUMMARY OF THE INVENTION In order to automate this work, it is an object of the present invention to provide a device that detects the electrode position in a non-contact manner and notifies the bonding machine of the electrode position coordinate values.

That is, according to the present invention, a binary measurement pattern obtained by photoelectrically scanning an area covering the positional deviation of the pellet is converted into a circular shape having a diameter three to four times the diameter of the wire used for bonding (several tens of microns). A position detecting device is obtained which is characterized in that it scans with a mask, stores the coordinates when matching is achieved, and calculates the pellet position based on these coordinates.

The present invention will be described in detail below with reference to the drawings.

FIG. 3 is a block diagram of a pellet position detection device according to the present invention, in which a pellet 2 bonded on a lead frame 1 is enlarged to an appropriate size by an optical system, an image is formed on a photoelectric scanner 6, and an electric signal is generated. Convert to The scanning area at this time is the part indicated by the dashed-dotted line 7 in FIG. 1, and when enlarged, it becomes as shown in FIG. 2. Scanning is done in left-to-right, top-to-bottom order. The electrical signal obtained by the photoelectric scanner is an analog signal, which is sliced at an appropriate level by the binarization circuit 8 and converted into a digital signal of 1 and 0. (The figure to be detected in the following explanation is 1,
It is assumed that other figures are sliced on the 0 side. ) The binarized figure is as shown in FIG. 4 as an example, and the figures of electrodes E and B appear as 12 and 13. This figure is scanned with a circular mask 14 that is slightly smaller than the inscribed circle of the electrode part figure, and when matching with the electrode part figures 12 and 13 is achieved, the center coordinates of the circular mask are extracted, the coordinate values are memorized, and the bonding position is determined. calculate. The circular mask matching circuit 9 and coordinate storage calculation circuit 10 shown in FIG. 3 are circuits that perform this function. The bonding tool control circuit 11 drives the bonding tool mechanism 15 based on the calculated electrode coordinate values to perform bonding.

Figure 5 shows an example of the configuration of a circular mask matching circuit.
It consists of one shift register that outputs to the right end of -11. The first grid corresponds to one picture element.
The data output to the right end of SR1-1 is input to the left end of SR1-2, the right end output of SR1-2 is input to the left end of SR1-3, and so on. Also, one row of shift registers corresponds to one horizontal pixel row in Figure 4, and as a two-dimensional pixel arrangement, it is completely the fourth row.
The shape will be the same as the figure. 2 in this shift register
The graphical information output from the value converting circuit is input sequentially for each picture element, and circular mask matching is performed. In other words, when the picture element of interest in FIG. 5 is a,
It is sufficient to extract signals from all picture elements (shaded areas in the figure) contained in a circle centered on a, and extract the coordinates of picture element a when the logical product is established in the AND circuit 50.

FIG. 6 is another example of a binarized figure, which includes a noise figure 16, a small missing figure 17, etc. in addition to the figures 12 and 13 of the electrode section. In such a case, first, a small noise removal circuit for correcting the minute missing figure is inserted before the circular mask matching circuit, and the minute missing figure 17 is corrected.

Figure 8 shows an example of the configuration of a small noise removal circuit.
The configuration of the shift register is connected in the same manner as in FIG. 5, except that it has five rows. When the picture element of interest is a, it is determined whether this picture element is 0 or 1 by looking at the picture elements around a and using the logical formula F=(a+
b+c+d+e)・(d+g+h+a+f)・(b+
h+i+j+a)・(e+a+i+k+e)・(c+
f+a++m), if F=0, a=0, if F=1, a=1, and output to the next circular mask matching circuit.

FIG. 9 shows an example of an improved configuration of the present invention which is effective when a lot of noise occurs, and the figure processed by the small noise removal circuit 18 is processed by the circular mask matching circuit 19, which When scanning is performed using a small mask and the central picture element of the circular mask is left unmatched, the figure shown by the black dot (-) in FIG. 6 is detected. When this figure is cut by the large noise removal and coordinate extraction circuit 20 using the method described later, figures in which a certain number of black points (about the same size as the electrode figure) or more are lined up consecutively will be cut, and the figure at the right end of Figure 6 will be cut. Furthermore, the center coordinates (P1, P2, P3 in FIG. 7) of the remaining black dot figures are determined.
The black dot group G1 in FIG. 7 is a noise figure, G2,
G3 provides information on the electrode shape, and the coordinate extraction circuit extracts the leftmost coordinate, rightmost coordinate, topmost coordinate, and bottommost coordinate for each remaining group of G1, G2, and G3. extracts the coordinate information of and stores it in coordinate storage circuit 2.
1 memory. The calculation identification circuit 22 calculates the center coordinates P1, P2, P3 of each sunspot group from the stored left, right, top, and bottom coordinates, and also calculates the lengths of the line segments between each center coordinate, 1-2, 2-3, 1-.
3 and calculate the slope. To verify which of these three center coordinates is the electrode coordinate, it is calculated by comparing it with the reference value that determines the range that can occur in the case of an actual electrode shape, which is set in the standard figure constant setting section 23 in FIG. It is determined by the identification circuit 22. The above explanation is for the case where two electrode figures and one noise figure are processed. However, if there are no noise figures or if there are multiple noise figures, the combination of all figure center coordinates P1, P2, and PN will be used to The length and slope of the line segment are calculated and the judgment is made. As a result of the judgment, if one pair of electrodes is detected, bonding is performed, but otherwise it is judged as an erroneous judgment, the threshold level of the binarization circuit is changed, the pellet shape is rescanned, and the same process as above is performed. repeat. If the detection is still unsuccessful even after rescanning several times, it is determined that detection is not possible, and an operation is performed to skip to the detection of the next pellet without bonding.

10 and 11 are examples of the structure of the large noise removal circuit and coordinate extraction circuit 20, and the structure of the shift register is the same as that shown in FIG. It is determined that a figure exists only when all five picture elements are 1 and all picture elements in grid line part B are 0. When a figure is detected under these conditions (for example, only the part marked with an x in the figure is 1) When a figure is inserted), information is detected about how far the figure extends left, right, up and down, centering on the pixel coordinate a of interest of this figure.
That is, the number of bits of the picture element inside the grid line picture element B in FIG. 10 is as shown in FIG. 11a, and if the logical product is calculated for each row and each column in this matrix, it becomes as shown in b and c. The information on b and c and the coordinates of the picture element of interest may be stored in the graphic coordinate storage circuit. When the storage is completed, the figure on the shift register becomes unnecessary and all picture elements in grid picture element B are cleared to 0. This is because if the data is not cleared, the same figure may be detected multiple times, and the memory of the figure coordinate storage circuit will be wasted.

When the pellet is die-bonded, the rotation from the reference position is actually within ±15 degrees, which is very small. In such a case, use the circular mask 14 in Figure 6.
Instead of -10 as shown in Figure 12 a, b, c
Scan with three pairs of circular masks rotated by ゜, 0゜, +10゜, memorize the coordinates when each mask is matched with the binarized figure, and calculate the pellet position based on these coordinates. It is also possible. In this case, since the electrode shape and the mask shape are closer to each other, most of the noise shapes are removed by the matching circuit, making subsequent calculation processing easier.

As explained above, according to the device of the present invention, storage of figure coordinates can be processed in real time in the same way as pellet figure scanning, and it is only necessary to calculate the pellet position based on the stored coordinate data after the completion of travel, resulting in high speed processing. The pellet position can be detected. Furthermore, since the basic method is to extract a figure that matches the circular mask, it can be applied regardless of the shape of the figure, such as circular, triangular, square, rectangular, etc., and various pellets can be put to practical use.

As described above, the electrode position on the pellet can be automatically detected without visual inspection by the operator.
It has become possible to fully automate wire bonding equipment.

[Brief explanation of the drawing]

FIG. 1 is a plan view showing the state of wire bonding of the pellet on the lead frame, FIG. 2 is an explanatory diagram showing the scanning area of the photoelectric scanner, and FIG. 3 is a block diagram of the pellet position detection device according to the present invention.
FIG. 4 is a diagram showing an example of a binary figure, FIG. 5 is a configuration example of a circular mask matching circuit, FIG. 6 is a diagram showing an example of a binary figure including noise, and FIG. Figure 8 shows an example of the configuration of a noise removal circuit after processing the figure in the figure, Figure 9 shows another configuration example of the present invention that is effective when there is a lot of noise, Figures 10 and 11
The figure shows an example of the configuration of a large noise removal and coordinate extraction circuit.
FIG. 12 is a diagram showing a modification of the circular mask. 1, 3, 4...Lead frame, 2...Pellet,
5...Gold wire, 6...Photoelectric scanner, 7...Scanning area, 8
...Binarization circuit, 9, 19...Circular mask matching circuit, 10...Coordinate storage calculation circuit, 11...Bonding tool control circuit, 12, 13...Binarization figure of electrode part, 14...Circular mask, 15...Bonding Tool mechanism, 16... Noise figure, 17... Minute missing figure, 18... Small noise removal circuit, 20... Large noise removal and coordinate extraction circuit, 21... Figure coordinate storage circuit,
22...Calculation identification circuit, 23...Standard graphic constant setting section, 50...AND circuit.

Claims (1)

[Claims] 1. A photoelectric conversion means, a circuit for binarizing the image detected by the photoelectric conversion means, and a predetermined circular area having a diameter inscribed in the figure whose position is to be detected in the binarized figure; a circuit for detecting a matching area and extracting a center position of the area, and a small noise removal circuit for correcting minute missing figures included in the binarized figure at a stage preceding the center position extraction circuit. a large noise removal circuit that removes the center position figure having a size larger than the predetermined circular area from among the figures formed by the center position extracted from the center position extraction circuit; A coordinate extraction circuit that extracts coordinate information indicating the size of the central figure from the large noise removal circuit, a figure coordinate storage circuit that stores the coordinate information, and a comparison between the stored coordinate information and a predetermined reference value. A position detection device further comprising a calculation identification circuit that makes a determination.