JPS5925375B2 - position detection device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a position detection device that can detect the position of minute bonding pads on a semiconductor chip with high precision.

With the development of integrated circuit technology, the size of semiconductor chips has become smaller, and it has become necessary to specify the bonding positions of lead wires connected to the semiconductor chips with high precision.For example, for semiconductor chips such as conventional LSIs and MSIs, In the past, there was a margin to set the bonding pad area sufficiently wide compared to the bonding area, but
In SI and the like, the bonding area and the bonding pad area are approximately equal, and the bonding pad positions are close to each other.

For this reason, the accuracy of bonding pad position detection is extremely important. If this position detection accuracy is insufficient, there is a risk that the lead wire will be bonded across adjacent bonding pads, leading to the production of defective products due to short circuits in the semiconductor chip circuit. Recently, various studies have been conducted on capturing and detecting bonding pad images and detecting the bonding pad position from the pixel position. There are many problems such as discrimination from unnecessary image signals, and it has not been put into practical use. The present invention has been made in consideration of the above circumstances, and its purpose is to provide a practical and simple configuration that can directly and accurately detect the position of minute bonding pads on a semiconductor chip. The object of the present invention is to provide a position detection device with high performance.

The present invention enables highly accurate position detection of minute bonding pads by optimizing the image detection level by, for example, a photoelectric converter of a semiconductor chip and removing signal components other than desired pattern information from the detected image signal. be.

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

FIG. 1 is a schematic diagram showing the general structure of this apparatus, and in the figure, reference numeral 1 indicates a position detection object used for position detection, that is, a semiconductor chip.

Opposed to this semiconductor chip 1 is an optical lens barrel 2 of an imaging system. This optical lens barrel 2 reflects light from a light source 3 provided on the side of the barrel with a half mirror 2a, and illuminates the semiconductor chip 1 through an objective lens 2b. A light image is formed on the imaging surface of the imaging device 4 via the lens 2b. This imaging device 4 is equipped with a scanning type solid-state imaging device (photoelectric converter) 4a consisting of, for example, a 100×100 pixel photoelectric conversion element, and is controlled by an imaging control device 5 to image the semiconductor chip 1. Image is being detected. The image signals of the semiconductor chip 1 detected by the imaging device 4 and sequentially output by scanning are individually quantized to a predetermined quantization level via the A/D converter 6, and converted into digital signals of several bits. It is input to the computer 7 as a pixel signal represented as data. This calculator 7 is composed of, for example, a microcomputer, which is configured with an internal memory, an accumulator, etc., and performs appropriate signal processing on the quantized pixel signals, and also performs appropriate signal processing on the quantized pixel signals, and performs appropriate signal processing on the quantized pixel signals. The detected image signals are sequentially accumulated (stored) in a memory 8 having a memory area. That is, the image pickup device 4 captures two images of the semiconductor chip 1.
The pixel signals are detected as 100×100 pixel signals arranged in a dimensional array, and the memory 8 stores the pixel signals corresponding to the pixel positions. Further, the computer 7 calculates an average value of detection levels, which will be described later, from the levels of the pixel signals, and feeds this back to the imaging control device 5.

The imaging control device 5 controls the imaging scanning speed in the imaging device 4 based on the feedback information, or controls the light source driving device 9.
The imaging signal level is optimized by energizing the light source 3 and variably setting the irradiation intensity of the light source 3. Note that A-E shown in the memory 8 in FIG. 1 indicates a detected pattern image. Therefore, the computer 7 sequentially extracts the quantized pixel signals for each frame, starting from the highest level, and obtains, for example, the average level EH of n1 high-level pixel signals.
The average level EL of N2 low-level pixel signals is calculated.

The above average level EH has a total number of pixels of 100 x 100 (=
10,000), for example, n1=64 are extracted in order from the highest level and are averaged. In addition, the average level EL on the low level side is the background level, and for example, the average level EL is N2 from the lowest level.
It is obtained by extracting =2048 pieces and averaging them. Note that the numbers Nl and n2 may be determined as appropriate according to the specifications. Each average level E calculated in this way
Based on H and eL, the calculator 7 calculates the threshold value S as follows.
I'm looking for. This calculation formula has been determined experimentally, and by comparing and determining the threshold value S and the pixel signal level, pixels corresponding to a bonding pad pattern or the like are extracted.

That is, pixels with a level below the threshold S are removed as background components, and pixels with a level above the threshold S are selectively extracted to characterize the detected image. by the way,
The image signal from the image pickup device 4 changes depending on the surface condition of the semiconductor chip 1, and the image signal level may exceed the ability of the photoelectric conversion element and become saturated.

It is also expected that the image signal level will be low and that a signal of sufficient level will not be obtained. For example, as shown in FIG. 2a, if the level of the desired pattern By comparison, the patterns X and Y can be easily distinguished. However, as shown in FIG. 2b, when the level of the desired pattern X is saturated and the level of the unnecessary pattern Y is near the saturation level, it becomes impossible to discriminate between the patterns X and Y based on the threshold value S. Therefore, in order to perform pattern discrimination well, it is necessary to optimize the detection gain so that the highest level of the detection pixel signal is below the saturation level of the photoelectric detection element. Therefore, in this apparatus, the imaging detection sensitivity is adjusted based on the previously determined average level EH of the high-level pixel signal and by comparing it with the saturation level Es of the photoelectric converter. For example, when the level of the saturation level E8 (5 to 1001) is k, calculate the deviation, and when this deviation P is negative, it is assumed that there is a margin in the detection signal level, and a signal for increasing the imaging detection sensitivity is captured by the computer 7. It is given to the control device 5. Also, the above deviation P
If is positive, it is assumed that the detection signal level is too high and most of it is saturated, and a signal for reducing sensitivity is given to the imaging control device 5 from the computer 7. The imaging control device 5 receives such a sensitivity control signal, controls the operation of the light source driving device 9, and adjusts the illumination intensity of the light source 3 by increasing or decreasing the driving power of the light source 3 by the device 9. This results in
When the illumination intensity of the light source 3 is reduced, the semiconductor chip 1
The image becomes dark and its detection sensitivity is suppressed, and when the illumination intensity is increased, a high brightness image of the semiconductor chip 1 is obtained and the detection sensitivity is set high. Incidentally, the illumination intensity of the light source 3 may be adjusted by variable control of the light transmittance of a filter (not shown) provided in the illumination section between the light source 3 and the semiconductor chip 1. In addition, in the case where the illumination intensity of the light source 3 is kept constant, the photoelectric converter 4 of the imaging device 4
Since a is a scanning type, the imaging control device 5 variably controls the imaging scanning speed according to the sensitivity control signal.

That is, since a scanning photoelectric conversion element accumulates and outputs a charge corresponding to the optical image intensity during the time allotted to the element, the allotted time can be increased or decreased by changing the scanning speed. , whereby the detection sensitivity may be adjusted. Therefore, to lower the detection sensitivity, the scanning speed can be increased, and to increase the detection sensitivity, the scanning speed can be decreased to adjust to the desired detection sensitivity. Note that the scanning speed (imaging charge accumulation time) may be controlled by varying the scanning retrace time of the photoelectric converter, or the scanning frequency may be directly varied. Thus, by controlling the sensitivity, the level of the detected image signal is adjusted to the saturation level Es of the photoelectric conversion element.
It is possible to suppress the contrast to below, and also set the contrast to be sufficiently high.

Therefore, signal processing for position detection, which will be described below, becomes very easy, and unnecessary image signals can also be easily removed. Also, in the above explanation, what is the control of the illumination intensity of the light source 3 and the imaging device? Although the control of the scanning speed according to No. 4 was performed independently, it goes without saying that these may be controlled variably in relation to each other. Furthermore, the average level EH of the high-level pixel signal, which serves as a reference for detection sensitivity adjustment, may be obtained for each semiconductor chip 1, but it is also possible to further average the average level EH of several semiconductor chips 1. , the detection sensitivity may be adjusted based on the average level EH. In this case, it is preferable to discard the information on the average level EH in order of age and instead adopt the average level EH of the new chip 1 one after another. In this apparatus, the position of the bonding pad is detected from the bonding pad image subjected to such pre-processing in the following manner.

The image of the semiconductor chip 1 detected with the detection sensitivity adjusted as described above is discriminated for each pixel by comparing the level with the threshold S in the computer 7, binarized, and stored in the memory 8. It is stored again in accordance with the position. At this time, for example, the unnecessary image shown in pattern B in FIG. 1 is removed because its detection level is low. Also pattern B
Even if pattern B is not removed by the above detection level comparison, since pattern B has an elongated shape and is significantly different from the desired bonding pad shape, it is removed by the computer 7 by comparing the characteristics of the shape. Ru. Therefore, the memory 8 stores pattern images A, C, D, and E having shapes having the desired pattern characteristics. Of course, the pattern images stored in the memory 8 are generally not only images of the desired bonding pads, but also images of electrode lead line patterns that are continuous with the bonding pads, and images of needle scratches, etc. that occur on the pads. Contains unnecessary ingredients. Therefore, there is a problem in terms of detection accuracy if the position of the pattern image detected by the level comparison is directly detected. Therefore, in this apparatus, the following signal processing is performed between the computer 7 and the memory 8 to correct the pattern image and accurately detect the position of the bonding pad.

FIG. 3 shows one pattern obtained by the above-described image detection and stored in the memory 8, in which each rectangular section represents a pixel.

The pixel signals stored at each pixel position each have a quantization level corresponding to the optical image level, and hatched patterns 11, 12, and 13 in the figure indicate unnecessary patterns such as electrode lead lines, and In the figure, 14 indicates a pixel missing portion due to a needle wound or the like. However, when a detection pattern composed of such two-dimensionally arranged pixels is presented, the computer 7 first detects the number of pixels in each row and each column in the detection pattern.

Then, when the number of pixels is smaller than the preset number,
For example, a desired bonding pad pattern 10×
In the case of 10 pixels, pixel components existing in rows or rows where the number of pixels is less than 4" are considered unnecessary protruding patterns and are removed. Here, the diagonal line on the upper right is The marked pixel 11 will be removed by checking the number of pixels in each column, and the pixel 12 marked with diagonal lines on the upper left will be removed by checking the number of pixels in each row.Also, as explained earlier, the desired pattern shape Since it is known in advance that is 10×10 pixels, the unnecessary pixels 11 and 12
The number of pixels in the row and column directions of the pixel pattern from which the pixel pattern has been removed is checked at the same time. This check detects row (column) lines having extra pixels in the row direction.
In other words, the extra pixels are detected as unnecessary at the end of the row (column) line. Therefore, in this case, the number of pixels in the columns (rows) at both ends is compared with each other, and the pixel 13 at the end with the smaller number of pixels is regarded as an unnecessary pixel and removed. Only the pixel information of the pattern can be extracted. By checking the number of pixels at the peripheral edge, it is possible to effectively remove protruding information such as electrode lead lines that are mixed in the desired pattern image.

Therefore, there is no possibility that the position detection accuracy will deteriorate due to the information on the protrusion pattern. In other words, since only the desired original pattern information can be detected, it is possible to detect the central pixel position of the bonding pad image and, by extension, the position of the bonding pad on the semiconductor chip with high precision. The center pixel position of the bonding pad image is determined, for example, from the number of pixels constituting the image and their arrangement, as a pixel at a position that divides the image into two in the row direction and column direction, respectively. In this way, the position of the bonding pad can be detected accurately without being affected by the electrode lead wire pattern, etc., so it is possible to bond the lead wire to the bonding pad with high precision. It does not introduce the conventional problems. Therefore, great practical effects can be achieved. Then, by checking the number of pixels of the pixelated image pattern in a two-dimensional array configuration, an image pattern having the desired characteristic pattern is found very easily, and unnecessary pixel components such as extraction electrode lines included therein are removed. The desired bonding pad image is extracted from the bonding pad image, and its center position is mechanically determined from the pixels that make up this bonding pad image, so the detected position is extremely accurate compared to the pixel size. becomes.

Moreover, even if there is a pixel missing part such as a scratch in the image pattern, the bonding pattern image is captured according to the limbal characteristics of the image pattern as described above, so accurate position detection is possible without being affected by needle scratches, etc. It is possible. Therefore, the position of a bonding pad having a small area such as a VLSI chip can be detected with high precision, so that the wire bonding position relative to the bonding pad can be accurately controlled, and a decrease in production yield due to poor bonding can be significantly reduced.

Furthermore, as described above, position detection can be performed by relatively simple signal processing, so it is very practical and exhibits excellent industrial effects. Note that the present invention is not limited only to the above embodiments.

For example, the signal processing form may be performed by software,
Alternatively, it may be implemented as a notebook using a dedicated circuit.
Further, the numerical data for pattern discrimination may be determined according to the specifications of the desired position detection pattern. Further, the quantization precision and the like may be determined as appropriate. Furthermore, similar position detection may be performed on a plurality of pad images, and these may be combined to form the position detection zeta. In short, the present invention can be implemented with various modifications without departing from the gist thereof.

[Brief explanation of the drawing]

The figures are for explaining one embodiment of the present invention. Figure 1 is a schematic diagram of the device configuration, Figure 2 A and b are diagrams showing optimization of pixel signal level, and Figure 3 is a two-dimensional array. FIG. 3 is a schematic diagram showing a pattern image shown by pixels. 1...Semiconductor chip, 3...Light source, 4
...... Imaging device (photoelectric converter), 5...
Imaging control device, 6...A/D converter, 7...
...Calculator, 8...Memory.

Claims (1)

[Scope of Claims] 1. Means for imaging a semiconductor chip to obtain a bonding pad image consisting of two-dimensionally arranged pixels of bonding pads on the semiconductor chip, and each row and each column of pixels forming this bonding pad image. and means for determining the number of pixels in the bonding pad image by comparing the information on the previously known shape and size of the bonding pad with the number of pixels in each row and each column. A position detection device comprising: means for removing pixels having a size other than that size as unnecessary pixels; and means for determining the center position of the bonding pad from a modified bonding pad image from which the unnecessary pixels have been removed. 2. The position detection device according to claim 1, wherein the unnecessary pixels removed from the bonding pad image are pixels forming an electrode lead line pattern connected to the bonding pad. 3. The position detection device according to claim 1, wherein the pixels forming the bonding pad image are expressed as pixel data that has been subjected to binarization processing.