In the wire bonding inspection process the inspectors focus on (a) the position of the bonding wire and (b) the contour and position of the bonding ball [2]. Khotanzad et al. [3-4] presented an automatic system for evaluating bonding ball quality. Sreenivasan et al. [5] presented a method to compute the bonding ball’s shape, size and location to determine its quality. All of their systems [2-5] can determine the location of the bonding ball from a 2D image of an IC wafer and extract bonding ball geometrical information. Tsukahara et al. [6] presented a vision inspection system for IC bonding wire that uses high contrast image capture and an accurate bonding-ball measurement algorithm based on sub-pixel and morphological techniques.
With the development of automatic optical inspection (AOI) technology, the IC packaging foundries attempted to improve the wire bonding position inspection using an AOI system. The AOI system has emerged as an effective inspection approach in PCB assembly [7-9], but is still not ready for wire bonding position inspection in IC packaging foundries. Ayoub [10] pointed out the AOI system problems with wire bonding position inspection. The commercial systems still need to enhance the signal to noise ratio, extend the defect coverage from single wire ICs to multi-layered wire ICs and increase the inspection speed. The ability to separate the wire from a complex varying background between the die and pad is an important aspect of post-wire-bonding inspection.
Accomplishing this task requires smart illumination and inspection algorithms to work together to increase the signal-to-noise ratio between the wire and its surroundings. Wang et al. [11] presented a machine vision inspection technique with a defect detection algorithm for the bonding ball and bonding wire. Their experimental results showed that a good lighting condition can improve the recognition rate.
Ngan and Kang [12] presented an algorithm for inspecting the bonding wire. They used Hough transformation to determine the straight-line equation of the bonding wire. A fiber optic ring light was used as the light source to highlight the bonding wires. Ye et al. [13] presented an
inspection system that applied a stereo vision technique to detect the defects related to the 3D profile of bonding wires. They proposed that the illumination system should maximize the light reflected from the bonding wires and minimize the light reflected from the surface of the chip.
Perng et al. [14–16] devised a vision inspection system equipped with a structured lighting system to highlight the bonding wire. Their system can be used for the on-line inspection of single layer wire ICs. In the multi-layered wire IC case, such as that shown in Figures 7 and 8, none of the existing 2D image inspection methods [2-6, 11-16] can be applied because the image of the wires in the lower layer would be hidden or shadowed by the wires in the upper layer.
Some other researchers [17-19] recovered the 3D shape using a photometric method and structured lighting system. However, multiple images are required in these researches to determine the surface normal. Kim and Koh [20] discussed the shadow problem for in-line shape inspection of LEDs using pattern projection techniques. Because of the high ratio of the outside wall to the width of the inside base area, conventional measurement techniques for projecting patterns in off-optical axis easily fail to perceive the entire shape of LEDs and produce noisy results due to the shadow problem. Kim and Koh [20] also presented a sensor system utilizes a dual projection system. Using two measurement results acquired from pattern projections switching with different incident angles, shadow-free results can be acquired. But this only resolves the shadow problem for the outside wall, not for the hidden or shadowed problem by the wires in the upper layer.
Lim et al. [21] presented an auto-focusing technique to measure the height and diameter of the bonding ball. Their method can also be used for the inspection of missing bonds and wire loop height measurements. Because a high-magnification microscope was used in their system, only the image of the wires in the focused plane could be captured clearly, while the wires in the other planes were hard to observe or inspect. KAIJO Corporation [22] used an auto-focusing technique to develop a wire bonding AOI system (WI-110). The KAIJO Corporation claimed that their system could correctly inspect multi-layered wire products. The WI-110 system searches for the bonding
ball position on the pad side first and defines it as the starting point. Next, the loop height of the wire is measured from the starting point until the ending point is found, and then converts the loop height of each point in the wire and converts this into the wire track. The auto-focusing technique inspection speed is very slow. The KAIJO machine took 8 seconds to inspect a single wire [23].
Since the available AOI machine for the wire bonding inspection is not sufficiently fast, only off-line inspection machines were adopted.
Pacheco et al. [24] presented an advanced time domain reflectometry (TDR) technique to detect the defects in multi-layered wire ICs. Only a laboratory prototype was implemented because of the high cost of this system.
Kulicke & Soffa Industries Inc. (KNS) presented a process program comparator (PPC) technique with their KNet system [25]. Image verification is not required because such an inspection approach is based only on the bonding sequence and the coordinate value of the ending point of each wire. The major drawback of the PPC method is that the PPC method does not have any actual visual inspection and does not utilize the actual lead position information. Some serious mal-detection or lost detection problems were encountered when their system was applied to a product composed of leadframe based material. Compared with the other substrate-based IC product material fabricated using a mask, the leadframe based material is fabricated by punching or etching and has the problem of high variation and low accuracy for each lead on it. Although KNS provided a fast and on-line inspection solution for multi-layered wire ICs, their method could only be applied to a product composed of substrate-based material. Until now, no commercial equipment existed that could meet all of the wire bonding position inspection requirements.
In this dissertation, a novel method is proposed to solve the wire image hidden problem for multi-layered wire ICs and eliminates the mal-detection and lost detection problems for leadframe-based material ICs. The proposed method integrates the image processing and wire bonding simulation techniques to automatically inspect the correctness of wire bonding positions.
The wire bonding position correctness is compared with the standard before the actual bonding process begins. Post-wire-bonding image identification is not required because such an inspection approach is based only on the information of the bonding sequence, the coordinate values of both end points of each wire and the base material image. In other words, there is no need to concern about whether the image of the wires in the lower layer will be blocked or hidden by the shadow of the wires in the upper layers. Therefore, the proposed approach can be used for inspecting the bonding position of multi-layer wire ICs. It uses the original CAD drawing as the comparison standard. When the CAD drawing is not available for the IC packaging foundries, the first setup machine will be used as the comparison standard.
A summary of the previous researches, available commercial systems and the proposed method for wire bonding position inspection is given in Table 1.
Table 1. Summary of the previous researches, available commercial systems, and the proposed WWBP system for wire bonding position inspection.
Attributes Method
2D image inspection method [11] Yes Post-bonding Bad Good Fast On-line Yes High Inspection with structure lighting [16] Yes Post-bonding Bad Good Fast On-line No High Auto-focusing [23] Yes Post-bonding Good Good Very slow Off-line Yes High TDR technique [24] Yes Post-bonding Good Good Very slow Off-line Yes Hightest
PPC method [25] No Pre-bonding Good Bad Fast On-line No Low
The proposed WWBP system Yes Pre-bonding Good Good Fast On-line No Low
Image
3. Wrong wire bonding prevention system based on CAD drawing to inspect