Computer-aided design and analysis of circuits and semiconductor – Nanotechnology related integrated circuit design
Reexamination Certificate
1999-04-30
2001-09-11
Phan, Trong (Department: 2818)
Computer-aided design and analysis of circuits and semiconductor
Nanotechnology related integrated circuit design
Reexamination Certificate
active
06289492
ABSTRACT:
FIELD
This invention relates to machine vision systems, and particularly to machine visions systems for use with semiconductor chip wire bonding devices, and similar bonding devices.
BACKGROUND
Semiconductor devices, such as integrated circuit chips, are electrically connected to leads on a lead frame by a process known as wire bonding. The wire bonding operation involves placing and connecting a wire to electrically connect a pad residing on a die (semiconductor chip) to a lead in a lead frame. Once all the pads and leads on the chip and the lead frame have been wire bonded, it can be packaged, often in ceramic or plastic, to form an integrated circuit device. In a typical application, a die or chip may have hundreds of pads and leads that need to be connected.
As circuit density of integrated circuit chips has increased, the configuration of the lead frames, which provide input and output to the chips, has also become increasingly complex. The leads to which wire bonds must be made have become increasingly smaller, more varied in shape, and more numerous and, therefore, more difficult to form wire bonds thereto.
Before bonding, the wire bonder system learns the expected positions and orientations of the die, the lead frame, the pads, and the leads. Typically, the location and orientation of the die are indicated with respect to two reference points, such as fiducials. The wire bonder learns the pad shape and the pad locations on the die. Similarly, the location of the lead frame is indicated by one or more reference points, and the wire bonder learns the lead positions with respect to the lead frame.
The position of the target bond points on the leads is determined in a number of ways. In one technique, operators manually place a location of a target bond point on each lead to train the location of the leads and the target bond points to the system. In another technique, such as that described in U.S. Pat. No. 5,119,436, (the '436 patent) a manually trained location of a target bond point is refined by the pattern recognition systems of the wire bonder. The system then automatically determines a location of target bond points on successive leads.
The wire bonder uses the trained data, during operation, to bond the wires. Before bonding, typically, the system relocates the leads. However, for various reasons, some of the bonds formed are unacceptable.
Unacceptable bonds include, for example, missed bonds, misshapen bonds, or partial bonds. A missed bond is positioned off the lead and fails to form the electrical connection required for proper operation of the chip. A partial bond passes electrical testing during quality inspection, but will typically fail in the field.
Among other reasons, unacceptable bonds occur because of improper placement of the target bond position along the normal axis of the leads. For instance, the trained target bond position can be too far to the side of a lead. Further, when the training information about target bond points is applied to subsequent parts, the variation among the parts can cause unacceptable bonds, particularly if a trained bond position was offset from the center of the lead.
Among other reasons, unacceptable bonds also occur because of errors in lead detection. Lead detection errors can occur because of non-robust techniques, such as the one-dimensional scanning of intensities employed in the '436 patent, cited above, poor image quality, or a combination of both. Poor image quality is caused by poor lighting, poor focusing, specular reflections off the leads and the background of the image, noise, and irregularities in the surface of the leads, among other causes, as is known in the art.
A further reason for unacceptable bonds is the failure or inaccurate resolution of the target bond position along the length of the leads. The manual method does not resolve the target bond position along the length of the leads. Although the manual method is reasonably accurate and robust for large leads with parallel sides, typically its failure rate is high for smaller leads and leads with non-parallel sides. The '436 patent does resolve the target bond position along the length of the leads for parallel leads. The lack of robustness of the '436 patent, either alone or combined with poor image quality, however, leaves unclear the effect of the length resolution on the number of unacceptable bonds.
Further, the '436 patent does not address non-parallel leads.
SUMMARY
The invention provides methods and apparatuses for defining one or more positions on elongated objects. In one embodiment, data is acquired that represents the elongated objects, where the data is used to determine at least a portion of the edges of the elongated objects. The edges are then used to define a locus of viable points on each of the elongated objects, where the locus is positioned offset from a portion of at least one edge by at least a minimum-offset distance. The locus has a longitudinal axis and optionally a reference position.
In a preferred embodiment, the locus defines the location for one or more target bond points on selected leads (elongated objects) of a lead frame. The longitudinal axis of the locus is the bisector of the lead.
The locus is a region on the leads that decreases the likelihood of the formation of unacceptable bonds of wires to the leads of the lead frame. Among other aspects, the invention recognizes that the likelihood of an unacceptable bond is decreased when a target bond position is offset from at least a portion of the lead edges by at least a minimum-offset distance. Thus, the bond will more likely be on the lead, and not be partially on the lead, or miss the lead entirely. In a preferred embodiment, the locus is offset from all lead edges except the base, that is, offset from the lead sides and the lead “tip”.
The minimum-offset distance can vary along edges of the lead. In one embodiment, the minimum-offset distance is equal to at least one-half the dimension of the bond. Using this dimension to define the locus, a bond centered at a target bond position within the locus is more likely fully on the lead because the bond is not large enough to reach a lead edge.
In a preferred embodiment, a “tip” of each of the selected leads is also detected to refine the definition of the locus to be a region between the sides that is offset from both the sides and the “tip” by at least a minimum-offset distance.
In one embodiment, the “tip” is defined as the intersection of a bisector and a portion of one edge of each of the leads, where the bisector is a line that is substantially equidistant between the sides of each lead near the “tip”. More particularly, the bisector is a line substantially equidistant between the lines that represent part of the lead sides. This definition of “tip” for a lead is substantially independent of the shape of the lead. Therefore, the definition can be used with leads having varied shaped sides, such as leads with one or more curved sides, tapered sides, or straight sides, whether the lead sides are parallel, converging, diverging or angled, where angled leads have an angled bisector comprised of more than one line, for example.
In an alternate embodiment, the “tip” is defined as a center point where a line normal to the bisector first touches the lead when moved toward the “tip” from a position outside of each lead. This definition can also be used with any shaped lead.
In a preferred embodiment, the method of the invention is divided into two parts being: training and run-time. The loci are defined for leads of a training device. Thereafter, the loci are used to position the target bond point on the leads of a run-time device. In this embodiment, the leads of the training device and the leads of the run-time device correspond one-to-one, and the training device is an instance of the run-time device or a model to be used with a plurality of run-time devices.
In one embodiment, at least a portion of each side of the run-time leads is detected by searching an image of the run-time leads. Once th
Dutta-Choudhury Paul
Picard Len
Tosa Yasunari
Calabresi Tracy
Cognex Corporation
Phan Trong
Weinzimmer Russ
LandOfFree
Methods and apparatuses for defining a region on an... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Methods and apparatuses for defining a region on an..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Methods and apparatuses for defining a region on an... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2497795