Computer-aided design and analysis of circuits and semiconductor – Nanotechnology related integrated circuit design
Reexamination Certificate
2002-04-01
2004-01-27
Garbowski, Leigh M. (Department: 2825)
Computer-aided design and analysis of circuits and semiconductor
Nanotechnology related integrated circuit design
C716S030000, C700S121000
Reexamination Certificate
active
06684380
ABSTRACT:
FIELD OF THE INVENTION
This invention pertains to the packaging of electronic components, particularly to a technique for facilitating the analysis of electronic packages by reducing their structural complexity.
BACKGROUND OF THE INVENTION
The present invention relates to the electrical analysis of electronic packaging structures, such as integrated circuit chip metalization cards, printed circuit boards, modules. More specifically, the present invention relates to modifying the structure so that it retains nearly the same electric parameters, but can be analyzed more efficiently by package analysis tools (for example capacitance. resistance, or inductance analysis tools).
At present, the computational power to analyze entire, truly realistic package structures does not exist. Entire integrated circuit chips or multi-chip modules, for example, consist of millions of conductive segments. For analysis purposes, each segment corresponds to one or more unknowns in the solution matrix of a package analysis tool. This number of unknowns greatly exceeds today's computational abilities. To avoid this problem, engineers either consider small sections of the entire structure, make simplifications through omission or gross distortion of the segments, or a combination of the two.
Accordingly, it is a primary object of this invention to achieve a technique for efficient analysis of the defined package structure. This invention describes a systematic approach for simplifying an existing structure so that the number of unknowns is dramatically reduced, but without substantially changing the structure's electrical parameters. The structure is decomposed into non-overlapping elemental shapes, such as rectangles or triangles (2D or 3D) and then these shapes are modified in such fashion that the structure's topology (number of nets, for example) remains the same and the electrical properties (resistance, capacitance and inductance) change only minimally, yet the number of unknowns is significantly reduced. For example, holes in conductors, small peninsulas of conductor, or small irregularities that add complexity to the structure but have minimal effects on electrical property, might.
The problem of analyzing more complex structures has been identified and has generally been handled by having the user simplify the structure by hand, by developing techniques that allow greater numbers of unknowns, by reducing the coupling between various pieces of the structure, or by model order reduction.
In the first approach, the user continues to remove pieces of the structure until it is simple enough to be handled by the analysis tool. The problem is that the approach is not systematic—there are often so many shapes that a user may miss many simplifications or may make simplifications that have undesirable consequences. Because typical designs consist of thousands or more shapes, which overlap or make contact to each other in a variety of ways, without such a systematic approach as described here, the user would not be able to consider all the possible simplifications. Even if done properly, simplification by hand is extremely time consuming and would be difficult to apply to a large number of structures within a reasonable time period.
In the second approach, any of a variety of techniques that involve FFT or wavelets, for example, are used to allow an increase in the number of unknowns (see, for example, E. Michielssen et al, “Fast algorithms for the electromagnetic simulation of planar structures,” IEEE International Symposium on Electromagnetic Compatibility, Vol. 1, pp. 172-176, 1998). Conceptually, the original physical structure is directly converted into the mathematical domain, where mathematical techniques are brought to bear on the matrix (which may not be explicitly generated) to effectively reduce the number of unknowns so that the problem can be solved. Such techniques, however, are usually selective, working on some structures and not on others. A large number of techniques would need to be developed to handle the entire set of structures most users need to analyze. The present invention will work on most structures, though perhaps the factors might need tuning to provide the most optimal results.
In the third approach, coupling between sections of the structure is omitted (see B. Rubin et al, “Electrical Modeling of Extremely Large Packages,” 1997 Electronic Components and Technology Conference, pp. 804-809) so that the resulting matrix has a more block-diagonal structure, allowing a larger number of unknowns to be handled. From a circuit point of view, this simply means removing the capacitive coupling and mutual inductances between the sections so that fewer circuit elements and thus larger structures can be handled. One problem is that the coupling may be too large to be omitted, or even if done, the resulting structure is still too large. Approaches that reduce structure complexity through other means often give rise to accuracy and numerical stability problems.
In the last approach, a circuit is generated to represent the structure. Through network analysis or other techniques, a circuit model with far fewer circuit elements is generated (see L. T. Pillage and R. A. Rohrer, “Asymptotic waveform evaluation for timing analysis,” IEEE Trans. on Computer-Aided Design, 9(4) pp 352-366, 1990). At present, this is still an area of scientific investigation with related issues of stability and accuracy.
The present invention relies on taking the original structure and simplifying it geometrically within known constraints. None of the prior art techniques mentioned above do this. The inventive technique does not rule out the subsequent use of the other techniques described above, but rather our invention might serve as a first step of simplification. The other techniques could be applied after the simplification to allow an even larger number of unknowns. This problem of reducing structure complexity in a systematic way, which is the basis of this invention, has received little attention on the outside.
SUMMARY OF THE INVENTION
The structure to be simplified will be referred to as original structure. The first step is to modify the shapes so that they continue to fully represent the structure but do not overlap; the shapes may be isolated, or may touch other shapes. Shapes conventionally used in package analysis such as rectangles, triangular prisms, tetrahedrons, can be handled as can shapes with zero thickness. For understanding purposes only, we assume that the structure is stratified in the z direction, consistent with card and other package structures. Thus, we assume the structure consists only of 3D rectangular shapes lying in the xy plane, with thickness in the z direction. It is to be understood that this invention does apply to more general shapes.
Accordingly, the primary feature of the present invention resides in the technique or process (as well as the means involved) for creating a new geometrical representation of an electronic package comprising: entering shapes that define the structure of the electronic package; modifying the shapes so that they do not overlap; for each edge of each shape, determining moves that neither alter structure topology, nor violate user directives, nor overlap other shapes; finding unity-based factors for each move; finding a move that gives maximum product of unity-based factors; obtaining the products of the factors and, if greater than one, continuing the process, otherwise discontinue; and if continuing the process, moving edge and thus defining the new structure.
In a specific form, the inventive process also includes: generating a set of non-overlapping shapes that represent the said electronic package; generating a table of all the allowed moves of the edges of each said shape so that the move does not result in the overlap of any shapes and does not cause shapes in different nets to touch each other; generating a set of subfactors that give the desirability of each of the said edge movements; generating a total factor that is t
Breiland Erik J
Rubin Barry J
Garbowski Leigh M.
Morris Daniel P.
Ohlandt, Greeley, Ruggiero & Ferle, L.L.P.
LandOfFree
Intelligent structure simplification to facilitate package... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Intelligent structure simplification to facilitate package..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Intelligent structure simplification to facilitate package... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3206883