PACKAGING/ASSEMBLY: CAE modeling for ultrafine pitch bonded ICs
11/01/1999
Eric Kuah, Gary P. Widdowson, Ravi Kandasamy, Narasimalu Srikanth, Charles J. Vath III, S.C. Ho, ASM Technology Singapore Pte. Ltd., Singapore
High-quality plastic package molding systems can be optimized with computer-aided engineering software tools. These tools have made possible the design of a mold gate-and-runner system, in this case for a QFP100L package with a leading-edge 50µm pad pitch, which provides balanced encapsulant flow to minimize wire sweep during the molding process.
The capabilities of wire bonding systems are expected to exceed the SIA Roadmap's 50µm in-line pad-pitch bonding by 2000. Indeed, such ultrafine pitch bonding is already available. Although this manufacturing breakthrough leads to fundamental improvements in IC functionality, it presents an intense challenge to plastic-package automolding processes. Customers are demanding new automolding capabilities with ever-lower percentages of wire sweep (i.e., movement of straight bond wires).
Numerous process variables affect the quality of molding. For example, intimate knowledge of wire, die, mold, and leadframe properties is essential to optimize any given process. Selection of the correct epoxy molding compound and process parameters such as compound transfer velocity, transfer pressure, and mold design are even more critical for 50µm in-line pad-pitch bonding. A mold designer must understand which of these variables has the greatest influence on the process for each package.
We have used computer-aided engineering (CAE) software tools to understand the interrelationship of the most critical variables in a plastic molding process, thus allowing achievement of an optimum process design. Specifically, our study applied CAE to the combination of an ASM AB339 automatic wire-bonding system and an ASM EM649 automatic molding system. Our data show that molding 50µm bond pad-pitch packages with low wire sweep is now possible.
Modeling the molding process
As wire sweep requirements become more stringent, process engineers can no longer rely on heuristic or empirical techniques to predict molding quality. The number of interrelated variables is so large that numerical analysis must be used to produce an
optimum mold design. CAE tools allow process engineers to study the effects of molding process conditions and the impact of molding compound flow on wire sweep in IC packages. With CAE the effects of altering wire properties can be simulated readily, thus allowing automolding manufacturers to offer customers alternatives that optimize wire properties for given applications.
 Figure 1. Finite element model of the top and bottom mold halves, for a QFP100L package, connected mathematically. |
We used C-Mold software to simulate the dynamics of fluid flow, heat transfer, and curing of resin during a molding process. This software models the mold-filling process by assuming flow of an incompressible viscous polymeric compound. The technique uses finite element and finite difference methods to solve both pressure and temperature fields. To judge the accuracy of the finite element analysis, the software uses an experimental technique called "short-shot," where the molding process is interrupted before completion so that intermediate steps can be analyzed. This technique determines important understanding of mold flow filling. Some of the more critical aspects addressed with "short-shot" experimentation include:
- jetting — when the molding compound enters the mold cavity in a jet form, usually leading to a degradation in the wire bond quality;
- spiking — a sudden increase in mold compound velocity toward the end of the filling process, leading to severe wire sweep; and
- racing — a difference in the velocity of mold compound in the upper and lower halves of the mold, usually resulting in die tilt and deformed loops.
QFP100L ultrafine pitch model
Our test vehicle for the finite element analysis was a QFP100L package. We created a finite element model to simulate the top and bottom halves of the mold, which are connected mathematically in the FEA model (Fig. 1). A factorial design of experiment analysis determined the optimal combination of epoxy molding compound flow transfer profile and transfer time. We used these optimum results throughout our investigation, in which we looked at two different brands of 22µm gold wire.
 Figure 2. Short-shot results after ~2.66 sec with FEA simulation (left) and actual mold filling (right) for top and bottom mold halves. |
We used CAE analysis to determine the quality of our mold design. One of the most important parts of the mold is the runner-and-gating system; its modeling is critical to the accuracy of FEA results. Think of the gate as a valve and the runner as a pipe. They regulate the flow of epoxy molding compound into the mold cavity, thus preventing jetting, spiking and racing. With the C-Mold software, the runner and gating systems are modeled as a series of discrete steps (C-Mold cannot accept sloped surfaces).
Proving the optimization
The control of wire sweep becomes more difficult as wire length increases; longer wires have to sustain more cross-sectional loading, but material strength is the same as with short wires.
We concentrated on 4mm long wires and measured 18 of the 100 wires in our test package, choosing wires spread evenly around the periphery of the IC. Some of these tested wires were orthogonal to the direction of the epoxy molding compound flow and thus were more vulnerable to large wire sweep (i.e., they experience a large shear force from mold flow advance). Others were parallel to the flow and less vulnerable.
To gauge the accuracy of the FEA results, Fig. 2 shows the "short-shot" results for the molding process after 2.66 sec, which represents approximately 50% of the molding time. We see that the FEA simulated results closely mimic the actual filling process for both top and bottom halves of the mold. An experienced process engineer can determine from these results that there is no significant jetting or racing.
 Figure 3. Melt front velocity vs. time, showing no spiking. |
We used the melt velocity result from the simulation to determine that there is no spiking of the epoxy molding compound (Fig. 3); the mold and epoxy molding compound flow were satisfactory.
Our results predicted and showed that the maximum wire sweep for 4mm orthogonal wires will be approximately 3% (i.e, the percent change from a baseline before molding to the maximum point of wire displacement after molding) for one of the tested types of gold wire and slightly higher for the second type (Fig. 4). Further, parallel wires had a wire sweep close to zero. Of course, wires oriented in the range from orthogonal to parallel to the mold flow have a predicted wire sweep in between these maximum and minimum values.
We found that the material properties of different gold wire do not affect the results as much as the direction of the bonded wires. The data in Fig. 4 show that the second type of wire has a higher wire sweep and higher wire sweep standard deviation than the first. Thus, a process engineer might recommend the first type of gold wire for optimization of this molding process.
Conclusion
Using mold experimental short-shot computer simulated results, we have demonstrated that it is possible to control molding of 50µm bonded packages to achieve <3% wire sweep, regardless of wire type. Indeed, we observed via x-ray analysis of a molded package a wire sweep from 0.5% to 3% for 4mm-long wire bonds.
As bonding pad-pitch continues to decrease to <40µm, process engineers responsible for automolding operations must address the fundamental issues affecting wire sweep because the interactions of the epoxy molding compound, wire and die material, and mold design become even more critical. This process can only be optimized efficiently with CAE tools.
In addition, the layout of wires and the placement of the long wire bonds need serious consideration in future packages. Close communications between IC designers, process engineers, and equipment manufacturers will become essential for the development of the next generation of semiconductor manufacturing equipment.
Acknowledgment
C-Mold, which is a registered trademark of Advanced CAE Technology Inc., is a commercial FEM (finite element method), FDM (finite difference method) and FVM (finite volume method) fluid flow package (see www.cmold.com).
Eric Kuah Teng Hock received his BSME and MSME from Ohio University and his MS in theoretical and applied mechanics from the University of Illinois. He is technical manager of the product reliability department at ASM Technology Singapore Pte. Ltd., 2 Yishun Avenue 7, Singapore 2776; ph 65/752-6311, fax 65/758-2287, e-mail [email protected].
Gary Widdowson received his BE in electrical engineering and PhD in computer-aided optimization from the University of Sheffield, England. He is technical manager at ASM Technology Singapore.
Ravi Kandasamy received his masters degree with first class honors from the department of mechanical engineering at Regional Engineering College, India. He is a CAE engineer at ASM Technology Singapore.
Narasimalu Srikanth received his masters degree from the department of mechanical engineering at the Indian Institute of Technology, Bombay, India. He is a senior CAE engineer at ASM Technology Singapore.
Charles J. Vath III received his BS in chemistry and MS in material science from San Jose State University. He is director of process and packaging technology at ASM Technology Singapore.
S.C. Ho received his degree in industrial engineering from the University of Hong Kong, a postgraduate diploma in information technology from Polytechnic University, Hong Kong, and an MS in industrial and system engineering from National University of Singapore. He manages the Automold product group at ASM Technology Singapore.