Research On Modeling, Control And Implementation Of Dispensing System For IC Packaging

In Advanced Integrated Circuit Encapsulation (AICE) process, the time-pressure dispensing system has been widely used to squeeze the adhesive fluid in a syringe onto boards or substrates with the pressurized air for wafer/substrate bonding. However, complexity of the process, which includes the air-fluid coupling and their nonlinear uncertainties, makes it difficult to obtain the desirable process model. The Non-Newtonian dispensing material (generally epoxy) only further deteriorates the case. Traditionally the dispensing process is based on the open-loop control, which heavily relies on the field engineer's experiences. Therefore, its performance is limited. For dispensing process, no systematic effort has been made and a desirable process model is not available. It's imperative to have an accurate control of the fluid volume dispensed in the AICE process.Based on the current packaging practice and aimed for the dispensing process performance improving, it is the purpose of this thesis to realize the accurate fluid manipulation in AICE process by using a modeling and control approach.A survey of the AICE processes and the current fluid dispensing technology is first given. The main workflow and development of the packaging process is introduced. It is noteworthy that process characteristics and some key technology are summarized. A comparative study of the fluid dispensing technology is also made. The advantage and disadvantage of the different dispensing technology and their application area are presented. In addition, some important factors influencing the dispensing quality are listed. Secondly, a physically accurate and mathematically simple process model is derived for the dispensing process. By taking into account nonlinear flow passing through the valve, syringe chamber dynamics, a pneumatic system model is first provided. For the Non-Newtonian fluid, Finite Element Method (FEM) is used to have an accurate description of the fluid flow details and this result can be used for further reduced-order modeling. Finally an analytic modeling approach is taken to solve the nonlinear partial differential equation that describes the fluid motion. By using the Spectral Method and a suitable basis function, an approximate fluid flow model for both the Newtonian and non-Newtonian fluid can be developed. Without significant accuracy loss, this approximate model is much simpler than the FEM one. By integrating the pneumatic and fluid model together, a simple but effective model is derived to describe the whole dispensing system in a reasonable accuracy.Then based on the derived process model a Run-to-Run control method is designed for the dispensing process. With the purpose of controller design, an ANFIS fuzzy modeling technique is used to simplify the process model and a control-oriented piecewise linear input-output model is deduced. An EWMA control architecture for the dispensing process can be easily designed with this linear model. The stability region of the control method is given. The controller parameter can be updated by the online estimation of the physical process. Simulation and experiment verify the proposed control method.Finally, to implement the proposed Run-to-Run control method on PC, some practical issues are addressed. After comparing the usual real time control implementation scheme, a high performance timer is constructed by utilizing the available dispensing hardware resources. The whole control software architecture is also presented as well as some application concerns. The final experiment verified the control method's effectiveness.