Design And Error Compensation Of Multi-DOF Flip-Chip Bonding Mechanisms

With the development of high density IC package, flip chip technology is applied more and more widely as it has many advantages such as higher package density, more excellent electric and heat dissipation performance. Flip chip bonders (FCBs), which can accomplish high density package, are all imported from abroad at present. To break market monopolization and technological block, it has significant meaning to research the key technology for developing a flip chip bonding mechanism (FCBM). This dissertation focuses on how to improve the placement accuracy of a FCBM. To achieve this aim, mechanism type synthesis, dimension synthesis, precision design and error compensation are investigated thoroughly and systemically. This dissertation covers whole product development cycle from design, assembly to test and provides theory foundation to develop an accurate FCBM with high quality.Mechanism type synthesis means establishing proper mechanism according to kinematic rule and attribute of technological manipulation during conceptual design. A FCBM must be with six DOFs at least to accomplish pick-and-place and leveling manipulation and can divide into two parts, namely flip chip attachment mechanism (FCAM) and leveling mechanism (LM). Typical structure of the FCAM is gantry type. The LM can adjust orientation of a chip and must be able to carry large load, respond rapidly and move accurately to satisfy the requirement of high density package. Parallel mechanism (PM) can deal with these problems roundly. This dissertation proposes a method for type synthesis of a2-DOF rotational decoupled PM based on constraint screw theory. Among these types, there are two fitted for LM. Taking one type as an example, mobility analysis, the selection of inputs and kinematics simulation are dealt with in sequence. The result shows this2-DOF rotational decoupled LM perfectly meets the design requirement.After conceptual design is accomplished through type synthesis, optimization of dimensional parameters should be carried out to improve performance indices, such as stiffness, workspace, dexterity, etc. This optimization process is just dimensional synthesis. Dimensional parameters, definition equations of performance indices and position solutions of kinematics are regarded as design variables, objective functions and constrained conditions that are the three key factors in optimization problem, respectively. Considering it is a rotational decoupled mechanism, LM can divide into two planar four-bar mechanisms. Direct and inverse closed-form kinematic solutions are deduced and two indices named as workspace and transmission angle are defined for dimensional synthesis. Performance atlases are figured to illustrate the relationship between normalized dimensional parameters and performance indices. The global optimum dimensional parameters are located in the intersection region of performance atlases represent different indices.As it is an important index to evaluate the performance of a FCB, precision must be improved by varieties of ways. Precision design and kinematic calibration are the two main methods to improve precision. Jacobian matrix is the foundation for evaluating precision of a machine and can be derived by velocity analysis, which is a key point of precision design. Virtual mechanism method is generally used to derive the first order kinematic influence coefficient matrix, namely Jacobian matrix, of PM with DOF fewer than six. As the LM is one of this kind of mechanisms, some pose errors of it can no longer be compensated by kinematic calibration. Therefore, a deep insight into error model, which is the foundation of precision design, must be gained to solve this problem. The error model, which is established based on perturbation method, clarifies the mapping relationship between the pose errors and geometric error sources, such as the dimensional errors of links, the assembly errors of joints and the deviation of inputs. Sensitivity analysis can reveal the error sources that have a great influence on pose errors. These error sources must be eliminated or at least be minimized in the design, manufacturing and assembly processes. Therefore, precision of the LM can be improved by paying least cost.Placement accuracy is the most important performance index of FCAM. If placement accuracy already reaches a certain level, the method for improving placement accuracy by improving the precision of manufacturing and assembly causes the cost exponential growth. Error compensation can eliminate or minimize the unfavorable influence caused by error sources on placement accuracy by additional movement under the condition of unchanged cost. Therefore, the research and application of error compensation gains wide attention by both scholars and engineers. Geometry error is the main error source of FCAM and the error items can be treated as constant supposing that all links are rigid. The main difference between FCAM and regular X-Y-Z-θ4-axis NC machine tool is whether machine vision system (MVS) exists. The MVS can provide a new method for error compensation. When FCAM implements pick-and-place manipulation, the two inside MVSs will locate the position of chip and substrate respectively. Based on this characteristic, FCAM can divide into two parts, namely θ-axis mechanism and X-Y-Z3-axis mechanism, to establish error model. The parameters of error model can be identified based on the rotation trajectory of θ-axis and the positioning data of a fixed cross mark measured by a moved MVS. After parameter identification, the error compensation equations can be derived which clarify the relationship between the position deviation of chip and substrate and the compensation displacement.At last, the proposed FCBM has been implemented in an experiment platform and a FCAM of a RFID inlay assembly line is calibrated with the error compensation method proposed in this dissertation. To prove the validity of this method, a pick-and-place experiment is carried out with the FCAM. After the positions of chips on substrate are measured, the placement accuracy and CPK index of FCAM are derived based on the measurement data. Result shows that the error compensation method is quite effective.