Semi-distributed manipulation and microassembly of optical fibers

A new manipulation paradigm has been sought to shift traditional sequential assembly processing to parallel processing. Simultaneous manipulation of multiple parts will make assembly processing faster yielding higher throughput. Flexible manipulation often requires individual positioning during parallel processing. In addition, manipulation faces more challenges when it deals with small parts which are usually difficult to grasp. Applications include handling small parts which are used in electronics, photonics, MEMS, etc. The paradigm shift is expected to make improvement toward higher throughput manufacturing.; This thesis studies a new manipulation method entitled “Semi-Distributed Manipulation (SDM)”, proposed for parallel manipulation. The new method achieves distributed manipulation by centralized actuation and creates less control load compared to previous prototypes aimed for distributed manipulation. Design for reduced control load will be a critical advantage when the method is extended to massively parallel manipulation. The method is implemented on a testbed, “Piezo Active Surface (PAS)” on which small planar parts, such as silicon dies, are translated and rotated. Several different manipulation modes are available by supplying different inputs. Manipulation mechanisms underlying those different modes are analyzed, and strategies to achieve desired positioning goal are developed. Experimental results show that parallel manipulation of multiple parts are possible also address coarse and fine motion capability in one single device.; As a microassembly application, fiber coupling to photonics components is explained. Fiber coupling is a ubiquitous process in photonics devices, and automating the process deals with either active fiber alignment or passive assembly. Active fiber alignment is treated as a geometric misalignment problem where transmitted light power is used as alignment feedback. Since the feedback power measurement does not carry enough information for multi-axes alignment, new methods developed here use power loss model and relative movement of the fibers. With these methods, alignment can be achieved with smaller number of measurements or can be constantly corrected in a disturbance situation. Compared to active alignment, passive fiber assembly relies its alignment on geometric constraints instead of feedback power. Relatively high throughput can be achieved but tolerances are less. An experimental automation setup has been constructed for the passive fiber assembly, and the results are favorable.