In-situ backplane inspection of fiber optic ferrules

• Main Author: Wilson, Andrew Kirk, 1977-
• Other Authors: Samir Nayfeh.
• Format: Others
• Language: English
• Published: Massachusetts Institute of Technology 2008
• Subjects: Mechanical Engineering.
• Online Access: http://hdl.handle.net/1721.1/35625

Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Mechanical Engineering, 2006. === Includes bibliographical references (p. 193-200). === The next generation of supercomputers, routers, and switches are envisioned to have hundreds and thousands of optical interconnects among components. An optical interconnect attains a bandwidth-distance product as high as 90 GHz.km, about 200 times higher than can be attained by a copper interconnect. But defects (such as dust or scratches) as small as 1 micron on the connector endfaces can seriously degrade performance. Therefore, for every mate and de-mate, optical connectors must be inspected to ensure high performance data transmission capabilities. The tedious and time consuming task of manually inspecting each connector is one of the barriers to adoption of optics in the backplanes of large card-based machines. This thesis provides a framework and method for in-situ automatic inspection of backplane optical connectors. We develop an inspection system that fits into the envelope of a single daughter card, moves a custom microscope objective in three degrees of freedom to image the connector endfaces, and detects and classifies defects with major diameter of one micron or larger. === The inspection machine mounts to the backplane in the same manner as a daughter card, and positions the microscope with better than 0.2 micron resolution and 15 micron repeatability in three degrees of freedom. Despite tight packaging constraints, the ultra-long working distance custom microscope objective attains 1 micron Rayleigh resolution via deconvolution. Several images taken at different exposures and focus settings are fused to extend the imaging sensor's limited dynamic range and depth of field. A set of machine-vision algorithms are developed to process the resulting image and detect and classify the fiber core, cladding and their defects. === by Andrew K. Wilson. === Ph.D.