טכניון מכון טכנולוגי לישראל
הטכניון מכון טכנולוגי לישראל - בית הספר ללימודי מוסמכים  
Ph.D Thesis
Ph.D StudentGreenberg Maxim
SubjectMultimode Optical Interconnects
DepartmentDepartment of Electrical Engineering
Supervisors Professor Moshe Nazarathy
Professor Meir Orenstein
Full Thesis textFull thesis text - English Version


Abstract

While optical transmission is mainly used for long and medium-haul communications,   it has recently emerged as the preferred means for short-reach data interconnects. Current research in optical communication schemes has targeted previously unutilized degrees of freedom, namely the multiplicity of modes of certain optical channels, such as Multi Mode Fiber (MMF). Potential applications to Very Short Reach Optical Interconnects (VSROI) and Local Area Networks (LAN) are being considered. Novel approaches are Modal Division Multiplexing (MDM), whereby each high-order fiber propagation mode serves as separate effective channel, and Multiple-Input-Multiple-Output (MIMO) transmission, whereby each of the multiple data channels is distributed over the multiple modes, followed by appropriate post-detection processing for effective decoupling of the data channels.

In the framework of the MDM approach, we have conceptualized and modeled a novel type of Modal Add/Drop Multiplexer device, based on adiabatic transmission, capable of uploading/downloading of particular channel (carried by certain high order mode) into/from the optical bus.

Data parallelization by means of optical MIMO transmission over dispersive MMF with a high degree of modal coupling was investigated by developing an analytical model for direct detection MMF MIMO transmission with mutually incoherent sources. The MIMO channel performance was derived in terms of a new formulation of a MIMO Channel Matrix for modal group powers accounting for modal coupling.

We have further extended the MIMO approach to operate with “binary” detection, in lieu of the more complex analog-to-digital-conversion-based soft conversion. The proposed system provides significant simplification of the Maximal Likelihood (ML) detector, while only slightly degrading the performance. In conjunction with this scheme, we have investigated several coding approaches: input spatial coding for improving the Bit Error Rate (BER) performance, and hierarchical spatial coding with list decoding for reducing ML complexity.

Finally we investigated a novel SIMO transmission scheme over multimode fiber, mitigating intermodal dispersion by combining coherent phase shift keying transmission with direct detection. Statistical characterization reveals the influence on the statistical properties of the electrical charge generated by each detector of factors such as detector type, dimension and offset position, ultimately determining overall system performance. In the frequency-selective case we identified additional temporal degrees of freedom resulting from non-overlapping time pulses further contributing to the decision statistics.