Optical mesh and hypercube perfect shuffle by space invariant multiple imaging

Abstract

Incoherent space invariant multiple imaging is used in a class of systems for free-space interconnections. They are classical imaging systems using mirror array, off-axis lens array, beam splitter, or holograms for multiple imaging. Their point spread function is a number of points shifted by tilting the mirrors. These systems can implement many types of regular interconnection. In mesh networks, the input nodes in a 2-D array have the same interconnection pattern. They are implemented by multiple imaging with the number of images equal to the fan out number. In the hypercube and perfect shuffle networks, the input nodes in different locations in a 2-D array require different interconnection patterns. They can also be implemented in space invariant systems. Only the number of multiple images has to be larger than the fan out number. This introduces a loss of power. We used eight mirrors for four-cube network and four mirrors for 2-D perfect shuffle. The space invariant systems can have a very large number of input nodes, high spatial density of interconnects, and potentially a very large fan out number for massive parallel interconnections.