| New methods to implement optical digital computation elements and processors are studied in this thesis. Various ultrafast nonlinear optical effects, such as Kerr, phase conjugation, second harmonic generation processes are used for optical computation. The ultrafast Sagnac interferometric logic switches which have a better mechanical stability over other two beam interferometric switching geometries are described. Using a variety of these switches, various binary and multiple-valued optical logic elements is detailed. As an alternative optical logic implementation technique, an optical phase conjugation pattern logic method is proposed and demonstrated. The major advantage of this type of optical logic elements together with a transparent/opaque logic code is it can perform all sixteen two-input Boolean logic functions. Using a polarization encoding methods, optical implementation of multiple-input binary as well as multiple-valued logic operations are also studied.;In this thesis, the implementation of ultrafast optical interconnection devices, i.e. an optical sampler, an A/D converter, a dynamic cross-bar, as well as various generalized perfect shuffles, is also described. Most of the proposed schemes uses a parallel processing geometry. Using a Sagnac interferometric sampler, the sampling frequency can be doubled with respect to other existing sampling geometry. With the theta-modulation-based optical A/D converter, a fast (subnanosecond), compact and flexible A/D conversion can be performed. Compared to other existing parallel data shuffling geometries, the implementation methods described in this thesis are more compact and efficient. Using the proposed phase-conjugation-based dynamic cross-bar, the parallel interconnection speed has been increased from microsecond to picosecond region.;In this thesis, for the optical numerical computation, various binary and non-binary computing structures are also investigated. In the binary case, using a number of cascaded Sagnac interferometers, optical implementation of a full adder and an array multiplier are studied. Using a combination of the digital multiplication via analog convolution algorithm and the ultrafast noncollinear second harmonic generation effect, a number of ultrafast data convolution pre-processing techniques are proposed. The proposed methods are generally suitable for ultrafast pre-processing involving scalar, vector and matrix multiplication operations. In the non-binary optical numerical processing case, a number of new processing methods are also presented. Using the Sagnac interferometer-based approach, the optical binary coded ternary adders are implemented. Based on residue arithmetic, using optical second harmonic generators, a number of ultrafast optical residue computing structures are described. Finally, a new approach to implement a content-addressable memory-based modified sign-digit addition and subtraction elements are detailed. |
