Nano anahtarlamalı dizinler için güvenilirlik ve hesaplama teknikleri

Lithographic top-down based production of integrated circuits are approaching the limits in a manner of both feasibility and commercial aspects. In spite of the fact that, Moore's Law keeps holding, emerging technologies need to be considered. Crossbar based nano switching arrays are shown to be a likely candidate to overcome shortcomings of current CMOS based paradigm or coexist as a complementary instrument. Abundant research papers in literature help to support this claim. Nano-arrays are produced with placing a group of nanowires aligned parallel to each other on another group of nanowires orthogonally. Crosspoints present between top and bottom nanowires act as a switching device. According to the preference, switches might show resistor, diode or FET like characteristics. Computing with nano-arrays are similar to the Programmable Logic Arrays (PLA). Every switch can be appointed to the corresponding logic element found in the boolean function which is realized with the crossbar in question. Nevertheless, the nature of nano-fabrication contains random elements and devices obtained from the process are prone to have faulty components. As a result, realization of target logic functions with nano-arrays differ from PLA due to the number of considerable faulty components. Since discarding faulty devices would not be practical and sustainable, fault tolerance and reliability of crossbar based nano switching arrays are extensively studied in this thesis. Most common faults occur in described switches can be categorized under two main titles which are permanent and transient. Also, two categories have subtitles such as stuck-open, stuck-closed and nanowire break-downs. Because of the immense effect of nanowire break-downs, they are excluded from the body of study. Permanent faults are taken into account by independently assigning stuck-open and stuck-closed defect probabilities into crosspoints. After obtaining defective array, following step is determining whether there is a valid mapping of a given logic function on defective array. In the presence of permanent faults, a heuristic algorithm using index sorting, backtracking and matrix multiplication techniques is proposed. The algorithm's effectiveness is demonstrated on standard benchmark circuits that shows 99\% accuracy in accordance with the results of an exhaustive search algorithm. Runtime and success rate of algorithm is presented with experimental results of simulation using standard industry benchmark circuits. In the presence of transient faults, tolerance analysis is performed by recursively constructing equivalent sets of implemented logic functions. It is demonstrated that transient faults causing OFF-to-ON state changes in crosspoints do not necessarily cause the array to produce an incorrect output; they can be discarded. Difference between the assumed and the actual fault tolerance performances, which is obtained with the proposed algebraic method, is presented with standard benchmark circuits for several fault rates.