Off-campus UMass Amherst users: To download campus access dissertations, please use the following link to log into our proxy server with your UMass Amherst user name and password.
Non-UMass Amherst users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Dissertations that have an embargo placed on them will not be available to anyone until the embargo expires.
Open Access Dissertation
Doctor of Philosophy (PhD)
Electrical and Computer Engineering
Year Degree Awarded
Month Degree Awarded
Dr. Dev Gupta
Dr. Joshua Yang
Dr. Patrick Kelly
Dr. Blair Perot
Electronic Devices and Semiconductor Manufacturing | Hardware Systems | Signal Processing | VLSI and Circuits, Embedded and Hardware Systems
Emerging applications in the field of machine vision, deep learning and scientific simulation require high computational speed and are run on platforms that are size, weight and power constrained. With the transistor scaling coming to an end, existing digital hardware architectures will not be able to meet these ever-increasing demands. Analog computation with its rich set of primitives and inherent parallel architecture can be faster, more efficient and compact for some of these applications. The major contribution of this work is to show that analog processing can be a viable solution to this problem. This is demonstrated in the three parts of the dissertation.
In the first part of the dissertation, we demonstrate that analog processing can be used to solve the problem of stereo correspondence. Novel modifications to the algorithms are proposed which improves the computational speed and makes them efficiently implementable in analog hardware. The analog domain implementation provides further speedup in computation and has lower power consumption than a digital implementation.
In the second part of the dissertation, a prototype of an analog processor was developed using commercially available off-the-shelf components. The focus was on providing experimental results that demonstrate functionality and to show that the performance of the prototype for low-level and mid-level image processing tasks is equivalent to a digital implementation. To demonstrate improvement in speed and power consumption, an integrated circuit design of the analog processor was proposed, and it was shown that such an analog processor would be faster than state-of-the-art digital and other analog processors.
In the third part of the dissertation, a memristor-CMOS analog co-processor that can perform floating point vector matrix multiplication (VMM) is proposed. VMM computation underlies some of the major applications. To demonstrate the working of the analog co-processor at a system level, a new tool called PSpice Systems Option is used. It is shown that the analog co-processor has a superior performance when compared to the projected performances of digital and analog processors. Using the new tool, various application simulations for image processing and solution to partial differential equations are performed on the co-processor model.
Athreyas, Nihar, "ANALOG SIGNAL PROCESSING SOLUTIONS AND DESIGN OF MEMRISTOR-CMOS ANALOG CO-PROCESSOR FOR ACCELERATION OF HIGH-PERFORMANCE COMPUTING APPLICATIONS" (2018). Doctoral Dissertations. 1215.
Available for download on Saturday, May 11, 2019