Author

John A. Frye

Abstract

Transistor technology scaling has been the driving force in improving the size, speed, and power consumption of digital systems. As devices approach atomic size, however, their reliability and performance are increasingly compromised due to reduced noise margins, difficulties in fabrication, and emergent nano-scale phenomena. Scaled CMOS devices, in particular, suffer from process variations such as random dopant fluctuation (RDF) and line edge roughness (LER), transistor degradation mechanisms such as negative-bias temperature instability (NBTI) and hot-carrier injection (HCI), and increased sensitivity to single event upsets (SEUs). Consequently, future devices may exhibit reduced performance, diminished lifetimes, and poor reliability.

This research proposes a variation and fault tolerant architecture, the inSense architecture, as a circuit-level solution to the problems induced by the aforementioned phenomena. The inSense architecture entails augmenting circuits with introspective and sensory capabilities which are able to dynamically detect and compensate for process variations, transistor degradation, and soft errors. This approach creates ``smart'' circuits able to function despite the use of unreliable devices and is applicable to current CMOS technology as well as next-generation devices using new materials and structures. Furthermore, this work presents an automated prototype implementation of the inSense architecture targeted to CMOS devices and is evaluated via implementation in ISCAS '85 benchmark circuits. The automated prototype implementation is functionally verified and characterized: it is found that error detection capability (with error windows from $\approx$30-400ps) can be added for less than 2\% area overhead for circuits of non-trivial complexity. Single event transient (SET) detection capability (configurable with target set-points) is found to be functional, although it generally tracks the standard DMR implementation with respect to overheads.

Library of Congress Subject Headings

Metal oxide semiconductors, Complementary--Design and construction; Metal oxide semiconductors, Complementary--Reliability; Fault tolerance (Engineering)

Publication Date

9-2015

Document Type

Thesis

Student Type

Graduate

Degree Name

Computer Engineering (MS)

Department, Program, or Center

Computer Engineering (KGCOE)

Advisor

Dhireesha Kudithipudi

Advisor/Committee Member

Sonia Lopez Alarcon

Advisor/Committee Member

James Moon

Comments

Physical copy available from RIT's Wallace Library at TK7871.99.M44 F79 2015

Campus

RIT – Main Campus

Share

COinS