Design and demonstration of integrated micro-electro-mechanical relay circuits for VLSI applications

Abstract

Complementary-Metal-Oxide-Semiconductor (CMOS) feature size scaling has resulted in significant improvements in the performance and energy efficiency of integrated circuits in the past 4 decades. However, in the last decade and for technology nodes below 90 nm, the scaling of threshold and supply voltages has slowed, as a result of subthreshold leakage, and power density has increased with each new technology node. This has forced a move toward multi-core architectures, but the energy efficiency benefits of parallelism are limited by the sub-thresahold leakage and the minimum energy point for a given function. Avoiding this roadblock requires an alternative device with more ideal switching characteristics. One promising class of such devices is the electro-statically actuated micro-electro-mechanical (MEM) relay which offers zero leakage current and abrupt turn-on behavior. Although a MEM relay is inherently slower than a CMOS transistor due to the mechanical movement, we have developed circuit design methodologies to mitigate this problem at the system level. This thesis explores such design optimization techniques and investigates the viability of MEM relays as an alternative switching technology for very-large scale integration (VLSI) applications. In the first part of this thesis, the feasibility of MEM relays for power management applications is discussed. Due to their negligibly low leakage, in certain applications, chips utilizing power gates built with MEM relays can achieve lower total energy than those built with CMOS transistors. A simple comparative analysis is presented and provides design guidelines and energy savings estimates as a function of technology parameters, and quantifies the further benefits of scaled relay designs. We also demonstrate a relay chip successfully power-gating a CMOS chip, and show a relay-based pulse generator suitable for self-timed operation. Going beyond power-gating applications, this work also describes circuit techniques and trade-offs for logic design with MEM-relays, focusing on multipliers which are commonly known as the most complex arithmetic units in a digital system. These techniques leverage the large disparity between mechanical and electrical time-constants of a relay, partitioning the logic into large, complex gates to minimize the effect of mechanical delay and improve circuit performance. At the component design level, innovations in compressor unit design minimize the required number of relays for each block and facilitate component cascading with no delay penalty. We analyze the area/energy/delay trade-offs vs. CMOS designs, for typical bit-widths, and show that scaled relays offer 10-20x lower energy per operation for moderate throughputs (<10-100MOPS). In addition to this analysis, we demonstrate the functionality of some of the most complex MEM relay circuits reported to date. Finally, considering the importance of signal generation and transmission in VLSI systems, this thesis presents MEM relay-based I/O units, focusing on design and demonstration of digital to analog converters (DAC). It also explores the concept of faster-than-mechanical-delay signal transmission.