Date of Award

1-1-2004

Degree Type

Thesis

Degree Name

Master of Engineering Science

School

School of Engineering and Mathematics

Faculty

Faculty of Computing, Health and Science

First Advisor

Dr Stefan Lachowicz

Abstract

The invention of the transistor in 1947 at Bell Laboratories revolutionised the electronics industry and created a powerful platform for emergence of new industries. The quest to increase the number of devices per chip over the last four decades has resulted in rapid transition from Small-Scale-Integration (SSI) and Large-Scale-lntegration (LSI), through to the Very-Large-Scale-Integration (VLSI) technologies, incorporating approximately 10 to 100 million devices per chip. The next phase in this evolution is the Ultra-Large-Scale-Integration (ULSI) aiming to realise new application domains currently not accessible to CMOS technology. Although technology is continuously evolving to produce smaller systems with minimised power dissipation, the IC industry is facing major challenges due to constraints on power density (W/cm2) and high dynamic (operating) and static (standby) power dissipation. Mobile multimedia communication and optical based technologies have rapidly become a significant area of research and development challenging a variety of technological fronts. The future emergence or 4G (4th Generation) wireless communications networks is further driving this development, requiring increasing levels of media rich content. The processing requirements for capture, conversion, compression, decompression, enhancement and display of higher quality multimedia, place heavy demands on current ULSI systems. This is also apparent for mobile applications and intelligent optical networks where silicon chip area and power dissipation become primary considerations. In addition to the requirements for very low power, compact size and real-time processing, the rapidly evolving nature of telecommunication networks means that flexible soft programmable systems capable of adaptation to support a number of different standards and/or roles become highly desirable. In order to fully realise the capabilities promised by the 4G and supporting intelligent networks, new enabling technologies arc needed to facilitate the next generation of personal communications devices. Most of the current solutions to meet these challenges are based on various implementations of conventional architectures. For decades, silicon has been the main platform of computing, however it is slow, bulky, runs too hot, and is too expensive. Thus, new approaches to architectures, driving multimedia and future telecommunications systems, are needed in order to extend the life cycle of silicon technology. The emergence of Truly Deep Submicron Technology (TDST) and related 3-D interconnection technologies have provided potential alternatives from conventional architectures to 3-D system solutions, through integration of IDST, Vertical Software Mapping and Intelligent Interconnect Technology (IIT). The concept of Soft-Chip Technology (SCT) entails integration of "Soft• Processing Circuits" with "Soft-Configurable Circuits". This concept can effectively manipulate hardware primitives through vertical integration of control and data. Thus the notion of 3-D Soft-Chip emerges as a new design algorithm for content-rich multimedia, telecommunication and intelligent networking system applications. 3•D architectures (design algorithms used suitable for 3-D soft-chip technology), are driven by three factors. The first is development of new device technology (TDST) that can support new architectures with complexities of 100M to 1000M devices. The second is development of advanced wafer bonding techniques such as Indium bump and the more futuristic optical interconnects for 3-D soft-chip mapping. The third is related to improving the performance of silicon CMOS systems as devices continue to scale down in dimensions. One of the fundamental building blocks of any computer system is the arithmetic component. Optimum performance of the system is determined by the efficiency of each individual component, as well as the network as a whole entity. Development of configurable arithmetic primitives is the fundamental focus in 3-D architecture design where functionality can be implemented through soft configurable hardware elements. Therefore the ability to improve the performance capability of a system is of crucial importance for a successful design. Important factors that predict the efficiency of such arithmetic components are: • The propagation delay of the circuit, caused by the gate, diffusion and wire capacitances within !he circuit, minimised through transistor sizing. and • Power dissipation, which is generally based on node transition activity. [2] Although optimum performance of 3-D soft-chip systems is primarily established by the choice of basic primitives such as adders and multipliers, the interconnecting network also has significant degree of influence on !he efficiency of the system. 3-D superposition of devices can decrease interconnect delays by up to 60% compared to a similar planar architecture. This research is based on development and implementation of configurable arithmetic primitives, suitable to the 3-D architecture, and has these foci: • To develop a variety of arithmetic components such as adders and multipliers with particular emphasis on minimum area and compatible with 3-D soft-chip design paradigm. • To explore implementation of configurable distributed primitives for arithmetic processing. This entails optimisation of basic primitives, and using them as part of array processing. In this research the detailed designs of configurable arithmetic primitives are implemented using TDST O.l3µm (130nm) technology, utilising CAD software such as Mentor Graphics and Cadence in Custom design mode, carrying through design, simulation and verification steps.

Share

 
COinS