Our burgeoning appetite for data relentlessly demands exponential scaling of computing and communications resources leading to an overbearing and ever-present drive to improve efficiency while reducing on-chip area even as photonic components expand to fill application spaces no longer satisfied by their electronic counterparts. With a high index contrast, low optical loss, and compatibility with the CMOS fabrication infrastructure, silicon-on-insulator technology delivers a mechanism by which efficient, sub-micron waveguides can be fabricated while enabling monolithic integration of photonic components and their associated electronic infrastructure. The result is a solution leveraging the superior bandwidth of optical signaling on a platform capable of delivering the optical analogue to Moore's Law scaling of transistor density. Device size is expected to end Moore's Law scaling in photonics as Maxwell's equations limit the extent to which this parameter may be reduced. The focus of the work presented here surrounds photonic device miniaturization and the development of 3D optical interconnects as approaches to optimize performance in densely integrated optical interconnects. In this dissertation, several technological barriers inhibiting widespread adoption of photonics in data communications and telecommunications are explored. First, examination of loss and crosstalk performance in silicon nitride over SOI waveguide crossings yields insight into the feasibility of 3D optical interconnects with the first experimental analysis of such a structure presented herein. A novel measurement platform utilizing a modified racetrack resonator is then presented enabling extraction of insertion loss data for highly efficient structures while requiring minimal on-chip area. Finally, pioneering work in understanding the statistical nature of doublet formation in microphotonic resonators is delivered with the resulting impact on resonant device design detailed.

FPGA clock frequencies are slow enough that only a fraction of the interconnect’s bandwidth is used. By exploiting this bandwidth, the transfer of large amounts of data can be greatly accelerated. Alternatively, it may also be possible to save area on fixed-bandwidth links by using on-chip serial signaling. For datapath-intensive designs which operate on words instead of bits, this can reduce wiring congestion as well. This thesis proposes relatively simple circuit-level modifications to FPGA interconnect to enable high-bandwidth communication. High-level area estimates indicate a potential interconnect area savings of 10 to 60% when serial links are used. Two interconnect pipelining techniques, wave pipelining and surfing, are adapted to FPGAs and compared against each other and against regular FPGA interconnect in terms of throughput, reliability, area, power, and latency. Source-synchronous signaling is used to achieve high data rates with simple receiver design. Statistical models for high-frequency power supply noise are developed and used to estimate the probability of error of wave pipelined and surfing links as a function of link length and operating speed. Surfing is generally found to be more reliable and less sensitive to noise than wave pipelining. Simulation results in a 65nm process demonstrate a throughput of 3Gbps per wire across a 50-stage, 25mm link.

With the advances in process technology, comes the domination of interconnect in the overall propagation delay in modern VLSI designs. Hence, interconnect synthesis techniques, such as buffer insertion, wire sizing and layer assignment play critical roles in the successful timing closure for EDA tools. In this thesis, while our aim is to satisfy timing constraints, accounting for the overhead caused by these optimization techniques is of another primary concern. We utilized a Lagrangian relaxation method to minimize the usage of buffers and metal resources to meet the timing constraints. Compared with the previous work that extended traditional Van Ginneken’s algorithm, which allows for bumping up the wire from thin to thick given significant delay improvement, our approach achieved around 25% reduction in buffer + wire capacitance under the same timing budget.

No description available.

Switching device power losses place critical limits on the design and performance of high-frequency integrated DC-DC converters. Especially, the layout of metal interconnects in lateral power MOSFETs has a profound effect on their on-resistances and conduction power losses. This thesis presents an analytical interconnect modeling and layout optimization technique for large-area power MOSFETs. The layout optimization of 24V LDMOS transistors in the area of 1 mm2 has achieved an improvement of 55 % in its on-resistance. The simulation result has been verified by experimental measurements on a test chip fabricated in TSMC 0.25 µm HV CMOS technology. In addition, this thesis presents an optimized output stage design methodology for the implementation of a 4 MHz, 12V to 1V integrated DC-DC converter. A segmented output stage scheme is employed to increase the converter efficiency at light load conditions. The peak efficiency of 84% was achieved at load current of 2 A.

Switching device power losses place critical limits on the design and performance of high-frequency integrated DC-DC converters. Especially, the layout of metal interconnects in lateral power MOSFETs has a profound effect on their on-resistances and conduction power losses. This thesis presents an analytical interconnect modeling and layout optimization technique for large-area power MOSFETs. The layout optimization of 24V LDMOS transistors in the area of 1 mm2 has achieved an improvement of 55 % in its on-resistance. The simulation result has been verified by experimental measurements on a test chip fabricated in TSMC 0.25 µm HV CMOS technology. In addition, this thesis presents an optimized output stage design methodology for the implementation of a 4 MHz, 12V to 1V integrated DC-DC converter. A segmented output stage scheme is employed to increase the converter efficiency at light load conditions. The peak efficiency of 84% was achieved at load current of 2 A.

L’évolution tendant à réduire la taille et augmenter la complexité des circuits électroniques modernes, est en train de ralentir du fait des limitations technologiques, qui génèrent beaucoup de d’imperfections et de defaults durant la fabrication ou la durée de vie de la puce. Les FPGAs sont utilisés dans les systèmes numériques complexes, essentiellement parce qu’ils sont reconfigurables et rapide à commercialiser. Pour garder une grande fiabilité de tels systèmes, les FPGAs doivent être testés minutieusement pour les defaults. L’optimisation de l’architecture des FPGAs pour l’économie de surface et une meilleure routabilité est un processus continue qui impacte directement la testabilité globale et de ce fait, la fiabilité. Cette thèse présente une stratégie complète pour le test et le diagnostique des defaults de fabrication des “mesh-based FPGA” contenant une nouvelle topologie d’interconnections à plusieurs niveaux, ce qui promet d’apporter une meilleure routabilité. Efficacité des schémas proposes est analysée en termes de temps de test, couverture de faute et résolution de diagnostique. / The evolution trend of shrinking feature size and increasing complexity in modern electronics is being slowed down due to physical limits that generate numerous imperfections and defects during fabrication steps or projected life time of the chip. Field Programmable Gate Arrays (FPGAs) are used in complex digital systems mainly due to their reconfigurability and shorter time-to-market. To maintain a high reliability of such systems, FPGAs should be tested thoroughly for defects. FPGA architecture optimization for area saving and better signal routability is an ongoing process which directly impacts the overall FPGA testability, hence the reliability. This thesis presents a complete strategy for test and diagnosis of manufacturing defects in mesh-based FPGAs containing a novel multilevel interconnects topology which promises to provide better area and routability. Efficiency of the proposed test schemes is analyzed in terms of test cost, respective fault coverage and diagnostic resolution.

FPGA clock frequencies are slow enough that only a fraction of the interconnect’s bandwidth is used. By exploiting this bandwidth, the transfer of large amounts of data can be greatly accelerated. Alternatively, it may also be possible to save area on fixed-bandwidth links by using on-chip serial signaling. For datapath-intensive designs which operate on words instead of bits, this can reduce wiring congestion as well. This thesis proposes relatively simple circuit-level modifications to FPGA interconnect to enable high-bandwidth communication. High-level area estimates indicate a potential interconnect area savings of 10 to 60% when serial links are used. Two interconnect pipelining techniques, wave pipelining and surfing, are adapted to FPGAs and compared against each other and against regular FPGA interconnect in terms of throughput, reliability, area, power, and latency. Source-synchronous signaling is used to achieve high data rates with simple receiver design. Statistical models for high-frequency power supply noise are developed and used to estimate the probability of error of wave pipelined and surfing links as a function of link length and operating speed. Surfing is generally found to be more reliable and less sensitive to noise than wave pipelining. Simulation results in a 65nm process demonstrate a throughput of 3Gbps per wire across a 50-stage, 25mm link. / Applied Science, Faculty of / Electrical and Computer Engineering, Department of / Graduate

With the rapid growth in consumer electronics, people expect thin, smart and powerful devices, e.g. Google Glass and other wearable devices. However, as portable electronic products become smaller, energy consumption becomes an issue that limits the development of portable systems due to battery lifetime. In general, simply reducing device size cannot fully address the energy issue. To tackle this problem, we propose an on-chip interconnect infrastructure and pro- gram storage structure for a coarse-grained reconfigurable architecture (CGRA) with emerging non-volatile embedded memory (MRAM). The interconnect is composed of a matrix of time-multiplexed switchboxes which can be dynamically reconfigured with the goal of energy reduction. The number of processors performing computation can also be adapted. The use of MRAM provides access to high-density storage and lower memory energy consumption versus more standard SRAM technologies. The combination of CGRA, MRAM, and flexible on-chip interconnection is considered for signal processing. This application domain is of interest based on its time-varying computing demands. To evaluate CGRA architectural features, prototype architectures have been pro- totyped in a field-programmable gate array (FPGA). Measurements of energy, power, instruction count, and execution time performance are considered for a scalable num- ber of processors. Applications such as adaptive Viterbi decoding and Reed Solomon coding are used for evaluation. To complete this thesis, a time-scheduled switchbox was integrated into our CGRA model. This model was prototyped on an FPGA. It is shown that energy consumption can be reduced by about 30% if dynamic design reconfiguration is performed.

Global on-chip interconnects in deep sub-micron CMOS present challenges in satisfying delay constraints in the presence of noise and dramatic temperature variations, while minimizing energy consumption due to leakage and static power. Although repeaters are typically used to reduce delay and maintain signal integrity in long interconnects, they introduce significant area, power (both dynamic and leakage), delay, noise and design overhead as well as exacerbating variations due to their local power supply noise and temperature. Current-Sensing is an alternative to repeaters that transfers signals with no intermediate circuits by sensing current rather than voltage at the end of a long interconnect. Among the current sensing circuits, Differential Current-Sensing (DCS), which uses conventional CMOS inverters to drive differential signal, is preferred because of its high common-mode noise rejection. The DCS circuit is fast and simple in layout compared to repeater insertion despite significant static and leakage power which remains a barrier for broad application. Temperature variation throughout the chip also causes the timing uncertainty on interconnects to increase. This thesis addresses current-sensing interconnect circuit design in several aspects. First, it provides an improved differential current-sensing circuit called the differential leakage-aware sense amplifier (DLASA), that uses local power gating that results in 39.6% reduced leakage and static power compared to conventional differential current sensing. Secondly, thermal impact on interconnect is studied and temperature sensitivity is analyzed for interconnect circuits. Theoretical analysis is discussed as a base design guideline, then accurate simulation based experiments in 65nm, 45nm and 32nm CMOS technologies are used for verification from 25OC to 150OC. Thus this project provides a view of the year of technology toward 2013.