With decreasing feature sizes, transistors are being added to ICs in consistently greater numbers, which is leading to dramatic increases in power consumption. Changing process parameters and redesigning circuits are complicated and expensive solutions to lower power dissipation. A simple and cost effective approach is to operate standard cell libraries at ultra-low power (ULP) supply voltages. By lowering the supply voltage, the transistor current drives decrease by orders of magnitude, resulting in dramatically lower power consumption. However, small transistor drive currents also result in slow operating frequencies, so a trade-off must be made between power and performance. <p>In this work, the use of ring oscillators as a single-event test structure is introduced for the first time. Ring oscillators allow for single-event characterization of technologies by finding the minimum energy at which harmonic oscillation occurs. By finding the minimum laser pulse energy at which the ring oscillator enters a state of sustained harmonic oscillation for a range of power supply voltages, the single-event susceptibility of a technology can be determined. Experimental two-photon, backside laser irradiation results show that ULP circuit operation is most susceptible to single-events, because the threshold laser pulse energy is the lowest in this region. <p>Additionally, the effect of supply voltage on single-event transient pulsewidth is seen both in 3D TCAD simulation and experimentally through the use of single-photon, topside laser irradiations. 3D TCAD simulations show that transients created by ion strikes on a device operating in the subthreshold region have longer pulsewidths than those from a device operating at nominal voltages due to the decreased currents of the pull-up and pull-down devices connected to the struck nodes. Additionally, when the devices are operating at ultra-low voltages, the single-event transient pulsewidths generated from strikes on PMOS transistors are smaller than those generated from strikes on NMOS transistors. Normally, at the nominal supply voltage, the opposite is true, but when operating ULP circuits, parasitic bipolar amplification does not occur. <p>The increase in pulsewidths as a function of power supply voltage also is shown experimentally, as well as the independence of pulsewidth on laser pulse energy. At voltages outside of the ULP region, the standard, expected response of increasing single-event transient pulsewidth with increasing laser energy is seen. These trends were seen after strikes on both NMOS and PMOS transistors.

This project is aimed at developing a system allowing Mobile Ad hoc Networks (MANETs) to perform distributed tasks within a fault-tolerant context, called the Distributed Apt Resource Transference System (DARTS). By creating a method by which individual, wirelessly-connected nodes can identify and respond to malicious intrusions and operational failures such as radio interference, DARTS provides a way to increase system uptime by allowing redundant resources owned by nodes to be reallocated to a task that would otherwise need to be suspended due to node failure. This work also attempts to understand the relationship between the efficiency of the system itself and differing network densities, defined by the geometric organization of nodes. Lastly, by integrating DARTS into the application of distributed disaster area assessment involving multiple unmanned aerial vehicles (UAVs), the speedup (up to 6-fold increase) and reduction in number of reallocation request messages (79% reduction) has shown that DARTS is a viable way to perform an accurate and timely survey of tornado storm damage using a distributed aerial platform.

Modeling physical systems is necessary to understand how they work and how the compositions of physical components (i.e., complex physical systems) behave. Graphical modeling languages are often used in engineering design. A graphical modeling tool such as Generic Modeling Environment (GME) helps managing complexity and supports the reuse of models to improve productivity and reduce the design time.<p>This thesis describes two results: the Functional Dependency Model (FDM), a Domain Specific Modeling Language (DSML) that can be used for modeling complex physical systems, and the FDM Model Interpreter (FDMMI), that generates Simulink models from the FDM models. The DSML is based on hybrid bond graphs with some extensions: simple controller elements, sensors, actuation elements, modulation functions, switching functions, information links, and power links. While bond graphs provide a domain independent modeling language, FDM supports to use of power ports from different domains (i.e., electrical, mechanical, hydraulic, and thermal) to allow for modeling the domain specific parts of a system.<p>The model interpreter checks some of the design rules, generates the causal graph of the system, and generates a Simulink model of the system. The modeler can then simulate and analyze the system using the simulation engine of MATLAB.

Single-event transients (SETs) are a significant reliability issue for space-based electronic systems. A single-event transient is a radiation-induced glitch in an electronic circuit caused by an ionizing particle. While there has been a large amount of research published on SETs, one key aspect that has been mostly ignored has been the impact of temperature on the time duration of these transients. Since the temperature ranges over which some space missions need to operate can be extreme, the role of temperature for all radiation effects is of vital importance for space systems. However, to understand fully how temperature will impact SETs, a complete understanding of SETs at room temperature is first needed. In this dissertation, heavy-ion data on ten SET test structures fabricated in a myriad of semiconductor technologies are presented. The data from these test structures give valuable insight into how the SET problem is changing with each technology node and how temperature affects the time duration of SETs.

A single event occurs when an ionizing particle strikes a microelectronic circuit and deposits charge in the semiconductor material. Charge deposited by the particle is often collected by the circuit, resulting in a current transient that can lead to soft errors, depending on the circuit response. In this thesis, single-event (SE) mechanisms in combinational logic are discussed in the context of a 90-nm digital cell library in bulk CMOS. Technology Computer Aided Design (TCAD) simulations are used to investigate various parameters across different cell designs. Knowledge of the effects of layout- and circuit-level parameters on SE response is used to perform a TCAD characterization of a 90-nm digital cell library. The results give insight into the relative SE vulnerability of different types of logic cells.

Scaling in VLSI systems leads to higher packing densities for transistors. As a result, they are more likely to be hit by an incident particle, such as neutrons or alpha particles. The interaction of neutron and alpha particles with semiconductor devices may lead to permanent, intermittent, or transient faults that result in an error. Thus, error detection becomes a greater concern for system reliability as transistor size decreases. To achieve the desired reliability, computer architects investigate new techniques to detect and correct soft errors caused by transient faults. Usually, a tradeoff is made between the performance of a processor and the area and power required for error detection. This thesis uses a 45-nm cell library to synthesize hardware description language (HDL) models for selected error detection techniques when used with Arithmetic Logic Units (ALUs). Key results include a drastic increase in power consumption for some techniques as the bit-width of the ALU increases. Results are also compared to the baseline of dual modular redundancy (DMR).

Scaling of gate oxides in bulk complementary metaloxidesemiconductor (CMOS) devices to thinner dimensions has reduced the significance of threshold-voltage shifts due to total-ionizing dose (TID) radiation-induced charge buildup in the thin oxides. As a result, the dominant TID effect in most CMOS technologies is now leakage current produced by charge buildup in the shallow-trench isolation (STI). In this thesis, the sensitivity of radiation-induced source-drain leakage to the amount of recess in STI of CMOS technologies is studied. The impact of the doping profile along the STI sidewall on the magnitude of the leakage current is quantified, and finally layout-related stress effects on TID-induced leakage current were investigated. The TID-induced leakage current of deep submicron MOSFETs increases with increasing active-to-isolation spacing. Mechanical stress reduces impurity diffusion at the STI sidewall, affecting the TID sensitivity. The sensitivity of the radiation-induced leakage current to these parameters provides insight into how process variability is manifested as variations in the radiation response.

Single-event transients (SETs) due to terrestrial or space radiation exposure have become a growing concern in modern high-speed data acquisition systems and space deployed electronics. Therefore the single-event vulnerability of DLL and phase-locked loop (PLL) circuits are a particular concern for space-deployed systems, since SETs in spacecraft clock circuits can result in systemic failure i.e. data loss throughout entire ICs data acquisition network. Recent work has enabled the understanding of SET effects in mixed-signal PLL, but no previous work on the DLL radiation sensitivity has been recorded. This thesis presents the single-event analysis of two analog delay-locked loops topologies, and shows that the SET sensitivity of the circuit depends on the sub-circuit stroked. Simulations and analysis of the DLLs show that the voltage controlled delay line (VCDL) and the charge pump (CP) are the most sensitive modules to SET, in terms of maximum phase displacement and missing pulses generated by ion strikes. Implementing a voltage-based charge pump (V_CP) in the analog DLL, in order to mitigate persistent inverted-lock error, has significantly hardened the DLL to SET effects.

The multiple sensitive volume model is a spatial and mathematical description that relates energy deposited by a quantum of ionizing radiation to charge collection at transistor nodes within semiconductor devices. The sensitive volume model is capable of capturing the spatial and directional dependence of ionizing energy loss on charge collected at transistor contacts. Monte Carlo estimation, rather than the calculation of closed-form expressions, is used to determine chord length distributions in an arbitrary number of sensitive volumes and sensitive volume groups. Physics based Monte Carlo simulation is ultimately used to abandon the chord length distribution approximations and to directly calculate energy deposition within the bounds of the sensitive volumes. The methodology of using multiple sensitive volumes, and sensitive volume groups, within a Monte Carlo framework, is applied to the problem of predicting SEU parameters such as cross sections and error rates for three common sequential logic devices.

Diffusion tensor imaging has been widely used to reconstruct neuronal fibers in the human brain. Studying these fibers often requires them to be grouped into bundles that correspond to coherent anatomic structures. Several fiber bundling methods are proposed and evaluated in this work. A unified fiber bundling and registration algorithm, which refers to a pre-built bundle template, is firstly proposed to provide fiber bundling consistent with well-defined major white matter pathways. Furthermore, a clustering algorithm, which is constrained by a cortex parcellation, is proposed to automatically segment connections between cortical/sub-cortical areas. Based on this framework, a group-wise fiber bundling method is further proposed to leverage a group of DTI data for improving across subject bundle consistency. The above methods have been rigorously evaluated with in vivo DTI data, demonstrating a potential of being used to better characterize white matter pathways and measure the connectivity.