No description available.

Two-dimensional free-space optical interconnects (2D-FSOIs) promise to deliver tremendous gains in bandwidth and architectural freedom for applications such as telecommunication switches and massively parallel computing systems. One major obstacle preventing the commercial deployment of 2D-FSOI systems is the problem of optical alignment, which is further exacerbated by the requirements that these systems be field-serviceable and able to sustain the harsh conditions of industrial environments. / This thesis proposes a broad range of solutions to alleviate this alignment problem. One important aspect of this work concerns the development of a generic packaging strategy, which consists of partitioning an optical system into separate modules in such a way that the loose tolerances are between the modules while the tight tolerances are between the components inside the modules. To accomplish this, novel alignment techniques are designed and demonstrated, including the use of integrated diffractive features, CMOS position detectors, ultrathick photoresist micro-structures, and semi-kinematic fixtures using dowel pins. In all cases, emphasis is placed on approaches that are amenable to low-cost manufacturing and high-volume production. / These techniques were developed in the context of a photonic backplane prototype experiment that demonstrated 1024 free-space interconnections between four optoelectronic-VLSI (OE-VLSI) chips. The design and implementation of a module integrating an OE-VLSI chip, a mini-lens array, a thermoelectric cooler and a heatsink is presented. Optomechanical, electrical and thermal characterization results are reported. / The other aspect of this work aims at identifying the types of optical designs that provide more generous misalignment tolerances. This is done by investigating various optical configurations for the design of the chip module. The central objective is to understand the underlying reasons that make one configuration more misalignment-tolerant than another. A significant outcome of this work is to show that the inherent misalignment tolerances of 2D-FSOI systems translates into an aspect-ratio limitation similar to the one found in electrical interconnects.

An integral equation formulation for the calculation of the capacitance of three-dimensional VLSI geometries is presented. A proper combination of 2D and 3D methods is used for efficient numerical computations. The method of moments is used for the solution of the integral equation. In addition, Green's functions that satisfy the boundary conditions at the dielectric interfaces are implemented in order to minimize the number of unknowns involved in the numerical solution. The mathematical formulation presented here and the associated computer program are appropriate for obtaining the capacitance matrix of complex three-dimensional multi-conductor configurations of the microstrip and the stripline type. Finally, numerical results for the per-unit-length capacitance and total capacitance of several interconnections are provided and compared with known results. Applications include the extraction of lumped capacitive elements used in the equivalent circuit representations of coupled conductor bends, vias and crossovers. In addition, calculations of per-unit-length capacitance of coupled flaring lines are performed.

The level of coupling noise or cross talk is an important factor in the design of reliable VLSI packages. Coupling noise can be reduced by modifying the underlying conductor geometry. An expert system is developed and integrated into the packaging design and simulation environment UAPDSE. The system is provided with rules to reduce the coupling noise, and create models that satisfy the constraints set by the package designer. The parameter calculator UAMOM and the circuit simulator UANTL are used to evaluate the electrical performance of the model under study. The expert system is an interactive and user-friendly software that provides efficient method for performing the coupling noise reduction in interconnect systems.

This thesis considers the application of the finite-difference time-domain (FD-TD) method to the electromagnetic characterization of multichip-module (MCM) interconnects with perforated (mesh) reference planes. The limitations of the method in finding the propagation characteristics, characteristic impedance, Z0, and phase constant, beta, are investigated. A different method for the characterization of MCM interconnects is suggested which exploits the capabilities of the FDTD method. This alternative approach uses the results from the FDTD method to extract a per unit length delay and an approximate impulse response of the system. These results can be used to identify the effects of the perforated reference plane on signal propagation. In particular the validity of the TEM approximation for signal propagation in realistic MCM structures is examined. For those cases where the TEM approximation is valid, a quasi-TEM approach is developed to find equivalent transmission line characteristics of MCM interconnects. This quasi-TEM method can be used to obtain effective transmission line parameters for both single and coupled interconnects, which in turn can be used directly in SPICE-like waveform simulation tools for an overall electrical analysis of complex MCM interconnect nets.

High cycle vibration fatigue tests are performed on Ball Grid Array (BGA) packages employing a specially designed Cyclic Controlled Curvature Cantilever Device (C4D) and an imaging system in which a stroboscope is used with an optical sensor to freeze the vibration of the critical solder interconnect in BGA specimens. The failure mode is identified as crack initiation and propagation on the component side along the nickel/solder interface. A primary crack starts from the inner edge, progressing stably until the secondary crack begins on the opposite edge. The crack growth is then accelerated till the complete crack has been formed. Averagely, the time spent in crack initiation, stable propagation and accelerated crack propagation are about 15%, 60% and 25%, respectively. Vibration tests at various frequencies were also performed. The cycles to failure is found to be frequency-independent from 50 Hz to 100 Hz Several commonly used fatigue life-prediction models, such as Solomon's model and Paris' law, are examined based on failure parameters computed from nonlinear finite element analysis. It is found that while the damage models usually show large discrepancy, the fracture model can correlate with the test data within a factor of 1.5 The High Sensitivity Moire Interferometry method is also used to capture the mechanical response in the BGA three-point bending and Flip Chip four-point bending tests. Both packages come in two forms: with and without underfill. The influence of underfill is shown to be obvious Underfill curing-induced shrinkage problem is investigated using a unit cell finite element model comprised of solder and underfill in cylindrical shape. A two-phase solution scheme is utilized: The shrinkage is modeled in the first phase by introducing a thermal contraction to the underfill; the solder is allowed to creep based on a hyperbolic creep law in the second phase for 1000 hours. A compressive residual steady state stress in the solder interconnects can be reached within several weeks time. This compressive stress can extend the fatigue life of solder interconnects undergoing Mode II cyclic shear according to a relative life model proposed by Larson and Verges (2003) / acase@tulane.edu

Accuracy of fatigue life prediction of solder joint materials using the finite element method (FEM) depends on realistic and reliable constitutive models and the FEM procedure itself. The effectiveness and accuracy of a constitutive model to simulate and predict the behavior of materials strongly depends on consistent, reliable and precise experimental data. In this research, a thermomechanical digital image correlation test device has been developed to investigate solder joint material behavior under thermal mechanical and cyclic loading. The test device is composed of two high precision actuators driven by two brushless servo-motors; a vacuum chamber used to prevent the heating and cooling plate from heating and cooling the air in the vacuum chamber and other components to save the energy needed to heat and cool the specimen; a heating and cooling unit composed of four Peltier devices; a digital correlation deformation and displacement measurement system and various load and displacement sensors. The test device is controlled by Labview. Two closed loop system control programs in Labview were developed to control the device and perform tests under specified loads, displacements and temperature profiles. The temperature, loads, and displacements can be applied manually and automatically. A series of tests such as normal load tests, temperature and rate dependent tests, isothermal mechanical fatigue and thermomechanical fatigue tests have been carried out on 63Sn/37Pb and 60Sn/40Pb solder joint materials. The rate and temperature dependent test results are consistent, which shows that newly developed test device is reliable. Material parameters for the HISS and DSC models were computed from the test data, which can be used to investigate the failure mechanism of solder joints in electronic packaging to improve the reliability of electronic packaging. Disturbed State Concept (DSC) constitutive model is a powerful tool for the simulation and prediction of material damage, failure and fatigue. Strain localization and material softening cause negative-semi definite finite element stiffness-matrix, which create numerical difficulties for the calculation of DSC. A new DSC finite element scheme was proposed and implemented to overcome the numerical difficulties. The finite element thermoplastic formulation was modified and implemented based on Drucker's postulate. Thermoplastic deformation and viscoplastic deformation usually happen sequentially or combined. An alternative finite element computational method for viscoplastic problem was proposed and the FE program is modified. Computational examples show the new procedures produce accurate and consistent results and the program is robust. Material parameters for the HISS and DSC models were used to back predict the rate and temperature dependent test results using the modified finite element program. These back predictions showed that the HISS and DSC constitutive models and the FE program can be used to accurately simulate solder joint material behavior and predict solder joint damage and failure.

Reliability quantification is one of the most important tasks in Reliability Engineering. During the design, development, and operational phase, this information can be very valuable to the product's designers, and users. The designers can use it to guide their design, find the design's weak points. Users can use it to setup maintenance plans and schedule. PCBs, as the major building block of electronic equipment, have been widely used in modern complex systems, such as aircraft, automotives, laptop computers, etc. Its reliability plays a vital role in the whole system's reliability. Thus, effective and accurate quantification of PCB's reliability becomes very essential to the whole electronic system's reliability quantification. Random vibration and thermal cycling are two common environments experienced by PCB's. As a result, quantifying the reliability of a PCB under these two environments becomes necessary. Currently, in industry, the commonly used methods to quantify the reliability of a PCB are the MIL-HDBK-217 and Bellcore type methods. However, the lack of accuracy and slow pace of updating the databases have limited the usage of these methods. In this dissertation, a Modified Stress-Strength Interference (MSSI) method is proposed to quantify the reliability of a PCB. In this method, not only the stress and the strength are assumed to be distributed, but also is the mean value of the strength, so that both the initial designed-in reliability and the reliability at any time can be quantified. Based on this method, a reliability quantification model for the PCB is developed. In this model, a PCB is divided into three parts: i.e., board, interconnects, and parts and modules mounted on the board. Seven (7) failure modes related to the board and the interconnects, and one (1) failure mode related to the module have been investigated. The dependence between these failure modes is studied and incorporated into the reliability quantification model. A three-step popcorn effect reliability quantification model is also proposed by means of considering the failure mechanism of the popcorn effect. Finally, a comprehensive example is given to demonstrate the usage of the methodology proposed in this dissertation.

Temperature cycling and random vibration have proven to be the two most effective environmental stress screens. This study presents an extensive research on the physical quantification and optimization of temperature cycling and random vibration screens. For temperature cycling screen, a general model has been proposed to describe a typical temperature response cycle and a typical power-temperature response cycle. The least-squares parameter estimates for the two modified Arrhenius models are derived. Two general models for quantifying the equivalent aging acceleration factor of a typical temperature cycle with or without power cycling, considering both reaction rate stress and temperature change rate stress and also incorporating the temperature dependence of the activation energy, are derived. A closed form solution under the mixed-exponential life distribution assumption and an iteration equation solution under the Weibull distribution assumption, of the optimum number of temperature cycles for a specified post-screen field Mean Residual Life (MRL) goal, are established. For the random vibration screen, the distributions of the cumulative damage and fatigue life, under both stationary narrow-band and stationary wide-band random stressings, are derived under both-normal, semi-normal, and Markov-process assumptions. A bimodal mixed P-S-N diagram is proposed, from the failure physics point of view, to describe the fatigue strength of a non-screened unit. The concepts of the threshold S-N curve and the screening probability for fatigue defect precipitation are proposed to facilitate the quantification of random vibration screens. Finally, the closed form solution of the optimum vibration duration for a specified screening probability is derived under both-normal, semi-normal and Markov-process assumptions, respectively.

Experiments were performed to characterize the conjugate heat transfer due to a square flush heat source mounted at the center of a square horizontal plate in a small horizontal enclosure. The plate area was six times larger than the heat source area. Three different plates with heat source facing upwards were considered: a 25mm balsa wood plate which provided a nearly adiabatic surface, a 1.57mm thick FR-4 plate with no copper, and a 1.57mm thick FR-4 plate with a single layer of 0.036mm thick copper cladding on source side. The back of the board was insulated for all cases. The experimental exploration included measurement of heat transfer coefficient over the heat source, plate surface temperature distribution and temperature distribution in the air volume above the plate. The heat transfer coefficients exhibited distinct behavior at high aspect ratios in which the dominant length scales were related to the source. At intermediate aspect ratios, length scales for both source and enclosure were relevant, and at small aspect ratios, a conduction limit was observed, which was dependent on board conductivity. The heat transfer coefficients at high aspect ratios exceeded by 14% the prior correlations for upward facing isolate plates, when the ratio of source area to perimeter was used as the significant length scale, and a stronger dependence than Ra1/4 was measured. Classical correlations for shallow differentially heated enclosure were not satisfactory in describing the dependence on enclosure height. With increasing board conductivity, board thermal spreading increased the effective source size so that the discretely heated board heat transfer coefficients tended towards the behavior of the classical uniformly heated board. New first-order thermal design formulae were derived for determining peak temperatures of sources on conducting substrates, and for determining the associated thermal "zone of influence" or "footprint." The board heat spreading was accounted for by using its effective "thermal footprint" radius and correlations for conjugate heat transfer based on this length scale were successful in describing the behavior of the average Nusselt number at large enclosure heights. Some qualitative flow visualization was also performed and representative results are shown.