Tamer Ali
Using Fragmented Parallel FDTD to Find the Magnitude of Negative Magnetic
Permeability Provided by Nanoparticles AggregatesAdvisor: Peter Nordlander
Project Committee:
The first goal of this project is developing a parallel FDTD code that can be segmented into successive stages. This will enable us to do the time propagation in smaller chunks of time and will enable us to make use of our computational resources more efficiently. This also will help speeding up FDTD simulations by using ready-made simulated parts instead of repeating common parts.
The second goal of this project is investigating the possibility of using small nanoparticle aggregates, such as the gold tetramer, in order to provide a negative magnetic permeability. It is expected to do, but we need to find the magnitude of this negative magnetic permeability. We will do so by calculating the magnetic fields induced when a light wave hits the nano-structure formed by the small nanoparticles aggregates. We will use the enhanced FDTD code to perform such kind of excessive calculations.
Marc Blanco
Advisor: Edward Knightly
Project Committee:
The variability of the wireless channel and delays due to multiple wireless hops and competing traffic are among the factors that degrade the quality of real-time multimedia applications in multi-hop wireless mesh networks (mesh networks). However, the success of municipal mesh networks, the earliest of which have recently come online and many more are in the deployment or planning stages, depends greatly on the ability to deliver popular streaming services, such as Voice-over-Internet-Protocol (VoIP). Thus, to make knowledgeable decisions regarding protocol design and network provisioning in future mesh networks, an understanding of VoIP and other real-time traffic is essential.
Currently, little is know about the capacity of outdoor urban mesh networks to handle the quality of service (QoS) requirements of Voice-over-Internet-Protocol (VoIP). Existing studies have been limited simulation results and experiments on indoor mesh test-beds, rather than outdoor urban networks. In this project, ns-2 simulations and experiments on the Technology for All network will be used to characterize VoIP traffic over mesh networks. Using the collected data, we extrapolate a model for the effects of physical and network layer factors on VoIP call capacity. In addition, we will investigate the impact of VoIP traffic on competing data traffic in the network.
Mian Dong
Advisor:
Project Committee:
As the VLSI technology is fast approaching nano-era, new opportunities arise for emerging solutions to compete with CMOS. The nanowire crossbar has been adopted by many as a leading candidate architecture for post-CMOS nanoelectronic circuits. However, there exists a huge technology gap between the associated bottom-up synthesis techniques and realistic circuits. In this project, we will explore the opportunities and challenges for constructing nanoscale computing system based on nanowire crossbar architecture. The work involves theoretical analysis and circuit simulation of crossbar circuits, logic synthesis and technology mapping based on crossbar architecture, and fabrication of demo circuits.
Mike Foss
Advisor:
Project Committee:
This 599 project has two major goals: implementing Binkert et al.'s the integrated network interface design and improving their work by adding functionality that efficiently controls multiple network interface cards (NICs). Binkert et al. have redesigned the NIC controller such that it resides almost entirely on the CPU die as a set of two FIFOs (transmit and receive) that communicate directly with the MAC on the NIC. In this arrangement, the device driver may access the packet data directly rather than wait for a DMA transfer. Since the NIC may now be primarily software-controlled, new protocols and capabilities can be added with a driver update instead of a firmware update or redesigned board. With their simplest controller, Binkert et al.'s simulations resulted in performance approximately equal to that of a standard NIC; however, they accommodated for only one NIC. We wish to expand this design to control multiple NICs by dynamically allocating the a set of FIFOs that sit near the CPU.
Chinmay Hegde
Compressive sensing of signal manifolds Advisor: Richard Baraniuk
Project Committee:
Much of signal processing can be viewed as a search for an appropriate model which can exploit the underlying structure of several classes of real world signals. One such model is sparsity , which assumes that a signal can be represented as a linear combination of relatively few building blocks drawn from a basis or dictionary. The field of Compressive Sensing (CS) has provided an exciting new approach to dealing with sparse signals, asserting that a sparse signal can be recovered from a small number of random measurements.
An extension of this model would be to examine signals which lie on a low-dimensional manifold in an ambient higher-dimensional signal space. For instance, images captured by a camera net observing a scene can be viewed as points lying on a low-dimensional hypersurface in an ambient space whose dimension equals the resolution of each camera. Recent results [Random Projections of Smooth Manifolds, R.Baraniuk and M.Wakin, ICASSP ‘06] have indicated that a small number of random measurements of signals from such classes preserves many aspects of manifold structure; thus, considerable information can be gained by analyzing and interpreting such measurements.
In this project, we shall focus on compressive sensing of signals which conform to manifold models as described above. We aim to develop a framework which allows us to handle signal processing problems from a manifold perspective. Issues such as manifold recovery and signal parameter estimation will be examined, with a special emphasis on geometrical understanding of manifolds. As an application, we shall analyze the lumigraph, a concept used in image-based rendering, and develop equivalent concepts and algorithms in a CS/manifolds setting.
Chris Hunter
Advisor: Ashutosh Sabharwal
Project Committee:
Research in multiple-input multiple-output (MIMO) communications has demonstrated tremendous gains in both reliability and bit-rate. However, for many applications, such as cellular phones and handheld computers, it is impractical to make use of a large antenna array. For these applications, researchers are focusing on cooperative communications as a means of achieving the spatial diversity gains of MIMO but using multiple single-antenna nodes. Nodes in an ad hoc network can help each other transmit information more reliably.
Research in the field has primarily assumed that all participating nodes are fully synchronized. Our preliminary research shows this assumption to potentially be problematic. Any error in the start time of a receiver can nullify the gains of the cooperation. This random-access cooperation problem will be the focus of this project.
The aim of this project is to propose a medium access algorithm capable of retaining the gains of cooperation in a random-access environment. To do this, we will need to enhance existing models to account for the losses described above. We can then use these new models to develop and test new algorithms that combat the problem. Analytical, simulation, and implementation results will demonstrate the algorithm's utility.
Roxana Ionutiu
Passivity Preserving Model Reduction Methods for Circuit Simulation Advisor: A.C. Antoulas
Project Committee:
Reduced order modeling is imperative in circuit simulation, as part of the physical verification stage of VLSI chips. Simulating complicated intreconnect systems in their full dimensions is computationally unfeasible. An efficient reduced order model for the interconnect is therefore needed to replace the original system. This approximate model is then co-simulated with other existing circuitry and the behavior of the overall system (nonlinear circut elements and the interconnect) is obtained. As interconnect systems are described by passive circuit elements, the reduced model should preserve the passivity of the original system. This guarantees overall system stability during simulation. Passivity preserving model reduction procedures have been developed, but were designed for systems with special internal structures. The aim of this research project is to first compare existing model reduction methods when applied on systems of such kind. Next, the new passivity preserving model reduction method: Lanczos with matching of spectral zeros, will be extened and applied to systems with arbitrary internal structures. In particular, interconnect circuit models resulting from real industrial applications, which obey more general internal structures, will be subject to reduction via the propsed method. The focus of our investigation will be the proper selection of spectral zeros in the above method, as to obtain reduced models that are passive and yield small approximation errors.
David T. Kao
Bandwidth Considerations in Cooperative Spectrum Sensing Advisor: Ashutosh Sabharwal
Project Committee:
Recently there has been a great deal of interest in the development of spectrally aware and spectrum sharing cognitive radios networks that are capable of more efficient utilization of spectral resources. However a common assumption regarding cognitive radios is that they are unlicensed spectrum users that should defer to (avoid interfering with) existing primary sources. Therefore effective sensing of primary users is a major focus of current research. Recently, it has been demonstrated that cooperation can greatly enhance the reliability of primary user detection in the presence of issues commonly dealt with in wireless environments (i.e. shadowing, fading, noise uncertainty, etc.).
A large portion of recent work focuses on centrally-managed spectrum sharing networks where a base station combines the observations of and coordinates individual spectrum sharing nodes. In this scenario cooperative sensing typically requires the use of a control channel which is assumed to be dedicated and free of other sources. This is similar to the topic of decentralized detection in sensor networks, where multiple nodes observe a phenomena through a degraded sensing channel and each node transmits a description regarding the phenomena to a fusion center via a corrupted reporting channel where a final decision is made [4][5]. However there is one issue that has yet to be considered that is directly related to the viability of centralized spectrum sharing networks. Any spectrum sharing network designed for spectral efficiency would have stringent constraints on control-channel bandwidth. Therefore it would be prudent to examine the relationship between the bandwidth of the control channel and the performance of the cooperative sensing system.
Performance in a spectrum sharing network involves evaluation of a number of system characteristics. Of primary importance is the tradeoff between minimizing interference with primary users and maximizing spectral efficiency, a relationship directly related to the receiver operating characteristic (ROC) curves of the cooperative sensing system. In addition to this, it is necessary to establish relationships between the bandwidth of the control channel and the spectrum sensing resolution and frequency agility, two additional characteristics that affect the efficiency of the network. It is hoped that this analysis will lead to the design of general protocols for unlicensed spectrum sharing networks.
Mark William Knight
Observation of Polarization-Dependent Plasmonic Coupling Between Silver Nanowires and Metallic Nanoshells Advisor:
Project Committee:
Nanoparticles positioned close to thin metallic wires should enable the efficient coupling of light into plasmons, effectively serving as plasmonic 'nanoantennas'. Initially, this project will focus on synthesizing silver nanowires and establishing a method for assembling nanoshell/nanowire systems. Plasmonic propagation in these systems will be induced by scattering photons from a diffraction-limited laser spot off the rough ends of the nanowires. Single-wire systems will be characterized through far-field measurements of remote, polarization-dependent plasmon-mediated emission. As time permits, the project may be extended to include near-field imaging of the nanowire/nanoparticle junctions to allow a quantitative characterization of localized field intensity enhancements.
Abhilash Krishna
Electrophysiological Model of a Rabbit Ventricular Cell Advisor:
Project Committee:
This modeling study focuses on the mechanisms of Ca2+ mediated cross-signaling that takes place between the DHP-sensitive and Ry-sensitive Ca2+channels. It serves as an important element in excitation-contraction coupling in cardiac muscle. Specifically, models of the two types of Ca2+ channels and the model of Ca2+ diffusion in the dyadic space is used to describe a graded Ca2+ - induced Ca2+ release with adequate calcium gain. These component models are embedded in a larger ventricular cell model, so that the character of model-generated myoplasmic Ca2+ transients might be compared with measured data. This model would then be used as a tool to simulate nano-pulse excitation given to a rabbit ventricular cell, producing an externally controlled Ca2+ release from the sarcoplasmic reticulum that can be compared with measured data.
Sanda Lefteriu
Model Order Reduction in Electromagnetic Simulation Advisor: A.C. Antoulas
Project Committee:
This project will address some issues in an area of research which has numerous utilizations in industry, namely that of model order reduction applied in electromagnetics.
Maxwell's curl equations are difficult to solve in their continuous version for complex geometries with dielectric media, so one must use a spatial discretization of the electric and magnetic fields. This is done by employing the Finite Element Method, which generates a discrete state-space model for the electromagnetic device.
The resulting model has large dimensions (in the range of millions), so SPICE simulation would be inadequate. Therefore, we are interested in finding a smaller system that would approximate the behavior of the original one but would be computationally more feasible. Model order reduction, in general, deals with designing an algorithm which, given the initial system in the state-space representation, produces a system of smaller dimension, which can be easier both to control and to simulate, and it approximates the original system with sufficient accuracy. Also, in VLSI circuits, it is important that, besides the standard properties of the reduced system, model reduction should preserve the passivity of the initial system. Passivity is an important property of systems. When connecting several macromodels, one may have to deal with instabilities if the single macromodels are nonpassive. In fact, stable but non-passive structures may become unstable depending on the termination networks. Unfortunately, no control over macromodel passivity is possible, since passivity is a global feature of the transfer matrix and requires its consideration as a whole. Conversely, passive models guarantee stability under any termination condition thus insuring successful system-level simulations.
This project aims at implementing several model order reduction techniques which are passivity-preserving and using them in reducing the systems obtained by applying the Finite Element Method to several electromagnetic devices.
Jiayang Liu
Cross-layer power analysis and optimization of wireless communication systems Advisor: Lin Zhong
Project Committee:
This project is aiming to optimize the power consumption of wireless network interface. In current networks based on IEEE802.11, the network interface card (NIC) consumes a considerable amount of power even in standby mode when there is no traffic going on, making it difficult for mobile devices to keep connected with the network all the time. One primary reason is that the standby mode implemented in most wireless NICs still has its PLL running, which drains out the battery quickly. Shutting down PLL, however, will introduce long latency to communication. In this project, we will explore new power saving mechanism which turns off PLL to save more energy and minimizes the latency at the same time. Our work will be based on WARP, an open source research platform developed by CMCLAB. We will implement MAC layer support for power saving mode (PSM) similar to that in 802.11 standard, then aggressively turn off PLL without interrupting communication, and evaluate the impact of this new PSM on network performance. Based on this new PSM, we will explore how to further optimize power consumption by combing cross layer mechanisms.
Lanyue Lu
DiskGroup: Energy efficient disk layout for RAID1 system Advisor: Peter Varman
Project Committee:
Energy consumption is becoming an important issue in storage systems, especially for high performance data centers and network servers. Much research has been conducted to save disk energy based on multi speed disks, cache strategies, and file systems. However, these solutions provide limited energy savings and degrade the performance significantly.
Conventional disk arrays (RAIDs) are not focused on energy efficiency, because the data layout distributes the loads evenly to all the disks, requiring all the disks to remain spun up. In this project, we will introduce a new disk layout for RAID1 system, DiskGroup, which skews the load to a subset of the disks, so that the other disks can be spun down to save energy without performance degradation.
Stanislav Miskovic
Resilience in two-tier wireless CSMA mesh networks Advisor: Edward Knightly
Project Committee:
In this project we will study and try to improve resilience in two-tier wireless mesh networks. Resilience is commonly desired objective in all types of communications. It reduces performance sensitivity to various networking effects. Using the examples from various fields of wireless networking research, we motivate our goal to better understand the resilience issue. We have a unique chance to gain additional insights in this area through observations of TFA network deployed in Houston's low income community. Finally, through this project we will start to develop solutions that improve resilience.
Hamid Nejati
Modeling and design of ultra-wideband low noise amplifiers with generalized impedance matching networks Advisor: Yehia Massoud
Project Committee:
In this project, we plan to generate an integrated modeling methodology for inductively degenerated cascade ultra-wideband low noise amplifiers (LNA) with generalized filter-based impedance matching networks. In this analytical model, we plan to capture the impact of device and passive component parasitics and transistor short channel effects to closely match circuit simulation results. The goal of this project is that we can accurately generate an ultra-wideband LNA in the 3.1 to 10.6 GHz band using nth-order Chebyshev filters as input matching networks. The power gain greater than 9dB, input and output impedance matching less than -10dB, and a noise figure less than 3dB is expected.
Pedro Santacruz
Advisor:
Project Committee:
The main objective of the research is to examine the performance of a multiple-node two-way channel network. The analytical model will require the development of new tools as well as the use of previously used techniques to provide the wanted results. In this model, each node receives information and has information to send. During some time, the network will be in broadcast mode from base station to mobile users, and at some other time, the network will be in multiple access mode from the mobile users to base station. The model to be analyzed features fading channels and non-coherent communications, i.e., the channel state information is not known. Previous work has analyzed and provided insight into broadcast channels and multiple access channels, including fading channels in both cases. Also there is literature in the analysis of two-way channels, including Shannon’s Two-Way Communication Channels. Some of the issues to be considered are the overall throughput of the system, the probability of outage, the optimal times for broadcast mode versus multiple access mode, the effect of knowledge of channel information, and the role of feedback.
Davood Shamsi
Advisor:
Project Committee:
Wireless sensor networks (WSNs) are an important and emerging class of distributed networked embedded systems. Addition of sensing to the system enables users to measure and analyze the properties of the physical environment. Analysis of data not only reveals the hidden properties of physical phenomena, but also creates an interface to the real world, promising great societal progress by enabling targeted and accurate manipulations of the physical world.
Various WSN research efforts make a wide range of differing assumptions that have produced vertically integrated models and algorithms. The solutions often can neither be integrated nor compared with the other algorithms that address the same problem. This exceedingly reduces the synergy between research efforts, and impedes progress in WSNs modeling and application optimization. To alleviate these problems, My project is concerned with developing a set of benchmark instances for a number of sensing applications that can be used for comparing the various models and algorithms.
The benchmarks will be built upon deployed WSNs traces that will be processed by combinatorial techniques which can control the difficulty and size of the instances. I will create algorithms to select instances or subsets of nodes in a WSN so that it has specified properties such as the percentage of missing and faulty data, or the number of neighbors for location discovery. The key observation that by selecting a subset of nodes or samples from actual traces we virtually create new actual WSN that corresponds to different deployed nodes or measurements conducted in different time intervals. The impetus for development and use of benchmarks is provided by their common use in many domains of embedded systems where benchmarks are utilized to drive new research efforts.
Yanbin Zhang
A tunable SERS cavity based on a mechanically controlled break-junction Advisor: Kevin Kelly
Project Committee:
There has been a great deal of interest in metal nanostructures for creating surface-enhanced Raman scattering sensors. The wavelength of this enhancement is based on the plasmon frequency of the metal and the geometry of the nanostructure. Many of the current attempts create fixed frequency resonances based on nanoparticles or by lithography. The goal of the project is to create a tunable SERS cavity based on a controllable break-junction with a gold or silver wire, which has been previously used to explore the electronic properties of single molecules. If successful, such a device could provide for a low-cost sensor for various biomolecules when incorporated in a flow-cell geometry.
Comments and questions to: www-ece@ece.rice.edu
Last modified: February 21, 2007