Summer Research Experience for Undergraduates (REU)
Robert Becker - Blood Volume Detection for a Venous Reservoir
Nathaniel Bird - 3D Watercolor
Joshua Braun - 5GHz Narrow Band Low Noise Amplifier Design
Jonathan Carlson - Multi-layer Antireflection Coatings for Semiconductor Optical Amplifiers
Jessie Cassada - Magneto-optical Isolators
John Foster - Analysis of Network Control System
Michael Friedman - 3D Medical Positioning
Greg Haider - Thermal Conductivity of Liquids
Sarah Handahl - Using Untrasound to Perform Non-Invasive Surgery
Kettly Joseph - Superthreaded Processor Architecture
Sung-Hoon Kim - Finite-State Markov Chain Model for Generalized Fading Channel
Jeremy Kiser - The Superthreaded Processing Architecture
Syreeta Knight - The Superthreaded Architecture Processor
Ethan Miller - Dielectric-Loss Characterization of SiO2 Films of Silicon Substrate
Michael Morse - SPEC CPU2000 Simulation for Evaluation of the Superthreaded Processor Architecture
Eric Nordberg - Techniques of Cobalt Electrodeposition for use in Nanoporous Alumina Structures
Jeremiah Remus - Wavelet Domain Reconstruction of JPEG2000 Images
Tiffany Williams - Power Flow Analysis
2001 Summer Program Staff
Professor Douglas Ernie Professor Lori Lucke
2001 Summer Program Abstracts
Heart-lung machines are medical devices used during open-heart surgery. A critical part of the usage of these machines is the continuous monitoring of the volume of blood contained in the machine's reservoir, insuring that the volume is never too high or too low. This task is currently the responsibility of a trained hospital employee, who carefully watches the reservoir. The goal of this research is to develop a computer system that can reliably monitor the volume of blood in a heart-lung machine's reservoir. This system must be simple to operate, reliable, and must not interfere with the normal use of the heart-lung machine.
-
- Participant: Nathaniel Bird
Home Institution: Ohio Northern Universitys
Faculty Mentor: Professor Victoria Interrante
Project: 3D Watercolor - Participant: Nathaniel Bird
- The goal of this project is to create a computer simulation of watercolor in three dimensions. Such a simulation can be used to render three dimensional computer models such as architectural CAD drawings in a way that is more fluid than a photo realistic method. Forms of non-photorealistic rendering like this are very useful for presenting graphical information such as building plans in a way that does not look too real to be useful. To implement this concept, I started by writing a program that simulates the physical properties of watercolor moving across a paper's surface. Unfortunately, it is a very system intensive simulation, even in two dimensions. But the way the simulation works makes it well suited to smooth animation.
As optical and wireless technologies continue to expand, so does the demand for inexpensive, high quality devices to manage these types of signals. Many current front-end amplifiers in electro-optical transceivers sacrifice either manufacture cost or signal quality, depending on the fabrication process used. Research involved the schematic design of a narrow band 5GHz low noise amplifier that could mediate noise and manufacture cost. Arranging the design for a standard CMOS process reduced production costs. The design also utilized resonance structures in a distributed amplifier architecture (typically a wide band structure) that realized a noise figure less than comparable CMOS microwave amplifiers. Computer simulations were a final test to verify performance in comparison to initial calculations.
Wavelength division multiplexing (WDM) allows a greater number of optical signals to be carried at the desirable 1300 nm and 1550 nm wavelengths. Due to transmission losses, these optical signals must be amplified after traveling long distances. One solution is a semiconductor optical amplifier (SOA). Unfortunately SOAs experience varying spectral gain or ripple due to facet reflectivity. Angled facets and multi-layer antireflection coatings (ARC) decrease undesirable facet reflectivity. Candidate coatings consisting of silicon nitride, oxynitride, and silicon dioxide were simulated using Code V optics software. Actual ARCs were fabricated on 1300 nm SOAs using plasma enhanced chemical vapor deposition (PECVD). Less ripple was measured on the coated SOAs than the uncoated SOAs.
Magneto-optical isolators are integral devices in photonic circuits. Isolators are necessary to protect light sources from back-reflected light and thus, electromagnetic wave interference. This back-reflection is created at component interfaces and by inhomogeneity and other agitations of light. The current isolation devices consist of at least five individual components and are inefficient and bulky. Integrating optical isolators into a single waveguide would reduce cost, space and transmission losses while eliminating the need for a polarizer and as well as the need for component alignment. Yttrium iron garnet (YIG) has proven to be a satisfactory Faraday rotator and can be integrated with semiconductors via a magnesium oxide buffer layer. Our group has previously proved that an integrated YIG waveguide is feasible and that the best method for growth of YIG is through the use of radio frequency sputtering. We are currently aiming to optimize the optical properties of the YIG film and are growing varying series of films in order to test different sputtering parameters. We will soon begin to explore the architectural design of the waveguides in search of the most effective method.
The issue of optimizing traffic flow, while considering such problems as limited channel capacity, network delays, bandwidth restrictions, and competition among sensors, actuators, and controllers for the network channel, is of primary importance for feedback control systems. Since it is difficult to adequately model all the complexities of a networked control system and confidently predict its performance, we are instead simulating the control system in a laboratory. Time delay, packet density, and the number of collisions are just a few of the parameters that are studied in the feedback system. While our objective is to simulate real time control, our hope is to exercise the network under various conditions and collect data which will allow us to understand the critical parts of control systems, and eventually gain insight into future development of such systems.
Three-dimensional positioning is extremely important in the medical industry. It allows physicians and scientists to use remote probes to determine the location and size of abnormalities such as cancer. Furthermore, it makes the medical industry more of an exact science, as abnormalities can be located to a known position. This allows future probing to the location of the abnormalities to monitor changes with treatments. This paper will focus its attention on differential capacitor accelerometers. The purpose of experiments performed is to see if accelerometers can replace current electro-magnetic positioning techniques for a probe to be used in a body cavity. In addition, other technologies will be discussed. Accelerometers can find two angles, which can be used for limited positioning. They are small and inexpensive due to MEM's technologies, while electro-magnetic positioning technology can find 6 degrees of position but is large and more costly.
In the past, thermal conductivity of liquids has been difficult to measure due to parasitic convective heat transfer effects. By confining a liquid to an extremely small gap, these effects can be minimized. If this gap is created using two thermally conductive plates, we can measure the thermal conductivity of the liquid. As current is passed through one of the plates, its temperature rises. As temperature rises, the resistance of the plate will also increase. Therefore, by measuring the change in resistance of the plate, the thermal conductivity of the liquid can be obtained. For this project the gap is created by use of etch releasing. Polysilicon acts as the resistive plate with a silicon substrate as the other plate.
Ultrasound is known for it's imaging capabilities. We are using this quality along with a new breakthrough to try to create a procedure which allows us to perform non-invasive surgery on the human body. Basically, we use the imaging quality of the ultrasound to search for a site in the body that needs operating on (such as a blood vessel that requires an incision). Once this site is found, using a phased array and computer guidance, the ultrasonic waves can be focused to that point. A phased array is a compilation of 64 elements that send and receive ultrasonic wave and can therefore be focused at a single point within a millimeter of accuracy. Once the focal point is achieved, the voltage that the waves are sent at is increased and then we send a "shot" of ultrasound towards that focal point for a specific amount of time. Because the waves are coming out of a phased array, their intensity isn't strong except for at the focus. This way, the skin and almost any other organ can be penetrated, unharmed, leaving no scarring and minimal pain while performing surgery at the focal point. In conclusion, it is as if we are creating a way to perform surgery with an invisible knife that leaves no wounds.
The Superthreaded Processor Architecture is a composite of SuperScalar applications and a multiprocessor on a chip. The processor is being tested both by a hardware based, Quickturn, and a software based, SIMCA. SIMCA, which is an execution-driven simulator, is being tested by the SPEC CPU2000 benchmarks. I have been working on both reducing the large input datasets of the 300.twolf benchmark from the SPEC CPU2000 suite and parallelizing the code of the same benchmark. Considering the fact that simulation could take years when using detailed execution-driven simulator, the reduction of the input datasets of the SPEC 2000 benchmark suite was required. After reducing and testing the input datasets of the 300.twolf, I compared a profile of my result with the original input and they were approximately equivalent. Now I am working on parallelizing my benchmark using superthreaded codes. The goal being to attempt to increase the performance of the processors that stress-out when executing numerically intensive application program such as those from scientific and engineering field.
The concept of finite-state Markov chain (FSMC) is applied to model fading channel. Specifically, the FSMC model is used over Rayleigh and Rician fading channel. In order to build the FSMC model for both fading channel, each state is partitioned in terms of signal-to-noise ratio (SNR), and all the necessary parameters are calculated. Particular attention is paid on two methods of SNR partitioning: equal average duration method and equal stationary probability method. Once all the variables are obtained, FSMC model is evaluated using average bit error rate (BER) and average outage duration (AOD). Finally, in order to increase the performance of this model, the concept of diversity is introduced to build three pure combining techniques, all of which a re evaluated for their performance and compared each other in terms of AODs.
The superthreaded processor consists of four to eight processing units on a single chip. As computer chips advance through design there's a greater possibility to have sixteen processing units on a chip. This same processor attempts to exploit parallelism in programs that are head scratching hard to parallelize by standard approaches. In other words, it basically attempts to speed up those programs. We use the SPEC2000 benchmarks in particular to evaluate the effectiveness of this processor on a chip. However, much of my time spent in this research has been dedicated to reducing the datasets of SPEC's benchmarks, because they run excessively long with highly detailed simulators for evaluating various computer architecture, including the superthreaded processor. Up to this point I've been working on multiple benchmarks. Currently, I'm doing two different things with two of SPEC's CINT2000 benchmarks. With one the main objective is reducing the reference dataset in a real careful way, because we want the function profile of the reference dataset fairly similar to the function profile of the reduced dataset. If they do happen to be different, the benchmark will not evaluate the computer design properly, which is why we must be careful when attempting to reduce the datasets. On the side, I'm attempting to run superthreaded code consistently and successfully with the other benchmark.
It is found that the superthreaded architecture can obtain better performance than a single-threaded wide-issue superscalar processor which exploits only the relatively small amount of instruction-level parallelism available in application programs. The superthreaded processor is a multithreaded architecture that exploits thread-level parallelism with multiple threads of control to increase the number of instructions executed per cycle. A detailed execution-driven superthreaded processor simulator, SIMCA, is used to evaluate the performance of the superthreaded architecture. This superthreaded processor uses compiler-directed thread-level speculation of control and data dependencies with run-time data dependence verification software. SIMCA is tested with the SPEC benchmarks to measure the performance of the computer's processor, memory architecture, and compiler. Majority of the datasets in the benchmarks have been reduced in a defendable way to reflect the execution profiles of the original dataset accurately. However, the large input datasets result in unreasonably long simulation times when using SIMCA for evaluation. I have been working on 197.parser, a benchmark from the CINT 2000 suite. After gathering function level profiling to determine where the program spends majority of its execution time, I have been attempting to manually parallelize into superthreaded code sections of my benchmarks code. Thus, to increase the number of instructions executed simultaneously and reduce the simulation time of the benchmark.
Oxidized porous silicon is being studied as an inexpensive alternative to GaAs and high-resistivity silicon as a low-loss substrate in microwave and optical integrated circuits. To fully characterize the losses of the porous silicon, the oxide capping layer is studied. We develop a fabrication process for coplanar waveguide circuits and a simulation and measurement regimen in order to estimate the loss-tangent of SiO2 films deposited by PECVD and LPCVD. Coplanar waveguide circuits are printed on the substrates using photolithography and plating. The line losses are measured with a network analyzer. A 2D-field simulator (Ansoft Maxwell 2D Extractor) is used to determine the line loss given different values for the conductivity of the film. A curve is fit to the simulation values to interpolate an estimated value for the conductivity. The loss tangent is then calculated.
The Superthreaded Processor Architecture has the potential to increase performance in general purpose applications by using multiple concurrent threads of execution. In order to predict the performance increase made possible by the Superthreaded Processor, the SImulator for Multithreaded Computer Architectures (SIMCA) is used. Programs from the SPEC CPU2000 benchmark suite are simulated to determine in a well-defined way the performance improvements that the Superthreaded Processor offers over other proposed and existing computer architectures. Simulating the CPU2000 benchmarks on SIMCA with the reference data sets provided by SPEC takes an unreasonable amount of time. Therefore, reduced data sets are needed that allow simulations to run faster, while keeping the same functional profile as those using the reference data sets.
Hard drive data storage today is contained on sandwiched layers of thin films. These layered films, however, have a limit to their capacity. It may be possible to use a similar type of layering involving cobalt columns in a nanoporous membrane in the same capacity as regions on the thin film, and achieve a greater storage density. Electrodeposition from a 2-electrode electrolytic cell has been chosen as the means of column creation. Since the entire process has not been finalized, test depositions on both Aluminum and Cobalt substrates have been conducted. These tests have been conducted using a Cobalt Sulfate solution in the afore mentioned cell. These depositions have tested for deposition content, texturing, thickness and roughness, in order to determine the best parameters for pore deposition.
Class E power amplifiers offer the potential to efficiently and inductively transfer power for implants with low coupling coefficients (k less than .2). The goal is to have the amplifier operate at a relatively high frequency (100 MHz) since the higher the frequency, the smaller the implant. Previous work by the group has moved the operating frequency to 21 MHz, but with poor power transfer (less than 1 mW). This paper considers limitations in both physical design and testing & measurement which proved to be a stumbling blocks. It then moves on to actual improvements in both frequency and power transfer performance. Finally it offers further suggestions for future improvements for the practical implementation of the amplifier's design.
In the wireless transmission of block-coded images, we expect that some of the data will be lost. As an alternative to requesting a retransmission of lost data, we investigate a method to interpolate lost wavelet coefficients from the surrounding blocks. Specifically, we were looking at images coded using the Discrete Wavelet Transform (DWT) as specified in the upcoming JPEG2000 standard. Due to the increased complexity and slower operation of the new JPEG2000 decoder, a reconstruction algorithm should be reasonably fast and perform a visually acceptable reconstruction.
Power flow studies are the backbone of power system analysis and design, and the power flow program is the visible evidence of the importance of these studies. Currently, the University of Minnesota uses the Cornell power flow program in much of its research. The output of this program is outdated, and has very poor efficiency compared to more modern programs. Using the programming language Matlab, My mentor and I plan to design an output routine that will increase the efficiency and comprehension of the Cornell power flow program. This has required the acquisition of new logic for several of the basic functions that are performed in the program. In our first design, some problems erected concerning double circuits, which made it unusable. We are now in the process of testing a theory using linked lists.
