<p>This thesis presents the noise characterization, modeling, and simulation of deep sub-100nm bulk MOSFETs and predicts the noise behavior for future technology nodes. There are two main subjects discussed in this thesis. First, we present the impact of scaling of MOSFETs on channel thermal noise. Second, we investigate how the technology development can affect noise performance of a single transistor.</p> <p>In the first topic, analytical MOSFET channel thermal noise expressions are presented and verified. We calibrate our model using experimental data from devices in 60 nm technology node. The technology scaling issue of MOSFETs on noise performance is also examined by applying the parameters predicted in the International Technology Roadmap of Semiconductor (ITRS).</p> <p>In the second topic, a new figure of merit, namely equivalent noise sheet resistance, is defined for the first time to demonstrate the impact of scaling. This new figure of merit represents the intrinsic part of the equivalent noise resistance that excludes the geometry information of the device, which captures the technology related parameters of transistors. By defining equivalent noise sheet resistance, we can provide process information not only for IC designers but also for process engineers.</p> / Master of Applied Science (MASc)

With the advance in nanotechnology, we are more interested in the "smaller worlds". One of the practical applications of this is to measure a very small displacement or the mass of a nano-mechanical object. To measure such properties, one needs a very sensitive detector. A quantum point contact (QPC) is one of the most sensitive detectors. In a QPC, electrons tunnel one by one through a tunnel junction (a "hole"). The tunnel junction in a QPC consists of a narrow constriction (nm-wide) between two conductors. To measure the properties of a nano-mechanical object (which acts as a harmonic oscillator), we couple it to a QPC. This coupling effects the electrons tunneling through the QPC junction. By measuring the transport properties of the tunneling electrons, we can infer the properties of the oscillator (i.e. the nano-mechanical object). However, this coupling introduces noise, which reduces the measurement precision. Thus, it is very important to understand this source of noise and to study how it effects the measurement process. We theoretically study the transport properties of electrons through a QPC junction, weakly coupled to a vibration mode of a nano-mechanical oscillator via both the position and the momentum of the oscillator. %We study both the position and momentum based coupling. The transport properties that we study consist of the average flow of current through the junction, given by the one-time correlation of the electron tunneling event, and the current noise given by the two-time correlation of the average current, i.e, the variance. The first comprehensive experimental study of the noise spectrum of a detector coupled to a QPC was performed by the group of Stettenheim et al. Their observed spectral features had two pronounced peaks which depict the noise produced due to the coupling of the QPC with the oscillator and in turn provide evidence of the induced feedback loop (back-action). Benatov and Blencowe theoretically studied these spectral features using the Born approximation and the Markovian approximation. In this case the Born approximation refers to second order perturbation of the interaction Hamiltonian. In this approximation, the electrons tunnel independently, i.e., one by one only, and co-tunneling is disregarded. The Markovian approximation does not take into account the past behavior of the system under time evolution. These two approximations also enable one to study the system analytically, and the noise is calculated using the MacDonald formula. Our main aim for this thesis is to find a suitable theoretical model that would replicate the experimental plots from the work of Stettenheim et al. Our work does not use the Markovian approximation. However, we do use the Born approximation. This is justified as long as the coupling between the oscillator and QPC is weak. We first obtain the non-Markovian unconditional master equation for the reduced density matrix of the system. Non-Markovian dynamics enables us to study, in principle, the full memory effects of the system. From the master equation, we then derive analytical results for the current and the current noise. Due to the non-Markovian nature of our system, the electron tunneling parameters are time-dependent. Therefore, we cannot study the system analytically. We thus numerically solve the current noise expression to obtain the noise spectrum. We then compare our noise spectrum with the experimental noise spectrum. We show that our spectral noise results agree better with the experimental evidence compared to the results obtained using the Markovian approximation. We thus conclude that one needs non-Markovian dynamics to understand the experimental noise spectrum of a QPC coupled to a nano-mechanical oscillator. / Physics

In a time when urban areas continue to expand, environmental noise pollution especially from road traffic remains a big challenge. This study was aimed at using open-source GIS tools to predict road traffic noise pollution using the mid-sized city of Gävle as a case study. The noise indicators measured were the equivalent day (Lday), evening (Levening), nighttime (Lnight), and the equivalent daily average (Lden). Traffic data (composition and flow of vehicles on selected roads), traffic source characteristics (road gradient, road surface type), and buildings (geometry) were integrated into Quantum GIS (QGIS) using the CNOSSOS-EU prediction method packaged in OpeNoise, a QGIS plug-in. The resultant noise levels at receiver points were interpolated using the Inverse Distance Weighting method to create noise maps for the city.  The results showed the maximum equivalent day, evening, nighttime predicted noise levels at 85 dB (A), 80 dB (A), 75 dB (A) respectively while the maximum for overall daily average noise level predicted was 85 dB (A). These limits far exceed population exposure threshold limits for the onset of annoyance (55 dB (A)) and sleep disturbance (40 dB (A)). This result is indicative of a poor sound acoustic environment. The pattern of noise level across the city was found to follow street connectivity and traffic intensity. The maximum noise levels were clustered around the highway. Within the city, areas with the highest noise levels were found close to main roads. Residential areas served by service roads were areas with the lowest noise levels. Predicted daytime noise levels (Lday) were compared with 60-second measurements of equivalent noise levels measured at 85 locations during the day in residential and mixed land use areas in the city. The mean of differences between predicted and observed noise levels was found at +1 dB for both residential and areas of mixed land use respectively. Correlation and regression analyses performed for observed and predicted values showed an initial weak positive association with a correlation coefficient of 0.21. However, when outliers were excluded, a correlation coefficient of 0.69 was observed indicating a strong association and linear relationship between the observed and predicted noise levels. Most outliers were underestimations recorded in residential areas at hidden facades. These were attributed to local effects at the measuring locations and assumptions made for building diffraction. The application of the CNOSSOS-EU method in this study did not consider attenuation from ground reflection and terrain effects. Despite these limitations, the results show that the CNOSSOS-EU has good predictive power. However, this study has only been exploratory in nature. It is recommended that further studies be performed with this model as well as in comparison with other models to find the one that best reflects the acoustic environment of the city. A wide application of the CNOSSOS-EU method across several cities will be integral in increasing our understanding of its strengths and weaknesses.

No description available.

The far field acoustics generated by turbulent flow over rough surfaces has been experimentally investigated in an acoustically treated wall jet facility. The facility allows direct measurement of the far field sound from small patches of surface roughness, without contamination from edge or other aerodynamic noise sources. The facility is capable of generating turbulent boundary layer flows with momentum thickness Reynolds numbers between 450 and 1160. The variation of surface conditions tested cover the range from hydrodynamically smooth surfaces through most of the transitional range, with h+ variations from 3 to 85. Single microphone narrow band acoustic spectra, measured in the far field, show sound levels as much as 15 dB above the background from 0.186 m2 roughness patches. The measurements revealed the spectral shape and level variations with flow velocity, boundary layer thickness, and roughness size; providing the first data set large enough to assess the affects of many aerodynamic properties on the acoustic spectra. Increases in the size of grit type roughness produced significant increases in acoustic levels. Patches of hydrodynamically smooth roughness generated measurable acoustic levels, confirming that acoustic scattering is at least one of the physical mechanisms responsible for roughness noise. The shapes of the measured spectra show a strong dependence on the form of the surface roughness. The acoustic spectra generated by periodic two-dimensional surfaces have a much narrower louder peak than that generated by three-dimensional grit type roughness. Measurements also show the orientation of the two-dimensional surface significantly affects the acoustic levels and directivity. The variation of sound levels with flow velocity and roughness size suggests the acoustic field is significantly affected by changes in the near wall flow due to the presence of the roughness. Current models of noise generated by rough surfaces predict the general trends seen in measurements for flows over grit and two-dimensional roughness in the range of 20<h+<50. However, in cases with a low Reynolds number or large grit size, where the roughness is likely to significantly affect the hydrodynamic pressure field, the scattering models did not perform as well. / Ph. D.

An innovative application of existing technology is proposed for attenuating the effects of transient phenomena, such as rotor-stator and rotor-strut interactions, linked to noise and fatigue failure in turbomachinery environments. A computational study was designed to assess the potential of Passive Porosity Technology as a mechanism for alleviating interaction effects and radiated noise by reducing the fluctuating forces acting on the vane surfaces. The study involved a typical high bypass fan stator airfoil immersed in a subsonic free field and exposed to the effects of a transversely moving wake. Time histories of the primitive aerodynamic variables obtained from Computational Fluid Dynamics (CFD) calculations were input into an acoustic prediction code to estimate noise levels at a radial distance of ten chords from the stator airfoil. This procedure was performed on the solid airfoil to obtain a baseline, and on approximately fifty porous configurations in order to isolate those that would yield maximum noise reductions without compromising the aerodynamic performance of the stator. It was found that, for a single stator immersed in a subsonic flow field, communication between regions of high pressure differential - made possible by the use of passive porosity - tends to induce a time-dependent oscillatory pattern of small inflow-outflow regions near the stator leading edge (LE), which is well established before wake effects come into play. The oscillatory pattern starts at the LE, and travels downstream on both suction and pressure sides of the airfoil. The amplitude of the oscillations seemed to be proportional to the extension of the porous patch on the pressure side. Regardless of this effect, which may not have occurred if the airfoil were placed within a stator cascade, communication between regions of high pressure differential is necessary to significantly alter the noise radiation pattern of the stator airfoil. Whether those changes result in noise abatement or enhancement depends primarily on the placement and extension of the porous patches. For most viable configurations, porosity reduced loading noise but increased thickness noise. Variations in nominal porosity were of secondary importance. In general, the best aerodynamic performers (i.e., those configurations that were able to reduce unsteady lift without severely altering the lift and/or drag characteristics of the solid airfoil) were also the best acoustic performers. As a result of using passive surface porosity, overall peak radiated noise was reduced by approximately 1.0 dB. This reduction increased to about 2.5 dB when the effects of loading noise alone were considered. / Ph. D.

Active structural acoustic control has previously been used to reduce low-frequency sound radiation from relatively simple laboratory structures. However, significant implementation issues have to be addressed before active control can be used on large, complex structures such as an aircraft fuselage. The purpose of this project is to extend decentralized structural control systems from individual bays to more realistic airframe structures. In addition, to make this investigation more applicable to industry, potential control strategies are evaluated using a realistic aft-cabin disturbance identified from flight test data. This work focuses on decentralized control, which implies that each control unit is designed and implemented independently. While decentralized control systems are relatively scalable, performance can be limited due to the destabilizing interaction between neighboring controllers. An in-depth study of this problem demonstrates that the modeling error introduced by neighboring controllers can be expressed as the product of the complementary sensitivity function of the neighboring control unit multiplied by a term that quantifies the diagonal dominance of the plant. This understanding can be used to improve existing control strategies. For instance, decentralized performance can often be improved by penalizing control effort at the zeros of the local control model. This stabilizes each control unit and reduces the modeling error induced on neighboring controllers. Additional analyses show that the performance of decentralized model-based control systems can be improved by augmenting the structural damping using robust, low-authority control strategies such as direct velocity feedback and positive position feedback. Increasing the structural damping can supplement the performance of the model-based control strategy and reduce the destabilizing interaction between neighboring control units. Instead of using low-authority controllers to stabilize the decentralized control system, another option is to modify the model-based design. Specifically, an iterative approach is developed and validated using real-time control experiments performed on a structural-acoustic system with poles close to the stability boundary, non-minimum phase zeros, and unmodeled dynamics. Experiments demonstrate that the iterative control strategy, which combines frequency-shaped linear quadratic Gaussian (LQG) control with loop transfer recovery (LTR), is capable of achieving 12dB peak reductions and a 3.6dB integrated reduction in radiated sound power from a rib-stiffened aluminum panel. / Ph. D.

A novel concept in noise control devices for landing gears is presented. These devices consist of elastic membranes creating a fairing around the major noise sources. The purpose of these devices is to reduce wake interactions and to hide components from the flow, thus, reducing the noise emission. The design of these fairings was focused on the major noise sources identified in a 777 main landing gear. To find the major noise sources, an extensive noise source identification process was performed using phased arrays. To this end, phased array technologies were developed and a 26%-scale 777 main landing gear model was tested at the Virginia Tech Stability Wind Tunnel. Since phased array technologies present some issues leading to misinterpretation of results and inaccuracy in determining actual levels, a new approach to the deconvolution of acoustic sources has been developed. The goal of this post-processing is to "simplify" the beamforming output by suppressing the sidelobes and reducing the sources mainlobe to a small number of points that accurately identify the noise sources position and their actual levels. To this end, the beamforming output is modeled as a superposition of "complex" point spread functions and a nonlinear system of equations is posted. Such system is solved using a new 2-step procedure. In the first step an approximated linear problem is solved, while in the second step an optimization is performed over the nonzero values obtained in the previous step. The solution to this system of equations renders the sources position and amplitude. The technique is called: noise source Localization and Optimization of Array Results (LORE). Numerical simulations as well as sample experimental results are shown for the proposed post-processing. / Ph. D.

Magnetic sensors are widely used in the field of mineral, navigational, automotive, medical, industrial, military, and consumer electronics. Many magnetic sensors have been developed that are generated by specific laws or phenomena: such as search-coil, fluxgate, Hall Effect, anisotropic magnetoresistance (AMR), giant magnetoresistance (GMR), magnetoelectric (ME), magnetodiode, magnetotransictor, fiber-optic, optical pump, superconducting quantum interference device (SQUID), etc. Each of these magnetic field sensors has their merits and application areas. For low power consumption (<10uW), quasi-static frequency (<10Hz) and high sensitivity (<nT) application, magnetoelectric laminate sensors offer the best potential capability and thus are the topic of my dissertation. Here, in this thesis, I have focused on designs and optimizations of magnetoelectric sensor units (i.e., sensors and circuit). To achieve my goals, I have developed some useful rules for ME sensor and detection circuit design. For ME sensor optimization, designs should consider both frequencies far away from resonance and at resonance. For the former one, both internal and external noise contribution must be considered, as one of them will limit practical applications. With regards to the internal noise sources, I have developed two design optimization methods, designated as ”'scale effect” and “ME array”. I showed that they have the ability to increase the magnetic field detection sensitivity, which was verified by experiments. With regard to external noise consideration, I have investigated how the fundamental extrinsic noise sources (temperature fluctuation, vibration, etc) affect ME laminate sensors. A concept of separating signal and noise modes into difference is put forward. Optimization with this concept in mind required us to redesign the internal structure of ME laminate sensors. At the resonant frequency, the ME voltage coefficient α_ME is the most important parameter. To enhance resonant gain in α_ME, I have developed a three phase laminate concept, which is based on increasing the effective mechanical factor Q while reducing the resonant frequency. A ME voltage coefficient of α_ME ~40V/cm.Oe has been achieved at resonance, which is about 2x higher than that of a conventional bending mode. Investigations of detection circuit optimization were also performed. Component selection strategies and a new charge topology were considered. Proper component values were required to optimize the charge detection scheme. It was also found, under some specific conditions to satisfy the circuit stability, that if the lowest required measurement frequency of the charge source was f1, then that it was not necessary to make the high corner frequency <i>f</i>_p of the charge amplifier lower than <i>f</i>₁: as doing so would decrease the system's signal-to-noise ratio (SNR). A high pass, high order filter placed behind the charge amplifier was found to increase the charge sensitivity, as it narrows the intrinsic noise bandwidth and decreases the output noise contribution, while only slightly affecting the signal's output amplitude. Prototype ME unit were also constructed, and their noise level simulated by Pspice. Experimental results showed that prototypes ME unit can reach their detection limit. In addition, a new magneto-electric coupling mechanism was also found, which had a giant ME effect. / Ph. D.

Research presented herein involved the use of a smart skin treatment used for the attenuation of turbulent boundary layer induced interior noise. The treatment consisted of several Smart Foam actuators each having a reference and error sensor along with a feed forward, filtered-x controller. Studies were performed to determine if the use of multiple instances of single input, single output (SISO) control systems could be implemented with success given the difficulty of actively suppressing turbulent boundary layer induced interior noise. Further, this research will lead to the development of an integrated Smart Foam element consisting of a Smart Foam actuator, reference sensor, error sensor and SISO controller in one complete, stand--alone unit. Several topics were studied during this effort: reference sensing, error sensing, actuator design, controller causality, correlation of turbulent flow and resulting plate vibration, and coherence between plate vibration and the interior noise field. Each study was performed with the goal of improving the performance of active attenuation of turbulent boundary layer induced interior noise. Depending on the configuration of the control system, control was performed using either experiments or simulations based on experimental data. Within the desired control band of 400--800~Hz, attenuation of up to -3.1~dB$_A$ was achieved at the error sensors and up to -1.4~dB$_A$ within the observer plane relative to the uncontrolled case. However, over a band of greater coherence from 480--750~Hz, attenuation of up to -4.8~dB$_A$ was achieved at the error sensors and up to -2.6~dB$_A$ within the observer plane. Further, peak attenuation of up to -12~dB$_A$ was achieved within the observer plane. Studies were also conducted to increase the low frequency performance of the Smart Foam treatment. These experiments used tuning masses placed on the tops of the integrated Smart Foam elements to tune them to the fundamental mode of the vibrating plate. This treatment was used to reactively attenuate plate vibration such that the radiated acoustic field would be minimized. These experiments resulted in -6~dB$_A$ global attenuation at the plate fundamental resonance. Further, it was shown that the reactive treatment did not inhibit active control. / Ph. D.