Dissertations / Theses on the topic 'Intrinsic noise'

Electromagnetic Interference (EMI) problems in switching power supplies have been traditionally treated with cut-and-try approaches. In recent years, advancement has been made to better understand the problems and minimize the cut-and-try portion of the design process. However, there are still phenomena difficult to explain in many practical design situations. Often, the problems may be solved by luck but many puzzles remain unsolved. If not fully understood, these puzzles are very likely to come back to haunt the designers. According to the conventional theory, there are two modes of noise: the Differential-Mode (DM) noise and the Common-Mode (CM) noise. Recently, a new noise-coupling mode called Non-Intrinsic Differential-Mode (NIDM) noise was uncovered accidentally in the process of explaining certain EMI filter action [1]. This phenomenon has never been thoroughly studied. The focus of the present thesis is to investigate the NIDM phenomenon and its implications to practical EMI filter design issues. The generation mechanism and basic characteristics of this phenomenon will be briefly reviewed, which is crucial to the understanding of the remaining parts of the research. Two essential diagnostic tools are introduced. One is the DM/CM noise separator and the other is the zero-span mode operation of a spectrum analyzer. The results of the investigation will be presented. The results will be presented using practical examples, which tie the phenomenon to filter design issues. In some examples, explanations are given to dispel the puzzles commonly encountered in the practice. A filter design procedure is suggested for off-line power supplies. This procedure incorporates the NIDM phenomenon into an existing design procedure. Only first-order and second-order filter topologies are included in the discussion.
Master of Science

So called discrete models have been successfully used in engineering and computational systems biology. This thesis discusses algebraic methods for modeling and analysis of gene regulatory networks within the discrete modeling context. The first chapter gives a background for discrete models and put in context some of the main research problems that have been pursued in this field for the last fifty years. It also outlines the content of each subsequent chapter. The second chapter focuses on the problem of inferring dynamics from the structure (topology) of the network. It also discusses the characterization of the attractor structure of a network when a particular class of functions control the nodes of the network. Chapters~3 and 4 focus on the study of multi-state nested canalyzing functions as biologically inspired functions and the characterization of their dynamics. Chapter 5 focuses on stochastic methods, specifically on the development of a stochastic modeling framework for discrete models. Stochastic discrete modeling is an alternative approach from the well-known mathematical formalizations such as stochastic differential equations and Gillespie algorithm simulations. Within the discrete setting, a framework that incorporates propensity probabilities for activation and degradation is presented. This approach allows a finer analysis of discrete models and provides a natural setup for cell population simulations. Finally, Chapter 6 discusses future research directions inspired by the work presented here.
Ph. D.

In spite of remarkable progress in molecular biology, our understanding of the dynamics and functions of intra- and inter-cellular biological networks has been hampered by their complexity. Kinetics modelling, an important type of mathematical modelling, provides a rigorous and reliable way to reveal the complexity of biological networks. In this thesis, two genetic regulatory networks have been investigated via kinetic models. In the first part of the study, a model is developed to represent the transcriptional regulatory network essential for the circadian rhythms in Drosophila. The model incorporates the transcriptional feedback loops revealed so far in the network of the circadian clock (PER/TIM and VRI/PDP1 loops). Conventional Hill functions are not used to describe the regulation of genes, instead the explicit reactions of binding and unbinding processes of transcription factors to promoters are modelled. The model is described by a set of ordinary differential equations and the parameters are estimated from the in vitro experimental data of the clocks’ components. The simulation results show that the model reproduces sustained circadian oscillations in mRNA and protein concentrations that are in agreement with experimental observations. It also simulates the entrainment by light-dark cycles, the disappearance of the rhythmicity in constant light and the shape of phase response curves resembling that of experimental results. The model is robust over a wide range of parameter variations. In addition, the simulated E-box mutation, perS and perL mutants are similar to that observed in the experiments. The deficiency between the simulated mRNA levels and experimental observations in per01, tim01 and clkJrk mutants suggests some differences in the model from reality. Finally, a possible function of VRI/PDP1 loops is proposed to increase the robustness of the clock. In the second part of the study, the sources of intrinsic noise and the influence of extrinsic noise are investigated on an intracellular viral infection system. The contribution of the intrinsic noise from each reaction is measured by means of a special form of stochastic differential equation, the chemical Langevin equation. The intrinsic noise of the system is the linear sum of the noise in each of the reactions. The intrinsic noise arises mainly from the degradation of mRNA and the transcription processes. Then, the effects of extrinsic noise are studied by means of a general form of stochastic differential equation. It is found that the noise of the viral components grows logarithmically with increasing noise intensities. The system is most susceptible to noise in the virus assembly process. A high level of noise in this process can even inhibit the replication of the viruses. In summary, the success of this thesis demonstrates the usefulness of models for interpreting experimental data, developing hypotheses, as well as for understanding the design principles of genetic regulatory networks.

Self-organizing patterns arise in a variety of ways in nature, the complex patterning observed on animal coats is such an example. It is already known that the mechanisms responsible for pattern formation starts at the developmental stage of an embryo. However, the actual process determining cell fate has been, and still is, unknown. The mathematical interest for pattern formation emerged from the theories formulated by the mathematician and computer scientist Alan Turing in 1952. He attempted to explain the mechanisms behind morphogenesis and how the process of spatial cell differentiation from homogeneous cells lead to organisms with different complexities and shapes. Turing formulated a mathematical theory and proposed a reaction-diffusion system where morphogens, a postulated chemically active substance, moderated the whole mechanism. He concluded that this process was stable as long as diffusion was neglected; otherwise this would lead to a diffusion-driven instability, which is the fundamental part of pattern formation. The mathematical theory describing this process consists of solving partial differential equations and Turing considered deterministic reaction-diffusion systems.   This thesis will start with introducing the reader to the problem and then gradually build up the mathematical theory needed to get an understanding of the stochastic reaction-diffusion systems that is the focus of the thesis. This study will to a large extent simulate stochastic systems using numerical computations and in order to be computationally feasible a compartment-based model will be used. Noise is an inherent part of such systems, so the study will also discuss the effects of noise and morphogen kinetics on different geometries with boundaries of different complexities from one-dimensional cases up to three-dimensions.

O objetivo principal deste trabalho é o estudo de transistores MuGFETs de porta tripla de Corpo de canal tipo-n com e sem a aplicação da configuração DTMOS. Este estudo será realizado através de simulações numéricas tridimensionais e por caracterizações elétricas. A corrente de dreno, a transcondutância, a resistência, a tensão de limiar, a inclinação de sublimiar e a Redução da Barreira Induzida pelo Dreno (DIBL) serão analisadas em modo DTMOS e em configuração de polarização convencional. Importantes figuras de mérito para o desempenho analógico como transcondutância-sobre-corrente de dreno, a condutância de saída, a tensão Early e o ganho de tensão intrínseco serão estudados tanto experimentalmente como através de simulações numéricas tridimensionais para diferentes concentrações de dopantes no canal. Os resultados indicam que a configuração DTMOS apresenta as características elétricas superiores (4 e 10 %) e maior eficiência dos transistores. Além disso, os dispositivos DTMOS com alta concentração de dopantes no canal apresentaram um desempenho analógico muito melhor quando comparados ao transistor de porta tripla de Corpo em modo de operação convencional. O ruído de baixa frequência (LF) é pela primeira vez experimentalmente analisado na região linear e saturação. A origem do ruído é analisada de maneira a compreender os mecanismos físicos envolvidos neste tipo de ruído. As medições mostraram que os espectros do sinal dos dispositivos de porta tripla de Corpo e DTMOS são compostos por flutuações referentes ao número de portadores devido ao ruído flicker e por ondas de ruído de geração e recombinação no dielétrico de porta que se torna maior com o aumento da tensão de porta. No entanto, o principal fato desta análise é que o dispositivo DTMOS apresentou praticamente a mesma magnitude do ruído LF na região linear e de saturação que o dispositivo de Corpo. A energia de 60 MeV na fluência de p/1012 cm-2 de radiações de prótons é também estudada experimentalmente em termos das características elétricas, desempenho do analógico e ruído LF nos dispositivos de porta tripla de Corpo e DTMOS. Os resultados indicam que combinado com as suas melhores características elétricas e um ótimo desempenho analógico do DTMOS, faz o transistor de porta tripla de Corpo um candidato muito competitivo para aplicações analógicas em ruído de baixa frequência antes e depois da irradiação. A vantagem da técnica DTMOS em transistores de porta tripla em ambientes onde os dispositivos têm de suportar alta radiação é devido à menor penetração do campo de dreno que reduz o efeito das cargas induzidas pelo óxido de isolação (STI). Finalmente, o transistor de Corpo de porta tripla de canal tipo-n é experimentalmente estudado como célula de memória, isto é, como 1T-DRAM (Memória de Acesso Aleatório Dinâmico com 1 transistor). Para escrever e ler 1 é utilizado um modo de programação que utiliza o efeito do transistor bipolar parasitário (BJT) enquanto a polarização direta da junção do corpo e do dreno é usada para escrever 0. As correntes de leitura e escrita aumentam com o aumento da tensão do corpo (VB) porque as cargas induzidas pelo efeito BJT é armazenada dentro da aleta. Quando o corpo do transistor está flutuante, o dispositivo retém mais cargas dentro da sua aleta. Além disso, transistor de Corpo pode ser utilizado como 1T-DRAM com eletrodo de porta e substrato flutuando. Neste caso, o dispositivo funciona como um biristor (sem porta).
The main goal of this work is to investigate the n-channel MuGFETs (triple-gate) Bulk transistors with and without the application of DTMOS operation. This work will be done through three-dimensional numerical simulation and by electrical characterizations. The drain current, transconductance, resistance, threshold voltage, subthreshold swing and Drain Induced Barrier Lowering (DIBL) will be analyzed in the DTMOS mode and the standard biasing configuration. Important figures of merit for the analog performance such as transconductance-over-drain current, output conductance, Early voltage and intrinsic voltage gain will be studied experimentally and through three-dimensional numerical simulations for different channel doping concentrations. The results indicate that the DTMOS configuration has superior electrical characteristics (4 e 10 %) and higher transistor efficiency. In addition, DTMOS devices with a high channel doping concentration exhibit much better analog performance compared to the normal operation mode. Low-Frequency (LF) noise is for the first time experimentally investigated in linear and saturation region. The origin of the noise will be analyzed in order to understand the physical mechanisms involved in this type of noise. Measurements showed that the signal spectra for Bulk and DTMOS are composed of number fluctuations related flicker noise with on top generation and recombination noise humps, which become more pronounced at higher gate voltage. However, the most important finding is the fact that DTMOS devices showed practically the same LF noise magnitude in linear and saturation region than standard Bulk device. Proton irradiation with energy of 60 MeV and fluence of p/1012 cm-2 is also experimentally studied in terms of electric characteristic, analog performance and the LF noise in Bulk and DTMOS triple gate devices. The results indicate that the combined of the better electrical characteristics and an excellent analog performance of DTMOS devices, makes it a very competitive candidate for low-noise RF analog applications before and after irradiation. The advantage of dynamic threshold voltage in triple gate transistors in environments where the devices have to withstand high-energy radiation is due to its lower drain electric field penetration that lowers the effect of the radiation-induced charges in the STI (shallow trench isolation) regions adjacent to the fin. Finally, the n-channel triple gate Bulk device is used for memory application, that is, 1T-DRAM (Dynamic Random Access Memory with 1 Transistor). Bipolar junction transistor (BJT) programming mode is used to write and read 1 while the forward biasing of the body-drain junction is used to write 0. The reading and writing current increases with increasing body bias (VB) because the load induced by the BJT effect is stored within the fin. When the body of the transistor is floating, the device retains more charge within its fin. In addition, transistor could also operate as 1T-DRAM with both gate and bulk contacts floating, which is similar to the biristor (gateless) behavior.

Une population d'individus génétiquement identiques partageant le même environnement présente une certaine variabilité phénotypique résiduelle. Cette hétérogénéité découle de la nature stochastique, ou aléatoire, de l'expression des gènes, également appelée bruit. Cette stochasticité résulte d'une part de la rencontre aléatoire d'espèces chimiques pendant la transcription et la traduction (bruit intrinsèque), et d'autre part des fluctuations dans la concentration de ces substances chimiques (bruit extrinsèque). Un modèle stochastique ne faisant intervenir que le bruit intrinsèque prédit que la force du bruit phénotypique varie linéairement avec l'efficacité de la traduction, mais qu'il ne dépend pas du taux de transcription. Cette prédiction s'est révélée compatible avec des données portant sur un nombre limité de souches et de conditions, mais n'a jamais été entièrement testée sur un grand nombre de souches ayant différentes efficacités de transcription et de traduction. Notre objectif est d'aller plus loin dans le test de cette prédiction en utilisant une collection d'une quarantaine de souches de la bactérie Bacillus subtilis où la protéine GFP est exprimée sous le contrôle de différents promoteurs, TSS et RBS. Pour chaque souche, l'hétérogénéité entre cellules est étudiée en quantifiant le signal de fluorescence au niveau de la cellule unique, à l'aide de techniques de cytométrie en flux et de microscopie en épifluorescence. Nos résultats montrent que, contrairement aux attentes, la force du bruit phénotypique montre une forte corrélation positive avec l'efficacité transcriptionnelle. Nous avons démontré que sur une large gamme d'expression couvrant la majeure partie du protéome de B. subtilis, le bruit d'expression est dominé par les sources de bruit extrinsèques. Par conséquent, les modèles stochastiques d'expression génique ne conviennent pas pour quantifier les effets de la traduction et de la transcription sur le bruit d'expression génétique
A population of genetically identical individuals sharing the same environment exhibits some residual phenotypic variability. Such heterogeneity arises from the stochastic, or random, nature of gene expression also referred as noise. This stochasticity results on the one hand from the random encounter of chemical species during both transcription and translation (intrinsic noise), and on the other hand from the fluctuations in the concentration of these chemicals (extrinsic noise). A stochastic model involving only intrinsic noise predicts that phenotypic noise strength varies linearly with translational efficiency but does not depend on transcriptional one. This prediction was shown to be compatible with data on a limited number of strains and conditions but has never been fully tested on a large collection of strains with different transcription and translation efficiencies. We aim to go further in the test of this prediction by using a collection of ~40 strains of the bacterium Bacillus subtilis where GFP is expressed under the control of different Promoters, TSS and RBS. For each strain, cell-to-cell heterogeneity is investigated by quantifying fluorescence signal at the single cell level, based on flow cytometry techniques and epifluorescence microscopy. Our results show that, contrary to expectations, phenotypic noise strength shows a strong positive correlation with transcriptional efficiency. We demonstrated that over a wide range of expression covering most of the proteome of B. subtilis, the expression noise is dominated by external noise sources. Therefore, stochastic models of gene expression are not suitable for quantifying the effects of translation and transcription on gene expression noise

La dynamique de formation des structures de l'Univers est habituellement décrite dans le cadre du modèle standard de Cosmologie. Cependant, pour que les observations cosmologiques soient cohérentes avec le modèle standard, il est nécessaire de supposer l'existence d'une grande proportion d'éléments de nature inconnue dans le contenu de l'Univers. Pour tenter de résoudre cette énigme, nous ne considèrerons pas d'autres sources dans le contenu de l'Univers que celles ordinaires et resterons dans le cadre de la Relativité Générale. Nous développerons néanmoins une description plus réaliste de la formation de structures dans le cadre de la théorie d'Einstein. Ainsi, contrairement au modèle standard de Cosmologie, nous ne supposerons pas que l'Univers moyenné est une solution homogène et isotrope des équations d'Einstein. Lors de mon travail sous la direction de Thomas Buchert, j'ai participé au développement d'un formalisme perturbatif permettant une description plus réaliste de la dynamique de l'espace-temps. J'ai également contribué à l'obtention de solutions relativistes à la partie gravitoélectrique des équations d'Einstein en généralisant les solutions perturbatives newtoniennes. Ces travaux ont été réalisés dans le cadre d'une approche lagrangienne intrinsèque, évitant ainsi de définir les grandeurs physiques sur un fond plat. L'approche gravitoélectromagnétique que j'ai adoptée m'a permis une interprétation nouvelle et performante des solutions des équations d'Einstein. Enfin, j'ai étudié l'impact de la topologie sur la dynamique des ondes gravitationelles à l'aide d'une description globale de l'hypersurface spatiale, permise par des théorèmes mathématiques puissants
The dynamics of structure formation in the Universe is usually described by Newtonian numerical simulations and analytical models in the frame of the Standard Model of Cosmology. The structures are then defined on a homogeneous and isotropic background. Such a description has major drawbacks since, to be self-consistent, it entails a large amount of dark components in the content of the Universe. To address the problem of dark matter and dark energy, we will neither suppose that exotic sources contribute to the content of the Universe, nor that General Relativity is obsolete. We will develop a more realistic description of structure formation in the frame of General Relativity and thus no longer assume that the average model is a homogeneous-isotropic solution of the Einstein equations, as claimed by the Standard Model of Cosmology. During my work under the supervision of Thomas Buchert, I contributed to the development of the perturbative formalism that enables a more realistic description of spacetime dynamics. In the framework of the intrinsic Lagrangian approach, which avoids defining physical quantities on a flat background, I contributed to the building of relativistic solutions to the gravitoelectric part of the Einstein equations from the generalization of the Newtonian perturbative solutions. Moreover, the gravitoelectromagnetic approach I worked with has provided a new understanding of the dynamics of the analytical solutions to the field equations. Finally, treating globally the spatial manifold, I used powerful mathematical tools and theorems to describe the impact of topology on the dynamics of gravitational waves

碩士
中原大學
應用物理研究所
95
In this thesis, we investigate the intrinsic noise in transcription and translation level of genetic regulation networks without environmental conditions. The genetic regulation networks can be mathematically described by rate equations. To obtain the stochastic features of the system, the macroscopic rate equation is first rewritten as the stochastic master equation, and then expansion method is used to obtain the linear noise Fokker-Planck equation. We use the linear noise Fokker-Planck equation to analyze the stochastic fluctuations in the regulation networks of single gene and toggle switch. The noise strength is measured by Fano Factor which is defined as variance over mean, and the correlation of noise is also analyzed. The effect of regulation strength on the characteristics of noises and discussed.

碩士
國立交通大學
電子工程系所
94
For sub-100nm MOSFETs with the gate length scaling to 80 nm and 65 nm, the unit current gain cut off frequency (fT) can achieve as high as 100 GHz and 165 GHz, respectively. However, the as-measured noise figure shows no much difference between 80 nm and 65 nm devices. The minimum noise figure (NFmin) is even higher than 5dB at 10GHz under gate bias responsible for the maximum fT. Strong finger number dependence of noise figure was also observed. All the mentioned phenomena can not be simply explained by gate resistance reduction through multi-finger structure. It suggests that noise de-embedding is required for the as-measured noise parameters. In this thesis, the basic noise theory of MOSFET, noise measurement principles and instruments will be covered in the first place. Conventional noise correlation matrix de-embedding method will be reviewed. Regarding the intrinsic MOSFET model, I-V and C-V model calibration have been done based on the measured I-V, transconductance, and admittance by Y-parameters. Then discussion of different probing pad effect on device characterization and the corresponding equivalent circuit model has been established and extensively verified. A new equivalent circuit de-embedding method was proposed. Modeling of as-measured S-parameters and noise parameters was done by incorporating the pad model with a well calibrated MOSEFT model. The lossy pad and lossy substrate de-embedding has been conducted to obtain the intrinsic characteristic. Finally, the intrinsic performance of the device will be analyzed and discussed.

碩士
國立交通大學
電信工程系所
97
As the dimension of complementary metal-oxide-semiconductor (CMOS) devices shrunk into sub-65nm scale, the threshold voltage Vth fluctuation is pronounced and becomes crucial for the design window, yield, noise margin, stability, and reliability of ultra large-scale integration circuits. Various randomness effects resulted from the random nature of manufacturing process have induced significant fluctuations of electrical characteristics in nanometer scale (nanoscale) devices and circuits. In this thesis, a three-dimensional “atomistic” coupled device-circuit simulation is intensively performed to investigate the impact of intrinsic parameter fluctuations on 16-nm-gate planar metal-oxide-semiconductor field-effect-transistor (MOSFET) static random access memory (SRAM) cells. For device with 140 mV threshold voltage, the static noise margin (SNM) of 6T SRAM with unitary cell ratio is 20 mV with 80% normalized SNM fluctuation (σSNM), which may not ensure correct operation of circuits. Thus, improvement and suppression approaches based on the circuit and device viewpoints are implemented to examine the associated characteristics in 16-nm-gate SRAM cells. From the circuit viewpoint, an 8T planar SRAM architecture is explored. Compared with the conventional 6T SRAM, under the same Vth = 140 mV, the SNM is enlarged to 233 mV and the SNM is reduced to 22 mV (around 9.5% normalized SNM) at a cost of 30% extra chip area. To prevent the increase of chip area, silicon-on-insulator fin-type field-effect-transistors (SOI FinFETs) replaced the planar MOSFETs in 6T SRAM is further examined. The SNM of 6T SOI FinFETs SRAM is 125 mV and the normalized σSNM is suppressed significantly to 5.3% (6.8 mV in σSNM). The 8T SRAM architecture can provide largest SNM and is promising in near future design; however, to prevent the increase of chip area and suppress the intrinsic parameter fluctuations, development of fabrication for SOI FinFET SRAM is crucial for sub-22nm technology.

碩士
國立交通大學
電子研究所
106
In this thesis, an extensive investigation has been performed on the three- and four-terminal (3T and 4T) multi-finger nMOSFETs for precise extraction of intrinsic parasitic RLC and development of truly accurate intrinsic MOSFET model aimed at high frequency and RF noise simulation for nano-CMOS RF circuits design. Gate resistance (Rg), source resistance (Rs,int), and gate sidewall as well as finger-end fringing capacitances (Cof and Cf(poly-end)) appear as most important intrinsic parasitic RC with critical impact on high frequency performance and RF noise, and bring tough challenges to the conventional extraction methods when applied to 3T and 4T multi-finger MOSFETs. For the first time, a new extraction flow has been established for precise determination of the intrinsic parasitic source and drain resistances (Rs,int and Rd,int) and channel resistance (Rch) in 3T and 4T multi-finger MOSFETs and enable accurate prediction of the asymmetric gate to source/drain capacitances, i.e. Cgs≠Cgd at VDS=0 with critical dependence on Rs,int, Rd,int, and Rch. Afterwards, a new method and analytical model have been derived for accurate extraction of Rg and prediction of Rg@Y-method incorporating the Rs,int and Rd,int coupled through the intrinsic gate to source and drain capacitances (Cgs,i and Cgd,i). In this thesis, one more innovation creates new structures, namely multi-finger field devices for direct and precise extraction of Cof and Cf(poly-end) from high frequency measurement, without resort to 3-D interconnect RC simulation like Raphael. The mentioned innovations lead to successful extraction of intrinsic parasitic RC with complicated layout dependence and the integration with intrinsic device parameters determined by our proprietary high precision device parameters extraction method (US patent 8,691,599 B2) can realize the actual intrinsic MOSFET model for 3T and 4T multi-finger MOSFETs with proven accuracy for layout dependent effects and sensitivity to lot-to-lot and die-to-die variations. The actual intrinsic MOSEFT model has been extensively verified and the accuracy is proven by good agreement with high frequency Y-parameters after openM1 and shortM1 deembedding for 3T and 4T nMOSFETs with various multi-finger layouts. Furthermore, the intrinsic device parameters and parasitic RLC with proven accuracy, when applied to analytical models can reach accurate prediction of the high frequency performance like fT and fMAX associated with various multi-finger layouts and facilitate layout optimization. One of the important findings and conclusions is that 4T multi-finger MOSFETs with sufficient freedom for various circuit topologies like common source, common gate, and common drain (CS, CG, and CD) under various body biases, generally suffer significant degradation of fT and fMAX due to drastic increase of Rs,int and Ls,int. The mentioned achievements provide a useful and efficient solution for high frequency simulation and design, without resort to BSIM-4 with limited accuracy for specified sample layouts. Finally, the actual intrinsic MOSFET models can be further integrated with our proprietary lossy substrate model to build up a full equivalent circuit model, which can accurately simulate the high frequency S-, Y- and noise parameters, prior to deembedding. Furthermore, the lossy substrate deembedding method can be applied to both 3T and 4T multi-finger MOSFETs as a reliable solution for accurate extraction of intrinsic RF noise, which can eliminate the problems of conventional noise correlation matrix method and successfully identify the layout dependent effects in the truly intrinsic RF noise for multi-finger MOSFETs optimization aimed at low noise circuits design.

碩士
國立交通大學
電子研究所
106
In this thesis, one of our new proprietary structures, namely multi-finger field device has been designed and implemented in nanoscale CMOS processes like TN90GUTM and TN40G to realize direct and precise extraction of the gate sidewall and finger-end fringing capacitances, denoted as Cof and Cf(polyend). For the first time, the experimental Cof and Cf(polyend) can be achieved to serve as useful database for a serious calibration on the 3-D interconnect RC simulation like Raphael. Moreover, the multi-finger field devices can enable another innovative application, such as a truly clean deembedding of the Cof and Cf(polyend), namely field deembedding for the extraction of ideally intrinsic device parameters and high frequency performance parameters like fT and fMAX. For TN90GUTM and TN40G devices with gate length (Lg) pushed to the scales of 50 nm and 30nm, how to precisely separate and extract the intrinsic channel length (Lch) and source/drain extension (SDE) to gate overlap length (LSDE) becomes a very challenging work. Unfortunately, the original approach based on openM1 deembedding and Raphael simulation led to the abnormal results, such as signficant variation of LSDE and Lch associated with various finger width (WF), and the trend of shorter LSDE but longer Lch corresponding to the smaller WF i.e. the larger NF (finger number). In this thesis, the adoption of field deembedding method can eliminate the mentioned problem and achieve nearly constant LSDE among various WF and NF, i.e. nearly independent of WF and NF. It becomes an important progress to realize accurate extraction of the effective mobility associated with the intrinsic (inversion) channel and SDE (accumulation) regions, such as eff,ch and eff,acc, respectively. Furthermore, the ideal intrinsic Y- and H-parameters achieved after field deembedding can yield the ideally intrinsic gm, Cgg, and Cgd, and more importantly the ideal intrinsic fT and gate delay, int. For TN90GUTM nMOSFET, the ideal intrinsic peak fT can reach 227 GHz, which is around 27.5% higher than the actual intrinsic peak fT of 178 GHz. As For TN40G nMOSFET, there is around 75~100% enhancement of the peak fT compared to that of TN90GUTM, for actual and ideal intrinsic conditions. The ideal intrinsic peak fT can reach up to 473GHz, which is around 52% higher than the actual intrinsic peak fT of 312 GHz. The results reveal dramatic impact from the intrinsic parasitic RC on the high frequency performance and the impact increases in more advanced technology with further scaled devices. Multi-OD (MOD) MOSFETs have been designed and fabricated in TN90RF as a potential solution for effective reduction of STI compressive transverse stress ⊥ and source resistance (RS), aimed at the increase of eff and gm and eventually fT improvement compared to the multi-finger MOSFETs with the same channel width (WOD). In this thesis, a new extraction flow has been developed for MOD MOSFETs with various WOD and NOD at the same WF=WOD×NOD and NF as those of multi-finger MOSFETs. The new features specific to MOD MOSFETs can be summarized as a new component of gate sidewall fringing capacitance originated from the gate on STI between adjacent OD (Cof,STI) and steeper STI sidewall profile due to the minimum OD-OD space, i.e. STI width. The basic device parameters extracted from the MOD MOSFETs indicate major differences in Lg and W and apparently smaller W compared to that of multi-finger MOSFETs, which is very critical for accurate determination of effective channel width (Weff) and extraction of eff. Through an extensive DC and RF characterization, the MOD nMOSFETs demonstrate some attractive features, such as the higher eff, smaller RS, larger gm, and most importantly the higher fT compared with the multi-finger MOSFETs with the same WOD. However, the MOD MOSFETs reveal two major drawbacks, such as larger DIBL and lower fMAX in comparison with the multi-finger MOSFETs. The former can be understood through an analysis of the finger-end and inter-OD fringing capacitances, i.e. Cf(poly-end) and Cof,STI, achieved from Raphael simulation. The results indicate that the significant increase of Cf(poly-end)NF in multi-finger MOSFETs with large NF becomes the dominant factor responsible for the effective suppression of DIBL. In comparison, the MOD MOSFET even with the addition of Cof,STINODNF but the combination of Cf(poly-end)NF and Cof,STINODNF with NF fixed at the minimum always keeps smaller than Cf(poly-end)NF with much larger NF and it leads to less suppression of DIBL. As for the much lower fMAX suffered by MOD MOSFETs even with higher fT, it accounts for the critical impact from the significant increase of gate resistance (Rg) and tough challenge to simultaneous optimization of fT and fMAX. Full equivalent circuit models have been established with the actual intrinsic MOSFET models for multi-finger and MOD MOSFETs by adopting intrinsic parasitic RLC and lossy substrate RLC network for accurate simulation of high frequency characteristics and RF noise. The actual intrinsic MOSFET models including layout dependent device parameters and parasitic RLC with proven accuracy for various layouts can fix the problem of conventional compact model like BSIM-4 with limited accuracy for specific sample layouts in the PDK but severe deviation for the customer designed layouts beyond the PDK. RF noise simulation and intrinsic noise extraction appear as one more serious challenge to the conventional compact model. In this thesis, our developed equivalent circuit model can provide an effective solution for accurate simulation of the RF noise prior to deembedding and precise extraction of the actual intrinsic noise by using lossy substrate deembedding. Finally, the simulation by equivalent circuit model can facilitate the development of analytical models for quick calculation of intrinsic RF noise and assessment of layout dependent effects associated with multi-finger and MOD devices and impact from the parasitic RLC.

碩士
國立交通大學
物理研究所
104
It is rather often that intrinsic and extrinsic noises coexist in a biological system. A typical example is the transcription and translation of DNA. This raises numerous works devoting to decomposing these two effects. Apart from the previous definition of intrinsic and extrinsic noises from biological aspect, this work adopts the definition of van Kampen from physical aspect. Based on that definition, we perturb the chemical master equation and derive a physical version of decomposition formula for intrinsic and extrinsic noises. We apply this theory to the network model of Yi-Der Chen. The derived exact solutions and numerical studies on that model reveal the possibility of “suppressing intrinsic noise induced stochasticity by extrinsic noises”. The condition for this possibility is derived in a low dimensional network. This suppression indicates that the fluctuations could decline when intrinsic and extrinsic noises are added together, which is a bit counter-intuition. We compare this physical version of decomposition formula with the biological version of formula derived by Swain and analyze the common and distinct features between these two formulas.

The degree by which a species can adapt to the demands of its changing environment defines how well it can exploit the resources of new ecological niches. Since the nervous system is the seat of an organism's behavior, studying adaptation starts from there. The nervous system adapts through neuronal plasticity, which may be considered as the brain's reaction to environmental perturbations. In a natural setting, these perturbations are always changing. As such, a full understanding of how the brain functions requires studying neuronal plasticity under temporally varying stimulation conditions, i.e., studying the role of plasticity in carrying out spatiotemporal computations. It is only then that we can fully benefit from the full potential of neural information processing to build powerful brain-inspired adaptive technologies. Here, we focus on homeostatic plasticity, where certain properties of the neural machinery are regulated so that they remain within a functionally and metabolically desirable range. Our main goal is to illustrate how homeostatic plasticity interacting with associative mechanisms is functionally relevant for spatiotemporal computations. The thesis consists of three studies that share two features: (1) homeostatic and synaptic plasticity act on a dynamical system such as a recurrent neural network. (2) The dynamical system is nonautonomous, that is, it is subject to temporally varying stimulation. In the first study, we develop a rigorous theory of spatiotemporal representations and computations, and the role of plasticity. Within the developed theory, we show that homeostatic plasticity increases the capacity of the network to encode spatiotemporal patterns, and that synaptic plasticity associates these patterns to network states. The second study applies the insights from the first study to the single node delay-coupled reservoir computing architecture, or DCR. The DCR's activity is sampled at several computational units. We derive a homeostatic plasticity rule acting on these units. We analytically show that the rule balances between the two necessary processes for spatiotemporal computations identified in the first study. As a result, we show that the computational power of the DCR significantly increases. The third study considers minimal neural control of robots. We show that recurrent neural control with homeostatic synaptic dynamics endows the robots with memory. We show through demonstrations that this memory is necessary for generating behaviors like obstacle-avoidance of a wheel-driven robot and stable hexapod locomotion.