• Journal of Sensor Science and Technology
• /
• v.23 no.3
• /
• pp.185-191
• /
• 2014

The width of depletion region in a varactor diode can be modulated by varying a reverse bias voltage. Thus, the preferred characteristics of depletion capacitance can obtained by the change in the width of depletion region so that it can select only the desirable frequencies. In this paper, the TV tuner varactor diode fabricated by hyper-abrupt profile control technique is presented. This diode can be operated within 3.3 V of driving voltage with capability of UHF band tuning. To form the hyperabrupt profile, firstly, p+ high concentration shallow junction with $0.2{\mu}m$ of junction depth and $1E+20ions/cm^3$ of surface concentration was formed using $BF_2$ implantation source. Simulation results optimized important factors such as epitaxial thickness and dose quality, diffusion time of n+ layer. To form steep hyper-abrupt profile, Formed n+ profile implanted the $PH_3$ source at Si(100) n-type epitaxial layer that has resistivity of $1.4{\Omega}cm$ and thickness of $2.4{\mu}m$ using p+ high concentration Shallow junction. Aluminum containing to 1% of Si was used as a electrode metal. Area of electrode was $30,200{\mu}m^2$. The C-V and Q-V electric characteristics were investigated by using impedance Analyzer (HP4291B). By controlling of concentration profile by n+ dosage at p+ high concentration shallow junction, the device with maximum $L_F$ at -1.5 V and 21.5~3.47 pF at 0.3~3.3 V was fabricated. We got the appropriate device in driving voltage 3.3 V having hyper-abrupt junction that profile order (m factor) is about -3/2. The deviation of capacitance by hyper-abrupt junction with C0.3 V of initial capacitance is due to the deviation of thermal process, ion implantation and diffusion. The deviation of initial capacitance at 0.3 V can be reduced by control of thermal process tolerance using RTP on wafer.

• Journal of Korean Society for Atmospheric Environment
• /
• v.13 no.2
• /
• pp.147-159
• /
• 1997

One of the most critical problems associated with filter-pack data comparisons from various field networks is the use of different sampling flow rates. In this field study, the effects of various filter-pack(FP) sampling conditions were examined. Experiments were conducted to evaluate the effects of varying sampling flow rates (1.5 to 10 sLpm) on measured concentrations of dry acid deposition species. Collocated FP samples were also collected to determine sampling and analysis data reproducibility. Ambient air samples were collected simultaneously for the seven day durations at varous flow rate. The chemical species measured were sulfur dioxide ($SO_2$), particulate sulfate(P-$SO_{4}^{2+}$), nitric acid ($HNO_3$), and particulate nitrate (P-$NO_{3}^{-}$). The results indicated that the collocated samples can be measured reproducibly and that sampling bias for the species measured is not significant. It was concluded that variations in the flow rates (1.5 to 10 sLpm) did not significantly affect the concentration of the species of interest. Although the results were not significantly different between different flow rates, artifacts were more likely to occur at high flow conditions (>5 sLpm) (e.g., via volatilization of particulate nitrates) than at low flow conditions(<5 sLpm).

• Korean Journal of Optics and Photonics
• /
• v.30 no.6
• /
• pp.249-254
• /
• 2019

An optical current sensor device that measures electric current by the principle of the Faraday effect was designed and fabricated. The polarization-rotated reflection interferometer and the quadrature phase interferometer were introduced so as to improve the operational stability. Complex structures containing diverse optical components were integrated in a polymeric optical integrated circuit and manufactured in a small size. This structure allows sensing operation without extra bias feedback control, and reduces the phase change due to environmental temperature changes and vibration. However, the Verdet constant, which determines the Faraday effect, still exhibits an inherent temperature dependence. In this work, we tried to eliminate the residual temperature dependence of the optical current sensor based on polarization-rotated reflection interferometry. By varying the length of the fiber-optic wave plate, which is one of the optical components of the interferometer, we could compensate for the temperature dependence of the Verdet constant. The proposed optical current sensor exhibited measurement errors maintained within 0.2% over a temperature range, from 25℃ to 85℃.

• Proceedings of the Korean Institute of Electrical and Electronic Material Engineers Conference
• /
• 2008.06a
• /
• pp.241-241
• /
• 2008

Macrofore formation in silicon and other semiconductors using electrochemical etching processes has been, in the last years, a subject of great attention of both theory and practice. Its first reason of concern is new areas of macropore silicone applications arising from microelectromechanical systems processing (MEMS), membrane techniques, solar cells, sensors, photonic crystals, and new technologies like a silicon-on-nothing (SON) technology. Its formation mechanism with a rich variety of controllable microstructures and their many potential applications have been studied extensively recently. Porous silicon is formed by anodic etching of crystalline silicon in hydrofluoric acid. During the etching process holes are required to enable the dissolution of the silicon anode. For p-type silicon, holes are the majority charge carriers, therefore porous silicon can be formed under the action of a positive bias on the silicon anode. For n-type silicon, holes to dissolve silicon is supplied by illuminating n-type silicon with above-band-gap light which allows sufficient generation of holes. To make a desired three-dimensional nano- or micro-structures, pre-structuring the masked surface in KOH solution to form a periodic array of etch pits before electrochemical etching. Due to enhanced electric field, the holes are efficiently collected at the pore tips for etching. The depletion of holes in the space charge region prevents silicon dissolution at the sidewalls, enabling anisotropic etching for the trenches. This is correct theoretical explanation for n-type Si etching. However, there are a few experimental repors in p-type silicon, while a number of theoretical models have been worked out to explain experimental dependence observed. To perform ordered macrofore formaion for p-type silicon, various kinds of mask patterns to make initial KOH etch pits were used. In order to understand the roles played by the kinds of etching solution in the formation of pillar arrays, we have undertaken a systematic study of the solvent effects in mixtures of HF, N-dimethylformamide (DMF), iso-propanol, and mixtures of HF with water on the macrofore structure formation on monocrystalline p-type silicon with a resistivity varying between 10 ~ 0.01 $\Omega$ cm. The etching solution including the iso-propanol produced a best three dimensional pillar structures. The experimental results are discussed on the base of Lehmann's comprehensive model based on SCR width.

• Journal of Korea Water Resources Association
• /
• v.48 no.6
• /
• pp.425-438
• /
• 2015

The objective of this study was to evaluate simulation of agricultural water supply considering yearly variation of irrigation efficiency. The water supply data of the Idong reservoir from 2001 through 2009 was collected and used for this study. Total 6 parameters including irrigation efficiency (Es), drainage outlet height, and infiltration, were used for sensitivity analysis, calibration, and validation. Among the parameters, the Es appeared to be the most sensitivity parameter. The Es was calibrated on a yearly basis considering sensitivity and time-varying characteristic, while other parameters were set to fixed values. The statistics of percent bias (PBLAS), Nash-Sutcliffe efficiency (NSE), and root means square error to the standard deviation of measured data (RSR) for a monthly step were 2.7%, 0.93, and 0.26 for the calibration, and 3.9%, 0.89, and 0.32 for the validation, correspondently. The results showed a good agreement with the observations. This implies that the modeling only with appropriate parameter values, apart from modeling approaches, can simulate the real supply operation reasonably well. However, the simulations with uncalibrated parameters from previous studies produced poor results. Thus, it is important to use calibrated values, and especially, we suggest the Es's yearly calibration for simulating agricultural water supply.

• Journal of the Korean earth science society
• /
• v.36 no.2
• /
• pp.139-157
• /
• 2015

We investigated the relative errors of satellite-observed Surface Skin Temperature (SST) data caused by sea ice in the northern hemispheric ocean ($30-90^{\circ}N$) during April 16-24, 2003-2014 by intercomparing MODerate Resolution Imaging Spectroradiometer (MODIS) Ice Surface Temperature (IST) data with two types of Atmospheric Infrared Sounder (AIRS) SST data including one with the AIRS/Advanced Microwave Sounding Unit-A (AMSU) and the other with 'AIRS only'. The MODIS temperatures, compared to the AIRS/AMSU, were systematically up to ~1.6 K high near the sea ice boundaries but up to ~2 K low in the sea ice regions. The main reason of the difference of skin temperatures is that the MODIS algorithm used infrared channels for the sea ice detection (i.e., surface classification), while microwave channels were additionally utilized in the AIRS/AMSU. The 'AIRS only' algorithm has been developed from NASA's Goddard Space Flight Center (NASA/GSFC) to prepare for the degradation of AMSU-A by revising part of the AIRS/AMSU algorithm. The SST of 'AIRS only' compared to AIRS/AMSU showed a bias of 0.13 K with RMSE of 0.55 K over the $30-90^{\circ}N$ region. The difference between AIRS/AMSU and 'AIRS only' was larger over the sea ice boundary than in other regions because the 'AIRS only' algorithm utilized the GCM temperature product (NOAA Global Forecast System) over seasonally-varying frozen oceans instead of the AMSU microwave data. Three kinds of the skin temperatures consistently showed significant warming trends ($0.23-0.28Kyr^{-1}$) in the latitude band of $70-80^{\circ}N$. The systematic disagreement among the skin temperatures could affect the discrepancies of their trends in the same direction of either warming or cooling.

• Journal of Intelligence and Information Systems
• /
• v.22 no.2
• /
• pp.127-142
• /
• 2016

Deep learning model is a kind of neural networks that allows multiple hidden layers. There are various deep learning architectures such as convolutional neural networks, deep belief networks and recurrent neural networks. Those have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. Among those architectures, convolutional neural networks and recurrent neural networks are classified as the supervised learning model. And in recent years, those supervised learning models have gained more popularity than unsupervised learning models such as deep belief networks, because supervised learning models have shown fashionable applications in such fields mentioned above. Deep learning models can be trained with backpropagation algorithm. Backpropagation is an abbreviation for "backward propagation of errors" and a common method of training artificial neural networks used in conjunction with an optimization method such as gradient descent. The method calculates the gradient of an error function with respect to all the weights in the network. The gradient is fed to the optimization method which in turn uses it to update the weights, in an attempt to minimize the error function. Convolutional neural networks use a special architecture which is particularly well-adapted to classify images. Using this architecture makes convolutional networks fast to train. This, in turn, helps us train deep, muti-layer networks, which are very good at classifying images. These days, deep convolutional networks are used in most neural networks for image recognition. Convolutional neural networks use three basic ideas: local receptive fields, shared weights, and pooling. By local receptive fields, we mean that each neuron in the first(or any) hidden layer will be connected to a small region of the input(or previous layer's) neurons. Shared weights mean that we're going to use the same weights and bias for each of the local receptive field. This means that all the neurons in the hidden layer detect exactly the same feature, just at different locations in the input image. In addition to the convolutional layers just described, convolutional neural networks also contain pooling layers. Pooling layers are usually used immediately after convolutional layers. What the pooling layers do is to simplify the information in the output from the convolutional layer. Recent convolutional network architectures have 10 to 20 hidden layers and billions of connections between units. Training deep learning networks has taken weeks several years ago, but thanks to progress in GPU and algorithm enhancement, training time has reduced to several hours. Neural networks with time-varying behavior are known as recurrent neural networks or RNNs. A recurrent neural network is a class of artificial neural network where connections between units form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior. Unlike feedforward neural networks, RNNs can use their internal memory to process arbitrary sequences of inputs. Early RNN models turned out to be very difficult to train, harder even than deep feedforward networks. The reason is the unstable gradient problem such as vanishing gradient and exploding gradient. The gradient can get smaller and smaller as it is propagated back through layers. This makes learning in early layers extremely slow. The problem actually gets worse in RNNs, since gradients aren't just propagated backward through layers, they're propagated backward through time. If the network runs for a long time, that can make the gradient extremely unstable and hard to learn from. It has been possible to incorporate an idea known as long short-term memory units (LSTMs) into RNNs. LSTMs make it much easier to get good results when training RNNs, and many recent papers make use of LSTMs or related ideas.