• Proceedings of the Korean Vacuum Society Conference
• /
• 1999.07a
• /
• pp.136-136
• /
• 1999

박막의 탄성특성을 평가하는 방법으로 nano-indentation, Brillouin light scattering measurement, ultrasonic surface wave measurement, bulge test, vibration membrane method 등 여러 가지가 제시되어 왔다. 최근에는 탄성특성을 평가할 수 있는 간단한 방법으로 기판 식각 기법을 이용한 freehang, bridge 방법이 제시되었다. 이중에서 bridge 방법은 간단한 식각 기법을 이용하여 얇은 박막에서도 탄성 특성을 평가할 수 있는 방법으로 제시되었다. 그러나 식각 과정에서 발생하는 patch 부분의 under-cut으로 인해 정확한 bridge의 길이를 측정할 수 없게 되어 오차가 발생하고 있다. 본 연구에서는 bridge 방법에서 발생하는 오차를 줄이기 위한 방법으로, patch 부분에 etch-stop을 제작해 줌으로서 식각 과정에서 발생하는 under-cut을 효과적으로 제거시켰다. Etch-stop은 2장의 mask를 align key를 이용하여 제작하였다. 먼저 산화막이 형성되어 있는 Si 기판위에 mask 1을 이용하여 patch 부분을 lithography 작업하고, 습식 식각 공정을 한 뒤 DLC 필름을 증착시킨다. 다음으로 mask 2를 이용하여 bridge pattern을 제작하고, DLC 필름을 증착시킨 후 lift-off 기술과 산화막 등방식각 공정을 통해 bridge를 제작하였다. 이렇게 제작된 bridge를 통해 필름이 기판에 부착되기 위해 필요한 변형률을 측정하고, 독립적으로 측정된 필름의 잔류응력과 함께 박막의 응력-변형률 관계식에 적용시켜 biaxial elastic modulus, E/(1-v)를 구할 수 있었다. Sidl 첨가된 DLC 필름은 rf-PACVD 장비를 이용하여 증착하였다. 이때 전극과 플라즈마 사이의 바이어스 음전압은 -400V로 합성압력은 10mTorr로 고정하였다. 사용한 반응가스는 벤젠(C6H6)과 희석된 실렌(SiH4:2H=10:90)이며, 희석된 실렌의 첨가량을 조절하여 필름 내에 함유된 Si의 양을 조절하였다. 각각의 조건에서 증착시간을 조절하여 필름의 두께를 조절하였다. 필름의 잔류응력은 압축잔류 응력에 의해 발생한 필름/기판 복합체의 곡률을 laser 반사법을 이용하여 측정하고, 이 결과를 Brenner 등에 의해 유도된 식을 대입하여 계산하였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.4
• /
• pp.1795-1811
• /
• 2019

In order to achieve rapid and accurate detection of vehicle objects in complex traffic conditions, we propose a novel vehicle detection method. Firstly, more contextual and small-object vehicle information can be obtained by our Joint Feature Network (JFN). Secondly, our Evolved Region Proposal Network (EPRN) generates initial anchor boxes by adding an improved version of the region proposal network in this network, and at the same time filters out a large number of false vehicle boxes by soft-Non Maximum Suppression (NMS). Then, our Mask Network (MaskN) generates an example that includes the vehicle occlusion, the generator and discriminator can learn from each other in order to further improve the vehicle object detection capability. Finally, these candidate vehicle detection boxes are optimized to obtain the final vehicle detection boxes by the Fine-Tuning Network(FTN). Through the evaluation experiment on the DETRAC benchmark dataset, we find that in terms of mAP, our method exceeds Faster-RCNN by 11.15%, YOLO by 11.88%, and EB by 1.64%. Besides, our algorithm also has achieved top2 comaring with MS-CNN, YOLO-v3, RefineNet, RetinaNet, Faster-rcnn, DSSD and YOLO-v2 of vehicle category in KITTI dataset.

• The Korean Journal of Applied Statistics
• /
• v.20 no.3
• /
• pp.551-559
• /
• 2007

The maximum likelihood estimation is not robust against outliers in the logistic regression. Thus we propose an algorithm for the robust estimation, which identifies the bad leverage points and vertical outliers by the V-mask type criterion, and then strives to dampen the effect of outliers. Our main finding is that, by an appropriate selection of weights and factors, we could obtain the logistic estimates with high breakdown point. The proposed algorithm is evaluated by means of the correct classification rate on the basis of real-life and artificial data sets. The results indicate that the proposed algorithm is superior to the maximum likelihood estimation in terms of the classification.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.26 no.10
• /
• pp.2187-2193
• /
• 2002

We present a novel method for 3D microfabrication with LIGA process that utilizes a deep X-ray mask in which a micro-actuator is integrated. The integrated micro-actuator oscillates the X-ray absorber, which is formed on the shuttle mass of the micro-actuator, during X-ray exposures to modify the absorbed dose profile in X-ray resist, typically PMMA. 3D PMMA microstructures according to the modulated dose contour are revealed after GG development. An X-ray mask with integrated comb drive actuator is fabricated using deep reactive ion etching, absorber electroplating, and bulk micromachining with silicon-on-insulator (SOI) wafer. 1mm $\times$ 1 mm, 20 $\mu$m thick silicon shuttle mass as a mask blank is supported by four 1 mm long suspension beams and is driven by the comb electrodes. A 10 $\mu$m thick, 50 $\mu$m line and spaced gold absorber pattern is electroplated on the shuttle mass before the release step. The fundamental frequency and amplitude are around 3.6 kHz and 20 $\mu$m, respectively, for a do bias of 100 V and an ac bias of 20 $V_{p-p}$ (peak-peak). Fabricated PMMA microstructure shows 15.4 $\mu$m deep, S-shaped cross section in the case of 1.6 kJ $cm^{-3}$ surface dose and GG development at 35$^{\circ}C$ for 40 minutes.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.3
• /
• pp.355-366
• /
• 2022

This paper proposes a microcode-based neural network accelerator controller for artificial intelligence accelerators that can be reconstructed using a programmable architecture and provide the advantages of low-power and ultra-small chip size. In order for the target accelerator to support various neural network models, the neural network model can be converted into microcode through microcode compiler and mounted on accelerator to control the operators of the accelerator such as datapath and memory access. While the proposed controller and accelerator can run various CNN models, in this paper, we tested them using the YOLOv2-Tiny CNN model. Using a system clock of 200 MHz, the Controller and accelerator achieved an inference time of 137.9 ms/image for VOC 2012 dataset to detect object, 99.5ms/image for mask detection dataset to detect wearing mask. When implementing an accelerator equipped with the proposed controller as a silicon chip, the gate count is 618,388, which corresponds to 65.5% reduction in chip area compared with an accelerator employing a CPU-based controller (RISC-V).

• Proceedings of the KIEE Conference
• /
• 1999.07g
• /
• pp.3276-3278
• /
• 1999

LIGA 기술을 제품의 대량생산에 적용하기 위해서는 한번에 넓은 면적을 노광할 수 있는 X-ray 마스크가 요구된다. 기존에 널리 사용되고 있는 SiN 멤브레인 마스크는 내구성이 좋지 않고 면적을 크게하기 어렵다. 따라서 본 연구에서는 이러한 단점을 보완하기 위해 상용 graphite sheet를 이용하여 X-ray 마스크를 제작하였다. 제작된 graphite 마스크와 SiN 마스크를 이용하여 동일한 조건에서 X-ray 노광 실험을 하였고 마스크의 외형변화를 관찰하였다. 그 결과 SiN 마스크는 에너지 2.3GeV, 평균 전류 110mA에서 약 18시간 만에 파괴되었으나 graphite mask는 60시간 경과 후에도 육안상의 변화는 관찰되지 않았다. 또한 graphite 마스크를 이용하여 제작된 미세구조물의 치수측정결과 오차가 $1{\mu}m$ 미만인 정밀한 구조물 제작이 가능함을 확인하였다.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.3
• /
• pp.490-493
• /
• 2021

This paper proposes an embedded system that detects mask and face recognition based on a microprocessor instead of Nvidia Jetson Board what is popular development kit. We use a class of efficient models called Mobilenets for mobile and embedded vision applications. MobileNets are based on a streamlined architechture that uses depthwise separable convolutions to build light weight deep neural networks. The device used a Maix development board with CNN hardware acceleration function, and the training model used MobileNet_V2 based SSD(Single Shot Multibox Detector) optimized for mobile devices. To make training model, 7553 face data from Kaggle are used. As a result of test dataset, the AUC (Area Under The Curve) value is as high as 0.98.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.45 no.5
• /
• pp.12-18
• /
• 2008

Ag paste was employed for source and drain electrode of OTFTs and for the data metal lines of OTFT-OLED array on PC(polycarbonate) substrate. We tested two kinds of Ag-pastes such as pastes for 325 mesh and 500 mesh screen mask to examine the pattern ability and electrical performance for OTFTs. The minimum feature size was 60 ${\mu}m$ for 325 mesh screen mask and 40 ${\mu}m$ for 500 mesh screen mask. The conductivity was 60 $m{\Omega}/\square$ for 325 mesh and 133.1 $m{\Omega}/\square$ for 500 mesh. For the OTFT performance the mobility was 0.35 $cm^2/V{\cdot}sec$ and 0.12 $cm^2/V{\cdot}sec$, threshold voltage was -4.7 V and 0.9 V, respectively, and on/off current ratio was ${\sim}10^5$, for both screen masks. We applied the 500 mash Ag paste to OTFT-OLED array because of its good patterning property. The pixel was composed of two OTFTs and one capacitor and one OLED in the area of $2mm{\times}2mm$. The panel successfully worked in active mode operation even though there were a few bad pixels.

• Journal of IKEEE
• /
• v.24 no.1
• /
• pp.161-169
• /
• 2020

In order to reduce the manufacturing cost of MCU chips used in applications such as wireless chargers and USB-C, compared to DP-EEPROM (Double Poly EEPROM), which requires 3 to 5 additional process masks, it is even more necessary MTP(Multi-Time Programmable), which is less than one additional mask and have smaller unit cell size. In addition, in order to improve endurance characteristics and data retention characteristics of the MTP memory cell due to E/P(Erase / Program) cycling, the distribution of the V_TP(Program Threshold Voltage) and the V_TE(Erase Threshold Voltage) needs to be narrow. In this paper, we proposed a current-type BL S/A(Bit-Line Sense Amplifier) circuit, WM(Write Mask) circuit, BLD(BL Driver) circuit and a algorithm, which can reduce the distribution of program and V_T and erase V_T, through compare the target current by performing the erase and program pulse of the short pulse several times, and if the current specification is satisfied, the program or erase operation is no longer performed. It was confirmed that the 256Kb MTP memory fabricated in the Magnachip semiconductor 0.13㎛ process operates well on the wafer in accordance with the operation mode.

• The Transactions of the Korean Institute of Electrical Engineers C
• /
• v.51 no.6
• /
• pp.229-235
• /
• 2002

This paper presents the fabrication of polycrystalline thin film transistor(TFT) using sequential lateral solidification(SLS) of amorphous silicon. The fabricated SLS TFT showed high Performance suitable for active matrix liquid crystal display(AMLCD). The SLS process involves (1) a complete melting of selected area via irradiation through a patterned mask, and (2) a precisely controlled pulse translation of the sample with respect to the mask over a distance shorter than the super lateral growth(SLG) distance so that lateral growth extended over a number of iterative steps. The SLS experiment was performed with 550$\AA$ a-Si using 308nm XeCl laser having $2\mu\textrm{m}$ width. Irradiated laser energy density is 310mJ/$\textrm{cm}^2$ and pulse duration time was 25ns. The translation distance was 0.6$\mu$m/pulse, 0.8$\mu$m/pulse respectively. As a result, a directly solidified grain was obtained. Thin film transistors (TFTs) were fabricated on the poly-Si film made by SLS process. The characteristics of fabricated SLS p -type poly-Si TFT device with 2$\mu\textrm{m}$ channel width and 2$\mu\textrm{m}$ channel length showed the mobility of 115.5$\textrm{cm}^2$/V.s, the threshold voltage of -1.78V, subthreshold slope of 0.29V/dec, $I_{off}$ current of 7$\times$10$^{-l4}$A at $V_{DS}$ =-0.1V and $I_{on}$ / $I_{off}$ ratio of 2.4$\times$10$^{7}$ at $V_{DS}$ =-0.1V. As a result, SLS TFT showed superior characteristics to conventional poly-Si TFTs with identical geometry.y.y.y.