Browsing by Author "Wu, MD"

Now showing 1 - 7 of 7

A novel fully self-aligned process for high cell density trench gate power MOSFETs
(2004)
Tsui, BY
;
Gan, TC
;
Wu, MD
;
Chou, HH
;
Wu, ZL
A novel self-aligned process for high cell density trench gate power MOSFETs with only four mask layers was proposed. The specific on-resistance can be as low as 0.21 mOmega-cm(2) with 1.5um cell pitch and 35V breakdown voltage. Because this process shrinks trench space but not trench width, the quasi-saturation phenomenon is lighter. After optimization the thickness of n- drift layer and n+ substrate, specific on-resistance lower than 0.1 mOmega-cm(2) with 0.6um technology could be expected.
Comments on "Constraining the optimization of a fuzzy logic controller"
(2001-08-01)
Wu, MD
;
Sun, CT
;
資訊工程學系
;
Department of Computer Science
Genetic algorithms (GAs) are a highly effective and efficient means of solving optimization problems. Gene encoding, fitness landscape and genetic operations are vital to successfully developing a GA. Cheong and Lai(1) described a novel method, which employed an enhanced genetic algorithm with multiple populations, to optimize a fuzzy controller, and the experimental results revealed that their method was effective in producing a well-formed fuzzy rule-base. However, their encoding method and fitness function appear unnatural and inefficient. This study proposes an alternative method of concise genetic encoding and fitness design.
Fuzzy modeling employing fuzzy polyploidy genetic algorithms
(2002-03-01)
Wu, MD
;
Sun, CT
;
資訊工程學系
;
Department of Computer Science
Fuzzy modeling generally comprises structure identification and parameter identification. The former determines the structure of a rule-base, whereas the latter determines the contents of each rule. Applying neural networks or genetic algorithms to identify the parameter sets and structures of a fuzzy system is increasingly popular owing to their ability to learn and adapt. However, most conventional approaches cannot integrate structure identification and parameter identification efficiently. This work presents a general approach to fuzzy modeling, i.e. fuzzy polyploidy genetic algorithms which integrate structure identification and parameter identification in a single evolution process. Capable of simulating the structural adaptation process of natural evolution, the proposed model is a generalized model for simultaneously optimizing both structure and parameters of fuzzy rule-bases. The structural adaptation proposed herein provides complete structural operations to simulate structural variation process and simple to complex life form of natural evolution. Illustrative examples involving typical FLCs, such as Mamdani and TSK models, demonstrate the effectiveness of applying the polyploidy scheme.
Impact of silicide formation on the resistance of common source/drain region
(2001-10-01)
Tsui, BY
;
Wu, MD
;
Gan, TC
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
Silicide had been used to reduce the sheet resistance of diffusion region for almost 20 years. However, as the silicided region becomes small, the contact resistance of silicide/silicon interface becomes higher than the resistance of the Si diffusion region such that current may not flow into the silicide layer. The effect of Mi silicide thickness and contact resistivity on the total resistance of the silicided diffusion region was studied by two-dimensional simulation. It is observed that below a threshold length, the resistance of silicided diffusion region is higher than the unsilicided diffusion region if the silicon consumption during silicide formation is taken into consideration. Thinner silicide and lower contact resistivity reduce total resistance and threshold length but the threshold length is still much longer than the typical design rule of poly-Si to poly-Si distance. It is thus recommended to inhibit silicide formation at the common source/drain region at the metal-gate generation.
Network tournament pedagogical approach involving game playing in artificial intelligence
(2003-07-01)
Wu, MD
;
Liao, YH
;
Sun, CT
;
資訊工程學系
;
Department of Computer Science
Game playing and genetic algorithms (GAs) are two important topics in artificial intelligence (AI). In this work we employ network tournament to assist in teaching these concepts associated with AI. Three exercises that implement a game-playing program are designed to help students learn relevant topics in AI. The first exercise involves game theory, e.g. mini-max search and alpha-beta pruning. The second exercise helps students understand the critical nature of a good heuristic function in game playing. And, the third exercise introduces GAs to learn a heuristic function. In addition to these exercises, this work also designs several programming toolkits to help students complete their exercises, such as a network tournament interface, a graphical man-machine interface, and a genetic algorithm based game-playing engine. These exercises encompass pertinent topics involving artificial intelligence and the network tournament. The network tournament provides a relatively easy means of Othello competition, and has merit in improving students' motivation for learning relevant topics.
Self-adaptive genetic algorithm learning in game playing
(1995)
Sun, CT
;
Wu, MD
;
資訊工程學系
;
Department of Computer Science
Trench gate power MOSFETs with retrograde body profile
(2004)
Tsui, BY
;
Wu, MD
;
Gan, TC
;
Chou, HH
;
Wu, ZL
Low specific on-resistance (R-ds,R-on) and low gate capacitance are essential for trench gate power MOSFETs. In this work, we propose a simple method of retrograde body profile to improve the two parameters simultaneously. The retrograde body MOSFET (RBMOS) is realized by high energy implantation. Because the highest channel concentration locates close to drain side, the depletion width at drain side is suppressed so that the channel length can be shortened greatly without sacrificing punch-through voltage. Low thermal budget and easy trench gate process are additional benefits. Power MOSFET with channel length of 0.4 mum, threshold voltage of 1V, and breakdown voltage of 32V is demonstrated. Comparing with the conventional device, the specific on-resistance and the figure-of-merit can be improved by 38% and 70%, respectively.