Browsing by Author "Liu, CF"
Now showing 1 - 2 of 2
- Results Per Page
- Sort Options
Item type:Item, A novel SiGe raised source/drain polycrystalline silicon thin-film transistor with improved on-current and larger breakdown voltage(2003-03-01) ;Peng, DZ ;Chang, TC ;Liu, CF ;Yeh, PHLiu, PTA novel poly-Si thin-film transistor with a self-aligned SiGe raised source/drain (SiGe-RSD TFT) has been proposed and fabricated. The SiGe-RSD regions were grown selectively by the ultra-high vacuum chemical vapor deposition (UHVCVD) process designed by us at 550degreesC. The resultant transistor structure features an ultra-thin active channel region (20 nm) and a self-aligned thick source/drain region (120 nm), and is ideally suited for optimum performance. Significant improvements in electrical characteristics, such as higher turn-on current, lower leakage current and higher drain breakdown voltage have been observed in the SiGe RSD TFT, compared to its conventional TFT counterpart. Moreover, the process is simple and no additional masks are necessary, which is consistent with conventional fabrication processes.Item type:Item, On the use of loss history for performance improvement of TCP over wireless networks(2002-11-01) ;Lee, CL ;Liu, CF ;Chen, YC ;資訊工程學系Department of Computer ScienceDue to the fast advances in wireless networking technology, there is an increasing number of hosts using TCP/IP to connect to the Internet via wireless links. However, it is known that TCP performs poorly on paths with wireless links. This paper presents an approach to address the problem. In the proposed approach, a sender controls the size of transmitted packets and observes the number of losses for each of the controlled packet sizes. It then estimates the number of congestion losses and random losses separately. The results are used to decide whether to reduce the window size or not when a packet loss is detected. We use ns-2 simulator to evaluate the proposed approach. The simulation results show that the proposed approach is able to achieve between 25% and 150% better throughput than FACK under the byte-error rate range of 5 x 10(-5) to 20 x 10(-5).