| With the chip design into the ultra-deep sub-micron and nano-technology domain, the integrity and metal layers of chip increase, P/G network becomes larger and more complex, and it carries bigger current. This makes the signal integrity and thermal reliability problems of P/G network more serious, so that it becomes an attractive topic in IC design area.This thesis focuses on the static IR-drop and the thermal reliability of Mesh-like P/G network, and makes a source code with C language to quickly solve node voltages and wire temperature. In the study of static IR-drop, the matrix auto-generation and compression technique is used to reduce memory occupation and increase the speed of solving node voltages. The algorithms of CG, BCG and ICCG are used to solve node voltages, and the results are compared each other. In the study of thermal reliability, considering the constraint of electromigration depending on the temperature, the maximum current density is determined by the electromigration and self-heating effects. Also an in-depth study of the temperature distribution of P/G network caused by the self-heating effect is presented. According to the P/G network structure, the thermal models from the conservative to the precise are developed, taking account of the cooling through visa, as well as the thermal coupling of adjacent wires. With introducing via modulation factor, the cooling of vias is contained and introducing equivalent thermal conductivity, the effect of adjacent wires is included. In order to search for hot spots of P/G network quickly, a strategy to solve the temperature is proposed in this thesis.The experiment proves that the source code can solve the node voltages quickly and precisely, and the wire temperature is acquired by proposed thermal models. The results suitably demonstrate the temperature distribution of P/G network. The algorithm strategies for solving temperature reduce the search time for hot spots, and save valuable time in the chip design. |
PDF Full Text Request