Capacitance Extraction Of Interconnect Structures In Large-Scale Integrated Circuits With Macromodels

With the development of large scale integrated circuits, the volume of integrated circuits is reduced, and connection between metal interconnects becomes more and more closely. This makes the coupling capacitance between the conductor become more and more obvious, and even more than the gate circuit itself. So it is very important to study how to extract the coupling capacitance between conductors. At present, there are a few of authoritative methods to extract the coupling capacitance. For example, Raphael RC3, and so on. The extracted result of the capacitor using these methods is very accurate, but they may need a lot of time, or they may have a huge memory consumption. Also, there are some situations where the coupling capacitance can’t be extracted using these methods. Therefore, it is imperative to propose a new method to solve these problems.The thesis uses a floating random walk algorithm (FRW) to extract the values generated by the conductor. At present, the FRW algorithm is a kind of algorithm with a very good development, it has advantages of good flexibility, low memory consumption, precision control and so on. But with the development of technology, there are some new requirements, which is FRW algorithm can’t handle or the processing result can not meet the requirements. For example the appearance of non Manhattan type geometry, Repetitive structure which often occurs in Memory chip (IC Memory) or FPGA and so on. For the above mentioned conditions, the FRW algorithm can not be used to accurately extract the capacitor. The thesis propose a new algorithm to solve the questions, it is the floating random walk algorithm with macromodels. A macromodel is established for the non Manhattan structure, repeated structures, complex substrates, and the need for secrecy, and then combines the macromodel with the FRW algorithm. The so-called macromodel is a matrix that reflects the relationship between electric potential and electric quantity within a sub region. We can use it to replace the details in the sub region. Two methods are proposed in this thesis, namely, the boundary element method and the finite difference method, they will be introduced in detail in this thesis. In addition, the thesis presents a new method to increase the accuracy of coupling capacitance extraction by modifying the Gauss point of Gauss integral. Finally, this thesis proposes a new form to storage the macromodel. In this way, the macromodel is symmetric, so we can reduce the memory consumption by storing only half of the macromodels.This thesis studies the floating random walk algorithm with macromodels, and proposes a new way to storage the macromodel, and using a new method to improve the accuracy of capacitance extraction. All programs in this thesis are implemented using C/C++ language under the Linux platform. Experiments were carried out using the new algorithm, and the obtained experimental result are compared with the results obtained by using the Raphael RC3 method. The comparison results show that the accuracy of the proposed method is in line with the requirement of the accuracy of the extraction., and the scope of the new proposed method is more widely, and the memory consumption is reduced. Moreover, the new method greatly shortens the time of capacitance extraction. For example, to repetitive structure cases, the acceleration can be up to 10 times.