PROMOTION OF TRADE AND FOREIGN DIRECT INVESTMENT (FDI) BY ASEAN TO PEOPLE'S REPUBLIC OF CHINA (PRC)

Abstract

Three strategies to help circuit designers improve the manufacturing yield of their designs has been presented in this study. These strategies of yield optimization are built on the Centres of Gravity Method ("CoG Method"). A study of the CoG Method is first presented in detail. The three parameters of the algorithm, mainly step length, sample size and starting solution, are introduced and their effects on the overall performance of the CoG Method are then studied. As a result, the strengths and weaknesses of the CoG Method are clearly determined. One key strength of the CoG Method is its ability to rapidly increase the manufacturing yield of a current design. Therefore, the current design reaches the region of high yield relatively quickly. However, a major drawback is its oscillatory behaviour, which leads to an inability of the CoG Method to converge on its own. As a result, users are required to stop the algorithm at their discretion. It is shown that the inefficiency of the CoG Method's performance can be compensated by combining it with one of three proposed strategies. They are the Normal Averaging Strategy ("NA Strategy"), Weighted Averaging Strategy ("WA Strategy") and the Response Surface Methodology ("RSM'). A comparison of the three strategies shows that the amount of desired yield improvement is proportional to computational cost, i.e. the amount of time a designer is willing to spend during the optimization process.