Use of a by-wafer yield model for improved signal to noise on split experiments targeting systematic yield loss mechanisms

A technique for evaluating split experiments targeting systematic yield loss mechanisms is presented. Often systematic yield loss mechanisms are evaluated using probe results. These data are clouded to a great degree by random yield loss mechanisms. Random yield loss may be effectively modeled on a by-wafer basis, given a portion of the wafer real estate is allocated to random defectivity test structures. This paper describes how such a model may be used to filter the random yield loss contributions from the total yield and thereby provide an excellent measure of the systematic yield loss for each wafer. Several examples demonstrating this technique are presented.