Conducting experiments across processing steps is important when it is
desired to study the interactions among factors where some factors ar
e from one step and the other factors are from the other step. The str
ip-plot design is a very efficient design, both on resources required
and time needed, for studying multiple-step processes. This paper desc
ribes the construction and analysis of the strip-plot and split-plot d
esigns. A case study is described where an experiment was run to impro
ve the quality of separation by implantation of oxygen silicon on insu
lator wafers while reducing costs. In this study the levels of one set
of factors are not replicated; thus the half-normal probability plot
is used to extract information so that a mixed model can be used to pr
ovide estimates of the parameters of the corresponding response surfac
e model. (C)1998 John Wiley & Sons, Ltd.