Q charts provide means for statistical process control in low-volume proces
ses and start-up phases of production. Concerns on their performance have l
ed to research into different types of enhancements and much discussion on
the appropriateness of these. Driven by the aim to implement control charts
in the low-volume production of advanced wafer steppers, we investigate th
e performance of additional run rules and tightening control limits on the
traditional Q chart compared with an exponentially weighted moving average
(EWMA). Furthermore, we develop an alternative QR chart based on the mean m
oving range as estimator of the process standard deviation and consider the
economics of low-volume processes by means of a specific cost model. The c
omparisons are based on the run length distributions after a permanent shif
t and trend, both with an onset early in the process. Real life examples ar
e given for various important variables in wafer stepper production. It is
concluded that the EWMA based on QR statistics provides the best performanc
e throughout. Competing alternatives with almost equal performance are the
EWMA of Q statistics and the combination of four tests of special causes (1
-of-1, 2-of-3, 4-of-5 and 8-of-8) applied on either the Q or QR chart. Over
all, the mean moving range performs better. Copyright (C) 1999 John Wiley &
Sons, Ltd.