Over the past few decades, rapid strides have been made in control the
ory, resulting in the solution of several important controller design
problems: LQR (Linear Quadratic Regulator), LQG (Linear Quadratic Gaus
sian Criterion), H infinity, to name a few. Many of these controllers
have been successfully implemented in industrial applications. However
, a disadvantage with most of these methods is that they are optimal i
n only a narrow sense, and actual engineering specifications (which ar
e usually stated as competing constraints) must be translated or reint
erpreted so as to fit into the narrow framework of these methods. In p
arallel with the theoretical developments in control theory, there hav
e been significant advances in optimization theory and algorithms, as
well as an almost exponential growth in computing power, so that numer
ical controller design methods, especially those based on convex optim
ization, have become increasingly relevant. Such numerical methods enj
oy the advantage that several commonly-encountered design requirements
can be specified directly in a natural manner, and the design interac
tion between various competing performance specifications can be readi
ly studied. In this report, we describe the application of one such CA
D method for controlling the NEC laser bonder. This design, called Q-d
esign, combines the Youla parametrization of the set of achievable sta
ble closed-loop maps with convex optimization to numerically design op
timal linear controllers under multiple design specifications.