Math. Systems Theory 22, 255-273 (1989)
Mathematical
Systems Theory
© 1989 Springer-Verlag New York Inc.
Optimal q-Markov COVER for Finite Wordlength Implementation
Darrell Williamson 1 and Robert E. Skelton 2
i Department of Systems Engineering, Research School of Physical Scien~s,
Australian National University, Canberra, ACT 2601, Australia
z School of Aeronautics and Astronautics, Purdue University,
West Lafayette, IN 47907, USA
Abstract. The existing q-Markov COVER realization theory does not take
into account the problems of arithmetic errors due to both the quantization of
states and coefficients of the reduced-order model. All q-Markov COVERs
allow some freedom in the choice of parameters. In this paper we exploit this
freedom in the existing theory to optimize the models with respect to these
finite wordlength effects.
Introduction
An asymptotically stable system can be characterized in terms of its impulse
response sequence (Markov parameters) and its output covariance sequence
(covariance parameters) due to a zero mean white-noise input process. A general
approach has been developed [KDS] for realizing a system which matches q-
Markov parameters and q-covariance parameters. Such a system is referred to as a
q-Markov COVER, and q-Markov COVERs may be generated from output data
FKDS], [SA] or from higher-order models FSC], [AS]. The Markov and covar-
iance parameters are not independent and consequently the q-Markov COVER is
not unique. In particular, all q-Markov COVERs are not related by state space
similarity transformations [SA-I. In this paper we exploit the remaining degrees of
freedom to optimize the q-Markov COVER realization with respect to an aspect of
its finite wordlength realization.
Specifically, when digital controllers are to be implemented, both the controller
coefficients and the controller states must be represented in finite wordlength
precision. This finite wordlength (FWL) representation (or quantization) causes
inaccuracies in the response when compared with the ideal (i.e., infinite precision)