Abstract. Lower bounds for the stability margins of 2-D digital systems are extended to n-D systems. These bounds are then improved f?)r n-D (including 2-D) systems which have characteristic polynomials with I-D factor polynomials. Stability analysis of n-D systems due to finite wordlength is considered, some tight lower bounds oll coefficient wordlength which guarantee the n-D system to be stable and/or globally asymptotically stable are presented. Improved and/or extended criteria for absence of overflow oscillations and global asymptotic stability of n-D systems are proposed as well. An example is presented to illustrate the theoretical results, and it is shown that the lower bound on coefficient wordlength could be considerably improved for the (partial) factorable denominator n-D digital systems. All the discussions are based on the n-D Lyapunov equation.Key Words: Multidimensional systems, Stability margins, Finite wordlength effect, Overflow oscillations, Limit cycles, Asymptotic stability
I n t r o d u c t i o nStability of multidimensional (n-D) digital systems is an essential requirement for the design and implementation of such systems. From a practical point of view, however, designers are not only interested whether an n-D system is stable or not in theory, but also interested in the following three important stability problems:1. How far is a stable n-D system from being unstable, i.e., what is the stability margin of the n-D system? 2. When a stable n-D system is implemented with finite wordlength machines, how many bits of coefficient wordlength should be used to guarantee the n-D system to be stable?3. Does the n-D digital system have the absence of limit cycles (or overflow oscillations) due to finite word length?Concerning the first problem, the stability margin of 2-D systems was defined in [ 1 ], then it was extended to the n-D case in [2]. Algorithms for calculating the stability margins of n-D systems were presented in [ 1 ]- [4]. These results, however, were carried out in the frequency domain. In [5], Agathoklis proposed a method, based on the 2-D Lyapunov equation, to estimate lower bounds for the stability margin of 2-D digital systems. In the first part of