On inverses and quasi-inverses of linear time-varying discrete systems☆

Abstract This paper presents a solution to the general problem of finding inverses and quasi-inverses of time-varying linear discrete systems. The results are applicable to systems defined over any abstract field. Unlike fixed systems, a time-varying system does not necessarily possess an inverse, even if a delay in the output of the inverse is allowed. This inverse with delay is called a quasi-inverse. The necessary and sufficient conditions for the existence of a quasi-inverse are established. These conditions are difficult to apply, since they often require an infinite number of calculations. However, a simple necessary condition and a simple sufficient condition are derived. The existence of a quasi-inverse does not insure that the inverse system can be represented by a unique canonical form or a unique difference equation. Necessary and sufficient conditions for the existence of a unique inverse are shown. Under these same conditions, the canonical form of the inverse is the same as the canonical form of the inverse of a fixed system with the same structure. The necessary and sufficient conditions for the special case of a coding system where the system is described by a transmission matrix of finite dimension are given.

[1]  B. Elspas,et al.  The Theory of Autonomous Linear Sequential Networks , 1959 .

[2]  Bernard Friedland,et al.  A technique for the analysis of time-varying sampled-data systems , 1957, Transactions of the American Institute of Electrical Engineers, Part II: Applications and Industry.

[3]  D. Huffman A linear circuit viewpoint on error-correcting codes , 1956, IRE Trans. Inf. Theory.

[4]  Alan B. Marcovitz Sequential generation and decoding of the P-nary Hamming code (Corresp.) , 1961, IRE Trans. Inf. Theory.

[5]  K. D. Tocher,et al.  Sampled‐Data Control Systems , 1959 .