INFORMATION AND CONTROL 34, 55-65 (1977)
A Sampling Theorem for Duration-Limited
Functions with Error Estimates
P. L. BUTZER AND W'. SPLETTSTOSSER
Lehrstuhl A fiir Mathematik, Rheinisch-Westfdlische
Teehnische Hochschule Aachen, Aachen, West Germany
In contrast to the classical Shannon sampling theorem, signal functions are
considered which are not band-limited but duration-limited. It is shown that
these functions can be approximately represented by a discrete set of samples.
The error is estimated that arises when only a finite number of samples is
selected.
1. INTRODUCTION
The well-known sampling theorem set up by Whittaker (1915), Kotel'nikov
(1933), and Shannon (1940/1949), plays a basic role in communication, control,
and data processing. It states that every real-valued deterministic signal f(t)
that is band-limited to [--zrW, ~W] can be reconstituted by
~.f(~_)sinTr(Wt--k)
~=_~ .(wt - k) (t ~ ~). (1.1)
Since by Fourier transform theory band-limited signals must necessarily
extend for an infinite time and, conversely, a signal which only exists for a
certain time must have a spectrum (Fourier transform) which contains fre-
quencies up to infinity, Wunsch (1963) and Kioustelidis (1969) have considered
another model for a sampling theorem, one which is just as practical, namely to
reconstruct duration- (or time-) limited functions from samples. Whereas in the
Wunsch paper no proof of the corresponding result is supplied, arguments in
the proof of the Kioustelidis paper are not complete; see also the review of the
latter paper. 1 Furthermore, neither author mentioned hypotheses upon the
signal function under which their sampling theorem would hold. Nevertheless,
the Kioustelidis paper inspired the present authors to establish a sampling
theorem for continuous duration-limited functions whose spectrum
f^(v) = (1/(27r) 1/2) f(t) e-i~* at (v c R) (1.2)
1 Doetsch, G. (1970), Math. Reviews 40, 4716.
55
Copyright © 1977 by Academic Press, Inc.
All rights of reproduction in any form reserved. ISSN 0019-9958