Suppose that m(xi) is a trigonometric polynomial with period 1 satisfying m
(0) = 1 and \ m(xi)\(2) + \ m(xi + 1/2)\(2) = 1 for all xi in R. Let <(phi)
over cap>(xi) = Pi(j-1)(infinity) m(2(-j)xi), phi(x) = integral(-infinity)(
+infinity) <(phi)over cap>(xi)(e2 pi ix xi) d xi. The orthogonality of phi(
x), i.e., integral(-infinity)(+infinity) phi(x-m) <(phi)over bar> (x - n) d
x = delta(m,n), is related to the zeros of m(xi). In 1995, A. Cohen and R.
D. Ryan, "Wavelets and Multiscale Signal Processing," Chapman & Hall, prove
d that if m(xi) has no zeros in [-1/6, 1/6], then phi(x) is an orthogonal f
unction. In (X. Zhou and W. Su, Appl. Comput. Harmon. Anal. 8, 197-202 (200
0)) we proved that if m(xi) has no zeros in [-1/10, 1/10] and \ m(1/6)\ - \
m(-1/6)\ > 0, then phi(x) is also an orthogonal function. A natural questi
on, then, is whether this procedure can be extended to arbitrarily small in
tervals, i.e., whether for any Delta is an element of (0, 1/2) there exists
a finite set Z(Delta) such that the orthogonality of cp(x) is ensured by t
he requirement that \ m(xi)\ > 0 for \xi \ less than or equal to Delta and
for some xi is an element of Z(Delta). In this paper, we show that this is
true if and only if Delta exceeds a strictly positive Delta(0) which we der
ive explicitly. (C) 2000 Academic Press.