Boris Tsirelson,
"Spectral densities describing off-white noises."
Annales de l'Institut Henri Poincare Probabilites et Statistiques 38:6, 1059-1069 (2002).
Available online from Elsevier (not free, sorry; if refused by Elsevier, use arXiv or ask me for reprint):
http://www.elsevier.com/gej-ng/10/37/26/51/57/44/abstract.html

Also available online (free of charge) from e-print archive (USA):
arXiv.org/abs/math.PR/0104027/
or its Israeli mirror:
il.arXiv.org/abs/math.PR/0104027/

A research paper, 12 pages, bibl. 6 refs.

For the white noise, the spectral density is constant, and the past (restriction to (-infinity,0)) is independent from the future (restriction to (0,+infinity)). If the spectral density is not too far from being constant, then dependence between the past and the future can be eliminated by an equivalent measure change; that is called an off-white noise. I derive from well-known results a necessary and sufficient condition for a spectral density to describe an off-white noise.

1. Introduction.
1. Analytic functions inside and outside the circle.
2. Generalized random processes in continuous time.
3. Off-white noises.