The recursive-approaching signal tilter (RASF) is a newly developed fi
ltering technique that combines many advantages of linear, nonlinear,
and adaptive filters. It passes step functions without altering them a
nd removes many types of noise, such as Gaussian and Laplacian distrib
uted noise. When applied to VSP data processing, the RASF emphasizes t
hose abrupt discontinuities that originate or terminate at discrete de
pth points and effectively accomplishes the separation of upgoing and
downgoing wave modes. The RASF may be transformed into a desired filte
r by simply changing a parameter to achieve the maximum usefulness of
VSP field data. In the tests with the synthetic VSP modeling data corr
upted by white Gaussian noise and real VSP data, the RASF compares fav
orably to f-k velocity and median filtering methods in removing noise,
preserving step functions, and computational simplicity.