We investigate stochastic resonance in a simple level-crossing detector at
the base frequency, as well as the higher harmonics, of a periodic stimulus
. Using quadratic approximations, we observe a stochastic resonance phenome
non with two maxims tin the presence of white Gaussian background noise) in
the first harmonic at the output power spectral density of the system. Tha
t means that there is an optimal strength of the input noise, where the sig
nal transfer is the most linear, as the first harmonic is concerned. This u
nique feature of this system cannot be found at the fundamental harmonic in
the output. The theoretical results are confirmed by numerical simulations
. This system is a good candidate for use in simple models of neurons, and
other (nonlinear) signal detection systems. (C) 1999 Published by Elsevier
Science B.V.