The mammalian cochlea is a remarkable organ that is able to provide up to 60dB
amplification of low amplitude sound with sharp tuning. It has been proposed that
in order qualitatively to explain experimental data, models of the basilar
membrane impedance must include an exponential term that represents a
time-delayed feedback. There are also models that include, e.g., a spatial
feed-forward mechanism, whose solution is often approximated by replacing the
feed-forward term by an exponential term that yields similar qualitatively
accurate results. This suggests a mathematical equivalence between time
delay and the spatial feed-forward models. Using a WKB approximation to
compare numerical steady-state solutions, we show that there is no such simple
equivalence. An investigation of the steady-state outputs shows that both
models can display sharp tuning, but that the time-delay model requires
negative damping for such an effect to occur. Conversely, the feed-forward
model provides the most promising results with small positive damping.
These results are extended by a careful stability analysis of both models.
Here it is shown that whereas a small time delay can stabilize an unstable
transmission-line model (with negative damping), that the feed-forward
model is stable when the damping is positive. The techniques developed in
the paper are directed towards a more comprehensive analysis of nonlinear
models.