Tracking capabilities of the LMS adaptive filter in the presence of gain variations

The tracking capabilities of the least mean square (LMS) adaptive filter when it is used to identify a nonstationary channel suffering from gain variations are studied. An analytical solution for the steady-state lag error is obtained, and its validity is verified by means of simulations. Depending on the magnitude and bandwidth of the variations, the total steady-state error is found either to increase monotically as a function of the adaptation parameter mu or to exhibit a global minimum for a finite, nonzero value of mu .<
>