Abstract: A method and corresponding apparatus for managing voice quality by optimizing the amount of attenuation placed in the echo return path based on the conditions observed during each telephone call. At the beginning of a telephone call, a pre-configured initial amount of attenuation is placed in the echo path of a communications connection that includes an echo canceller. After the call has begun, ERL is measured at a point in the communications connection that includes the attenuation. If the measured ERL is higher than necessary for the echo canceller to accurately perform double-talk detection, the attenuation in the echo path is slowly reduced from the initial amount. As the attenuation is reduced, the measured ERL reduces accordingly. When the measured ERL reduces to a minimum value required to perform double-talk detection, the reduction in attenuation is stopped.