A study on finite wordlength effects in FIR digital echo cancellers using learning identification algorithm
Hiroshi Yasukawa, Youjirou Ara · Electronics and Communications in Japan (Part III Fundamental Electronic Science) · 1990
Abstract Finite wordlength effects on the steady‐state characteristics of finite impulse response adaptive digital echo cancellers are considered. Focusing on an FIR‐type echo canceller based on the learning identification algorithm, the so‐called normalized least mean square algorithm was investigated. Both fixed point and floating point arithmetic systems are discussed. The equivalent noise models for both systems are given, and the output errors caused by quantization and rounding noises were analyzed to allow explicit representation of echo canceller parameters (particularly wordlength) when the input signal is white Gaussian. Further, the relationship between echo cancellation and circuit parameters such as number of filter tap, loop‐gain coefficient, etc., the optimal parameter values to minimize output error are also discussed. Computer simulation results were compared with the analytical values and found to be in good agreement. The proposed method is valid for obtaining the performance of echo cancellers.