On the Joint Adaptation of Memory-Based Nonlinear Adaptive Filters Used in Echo Cancellation
Echo cancellation is one of the key challenges in modern full-duplex digital data transmission systems. To achieve the necessary echo loss, not only a large amount of computing power is required but also highly linear data converters. In this paper, we present a concept which reduces the dynamic range of the A/D converter in the receive path by substracting a coarse analog echo estimate from the received signal. Such a coarse estimate is built with the help of the most significant bits of a conventional digital adaptive echo canceler (AEC) and a low-cost D/A converter. To compensate for its nonlinearities, a memory-based nonlinear adaptive filter (MBAF) is used. Unfortunately, this MBAF interferes with the AEC. Using simplified models of both adaptive filters, it is possible to analyze the joint adaptation behaviour. It is shown that the joint adaptation of two adaptive filters is stable in a real system such as an HDSL modem. It adapts very slowly but is still capable of tracking small variations of the system.
KeywordsMinimum Mean Square Error Adaptive Filter Error Surface Error Power Adaptive Gain
Unable to display preview. Download preview PDF.
- 1.D. Mueller. Hybride Echokompensation mit Anwendung in der digitalen Datenübertragung über Kupferleitungen, volume 1 of Series in Adaptive Signal Processing. Härtung Gorre Verlag Konstanz, 1997.Google Scholar
- 5.K. Yamazaki, S. Aly, and D. Falconer. Convergence behaviour of a jointly adaptive transversal and memory-based echo canceller. IEE Proceedings-F, 138 (4): 361–370, August 1991.Google Scholar
- 6.D. Mueller and A. Kaelin. A hybrid HDSL echo canceler. In Proc. IEEE Int. Symp. on Circuits and Systems, Seattle, pages 1078–1081, 1995.Google Scholar