Adaptive DFE with Regularization for Indoor Wireless Data Communications

Recursive least-squares adaptive decision-feedback equalizers (DFE) can be classified into the direct adaptation (DA) and channel-estimation-based adaptation (CEBA) approaches. CEBA provides better performance than DA when the channel length is known exactly. However, for wireless channels, the channel length is a random unknown quantity. In this paper, we evaluate the performance of the DA and CEBA DFE’s without assuming knowledge of the channel length. We find that the DA DFE is more robust with respect to delay-spread variations than the CEBA DFE. We also use a regularization technique for the fractionally-spaced CEBA DFE, which achieves a gain as large as 6dB. Since we did not assume knowledge of the channel length, these results apply for wireless communication systems.