Amnart Chaichoet and Chalie Charoenlarpnopparut

pp. 775 - 785

Abstract

For effective control of transmission power in WCDMA mobile systems, a good estimate of signal-to-interference ratio (SIR) is needed. Conventionally, an adaptive SIR estimator employs a moving average (MA) filter (Yoon et al., 2002) to encounter fading channel distortion. However, the resulting estimate seems to have high estimation error due to fluctuation in the channel variation. In this paper, an additional post-processing stage is proposed to improve the estimation accuracy by
reducing the variation of the estimate. Four variations of post-processing stages, namely 1) a moving average (MA) postfilter, 2) an exponential moving average (EMA) post-filter, 3) an IIR post-filter and 4) least-mean-squared (LMS) adaptive post-filter, are proposed and their optimal performance in terms of root-mean-square error (RMSE) are then compared by simulation. The results show the best comparable performance when the MA and LMS post-filter are used. However, the
MA post-filter requires a lookup table of filter order for optimal performance at different channel conditions, while the LMS post-filter can be used conveniently without a lookup table.