Optimization of SNR estimation in decibel domain over communication channels

A novel signal estimator is proposed to estimate the signal value in the decibel (dB) domain. The basic idea of the proposed estimator is to combine the averages of the linear domain and log domain with certain weights on averaging of each domain. In this paper, we show that the proposed method can cancel the bias resulting from the linear/log domain estimators alone and we obtain the analytical forms for the optimum combination weights for static and single path fading cases. We apply the proposed method to average signal-to-noise ratio (SNR) estimation in WCDMA systems where samples are no longer i.i.d. In that case, while the optimum estimator does not work well, the proposed method is very robust to the channel models in terms of number of multipath and different Doppler spreads: the proposed method provides a much larger linearity range of average SNR compared with the most widely used sample mean estimator.