Minimum selection GSC and adaptive low-power rake combining scheme

Abstract

In this paper, we investigate a new generalized selection combining (GSC) technique and an adaptive rake combining scheme to save the power consumption of mobile rake receivers for wideband CDMA systems. The new GSC technique called minimum selection GSC (MS-GSC) selects a minimum number of rake fingers, while maintaining the combined SNR larger than a given threshold. The proposed adaptive rake combining scheme dynamically adjusts the threshold value to maintain the desired BER. Therefore, our rake combining scheme can be integrated into any GSC, which select fingers based on a given threshold value. The proposed MS-GSC shows a low standard deviation in bit error statistics and is advantageous in practical implementation. Our experiments indicate that the proposed adaptive combining scheme reduces the power consumption of a mobile rake receiver up to 67.8 % by turning off unselected rake fingers.

Generalized Selection Combining with Log-Likelihood Ratio Threshold Test per Path for Rake Reception in Ultra-Wideband Communications

“…Since path gains are estimated separately and paths with smaller delays are generally stronger [9], the path search process starts with the path arriving with the smallest delay and proceeds to later arriving paths one by one. Once L c paths have been selected, the path search process as well as the channel estimation process stops.…”

“…According to (9), the estimated power of each path is biased by ( ) l e 2 σ (l = 0, …, N r -1). As a result, weak paths that contribute little energy to the combiner are most susceptible to estimation errors, which can degrade the receiver detection performance [4].…”

“…In [7], [8], the adaptive threshold for path selection takes into account the noise level, which is obtained by non-coherently averaging over multiple instantaneous power delay profiles measured using pilot symbols. In [9], a minimum number of paths are selected as long as the combined SNR maintains the bit error rate (BER) below a certain level, which relies on periodically measuring the BER to provide feedback information. In [10], performance comparable to MRC is obtained by dividing all received paths into two complementary groups and choosing the group with the larger magnitude of log-likelihood ratio (LLR).…”

“…However, the above schemes do not save hardware, because the number of selected paths may change with time as well as location in the range of 1 to N r (i.e. the number of resolvable multipath components) [9]. For a selective Rake receiver with L c fingers, where L c is fixed and less than N r , the SNR-GSC [12] selects the L c paths with the largest path SNR, and the LLR-GSC [11] selects the L c paths with the largest magnitudes of LLR.…”

“…The use of LLR in path selection and data detection is motivated by the fact that the hard decision based on LLR is optimal in the sense of minimizing BER [11]. Since channel taps are estimated separately [4] and paths with smaller delays are generally stronger [9], the path search for LLR-T-GSC starts with the path arriving with the smallest delay and proceeds to later arriving paths one by one. A path is selected if its estimated magnitude of LLR exceeds the threshold.…”