Realizing Data driven and Hampel preprocessor based Adaptive filtering on a Software Defined Radio testbed: A USRP case Study

Abstract

Adaptive filtering is a critical component of modern Direct Sequence Code Division Multiple Access (DS-CDMA) systems that operate in a highly dynamic environment. Adaptive filtering algorithms such as Sample Matrix Inversion (SMI) have been widely studied for Multiple Access Interference (MAI) suppression but is infamous for its strenuous matrix operations. The limited processing time and resources call for Auxiliary Vector (AV) filtering, a computationally efficient alternative that has been studied well in literature. Due to the extreme challenges associated with realization of these techniques on hardware, most of the work has been limited to simulation based analysis. Corroborating the simulation results through hardware implementation is necessary before these filtering techniques can be adopted by different tactical and commercial radios. Therefore, the objective of this work is to surpass the hurdles of implementation and demonstrate the significance of these adaptive receivers on an actual radio framework. Accordingly, we analyze the computationally exhaustive SMI against the mathematically efficient AV to examine the tradeoffs. In our analysis, the supervised and blind J-divergence rules showed remarkable performance at good signal strength scenario while cross-validated minimum output variance rule outperformed at low signal strength cases. Blind J-divergence rule proved to be a better choice for DS-CDMA systems with limited data record and binary phase shift keying modulation (or its variants). Additionally, we also examine the improvement achieved by introducing the Hampel preprocessor for both AV and SMI receivers.