Investigation of blind CFO estimation for GFDM system using universal software radio peripheral: theory, simulations and experiments

Generalised frequency division multiplexing (GFDM) utilises a circular pulse shaping mechanism to hinder out-of-band emissions and for achieving low latency. However, fallacious synchronisation of this multicarrier technique would lead to severe inter carrier interference and inter symbol interference. Conventional training sequence-based methods alleviate the effect of time and frequency misalignments by using the concept of data repetition in the pilot preamble, which recedes data efficacy. This study enlightens various aspects in real-time implementation of blind GFDM signal transceiver using oversampling for carrier frequency offset (CFO) correction in indoor environments. It is observed that the phase shift between neighbouring sample points in an oversampled GFDM symbol is constant throughout the symbol duration. This phase shift is found to be unassociated with either subcarrier or subsymbol indices. Using this property, the maximum likelihood estimate of the CFO is deduced and is analysed in various channel environments. The Cramer–Rao lower bound for the oversampling-based GFDM signal model is derived and the authors prove through simulation that the proposed algorithm results in a little error floor. The results obtained in simulations have been validated with real-time test bed built with universal software radio peripheral 2953R as hardware and LabVIEW as software.