Carrier frequency offset estimation in a frequency-hopping ultra-wideband OFDM system: Licenciate thesis

In this thesis the performance of selected carrier frequency offset (CFO) estimators is studied in an ultra-wideband (UWB) slow frequency-hopping orthogonal frequency division multiplexing (FH-OFDM) system. A challenge in a very high data rate FH-OFDM system is to minimize self interference such as intercarrier interference (ICI) and constellation rotation due to the CFO under several nonidealities. Various transmission techniques are first represented emphasising UWB transmission, orthogonal frequency division multiplexing (OFDM), frequency-hopping spread spectrum (FHSS) and the combination of these schemes. The synchronization requirements in wireless communications are considered next, followed by the review of the existing CFO estimation methods. After presenting the characteristics of the target system model the simulation results are shown with the appropriate analysis. First, the basic algorithms from each defined estimation category are compared in additive white Gaussian noise (AWGN) channel. The effects of nonidealities in despreading the frequency-hopped signal are considered next in a frequency-selective fading channel. The nonidealities may introduce a residual frequency-hopping timing offset (FHO) and phase offset (PHO) due to the nonideal hop-timing estimator, frequency-hopping synthesiser and frequency-selective channel. The main class of the studied CFO estimators is based on the autocorrelation of the received samples. The algorithms have two parameters to adjust for mitigating the noise. These include an averaging length and a distance between correlative samples. The simulation results indicate the kind of hop timing estimator, preamble structure, frequency-hopping synthesiser and signal-to-noise ratio (SNR) operation point that are required before the CFO estimation can be performed with the given estimators. Simulations show that Lank’s CFO estimator is suitable for the target scenario and requirements when using relatively high SNR. For the lower SNR operation points Fitz’s, Luise’s or Mengali’s algorithms give better performance at the cost of increased complexity or reduced estimation range. Furthermore, the CFO estimators are quite robust to the FHO less than half the hopping duration, which causes a slight increase in standard deviation. It is shown that this degradation can be compensated using the CFO algorithm itself and utilizing the available preamble. The estimators are, however, sensitive to the PHO caused by the possible lack of phase memory in the frequency-hopping synthesizer or frequency-selective channel, resulting in a significant bias to the CFO estimate.