Symbol Timing Recovery Algorithms and Their Evaluation for Burst Communication

Abstract

This paper reviews and presents the simulationresults on the performance of four alternative symboltiming offset estimation algorithms for burst modecommunication. Two basically different approaches are taken to the problem. The first approach isbased on the concept of maximum likelihood estimation,of which two basic varieties are derived and evaluated.These are referred to as the data-driven and the non-data-driven versions. In thedata-driven mode, either the decision symbols from theoutput of the demodulator or a number of known referencesymbols may be used to drive the algorithm. The maximum likelihood algorithms generate an error signalthat can be used in a closed-loop control system toadjust the phase of the sampling clock to the optimumsymbol sampling instant. The second approach uses the basic signal processing technique of tonefiltering using effectively the discrete Fouriertransform. The tone filtering approach results in thedirect estimation of the optimum symbol sampling phase and hence is more of an open-loop approach. Thesimulation results indicate that excellent performanceis achieved with this kind of open-loop approach, evenat low signal to noise ratios and in the presence of some frequency offset as long as the symboltiming offset does not exceed 90% from its optimumpoint. It is also shown that by choosing the propersampling rate, this algorithm results in a simple IIR filtering implementation with one pole that isvery efficient from a processing point ofview.