Block codes for statistical synchronization

The performance of a class of block codes with unbounded synchronization delay, though finite average synchronization delay, is analyzed. Basically the synchronizer inspects the code symbol stream for the first occurrence of one of a set of sequences that appear in only one timing position relative to true sync. The receiver can be implemented easily with shift registers and associated logic. The probability that the sync process will exceed any given number of code symbols is investigated and bounds on this probability are established. The average sync delay to the first occurrence of a synchronizing sequence is determined and the optimal encoding procedure for a memoryless message source is presented. If sync is established by observing an -tuple, being the codeword length, the optimal structure of the synchronizing -tuple is found and the associated dictionary size is specified.