The main objective of this dissertation is to find a design method for block-data translation codes with spectral nulls at DC and at the Nyquist frequency (NF).
First, we investigate the codeword distribution in the (CDS, CAS) domain, where CDS (codeword digital sum) and CAS (codeword alternate sum) of a codeword are associated with its spectral components at DC and at NF, reapectively. Based on the codeword distribution we present a design method for run-lengh-limited (RLL) block-data translation codes with two nulls at DC and at NF. We also propose a design technique for DC-and NF-free block codes with small minimum-path multiplicity, where Euclidean distances between codeword vectors are considered. The codeword distribution in the (CDS, CAS) domain, combined with the interleaving technique of DC-free codes, is applied to designing the DC-and NF-free codes of large number of levels and small block length.
Next, the performances of new codes based on the proposed method are evaluated in terms of data rate losses, bit error rates, and coding efficiencies. The results show that the (CDS,CAS)-based codes have small data rate losses for moderate numbers of levels and block lengths. It is found that the newly designed DN4B4T code outperforms the previous variants of modified duobinary code (MDB), such as TIB5 and VMDB5, in decision delays and in bit error rates. A procedure to obtain the upper bounds in coding efficiencies for DC-ant NF-free sequences is presented, and the efficiencies of realized codes are compared with theoretical bounds.
Finally, we present two applications in digital transmission systems, such as digital subscriber lines and multi-level QAM systems. It is shown that the performances of the new ID3B2Q code are comparable to other codes for subscriber lines. Spectral nulls at DC and NF of the new ID5B20 code are shown to be useful for the carrier and clock revovery of pilot-aided 64-QAM systems.