An improved interleaver design technique for parallel concatenated convolutional codes

This paper is aimed at the problem of designing optimized interleavers for parallel concatenated convolutional codes (PCCC) that satisfy several requirements simultaneously: 1) designing interleavers tailored to the constituent codes of the PCCC; 2) improving the distance spectra of the resulting turbo codes which dominate their asymptotic performance; and 3) constructing optimized interleavers recursively so that they are implicitly prunable. Two more modifications of a previously developed iterative interleaver growth algorithm (IGA) of polynomial complexity [F. Daneshragan et al., Sept. 1999] are presented to improve the performance of the optimized interleavers at a reduced complexity: 1) a growing window is used to trap error patterns of proper length in order to form the cost function; and 2) we employ error feedback to further improve the distance spectrum, of the optimized codes and to reduce complexity. The optimization is achieved via constrained minimization of a cost function closely related to the asymptotic bit error rate (BER) or frame error rate (FER) of the codes.