Convolutional codes from ''optimal'' linear block codes

Both non-recursive and recursive (i.e. turbo codes) convolutional codes (CCs) require careful choicer of generators to obtain coder with large minimum distance and ultimately low bit error rate for a given channel characteristic. In this paper, we first give a constructive method for building a generator matrix of either type of convolutional code based on certain low rate high distance linear block coder. For non-recursive CCs, these linear block codes have optimal dimension given block length n and minimum distance d if 2d > n. For recursive CCs, the linear block codes are constructed to appear random as determined by certain randomness tests.