Optimum non-iterative turbo-decoding

Abstract

By observing the structure of the decoder's trellis a new, non-iterative turbo-decoder based on a super-trellis structure is proposed, which exhibits the same decoding complexity as a conventional convolutional decoder possessing an identical number of trellis states. For the investigated half-rate, memory-length two code the proposed algorithm requires about 0.5 dB lower Gaussian channel signal-to-noise ratio (SNR) than the maximum a posteriori (MAP) algorithm using 16 iterations.